6.1. Ancistrus Cf. Triradiatus

Universiteit Gent Evolutionary

Faculteit Wetenschappen Morphology of Vakgroep Biologie Vertebrates

Ontogeny and functional morphology of a highly specialized trophic apparatus: A case study of neotropical suckermouth armoured catfishes (Loricariidae, Siluriformes)

I – Text

Tom Geerinckx

Thesis submitted to obtain the degree of Academiejaar 2006-2007 Doctor in Sciences (Biology)

Proefschrift voorgedragen tot het bekomen Rector: Prof. Dr. Paul Van Cauwenberge van de graad van Doctor in de Decaan: Prof. Dr. Herwig Dejonghe Wetenschappen (Biologie) Promotor: Prof. Dr. Dominique Adriaens

EXAMENCOMMISSIE

Prof. Dr. Wim Vyverman (Universiteit Gent, voorzitter) Prof. Dr. Dominique Adriaens (Universiteit Gent, promotor)* Dr. John P. Friel (Cornell University, USA)* Dr. Anthony Herrel (Universiteit Antwerpen)* Prof. Dr. Ann Huysseune (Universiteit Gent) Prof. Dr. Walter Verraes (Universiteit Gent) Prof. Dr. Wim Van den Broeck (Universiteit Gent)

* leescommisie

DANKWOORD

Van de mensen die ik wil bedanken voor hun hulp, steun of welke bijdrage dan ook die deze doctoraatsthesis positief beïnvloed heeft, komt mijn promotor, Prof. Dr. Dominique Adriaens, vanzelfsprekend op de eerste plaats. Na mijn licentiaatsthesis, toen hij mijn begeleider was, en na een jaartje andere leuke bezigheden op de universiteit, kwam ik met veel plezier en vertrouwen weer ‘onder zijn hoede’. Al wie Dominique reeds als begeleider of promotor gehad heeft zal het kunnen beamen: proffen met zo’n kennis van zaken, en mensen met zulke gedrevenheid, vind je zelden. Uit zijn ervaring en enthousiasme heb ik gretig en dankbaar geput.

Het fijne gezelschap van vrienden en mede-doctoraatsstudenten Stijn Devaere, Natalie De Schepper, Frank Huysentruyt, Paul Van Daele en Soheil Eagderi heeft de vier jaren die ik tijdens het werk op “‘t labo” heb doorgebracht aangenaam gekleurd. De eerste drie in dit lijstje hebben bovendien de twijfelachtige eer gehad mijn lokaalgenoten te zijn. Iets wat vele interessante, leuke, en bij momenten zelfs gezellige taferelen opleverde. Niet minder cruciaal voor dit doctoraat waren het discussiëren en palaveren met Frank over alle mogelijke raakpunten van onze beide projecten, het gretig afnemen van Franks en Natalie’s kennis in de 3D-reconstructie- programma’s, en het met z’n allen zoeken naar dat ene Engelse woord.

Aan het werk en de inzet van Marleen Brunain heb ik zeer veel gehad. Ze heeft duizenden en duizenden coupes gemaakt, zowel voor mezelf als voor mijn thesisstudenten. Ik kan haar niet genoeg bedanken voor haar inzet en betrokkenheid in het wetenschappelijk werk. Danku!!! Barbara De Kegel sneed met plezier o.m. de paraffine-coupes van alle rare weefsels die men zoal in katvissen terugvindt. Praktisch werk, praktische vragen, zonder hen was ik nu nog steeds op zoek geweest naar een of ander potteke of protocol, of het knopje om de histokinette te doen stoppen met piepen. Hetgeen ik nu kan qua histologisch werk in de praktijk, hebben Barbara en Marleen me allemaal aangeleerd.

De gelegenheid om de high-speed video-apparatuur van het Labo voor Functionele Morfologie (UA) te gebruiken, was een zegen voor dit onderzoek. Met uiteraard zeer veel dank aan Prof. Dr. Peter Aerts en Dr. Anthony Herrel. Het biomechanisch onderzoek, weliswaar geen deel van dit doctoraat zelf, had niet kunnen gebeuren zonder het enthousiasme van Anthony. De vele uren gespendeerd aan het fanatieke vis-kijken krijgen nog een staartje, dat hoop ik zo! Mijn dank gaat ook uit naar de assistentie door Jeannine Fret, en de hulp van Sam Van Wassenbergh.

Rita Van Driessche en, gedurende het laatste jaar, Marjolein Couvreur, hebben prachtig werk geleverd tijdens de regelmatige fotosessies aan de scanning electronen microscoop. Verschillende ‘details’, door mij over het hoofd gezien, maar door Rita’s scherpe blik ontdekt, zijn nadien zeer waardevol en relevant gebleken. Dankjewel!

Prof. Dr. Walter Verraes en wijlen Prof. Dr. Guy Teugels verdienen een bijzonder woord van dank. Het zijn immers zij die me voor het eerst hebben laten kennis maken met de toch wel boeiende en vooral gevarieerde wereld van de ichthyologie.

Gedurende deze vier jaren kregen we het afwisselende gezelschap van thesisstudenten die niet vies waren van een visje (of muisje) meer of minder. Vele verschillende persoonlijkheden, wiens stimuli op het onderzoek en het labo, en ja soms zelfs zeer geestige kantjes ik niet vergeten ben: o.m. Kim Nijs, Dré Maes, Kevin Lambeets, Gunther Jansen, Natalie Dirckx, Tine Debrauwere, Yves Verhaegen, Joni De Puysseleir, Heleen Leysen, Celine Ide, Eva Bequé, Evelien Meijfroidt, Marjolein Peys en Sara Laceur. In het bijzonder denk ik aan Kim, Dré en Joni, wiens thesis-

belevenissen ik van dichtbij heb meegemaakt. Met name het ‘roze boekje’ van Kim over Otocinclus en het al even fraaie (weliswaar niet roze) werk van Dré over Farlowella, en zo mogelijk nog meer de vele babbels over die o zo toffe Loricariidae zijn enorm constructief en stimulerend geweest voor m’n eigen werk.

De besprekingen met Prof. Dr. Ann Huysseune over, in het bijzonder, deel zeven van deze thesis (tandjes!), zijn zeer zinvol geweest. Zo zijn weer enkele misvattingen en onzekerheden de wereld (met name de mijne) uit geholpen. Dankjewel hiervoor!

Yves Verhaegen en Joris De Poorter bedank ik voor het noeste labeur dat zij leverden tijdens hun zelfstandig practicum over ontogenetische vormveranderiningen en schraapstructuren bij Loricariidae: hun resultaten en ideeën hebben volop weerklank gevonden in dit werk. Het enthousiasme en de ijver (niet gelogen!) van Annelies Genbrugge en Karen Bekaert tijdens hun zelfstandig practicum over het gruwelijke wapenarsenaal van katvissen was fenomenaal, en de resultaten idem dito; ik wens hen een mooi thesisjaar toe! (En ik vergeef het hen dat ze overgestapt zijn op de toch wel schattigere zeepaardjes en aanverwanten.) Wat dit minder lieve kantje van de Zuid-Amerikaanse katvissen betreft, heb ik overigens inspiratie en vergelijkingsmateriaal gevonden in de thesis van Diederik Maebe.

Je remercie Dr. Sonia Fisch-Muller, qui a exécuté une tâche bien importante: évaluer l’identité de l’espèce d’Ancistrus employée dans cette thèse, sans pouvoir savoir l’origine des spécimens. Ancistrus reste toujours un des genres les plus nombreux et complexes des Loricariidae.

Dit onderzoek is slechts kunnen gebeuren dankzij een doctoraatsbeurs van het Vlaams Instituut voor de Bevordering van het Wetenschappelijk-Technologisch Onderzoek in de Industrie (IWT). Financiële steun van het Fonds voor Wetenschappelijk Onderzoek Vlaanderen (FWO) is het doctoraatsproject ook zeer ten goede gekomen.

De sfeer van de omgeving waarin ik vier jaar gewerkt heb werd voor een groot deel bepaald door al de medewerkers van Prof. Dr. Dominique Adriaens, Prof. Dr. Ann Huysseune en Prof. Dr. Gaetan Borgonie, gezellig samenlevend op de derde verdieping van de Ledeganck. Wie zeker nog met naam vermeld moeten worden zijn Joachim Christiaens en Hilde Van Wynsberge, voor de hulp bij voornamelijk praktische zaken.

Een apart paragraafje wijd ik aan de ex-biologie-klasgenootjes die ik maar blijf tegen het lijf lopen: Dirk, Wouter, Maarten, Annelies, Bob, Wi(e)m, Jan, Tjörven, Liesbeth, Imke, enzovoort. Aan dit lijstje voeg ik ook zeker Catherine, Karen, Wim, David en de Gentse Universitaire Duikers toe. Waarom dit apart paragraafje? Voor de onvoorwaardelijke steun, of het onvoorwaardelijk maar goedbedoelde ontbreken daarvan, voor het lachen met vissen, het eten van vissen, het samen zoeken naar vissen in de Oosterschelde, het vergelijken van vissen met nematoden (1-0!), het plannen van zes zalige maanden in het verre Azië (nu toch wel zeer dichtbij…), het doen van vele leuke dingen die hoegenaamd niets met vissen te maken hebben.

Tenslotte gaat een warm woord van dank uit naar mijn ouders, die me toch maar mooi de kans gegeven hebben om de biologie-studie te voltooien. En een al even warm woord van dank bovendien, aan hen en aan mijn broer en zus, voor hun vriendschap, respect en begrip voor die jongen met zijn vissen.

Tom Gent, 24 november 2006

Een rode draad doorheen dit werk, soms sluimerend op de achtergrond, soms de ware drijfveer van het onderzoek, was de verwondering.

Oprechte blije verwondering over hoe energiek jong leven ontwikkelt en groeit, hoe subliem het leven zich aanpast, en wat een belevenis het is dit alles te bestuderen.

TABLE OF CONTENTS i

TABLE OF CONTENTS

PART 1 — GENERAL INTRODUCTION 1.1. General context and aims ...... 1 1.1.1. General frame...... 1 1.1.2. Introduction ...... 2 1.1.3. Aims ...... 4 1.2. The suckermouth armoured catfishes...... 7 1.2.1. Ostariophysi ...... 7 1.2.2. Siluriformes...... 7 1.2.3. Loricarioidea ...... 9 1.2.4. Loricariidae ...... 11

PART 2 — MATERIAL AND METHODS 2.1. Material...... 15 2.1.1. Choice of species...... 15 2.1.2. Material examined...... 16 2.2. Methods...... 20 2.2.1. Keeping and breeding of loricariid species...... 20 2.2.2. Live observations and high-speed filming ...... 20 2.2.3. Preparation of specimens for study ...... 21 2.2.4. Metrics...... 21 2.2.5. In toto clearing and staining...... 22 2.2.6. Dissections ...... 23 2.2.7. Serial sectioning...... 23 2.2.8. 3D-reconstructions ...... 24 2.2.9. Scanning electron microscopy ...... 25 2.3. Notes on terminologies...... 26 2.3.1. Terminology of anatomical structures...... 26 2.3.2. Use of the terms embryo, larva and juvenile...... 26 2.3.3. Use of age or size in the study of ontogeny ...... 27

PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY 3.1. Egg characteristics ...... 29 3.1.1. Introduction ...... 29 3.1.2. Brief material and methods ...... 29 3.1.3. Results ...... 29 3.1.4. Discussion ...... 30 3.2. Early life history...... 32 3.2.1. Introduction ...... 32 3.2.2. Brief material and methods ...... 33 3.2.3. Results ...... 33 3.2.4. Discussion ...... 35 ii TABLE OF CONTENTS

PART 4 — ONTOGENY OF THE SKULL 4.1. Ontogeny of the chondrocranium...... 37 4.1.1. Introduction ...... 37 4.1.2. Brief material and methods ...... 39 4.1.3. Results ...... 39 4.1.4. Discussion ...... 47 4.2. Ontogeny of the osteocranium ...... 60 4.2.1. Introduction ...... 60 4.2.2. Brief material and methods ...... 61 4.2.3. Results ...... 61 4.2.4. Discussion ...... 73

PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 5.1. Ontogeny of the jaw and maxillary barbel musculature...... 83 5.1.1. Introduction ...... 83 5.1.2. Brief material and methods ...... 85 5.1.3. Results ...... 85 5.1.4. Discussion ...... 93 5.2. Ontogeny of the intermandibular and hyoid musculature...... 104 5.2.1. Introduction ...... 104 5.2.2. Brief material and methods ...... 106 5.2.3. Results ...... 106 5.2.4. Discussion ...... 110 5.3. Ontogeny of the suspensorial and opercular musculature...... 117 5.3.1. Introduction ...... 117 5.3.2. Brief material and methods ...... 118 5.3.3. Results ...... 119 5.3.4. Discussion ...... 122

PART 6 — ADULT MORPHOLOGY 6.1. Ancistrus cf. triradiatus...... 127 6.1.1. Introduction ...... 127 6.1.2. Brief material and methods ...... 128 6.1.3. Results ...... 129 6.1.4. Discussion ...... 140 6.2. Farlowella acus ...... 148 6.2.1. Introduction ...... 148 6.2.2. Brief material and methods ...... 148 6.2.3. Results ...... 149 6.2.4. Discussion ...... 154 6.3. Otocinclus vestitus...... 156 6.3.1. Introduction ...... 156 6.3.2. Brief material and methods ...... 156 6.3.3. Results ...... 157 6.3.4. Discussion ...... 161 TABLE OF CONTENTS iii

PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

7.1. Introduction ...... 165 7.2. Brief material and methods ...... 166 7.3. Results ...... 166 7.3. Discussion ...... 170

PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS

8.1. Introduction ...... 175 8.2. Brief material and methods ...... 176 8.2. Results ...... 176 8.3. Discussion ...... 181

PART 9 — GENERAL DISCUSSION

9.1. Morphology and function in Loricariidae ...... 185 9.2. Evolution of the suckermouth apparatus...... 190 9.3. Considerations regarding loricariid ontogeny...... 198 9.4. Synopsis...... 203

PART 10 — SUMMARY & SAMENVATTING

10.1. Summary...... 207 10.2. Samenvatting ...... 213

PART 11 — REFERENCES

...... 219

PUBLICATION LIST

...... 243

PART 1

GENERAL INTRODUCTION

PART 1 — GENERAL INTRODUCTION 1

1.1. GENERAL CONTEXT AND AIMS

1.1.1. GENERAL FRAME

The research presented in this dissertation is part of the ongoing FWO project ‘Evolutionary adaptiveness for a highly specialized feeding niche: algae scraping in tropical catfishes’, led by the Evolutionary Morphology of Vertebrates1 group at the Ghent University (UGent), and in partnership and close collaboration with the Laboratory for Functional Morphology2 of the Antwerp University (UA) and the Ichthyology Department3 of the Africa Museum at Tervuren (KMMA/MRAC).

The major aspects that are treated in the general frame of this project are:

- The ontogeny of Corydoras aeneus, a representative of the Callichthyidae (Frank Huysentruyt). Callichthyidae are neotropical catfishes related to the suckermouth armoured catfishes, but lacking the suckermouth and scraping feeding apparatus.

- The ontogeny of Ancistrus cf. triradiatus, a representative of the Loricariidae (Tom Geerinckx).

- The functional morphology of Corydoras aeneus and related species (Frank Huysentruyt).

- The functional morphology of Ancistrus cf. triradiatus and other loricariid species (Tom Geerinckx, Kim Nijs, Dré Maes).

- A biomechanical study using X-rays and EMG of Pterygoplichthys lituratus (Loricariidae) (Tom Geerinckx, Anthony Herrel).

- A comparative study of non-algae scraping (Synodontis sp.) and algae scraping (Atopochilus savorgnani) African Mochokidae (Celine Ide, Joni De Puysseleir, Frank Huysentruyt, Tom Geerinckx), because of the convergent evolution towards algae scraping and mouth suction that has occurred in Mochokidae and Loricariodea.

1 http://www.fun-morph.ugent.be/ 2 http://webhost.ua.ac.be/funmorph/ 3 http://www.africamuseum.be/research/zoology/vertebrates/research/zoology/vertebrates/ichthyol 2 PART 1 — GENERAL INTRODUCTION

1.1.2. INTRODUCTION

This doctoral thesis treats the morphology and ontogeny of the cranial structures of Ancistrus cf. triradiatus, a representative of the neotropical family Loricariidae or suckermouth armoured catfishes. It is the most speciose family within the order Siluriformes (catfishes), containing over 700 species (see paragraph 1.2.4). These fishes are characterized by the peculiar trophic niche they occupy: many loricariids scrape epilithic or epiphytic algae and other food items from submerged substrates. This specialized feeding type is possible thanks to the remarkably formed, ventrally placed suckermouth that allows the animals to attach to a surface while scraping and eating the food attached to it. In spite of this highly specialized feeding apparatus (or thanks to it), a wide diversity of both body size and shape exists, and many loricariids actually feed on a broad range of food. As such, Loricariidae are the most specialized and succesful catfish family within the superfamily Loricarioidea. More basal families within this superfamily mostly display a more general (non-specialized) feeding mode, with a typical siluriform feeding apparatus suitable for finding and processing insects and other food items that abound in the water column or on or in the bottom.

The family-level phylogenetic relationships within the Loricarioidea are relatively well known (de Pinna, 1998; see paragraph 1.2.3). Some knowledge exists on the trophic, evolutionary trend in the group, but detailed studies dealing with the morphology of the feeding apparatus are few and often fragmentary. Studies including cranial morphology of loricariids were also restricted to adults (e.g., Howes, 1983a; Schaefer, 1987, 1988, 1997; Schaefer & Lauder, 1986, 1996). Especially the paper of Schaefer & Lauder (1986) has provided insight, though, in the evolutionary transformation of the suckermouth, jaw apparatus and associated muscle complex, from a basal to a more specialized configuration. Their research yielded a list of evolutionary transformations that characterized the lineage leading towards the loricariids. This list includes couplings and decouplings of structures, and the acquisition of new structures4. For an overview and reappraisal of these proposed transformations, I refer to chapter 9.2 in the general discussion of this dissertation. A verification of the findings of Schaefer & Lauder (1986) is integrated in this work. Still, a comprehensive study on, e.g., loricariids, including all muscles, skeletal interactions and tissue characteristics, is lacking. Functional hypotheses based on preserved specimens can be misleading, as preservation alters tissue characteristics and may well

4 A decoupling of biological components refers to the unlinking of developmental pathways, tightly linked functions, aspects of stereotyped behaviour patterns, mechanical associations of bones, ligaments and muscles, or the reduction of a high genetic correlation (Lauder et al., 1989). The recognition of ‘new structures’ (sometimes termed novelties) varies among literature, but in many cases so-called new structures are actually the result of profound decoupling events. The evolutionary appearance of, e.g., a sesamoid bone in a tendon or a ligament, can also be considered the evolution of a new structure. PART 1 — GENERAL INTRODUCTION 3 modify the observed effect when manipulating muscles. The complete mechanism and kinetics of respiration in loricariids is not understood [a limited study has been done by Vandewalle et al. (1986)]. Furthermore, the complex array of anatomical structures involved in respiration (function 1) also operates the feeding mechanism (function 2), and must allow the animal to attach to a substrate simultaneously (function 3). Many loricariids inhabit fast- flowing river systems. In anticipation of biomechanical studies (see paragraph 1.1.1), functional hypotheses can be formulated, relying on sufficient knowledge of the nature and interactions of the components of the musculo-skeletal system. However, function analysis has to be performed before any hypothesis can be tested properly. It should be noted that, first of all, a detailed morphological knowledge is needed to understand and support biomechanical and electromyographical data.

As stated above, previous studies have been based on adult specimens only. A result is that no insight at all exists on the development of the feeding and respiratory mechanism in loricariids (and most loricarioid families), nor on the degree of presence and functionality of structures during early embryonic stages. Nevertheless, study of ontogeny is an obvious necessity, if one wants to understand the evolution and integrated complexity of a feeding and respiratory mechanism (Lauder et al., 1989). Except for a few aspects in some species of Trichomycteridae, that have a different feeding habit (Arratia, 1990), and the Callichthyidae (also a basal family) (Hoedeman, 1960a, b, c; Howes & Teugels, 1989), no taxon within the Loricarioidea has been studied from a morphological-ontogenetic point of view (Arratia, 2003). Those studies incorporating loricariid ontogeny mention mere external habitus (no internal morphology) and growth of early life stages (e.g., Riehl & Patzner, 1991; Nakatani et al., 2001). This is a main reason why several fundamental questions concerning the evolution towards the scraping feeding mode of loricariids have remained unanswered. For example, no answer exists on the question how the shift from a general trophic type (i.e., the feeding on small invertebrates etc.) towards a scraping type might have occurred (both ontogenetically and evolutionarily) (Adriaens, 2003). Such questions are highly relevant, in view of the successful loricariid radiation and the vast diversity in shape and size of jaws and teeth (Schaefer & Lauder, 1986; Muller & Weber, 1992; Delariva & Agostinho, 2001). Because of several characteristics of the family, Schaefer & Stewart (1993) compared the Loricariidae to the Cichlidae of the great African lakes, a family that has been the focus of extensive and profound anatomical, ecomorphological, ecological and genetic studies (see also paragraph 1.2.4). Except for providing material that can be used in a comparative evolutionary perspective, ontogeny forms an essential part in the field of ecomorphology. Fishes have provided fertile 4 PART 1 — GENERAL INTRODUCTION grounds for examination of the relationship between form and ecological role (Motta et al., 1995), but ecological and distributional studies are often based on adults only, and the possibility that ecological, and thus evolutionary, success may be determined by events in early development is overlooked (Orton, 1955). Trophic and other ecological needs of free- living embryos, larvae and juveniles can differ substantially from those of adults. In addition, the developing feeding apparatus must be functional during early stages from the moment exogenous feeding starts, and must meet the varying trophic needs. Attachment onto substrates probably is important as well, already soon after hatching (Riehl & Patzner, 1991; Sabaj et al., 1999). Thus, in loricariids it appears that the respiratory mechanism is functional even earlier than the feeding mechanism. The importance of the interaction between the Bauplan of an organism at any moment during life, and the environment (natural selection) is widely recognized (Orton, 1955; Lauder et al., 1989). Without any doubt an organism should not only be considered in the adult, ‘fully developed’ stage, but also (and probably even more) in the young, ‘unfinished’ stage, in which the morphological design is not yet complete. It is this stage that is the most critical one, often facing other functional needs and restrictions than those that ‘mold’ the adult Bauplan. In this view, the present thesis might contribute to the insight of processes that are at the foundation of radiations, specializations, and the strikingly high diversity of certain animal taxa.

The study of ontogeny may yield other, more direct results. Many synapomorphies that identify loricarioid taxa are based on the development, reduction, loss or fusion of bones, muscles and ligaments, without any actual proof of these processes, that could be demonstrated by ontogenetic research. Ontogeny can equally well clarify morphological uncertainties, like the presence of an interopercle (Regan, 1904; Howes, 1983a; Schaefer, 1988), or the homology of jaw muscle divisions (Howes, 1983a; Diogo, 2005). Finally, the existing phylogeny, only based on characters of in- and outgroups (outgroup criterion), can be tested with ontogenetic evidence of different taxa (ontogeny criterion).

1.1.3. AIMS

The problems as formulated in the previous paragraph are at the basis of the general aim of this doctoral dissertation. The principal hypothesis of this dissertation is that the ontogenetic-evolutionary pattern of the cranial system of loricariids is strongly dominated by the formation of a scraping suckermouth. This is most likely as an adaption to the peculiar niche of scraping food from submerged substrates, while being able to attach to these substrates, and breathe while being attached with the suckermouth. I will try to define and characterize these specializations (i.e., evolutionary transformations, see above), by PART 1 — GENERAL INTRODUCTION 5 examining the loricariid head morphology from functional and evolutionary points of view. I will assess which evolutionary transformations can truly be considered as key innovations in the loricariid lineage5. By studying the early life history stages I will try to verify how this peculiar morphology arises ontogenetically, and whether and when the cranial elements involved in the acts of feeding, breathing and attaching, transform and become functional during the earliest free-living stages. The study of ontogeny, and the comparison with related (siluriform) families may provide insight on the evolutionary origin of the typical structure of the loricariid head.

The various aspects of this main aim are more concretely formulated as follows:

- Description of eggs and early life history stages in the hypostomine loricariid Ancistrus cf. triradiatus (from the moment movements of the embryos are observed), including a study on the ontogenetic allometries and head transformations and shape changes during early life history (Part 3). - An examination of the ontogeny of the skeletal elements of the skull of A. cf. triradiatus, focussing on both the early chondrocranium and later osteocranium (Part 4). Identification and homology of ‘problematic’ bones. - An examination of the ontogeny of the cranial muscle groups of A. cf. triradiatus, with a discussion on homologies, functional hypotheses, and (especially for the jaw musculature) evolutionary origin (Part 5). - A detailed study of the adult skeleton, musculature and ligaments in A. cf. triradiatus, combined with a limited kinematic analysis of the suction, breathing and feeding apparatus. First conclusions and hypotheses concerning functional morphology (Part 6, Chapter 6.1). - A comparative study on the cranial morphology of the loricariine loricariid Farlowella acus and the hypoptopomatine loricariid Otocinclus vestitus. A preliminary assessment of morphological-functional diversity within the Loricariidae (Part 6, Chapters 6.2 and 6.3). - Research on the morphology and development of teeth and epidermal brushes or unculi in several loricariid species (Part 7).

5 An innovation is defined as a phenotypic change that has important consequences for the dynamics of evolution. Innovations that have a major impact on adaptive radiation are called key innovations (Müller & Wagner, 2003). Well known examples of key innovations in vertebrates are the versatility of the cichlid pharyngeal jaw apparatus, and the anterior separation of the left and right lower jaw bones and enhanced jaw mobility in snakes. Innovations that introduce new (de novo) entities, units, or elements into phenotypic organization are sometimes called novelties (e.g., eyes, insect wings; Müller & Wagner, 2003), although novelty is often used as synonym of innovation (Lauder et al., 1989). 6 PART 1 — GENERAL INTRODUCTION

- An analysis of morphology and function of the cheek-spine apparatus of A. cf. triradiatus, and an assessment of the impact of this apparatus on the cranial morphology (Part 8). - A synthesis of acquired morphological and ontogenetic insights, reappraisal of the major transformations during ontogeny, and relations between morphology and function. A brief discussion on the developmental and evolutionary origin of the loricariid head design. Finally, a synopsis of the dissertation (Part 9).

PART 1 — GENERAL INTRODUCTION 7

1.2. THE SUCKERMOUTH ARMOURED CATFISHES

As an introduction and background to the taxa discussed in this thesis, an overview is given of the systematic situation of the Loricariidae and related loricarioid catfish families, as well as the position of the catfishes within the teleostean ostariophysan clade.

1.2.1. OSTARIOPHYSI

A recent classification of high-level teleostean taxa is given by Arratia (1997) (Fig. 1). Ostariophysi, sistergroup to the Clupeomorpha, are worldwide present, and include 25% of all known teleosts and 64% of all known freshwater fishes (Nelson, 1994). Five orders are recognized (Nelson, 1994; Fink & Fink, 1996; Teugels, 1996; Arratia, 1997): Gonorhynchiformes, Characiformes, Gymnotiformes, Cypriniformes and Siluriformes. The status of the newly proposed anotophyse fossil order Sorbininardiformes (Taverne, 1999) is uncertain (Diogo 2005). Recently, molecular studies substantially contribute to the knowledge of ostariophysan relationships (e.g., Lavoué et al., 2005). Many authors have elaborated on the features that group and characterize the Ostariophysi, distinguish them from other teleost taxa, or support their relationships (e.g., Regan, 1911a, b; Roberts, 1973; Greenwood et al., 1979; Fink & Fink, 1981, 1996; Gayet, 1986, 1993). It is not the aim of this chapter to list all ostariophysan characteristics; only a few are mentioned here. Ostariophysi are provided with a Weberian apparatus, a mechanism involving the anterior, specialized vertebrae and the swimbladder, greatly enhancing the sound perception of the fishes (Regan, 1911a; Alexander, 1964; Gayet & Chardon, 1987). This might be one of the factors promoting the success and diversity of the group. Ostariophysi all display a fright reaction in reaction to alarm substances released by injured skin of other individuals; a wide range of reactions exists (Pfeiffer, 1977). Unicellular horny projections or unculi are present in most Ostariophysi except Gymnotiformes, but are unknown in non-ostariophysan teleosts (Arratia & Huaquín, 1995). Among loricarioids, these structures are found in Nematogenyidae (Arratia & Huaquín, 1995), Trichomycteridae (Arratia & Huaquín, 1995) and Loricariidae (Ono, 1980).

1.2.2. SILURIFORMES

Siluriformes or catfishes are the only cosmopolitan ostariophysan order (Nelson, 1994). They have been found as fossils on Antarctica (Grande & Eastman, 1986). A recent count 8 PART 1 — GENERAL INTRODUCTION includes 3016 extant species in 458 genera (C. Ferraris, pers. commun.). Even recently, a considerable number of species discoveries are the rule (Lundberg et al., 2000). Progress in siluriform phylogeny might further increase the number of families (Diogo, 2003): for example, the erection of the family Lacantuniidae (Rodiles-Hernández et al., 2005), and the recent work of Sullivan et al. (2006). One of the most accessible overviews of the siluriform families was given by Burgess (1989). Only a few siluriform families have marine representatives (Ariidae, Aspredinidae, Plotosidae) (Roberts, 1975; Nelson, 1994; Teugels, 1996). Some generally freshwater families are encountered in brackish water as well (including some Loricariidae; Teugels, 1996). The following is a brief overview of some striking or relevant siluriform characteristics. The dorsoventrally flattened siluriform skull is platybasic, sometimes with a wide hypophyseal fenestra, and often with small eyes (Daget, 1964). This, and the fact that catfishes often live in murky waters, or are nocturnal, makes smell, taste and touch important sensory organs (Alexander, 1965). The highly sensorial oral barbels are characteristic for catfishes: most catfishes have up to four pairs of barbels bearing taste buds. The mouth is generally not protractile; the palato-maxillary mechanism enables a controlled movement of the maxillary barbels (Alexander, 1965; Singh, 1967; Gosline, 1975; Ghiot, 1978; Ghiot et al., 1984; Adriaens & Verraes, 1997a, b). Another important sense organ is the ostariophysan Weberian apparatus, first described by Bridge (1890). It efficiently transports sound vibrations from the swimbladder to the inner ear. The siluriform Weberian apparatus is more specialized than in other Ostariophysi (Chardon, 1968; Chardon et al., 2003): the reduced swimbladder is almost completely enclosed by the extended parapophyses of the fourth and fifth vertebrae. Only laterally it usually remains uncovered, allowing external vibrations to enter the swimbladder. Alexander (1964) presumed a relation between a reduced, encapsulated swimbladder and benthic behaviour. Within the order, variation exists in the presence of a claustrum (Britz & Hoffmann, 2006). Not only sound perception, but also sound production has developed in several catfish families (e.g., Bridge & Haddon, 1894; Tavolga, 1962; Ladich & Bass, 1996). In some families grunting sounds are also produced by pectoral spine stridulation (Fine et al., 1997; Kaatz & Stewart, 1997), an effective defense mechanism, but also used in the intraspecific communication of some Callichthyidae (Kaatz & Lobel, 1999). The dorsal and pectoral spines can be locked in the erected position, anchoring the fishes in narrow cavities, or making it difficult for a predator to handle or swallow the fishes (Alexander, 1965; Schaefer, 1984; Fine et al., 1997). The indirect bony connection of these spines to the skull (via the nuchal plates or the pectoral girdle), as well as bony scutes surrounding the body, substantially increase the effectiveness of the spine locking mechanism PART 1 — GENERAL INTRODUCTION 9 in various families, including Loricariidae. The production of toxic substances sometimes adds to the defensive function of the spines (Birkhead, 1972; Burgess, 1989). In South American freshwater systems, catfishes are the largest fish group, outnumbering the Characiformes in species number (Lowe-McConnell, 1969).

1.2.3. LORICARIOIDEA

The superfamily Loricarioidea is one of the first clades that separated during siluriform evolution. The clade might have originated well before 112 Ma ago (Lundberg, 1998). Until recently, it was almost generally accepted that the Diplomystidae are the sister group of all other catfishes, representing the most plesiomorphic condition (Arratia, 1987; Teugels, 1996; Diogo, 2005). The Loricarioidea are then considered the sister group of all remaining catfishes (Diogo, 2005). Opposed to this view, de Pinna (1993) regarded the Cetopsidae as the sister group of all non-diplomystid catfishes. A recent molecular study by Sullivan et al. (2006), however, placed the loricarioid clade at the basis of al siluriforms including Diplomystidae. The latter authors elevated the group to suborder-level (Loricarioidei). In this dissertation, the term Loricarioidea is still used, awaiting more phylogenetic studies that might or might not confirm the results of Sullivan et al. (2006). The relationships of the other catfish taxa remains a hot topic, with several phylogenetic studies having been published recently (e.g., de Pinna, 1996, 1998; He et al., 1999; Diogo, 2005), or still going on. All loricarioid families [except for a few trichomycterid species (de Pinna, 1989)] have odontodes, that have most probably evolved independently in this superfamily (Bhatti, 1938; Reif, 1982; Huysseune & Sire, 1996; see Part 8). They might well function as both protective devices and hydrodynamical structures (Huysseune & Sire, 1996). Odontodes were first defined by Ørvig (1977), but a more recent, adapted definition can be found in Reif (1982) and Huysseune & Sire (1998): an odontode is an isolated hard superficial structure of the skin which consists of a core of dentine or dentine-like tissue surrounding a pulpa cavity, is either or not covered by a hypermineralized cap of enamel or enameloid, and has a vascularized base consisting of bone that functions as an attachment tissue6. In fact, the presence of odontodes is the most important of the few characters supporting the monophyly of loricarioids (de Pinna, 1998). Family-level relationships within the Loricarioidea are relatively well known (de Pinna, 1998; Fig. 2). The diversity in form, size and habitat within the Loricarioidea is exceptional (Reis, 1998; Fig. 3). They exhibit a wide range of feeding behaviours. Linked to these

6 The term odontode has more recently been reserved for those denticles that gave rise to the first teeth in jawless vertebrates, and are currently only present extra-orally in Chondrichthyes (Sire & Allizard, 2001). According to these authors the ‘odontodes’ in catfishes should be termed denticles. 10 PART 1 — GENERAL INTRODUCTION behaviours, there is a notable variety in their trophic apparatuses and their ecologic niches and places in the food web (see below). Nematogenyidae are mainly carnivorous; Trichomycteridae and Callichthyidae are carnivorous as well as detritivorous, with a few trichomycterid genera being blood and tissue parasites of other fishes. Virtually nothing is known of the feeding ecology of Scoloplacidae. Astroblepidae feed largely on insects. The diet of Loricariidae is mainly herbivorous (see paragraph 1.2.4).

Nematogenys inermis is the sole representative of the Nematogenyidae (Eigenmann, 1928). Little is known of the biology of this Andean species (Berra, 2001). It lacks the opercular odontodes of the related trichomycterids. The skin and sense organs have been described by Arratia & Huaquín (1995). Trichomycteridae, having the widest distribution of all loricarioids, include generalist carnivorous feeders (Arratia & Menu-Marque, 1981; Oliveira Ribeiro et al., 1996; Trajano, 1997; Román-Valencia, 2001), but also mucus, scale and blood feeding parasitic species in the subfamilies Stegophilinae and Vandelliinae (Winemiller, 1989; Spotte et al., 2001). The latter group includes the infamous candiru, a collective name for several species feeding on blood from the gills of other fishes by swimming inside their gill opening (Kelley & Atz, 1964; Baskin et al., 1980; Spotte et al., 2001). Incidental penetration of the human urethra has been reported (whence their infamous reputation; Breault, 1991). Odontodes on the opercle and interopercle provide attachment to the substrate or host animal (Eigenmann, 1918). Trichomycterids include many miniaturized species (Weitzman & Vari, 1988). Callichthyidae or armoured catfishes are a speciose family, and have two rows of overlapping bony plates on each side of the body. Most species are bottom-scavenging generalists (Burgess, 1989). Ontogenetic data of several species are provided by Hoedeman (1960a, b, c). Much of their known anatomy is published by Reis (1998). The detailed morphology and ontogeny of Corydoras aeneus is the focus of Frank Huysentruyt and co- workers, and forms part of the present project on loricarioid catfishes at the Ghent University (see paragraph 1.1.1). Scoloplacidae or spiny dwarf catfishes are one of the ‘youngest’ catfish families, with the first species, Scoloplax dicra, described in 1976 (Bailey & Baskin, 1976). Only four miniaturized species are known (Schaefer et al., 1989; Berra, 2001). Some remarkable features that characterize scoloplacids are the absence of a prevomer and an exoccipital, and the presence of a prominent rostral plate with recurved odontodes (Schaefer, 1990). Very little is known about the ecology of the family (Armbruster, 1998; Berra, 2001). Astroblepidae form the sistergroup to the Loricariidae, and are distinguished by several characters, including expanded dorsolateral and anterolateral premaxillary processes, and the presence of lateral ethmoid-palatine and operculo-interhyal ligaments (Schaefer, 1990). They PART 1 — GENERAL INTRODUCTION 11 are predominantly insectivorous (Román-Valencia, 2001), and famous for their orobatic or mountain-climbing abilities. By the use of their suckermouth (which they share with loricariids) and pelvic girdle, they can climb overhanging rocks up to an inclination of 30% (Johnson, 1912; Shelden, 1937), thereby surpassing other climbing fish families [e.g., Mochokidae, Amphiliidae (Roberts, 1975)].

1.2.4. LORICARIIDAE

Loricariidae or suckermouth armoured catfishes are the largest catfish family, numbering about 717 valid living species in 96 genera (C. Ferraris, pers. commun.). The body is covered with bony plates, whence the name mailed or armoured catfishes originates (Buck & Sazima, 1995). Besides this armour and their ventral suckermouth, loricariids are easily recognized by their ventrally compressed head, the often elongated body form, and the absence of nasal and mandibular barbels (Alexander, 1965; Teugels, 1996; Diogo, 2005). A relation has been found between body shape and (micro-)habitat use (Jégu et al., 2004). The maximal total length varies from 1.8 cm in the miniaturized Microlepidogaster lophophanes (Weitzman & Vari, 1988) to 90 cm in Pseudacanthicus histrix (Ferraris et al., 2003). Schaefer (1987, 1990) lists synapomorphic characters that support the monophyly of the family. Among these, the following are most conspicuous: the presence of an expanded ventral mesethmoid disc (present, but smaller in astroblepids), asymmetrically bifid tooth cusps, the presence of a retractor palatini (retractor veli, Chapter 6.1), and two distinct subdivisions of the extensor tentaculi muscle. Gregory (1933) and Eaton (1948) mentioned the Loricariidae as one of the most specialized catfish families. Loricariids are found in tropical and subtropical South America, with species described from northern Argentina to Costa Rica in Central America (Miller, 1966; Myers, 1966; Isbrücker, 1980). Six subfamilies are currently recognized: Delturinae, Lithogeninae, Neoplecostominae, Hypoptopomatinae, Loricariinae and Hypostominae (de Pinna, 1998; Armbruster, 2004; Reis et al., 2006a; Fig. 4). Neoplecostominae probably are paraphyletic (Reis et al., 2006). Few molecular studies exist on loricariid phylogeny (e.g., Montoya- Burgos et al., 1997; 1998). Moreover, loricariid systematics continue to change as newly discovered species are added to the known taxa (e.g., Schaefer, 1998).

Several studies have discussed the trophic niche of loricariid species. Although many loricariids are known to feed mainly on algae (Angelescu & Gneri, 1949; Angermeier & Karr, 1983; Power, 1984; Buck & Sazima, 1995; Aranha et al., 1998), ecological data concerning their diet are still scarce. 12 PART 1 — GENERAL INTRODUCTION

The algae grazing of some loricariids is well documented (e.g., Buck & Sazima, 1995). Power (1984) found a close response of loricariid density to algal growth rates in neotropical rivers. The intestine is long [e.g., ca. 7 times the standard length in some specimens of Ancistrus cf. triradiatus (pers. ob.), or 14 times or longer in Hypostomus commersoni (Angelescu & Gneri, 1949 in Alexander, 1965)]. Compared to non-herbivirous taxa, intestine length and mass are strongly positively allometric with standard length [Ancistrus spinosus (Kramer & Bryant, 1995). Not only algae, but also epilithic detritus can be scraped, or loose detritus and other small food items or even wood can be fed on (Saul, 1975; Schaefer & Stewart, 1993; Fugi et al., 1996; Grosman et al., 1996; Nelson et al., 1999; Delariva & Agostinho, 2001; Nelson, 2002). Species with reduced jaws and/or teeth (e.g., Dentectus barbamatus; Salazar et al., 1982) probably rely much more on non-encrusted organic matter, much of it being detritus (dead organic matter that has been altered in some way that renders it unlike its original living form, i.e. organic matter that has undergone some diagenesis; Bowen, 1984). Together with Prochilodontidae and Curimatidae, Loricariidae are the best known detritivorous fish family in the neotropics (Delariva & Agostinho, 2001). In the tropics, detritivorous fishes can dominate ecosystem ichthyomass (Bowen, 1984).

Loricariidae are found in many different freshwater habitats, ranging from almost still water bodies to very fast flowing mountain streams (Suzuki et al., 2000; Jégu et al., 2004). The family includes fishes that are able to survive in relatively hypoxic conditions, displaying physiological, morphological and behavioural adaptations (Favaretto et al., 1981; Perna & Fernandes, 1996; Val et al., 1998). Air breathing has been described in more catfish families, with a variety of internal modifications to this function (Roberts, 1975; Browman & Kramer, 1985; Olson et al., 1990). Loricariids use their stomach as an accessory respiratory organ (Carter & Beadle, 1931; Gradwell, 1971a; Graham, 1997; Armbruster, 1998; Podkowa & Gonialowski-Witalinska, 2003). Silva et al. (1997) described oesophageal respiratory purses in Loricariichthys platymetopon. Synchronous air breathing has been observed in several species (Kramer & Graham, 1976). The shift from aquatic to air breathing involves both the swallowing of air and a decrease in aquatic respiration rate (Takasusuki et al., 1998). Kramer et al. (1978) reported the ability of some species (including some Ancistrus sp.) to survive briefly on land. The reduction in swimbladder volume might be related to a demersal mode of life (Alexander, 1965; Gee, 1976).

Loir et al. (1989) and Mazzoni & Caramaschi (1997) have included data on the reproductive system of loricariids. The number of (large) eggs is relatively low; maturation of eggs can occur throughout the year. The reproductive biology of loricariids appears to be diverse. Parental care has been observed in many species (see Chapter 3.2 for details). As in PART 1 — GENERAL INTRODUCTION 13 the majority of teleosts displaying parental behaviour, the father usually guards the nest (Keenleyside, 1979). He even carries the eggs in some cases [Lowe, 1964; Lopez-Rojas & Machado-Allison, 1975; Taylor, 1983; Suzuki et al., 2000; both sexes carry eggs in Loricaria cataphracta (Schmidt, 2001)]. Mouth-brooders exist as well (e.g., Strauss, 1995). This parental behaviour and a low fecundity (Suzuki et al., 2000), as well as several gradations in sexual dimorphism, in combination with the high diversity within the family, have evoked a comparison with the African cichlids (Schaefer & Stewart, 1993). Both the loricariid and the cichlid families are good examples of evolutionary radiation in tropical freshwater systems (lakes or rivers). De Pinna (1998) stated that these factors might well suggest parallel causes in the successful radiation of both families. Feeding specializations are widely known among cichlids; specializations among those cichlids that feed on algae (Liem, 1980a; Witte & Van Oijen, 1990) might well be related to the diversity of tooth forms in the group (Yamaoka, 1983). In a broad ecological study on freshwater fishes in Guyana, Lowe (1964) found that, while breeding of most fishes coincided with the main rainy season, cichlids and loricariids produced eggs at more frequent intervals during the year. Camouflage, both morphologically and behaviourally, has been observed in some loricariid species (Retzer & Page, 1996); even mimicry has been suggested (Retzer et al., 1999). The remaining ecological data include the presence of parasitic hirudineans in jaw tissue Weber (1987) and commensal chironomid midge larvae between the cheek spines (Freihofer & Neil, 1967). Some cave-dwelling species (including Ancistrus sp.) have reduced eyes or are blind (Ufermann, 1998; Reis et al., 2006b). Loricariids are among those catfish groups that are of considerable economic importance (Lowe-McConnell, 1984; Teugels, 1996), not only because of human consumption, but even more for the aquarium trade (Lachner et al., 1970; Chao, 2001). 14 PART 1 — GENERAL INTRODUCTION

PART 2

MATERIAL AND METHODS

PART 2 — MATERIAL AND METHODS 15

2.1. MATERIAL

2.1.1. CHOICE OF SPECIES

The loricariid genus Ancistrus (‘bushymouth catfish’, ‘bristlenose catfish’, ‘bluechin xenocara’, ‘silure bleu’) was chosen as the primary research target of this study (Fig. 5). It has a fairly typical loricariid habitus and a medium size. It belongs to the largest loricariid subfamily [(Hypostominae (Armbruster, 2004)], and is known to feed on algae as well as other food items. It is very easy to breed Ancistrus in aquarium conditions (Ufermann, 1998), which is, of course, of vital importance for an ontogenetic study. Among aquarists a broad knowledge exists on how to culture and breed Ancistrus. In natural environments Ancistrus prefers biotopes with lots of crevices in both wood and rock (Lowe, 1964). Until recently, the Ancistrinae were recognized as one of six subfamilies in the Loricariidae (Montoya-Burgos et al., 1997; de Pinna, 1998). Armbruster (2004) lowered the Ancistrinae to the tribe Ancistrini in the subfamily Hypostominae. The systematics within the Ancistrini remain largely unresolved (Montoya-Burgos et al., 1997; de Pinna, 1998; Armbruster 2004). Complete determination keys of Ancistrus itself are nonexistent, and the genus is in need of revision. Only partial determination keys exist (Miquilarena et al., 1994). It is almost impossible to identify Ancistrus species without knowledge of the locality where specimens were caught. Original species descriptions are often useless. The specimens of this study were identified as most probably belonging to Ancistrus triradiatus Eigenmann, 1918 by Sonia Fisch-Muller, who is a leading expert in Ancistrus taxonomy (e.g., Muller, 1990; Muller et al., 1994; Fisch-Muller, 1999). The specimens used in this dissertation are thus named Ancistrus cf. triradiatus throughout the text. Males of all Ancistrus species exhibit conspicuous fleshy tentacles or dermal cirri on the snout (Muller et al., 1994). Females only have a modest series of rostral tentacles. The tentacles might be related to the reproduction (Sabaj et al., 1999). Ancistrus triradiatus is an important algae-feeder in its natural habitat (Winemiller & Jepsen, 1998).

Using one species as a representative of a family containing over 700 species could lead to misleading generalizations. Therefore several other loricariid species were studied as well. An additional advantage of including more species is that one obtains some more insight in the morphological diversity within the family. - Within the genus Ancistrus two other species were examined (Fig. 6A-D): A. ranunculus is a well described, rather recently discovered species; one of the major 16 PART 2 — MATERIAL AND METHODS

differences with A. cf. triradiatus is the flattened head (Muller et al., 1994). A. dolichopterus more closely resembles A. cf. triradiatus. - Within the subfamily Hypostominae two other species were studied (Figs 6E-F, 7A-B): Pterygoplichthys lituratus is a large species, reaching 37 cm SL (Ferraris et al., 2003). It has a weakly armed opercular region (Weber, 1992; Armbruster, 2004; see Part 8). It is currently being used for a biomechanical study of the respiration and feeding, carried out in cooperation with Anthony Herrel (Laboratory for Functional Morphology, Antwerp University). Panaque nigrolineatus is renowned for its large, spatulate teeth, and its habit of feeding on wood. Within the loricariid family representatives were studied of the two other speciose subfamilies (Figs 7C-I, 8A-B). The almost miniaturized Otocinclus vestitus (Hypoptopomatinae) and the stick catfish Farlowella acus (Loricariinae) were studied in detail, as they reflect the broad diversity in loricariid body form. Other examined loricariine species are Sturisoma aureum and Rineloricaria parva.

Two non-loricariid outgroup species have been examined as well. Corydoras aeneus, currently studied by Frank Huysentruyt (see paragraph 1.1.1) is a member of the Callichthyidae, a loricarioid family related to Loricariidae. For several purposes, especially the homology and evolutionary origin of the jaw muscle complex, it was also included in the present dissertation. The clariid Clarias gariepinus is the only siluriform of which ample ontogenetic data already exist [several papers, especially the dissertation of Dominique Adriaens (1998)], and thus represents an ideal reference species to which the current results can be compared. It is a ‘typical’ catfish, with a broad, flattened head, long barbels, and is omnivorous (but largely carnivorous), and one of the best known siluriform species (Nawar, 1954, 1955a, b; Haylor, 1992; Legendre et al., 1992; Vandewalle et al., 1985; Radermaker et al., 1989; Surlemont et al., 1989; Surlemont & Vandewalle, 1991; Adriaens & Verraes, 1994, 1996, 1997a, b, c, d, e, 1998; Adriaens et al., 1997, 2001; van Snik et al., 1997).

2.1.2. MATERIAL EXAMINED

All specimens were commercially obtained (Tables I and II). Some specimens of the three main species studied have been deposited in the Zoology Museum of the Ghent University (UGMD 175370-373; Table III). Most cleared and stained specimens that have been used for part 4 have been deposited there as well (UGMD 175351-369; Table III). The other specimens are stored at the research group Evolutionary Morphology of Vertebrates of the Ghent University.

PART 2 — MATERIAL AND METHODS 17

TABLE I. Specimens of Ancistrus cf. triradiatus used in the present study, excluding specimens that were only measured for use in chapter 3.2.

No. SL Age Method Staining Used for (mm) (days PF) 1 4.3 2 - - Drawing 2 4.7 2 - - Drawing 3 4.8 2 Serial sections (2 µm) T Observation 4 5.2 3 Serial sections (2 µm) T 3D-reconstruction 5 5.6 4 Clearing AB + AR Drawing 6 5.8 4 - - Drawing 7 6.0 4 Clearing AB + AR Drawing 8 6.1 4 Serial sections (2 µm) T 3D-reconstruction + drawing 9 6.3 5 Clearing AB + AR Observation 10 6.6 5 - - Drawing 11 6.7 4 SEM - Photographs 12 6.8 5* Clearing AB + AR Observation 13 6.9 5 - - Drawing 14 7.0 6 Serial sections (2 µm) T Drawing 15 7.4 6 Clearing AB + AR Drawing 16 7.5 7 - - Drawing 17 7.7 6 Clearing AB + AR Observation 18 8.0 7 Clearing AB + AR Drawing 19 8.0 7 Serial sections (2 µm) T 3D-reconstruction + drawing 20 8.2 6 SEM - Photographs 21 8.5 7 Clearing AB + AR Observation 22 8.6 6 - - Drawing 23 8.7 7 Clearing AB + AR Observation 24 8.9 8 Clearing AB + AR Observation 25 9.1 8 Clearing AR Drawing 26 9.8 8 SEM - Photographs 27 9.8 10 Clearing AR Drawing 28 9.9 9 - - Drawing 29 9.9 10 Clearing AB + AR Drawing 30 10.2 10 SEM - Photographs 31 10.2 14 Serial sections (2 µm) T Observation 32 10.7 14 SEM - Photographs 33 10.8 18 Clearing AR Drawing 34 10.9 15 - - Drawing 35 11.5 30 Clearing AB + AR Observation 36 11.7 30 Clearing AB + AR Drawing 37 12.4 43 Serial sections (2 µm) T Drawing 38 12.7 48 Clearing AB + AR Observation 39 14.4 45 Clearing AB + AR Drawing 40 16.4 67 Clearing AB + AR Observation 41 20.7 96 Clearing AB + AR Drawing 42 25.0 160 Clearing AB + AR Observation 43 31.0 160 Clearing AB + AR Observation 44 33.5 160 Serial sections (5 µm) T Drawing 45 36 - (subadult) Clearing AB + AR Observation 46 44 (f) - (subadult) Clearing AB + AR Drawing 47 68 (f) - (adult) Dissection - Observation 48 70 (f) - (adult) Dissection - Observation 49 71 (f) - (adult) Serial sections (10 µm) various Observation 50 74 (f) - (adult) Dissection - Observation 51 74 (f) - (adult) SEM - Photographs 52 77 (f) - (adult) Clearing AB + AR Observation 53 86 (m) - (adult) Dissection - Drawing 54 88 (m) - (adult) Clearing AB + AR Observation (continued on next page)

AB: alcian blue, AR: alizarin red S, f: female, m: male, PF: post-fertilization, SEM: scanning electron microscopy, SL: standard length, T: toluidine blue, *: immediately after hatching. 18 PART 2 — MATERIAL AND METHODS

TABLE I (continued).

No. SL Age Method Staining Used for (mm) (days PF) 55 90 (m) - (adult) Clearing AB + AR Observation 56 94 (m) - (adult) Dissection - Drawing 57 95 (m) - (adult) Clearing AB + AR Drawing 58 102 (m) - (adult) Clearing AB + AR Observation 59 108 (m) - (adult) Manual sectioning AR Drawing 60 110 (m) - (adult) SEM - Photographs

AB: alcian blue, AR: alizarin red S, m: male, PF: post-fertilization, SEM: scanning electron microscopy, SL: standard length.

TABLE II. Specimens of other species used in the present study.

Species (family or SL Age Method Staining Used for subfamily) (mm) (days PF) Ancistrus dolichopterus (Hypostominae) 93 - (adult) Clearing AB + AR Observation Ancistrus ranunculus (Hypostominae) 58 - (adult) Clearing AB + AR Observation Pterygoplichthys lituratus (Hypostominae) 63 - (subadult) Serial sections (5 µm) T Observation 94 - (adult) SEM - Photographs 150 - (adult) Clearing AB + AR Observation 235 - (adult) Dissection - Obs. / Draw. Panaque nigrolineatus (Hypostominae) 71 - (adult) Clearing AB + AR Observation 76 - (adult) SEM - Photographs Farlowella acus (Loricariinae) 105 - (adult) Clearing AB + AR Drawing 109 - (adult) Clearing AB + AR Observation 115 - (adult) Dissection - Drawing 124 - (adult) Clearing AB + AR Observation 125 - (adult) SEM - Photographs 155 - (adult) Serial sections (5 µm) T Observation Rineloricaria parva (Loricariinae) 75 - (adult) Clearing AB + AR Observation 76 - (adult) SEM - Photographs Sturisoma aureum (Loricariinae) 83 - (adult) Clearing AB + AR Observation 85 - (adult) Clearing AB + AR Observation 86 - (adult) SEM - Photographs Otocinclus vestitus (Hypoptopomatinae) 22 - (adult) Serial sections (5 µm) T Observation 23 - (adult) Clearing AB + AR Observation 24 - (adult) Clearing AB + AR Observation 25 - (adult) Clearing AB + AR Drawing 26 - (adult) Dissection - Drawing 28 - (adult) SEM - Photographs Corydoras aeneus (Callichthyidae)* 4.9 6 Serial sections (2 µm) T Drawing 9.3 16 Serial sections (2 µm) T Drawing 36 - (adult) Dissection - Observation 39 - (adult) Serial sections (5 µm) T Drawing Clarias gariepinus (Clariidae)* 5.6 2 Serial sections (2 µm) T Drawing 7.2 8 Serial sections (2 µm) T Drawing 8.4 14 Serial sections (2 µm) T Drawing 18.7 32 Serial sections (5 µm) T Observation AB: alcian blue, AR: alizarin red S, PF: post-fertilization, SEM: scanning electron microscopy, SL: standard length, T: toluidine blue. *: see figure 8C-F. PART 2 — MATERIAL AND METHODS 19

TABLE III. Specimens used in the present study that have been deposited in the Zoology Museum of the Ghent University (Universiteit Gent Museum voor Dierkunde – UGMD).

UGMD coll. nr. and species Specimen nr. (Table I) SL Method 175351 Ancistrus cf. triradiatus 5 5.6 Clearing and staining 175352 Ancistrus cf. triradiatus 7 6.0 Clearing and staining 175353 Ancistrus cf. triradiatus 9 6.3 Clearing and staining 175354 Ancistrus cf. triradiatus 12 6.8 Clearing and staining 175355 Ancistrus cf. triradiatus 15 7.4 Clearing and staining 175356 Ancistrus cf. triradiatus 17 7.7 Clearing and staining 175357 Ancistrus cf. triradiatus 18 8.0 Clearing and staining 175358 Ancistrus cf. triradiatus 21 8.5 Clearing and staining 175359 Ancistrus cf. triradiatus 23 8.7 Clearing and staining 175360 Ancistrus cf. triradiatus 24 8.9 Clearing and staining 175361 Ancistrus cf. triradiatus 25 9.1 Clearing and staining 175362 Ancistrus cf. triradiatus 27 9.8 Clearing and staining 175363 Ancistrus cf. triradiatus 33 10.8 Clearing and staining 175364 Ancistrus cf. triradiatus 36 11.7 Clearing and staining 175365 Ancistrus cf. triradiatus 39 14.4 Clearing and staining 175366 Ancistrus cf. triradiatus 40 16.4 Clearing and staining 175367 Ancistrus cf. triradiatus 41 20.7 Clearing and staining 175368 Ancistrus cf. triradiatus 42 25.0 Clearing and staining 175369 Ancistrus cf. triradiatus 43 31.0 Clearing and staining 175370 Ancistrus cf. triradiatus - (unused specimen) 74.0 - 175371 Ancistrus cf. triradiatus - (unused specimen) 65.0 - 175372 Farlowella acus - (unused specimen) 117.0 - 175373 Otocinclus vestitus - (unused specimen) 21.5 -

20 PART 2 — MATERIAL AND METHODS

2.2. METHODS

2.2.1. KEEPING AND BREEDING OF LORICARIID SPECIES

The presence of live specimens of several loricariid taxa was essential for this dissertation: - observation (e.g., respiration, feeding, parental behaviour, use of the cheek-spine apparatus, activity and movements of embryos) and filming of specimens; - the presence of fresh specimens at the research group, so that sacrification and fixation could be done optimally; - the breeding of Ancistrus cf. triradiatus.

Ancistrus cf. triradiatus was bred in 24-27° C aquarium tanks of ca. 30 to 130 cm at the research group Evolutionary Morphology of Vertebrates (Ghent University). Water conditions were controlled regularly, and partial water renewments were done approximately every one to two months. Up to 18 nests were successfully bred by the adults. To minimize risks of disease three separate aquarium tanks were in use at all times. Each tank contained adequate shelter: plants, wood, holes formed by stones and broken terracotta pots. Such shelter is necessary, as Ancistrus males are territorial and care for the eggs and free-living embryos in these holes, fanning the young for aeration and the prevention of infections. All species were fed with vegetable fish tablets. From time to time general pellet food (containing potentially essential non-vegetable ingredients) and pieces of zucchini were added. The wood in the tank was fed on primarily by Ancistrus and Panaque. Wood is supposed to be an essential dietary part for several loricariid species (Schaefer & Stewart, 1993).

The commercially obtained Corydoras aeneus specimens were bred and provided by Frank Huysentruyt. Clarias gariepinus specimens were obtained from Adriaens (1998).

2.2.2. LIVE OBSERVATIONS AND HIGH-SPEED FILMING

The presence of live specimens at the research group provided the opportunity to observe the respiration and feeding movements. Observations were carried out in the aquaria or small experimental tanks (including experiments with milk used as dye to visualize water flows during respiration). Respiration and feeding in embryos (free-living or removed from the egg scale) was observed with an Olympus SZX9 stereoscopic microscope. PART 2 — MATERIAL AND METHODS 21

Respiration and feeding of Ancistrus cf. triradiatus and Pterygoplichthys lituratus was filmed with a Redlake Motionscope digital high-speed camera set at 200 frames s-1 (made available by the Laboratory of Functional Morphology of the Antwerp University, see paragraph 1.1.1). Respiration of these species and Farlowella acus and Otocinclus vestitus was filmed with a JVC DVL-9800 digital video camera as well. The Redlake Motionscope camera was also used for the filming of the fast movements of the cheek-spine apparatus in A. cf. triradiatus.

2.2.3. PREPARATION OF SPECIMENS FOR STUDY

Specimens were anaesthetized in a watery solution of MS 222 (ethyl 3-aminobenzoic acid methanesulfonate salt), and then sacrified by an overdose of MS 222. Fixation of ‘large’ specimens (over 2 cm SL) was done with a 4% buffered formalin solution (at neutral pH). For small specimens, a paraformaldehyde/glutaraldehyde mixture, buffered at pH 7.4, was preferred. The latter fixative is known to reduce (but not completely prevent) shrinkage of specimens and deformation of the different tissues (DeLeon et al., 1991; Reese et al., 1991). For pre-hatching stages, egg membranes were usually removed prior to fixation. As a rule, standard length of specimens was always measured before fixation. The shrinkage caused by fixation doesn’t only cause a shortening of the standard length, but also deforms the specimen (i.e., allometric shrinkage) (Drost & Van Den Boogaart, 1986a). Thus, often pictures were taken of the anaesthesized specimens. After fixation and washing, specimens were gradually transferred to a 75% ethanol solution, for preservation. For the study of the ontogeny, specimens were arbitrarily selected, representing the complete ontogenetic period in which the skeleton and muscles arise and transform.

2.2.4. METRICS

For chapter 3.2, digital measurements were made with AnalySIS 5.0 for small specimens, and with a digital caliper for larger specimens (over 20 mm SL). The following lengths were measured: total length (TL), standard length (SL), head length (HL), head width (HW), trunk length (between head and anal opening; TrL), tail length (length posterior to anal opening; TaL), pre-dorsal fin length (PDL), pre-pectoral fin length (PPcL), pre-pelvic fin length (PPvL), snout length (anterior to eye; SnL), caudal peduncle depth (CPD), and yolk sac length, width and depth (YSL, YSW, YSD) (Fig. 9). The angle between the body axis (along the notochord) and the upper lip surface (in lateral view) was measured in all specimens above 5 mm SL, as a straight body axis was not present in smaller specimens (angle α on 22 PART 2 — MATERIAL AND METHODS figure 9). A total of 132 specimens were so measured, ranging from pre-hatch embryos to fully grown adults. Those specimens that were only measured, and not used for further examination (e.g., dissection, staining, photographs, drawings), are not listed in Table I.

2.2.5. IN TOTO CLEARING AND STAINING

Osteology was studied on in toto cleared and stained specimens of all loricariid species included; alizarin red and alcian blue were used, following the (slightly adapted) method of Taylor & Van Dyke (1985) (see Table IV). Examination of the specimens was done using an Olympus SZX9 stereoscopic microscope, equipped with a camera lucida for drawing. As an aid to the drawings, dissections (e.g., removal of pectoral girdle or part of the splanchnocranium) were performed in the larger specimens. Drawings figure all cartilaginous and bony elements of the skull that are visible on the cleared and stained specimens. However, the study of serial sections of specimens demonstrates that sometimes early ossification is not visible in stained specimens of the same or earlier stage, a known artefact in the in toto staining techniques (Vandewalle et al., 1998). In such cases this is clearly mentioned in the text.

TABLE IV. Bone – cartilage stain for whole specimens, adapted from Taylor & Van Dyke (1985).

Step Solution Duration Dehydration 50% alcohol 12h 75% alcohol 12h 96-100% alcohol 12h 96-100% alcohol 12h Cartilage staining alcian blue7 8-24h Neutralization saturated borate solution8 48h Bleaching 3-10% H2O2 in 0.5% KOH 0.5h-... 9 Clearing 1-4% KOH / trypsin (0.6g in 400 ml 30% NaBO3) 12h-... Bone staining alizarin red10 24h Further clearing 0.5-4% KOH 12h... Preservation 25% glycerin + 75% 0.5% KOH 12h 50% glycerin + 50% 0.5% KOH 12h 75% glycerin + 25% 0.5% KOH 12h 100% glycerin storage

The protocol of Taylor & Van Dyke (1985) differs from that of Hanken & Wassersug (1981) in, a.o., avoiding extensive damage from the acid cartilage stain on the bony structures by the alcohol series and the use of borax, and by a more extensive use of KOH instead of

7 Alcian Blue 8GX (Sigma) 8 Dinatriumtetraborate (Vel) 9 Trypsin 1-300 (ICN Biomecdicals Inc.) 10 Alizarin red S (Sigma) PART 2 — MATERIAL AND METHODS 23 trypsin for the clearing step. Trypsin was used for small specimens though, as these benefit from trypsin instead of KOH. Cartilage or bone staining alone is also possible. In some cases cartilage staining was omitted, as the glacial acetic acid in the alcian blue stain decalcifies bone somewhat and thus masks early bone formation in small specimens.

2.2.6. DISSECTIONS

Dissections were performed for the study of both hard and soft tissues (muscles, ligaments, major nerve branches) of most loricariid species used in this study and Corydoras aeneus. Whenever necessary, visualization of muscle fibre arrangement was enhanced by the use of iodine (Bock & Shear, 1972). Examination of the specimens was done using an Olympus SZX9 stereoscopic microscope, equipped with a camera lucida for drawing. The combination of (short) in toto staining (see paragraph 2.2.4) and dissection proved to be a highly valuable tool. Dissection of freshly killed specimens is best to study the mobility of articulations and ligaments, as well as to infer possible muscle functions. Only when using fresh specimens, cautious interpretations can lead to functional hypotheses.

2.2.7. SERIAL SECTIONING

Eight specimens of Ancistrus cf. triradiatus were selected for serial sectioning, using a Technovit 7100 plastic embedding (Table V), and a Reichert-Jung Polycut microtome. Slice thickness was 2 µm for the embryonic and juvenile specimens (Table I), and 5 µm for the 33.5 mm A. cf. triradiatus specimen and the specimens of Pterygoplichthys lituratus, Otocinclus vestitus and Farlowella acus (Tables I-II). Slices were mounted on microscopic glass slides, stained with toluidine blue, and covered. Specimens selected for serial sectioning were always decalcified, to avoid tearing while sectioning. Slices were examined using a Reichert-Jung Polyvar light microscope, equipped with a camera lucida and a digital camera (Colorview 8) for easy tracing of, e.g., nerve paths and ligaments on computer, as well as for the study of cell types and other histological characteristics. The digital images, taken with the AnalySIS 5.0 software (Olympus) were also used as basis for the 3D-reconstructions (see paragraph 2.2.7). In the case of large specimens (and thus sections), several photos had to be taken to include the whole section; the Multi-Image-Arrangement (MIA) function of AnalySIS was used to make a composite photo of the whole section. The serial sections of Corydoras aeneus and Clarias gariepinus used in this study were provided by Frank Huysentruyt and Dominique Adriaens, respectively. See Table II for slice thickness of the sectioned specimens of these both species. 24 PART 2 — MATERIAL AND METHODS

TABLE V. Technovit 7100 embedding protocol for serial sectioning.

Step Solution/Action Duration Vacuum fixation 4% buffered formalin days to weeks Washing tap water 8h Decalcification Decalc11 36h Washing tap water 5h Dehydration 30% alcohol 12h 50% alcohol 12h 70% alcohol 12h 96% alcohol (two times alcohol renewal) 36h Embedding Technovit 7100 solution A12 min. 24h Technovit 7100 solution A renewal min. 48h add Technovit 7100 Harder II 12h place in deepfreeze 12h Polymerization place at room temperature (check progress) approx. 2h place in oven (approx. 40° C.) 1h

Manually (scalpel) sliced sections (ca. 2 mm thickness) were produced from one large Ancistrus cf. triradiatus male (108 mm SL), in order to provide detailed views of the skull modifications related to the cheek-spine apparatus (Part 8). Sections were coloured with alizarin red.

For the histological study of selected key tissues of Ancistrus cf. triradiatus (both upper and lower lip, oral valve and tissue connecting it to the upper jaws, cartilage plug between lower jaws) Verhoeff-Van Gieson’s stain procedure was applied on 10 µm paraffin sections. Sectioning was done with a Microm HM 360 microtome. Verhoeff’s stain was used for elastin and nuclei, and Van Gieson’s stain for collagen (Pearse, 1985).

2.2.8. 3D-RECONSTRUCTIONS

The techniques of 3D-reconstructions have advanced significantly during the last decades (e.g., Vanden Berghe et al., 1986; Haas & Fischer, 1997). For this study, computer-generated 3D-reconstructions were generated to visualize musculo-skeletal structures of small embryonic specimens, that would otherwise be impossible to grasp, and then to compare these structures to larger specimens. It is especially useful in the study of soft structures (muscles and nerves), as no quick whole-mount staining procedures exist. 3D-reconstructions also benefit from the high resolution that is achieved using serial sections. Digital images of serial sections are the source material of the computer-based reconstruction-method used for this thesis. Every eighth section was used; the resulting slice

11 Decalc (Histolab) 12 Technovit 7100 (Hereaus Kulzer) PART 2 — MATERIAL AND METHODS 25 interval was 16 µm, which proved to be sufficient to visualize the details needed in this study. Although the applied fixation techniques were chosen to minimize shrinkage, and Technovit 7100 is among the best embedding media, certain deformations were sometimes observed: shrinking of soft external parts (skin) during preparation or embedding, minor distorting of whole slices by imperfect stretching. To be able to adequately detect and correct such errors, the images were imported in CorelDraw 9.0. Here, surfaces of target structures (bones, cartilage, muscles, some nerves, brain, eyes) were manually traced (Fig. 10A. In the same process, alignment of subsequent slices was done. The resulting black-and-white sections (Fig. 10B) were then imported in the software package Amira 3.1.1 (T.G.S.). Each single anatomical structure of interest was reconstructed digitally (Fig. 11A). After rendering, the object was smoothed to produce a less artificial, even surface (Fig. 11B). Finally, all separate structures were imported in Rhinoceros 3.0 (McNeel), which allows to generate composite images of several or all structures from different view angles (Fig. 11C-D). 3D-reconstructions were made of the 5.2 mm (skeleton only), 6.1 mm and 8.0 mm stages of Ancistrus cf. triradiatus (Table I). A comparison between the cleared and stained 8.0 mm specimen used for the study of the skeleton (Part 4), and the serially sectioned and 3D-reconstructed 8.0 mm specimen used for the study of the myological ontogeny (Part 5) shows that there is little or no difference in the visualization of the skeletal parts (e.g., Figs 22, 41).

2.2.9. SCANNING ELECTRON MICROSCOPY

Scanning electron microscopy (SEM) was used for several purposes. Pictures were taken of eggs and embryos of Ancistrus cf. triradiatus, and the technique was also applied for the teeth and lip tissues of most of the loricariid species that were studied, as well as five ontogenetic stages of A. cf. triradiatus (Table I-II). After dehydration the specimen or tissue sample was critically point-dried with CO2 using a Balzers CPD 020, and gold coated using a Balzers SCD 040. The material was then examined using a Jeol JSM-840 scanning electron microscope. Length measurements of particular structures (e.g., epidermal lip brushes) were done on-screen. Pictures were taken using a magnification up to 15000x, but usually 4000x or less. 26 PART 2 — MATERIAL AND METHODS

2.3. NOTES ON TERMINOLOGIES

2.3.1. TERMINOLOGY OF ANATOMICAL STRUCTURES

Bone terminology is mostly based on Harrington (1955), Patterson (1975, 1977), Schaefer (1987) and Arratia (2003). I refer to de Beer (1937) and Adriaens & Verraes (1997c) for terminology of chondrocranium parts. Myological terminology largely follows Winterbottom (1974) except where noted.

2.3.2. USE OF THE TERMS EMBRYO, LARVA AND JUVENILE

Ancistrus cf. triradiatus has a direct development (Chapter 3.2). This implies that it has no real larval phase or stage (Balon, 1975, 1999). In the most consistent definition, the larval phase in fishes starts at the onset of exogenous feeding (Balon, 1975, 1979; Bartsch et al., 1997; Adriaens, 1998; Adriaens & Vandewalle, 2003), while it ends at the moment at which no real increase in complexity of the Bauplan, but only growth continues to occur (i.e. end of metamorphosis and start of the juvenile phase; Fostner et al., 1983; Balon, 1984, 1986; Adriaens & Vandewalle, 2003). The start of the larval phase should not necessarily coincide with the end of the embryonic phase. Free-living embryos [or eleutherembryos (Balon, 1975)] could start feeding exogenously while still (partly) relying on the endogenous yolk. In species without a real metamorphosis, the larval phase is absent, as well as in species with more advanced hiding strategies and/or yolk sacs large enough to allow all juvenile traits to develop while feeding is still endogenous (Balon, 1979). There are, however, several authors stating that the moment of hatching marks the transition from embryo to larva (e.g., Osse & van den Boogaart, 1995; Helfman et al., 1997). Also, the term larva is far more generally established than the term free-living embryo (Osse & van den Boogaart, 1995). As such, a larva is defined as a free- swimming organism with mixed or only external feeding and undeveloped adult characters (Pavlov, 1999). Following the latter terminology, Hunt von Herbing et al. (1996a, b) used ‘yolk-sac larva’ to denominate the newly hatched phase. Hatching, however, is rarely a developmental threshold in fishes. The transition to exogenous feeding, rather than hatching, is the decisive threshold of ultimate survival value (Balon, 1984). Features that indicate the moment of metamorphosis (e.g., complete differentiation of the medial finfold, pigmentation, sudden change in growth rate (Balon, 1975; Haylor, 1992; Holden & Bruton, 1994; Osse & Van Den Boogaart, 1995; Copp & Kováč, 1996; Helfman et PART 2 — MATERIAL AND METHODS 27 al., 1997) into the juvenile phase more or less coincide with the complete resorption of the yolk sac in Ancistrus cf. triradiatus (Chapter 3.2). Thus the embryonic phase is immediately followed by the juvenile phase. As the moment of hatching is variable among species, and morphological development might be very different at this moment (Balon, 1975; Surlemont & Vandewalle, 1991), the term larva is avoided in this thesis. As such, I prefer to follow Balon’s (1984, 1986, 1999) argumentations. I consistently use the terms embryo, free-living embryo, juvenile and adult. In fact, if publications mention the key moments (hatching, onset of exogenous feeding, resorption of yolk sac), as well as the age in days post-fertilization (not hatching), sufficient information is provided to avoid any misconception or debate.

2.3.3. USE OF AGE OR SIZE IN THE STUDY OF ONTOGENY

Both size (standard length) and age (after fertilization) are used to compare specimens or ‘stages’ in this dissertation. Size, however, rather than age, is used as primary standard to compare ontogenetic stages, as in many cases size has been found to be a better measure of biological time and morphological development than age (Osse & van den Boogaart, 1995; Fuiman, 1997; Adriaens & Verraes, 2002). Almost all anatomical, physiological, and behavioural attributes are size-related in some way (Strauss, 1984), and survivorship of young fishes is strongly size-dependent (Olson, 1996). In this study, the exact moment of the spontaneous fertilization was usually difficult to know (it typically happened late at night or during early morning), and embryos of a single clutch did not hatch simultaneously, but usually during a time range of up to 18 hours, and sometimes up to 48 hours. De Beer (1927) found that the majority of embryos of a certain clutch of Salmo trutta fario hatched within a period of 48 hours, and some remained unhatched for several days longer (pre-hatching time for S. trutta fario is 43 days). If using age, age post-fertilization (PF) is more useful than age post-hatching (PH), as (1) hatching doesn’t occur simultaneously (see above), (2) the standard using age post-hatching omits any ontogenetic event occurring before hatching, and (3) the morphological development at hatching differs between species, making it a misleading ‘milestone’ in comparative ontogenetic research (see also paragraph 2.3.2). The following is an example from this dissertation promoting the use of size rather than age. During the study of the chondrocranium it was seen that the 31.0 mm specimen is clearly more developed (i.e., has developed more structures) than the 25.0 mm specimen, although both have the same age. Also, temperature differences affect size and morphological growth, but not age, making it risky to use the age to compare specimens of batches that grew at different moments (and thus, probably, somewhat different temperatures). Environmental conditions were however found to affect size and morphological growth differently or 28 PART 2 — MATERIAL AND METHODS unequally (Gozlan et al., 1999); it is impossible to predict the exact morphological state of development from knowledge of age nor size. Thus size is a better, although not ideal criterion for the degree of development. Arguments for the (careful) use of both criteria exist (e.g., Blackstone, 1987; Strauss, 1987). Studies were embryos (and larvae) were grown at different temperatures, have shown that in some cases a higher temperature accelerates the rate of morphological growth somewhat more than the rate of growth (Fuiman et al., 1998). Irrespective of temperature, de Beer (1927) noted that in some rare cases younger and shorter embryos were found to be further developed than older and longer ones (of the same clutch), but these irregularities did not extend beyond certain limits. In the present study a few specimens were observed that appeared to be slightly more or less developed than the majority of specimens of comparable length and/or age, but these were not used in this dissertation. It seems plausible that this could be a more general phenomenon, due to natural phenotypic plasticity, although no consulted literature sources appear to mention it.

PART 3

EGG CHARACTERISTICS AND EARLY LIFE HISTORY

PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY 29

3.1. EGG CHARACTERISTICS

3.1.1. INTRODUCTION

As already mentioned in the general introduction (paragraph 1.2.4), the reproductive biology of loricariids is diverse, with a generally low fecundity, large egg size and different degrees of parental behaviour (e.g., Riehl & Patzner, 1991; Suzuki et al., 2000; Nakatani et al., 2001). This first chapter describes and discusses the egg characteristics of Ancistrus cf. triradiatus. Details on micropylar structure are not included.

3.1.2. BRIEF MATERIAL AND METHODS

About twenty eggs of various clutches were measured (AnalySIS software, see paragraph 2.2.4); SEM pictures (cf. paragraph 2.2.9) were taken of whole eggs and partial egg scales (with some measurements carried out on-screen).

3.1.3. RESULTS

Breeding Ancistrus cf. triradiatus in aquarium conditions resulted in almost 20 egg clutches in three years, many of these fathered by the same, dominant male. Egg number was usually between 50 and 90 eggs per clutch. The male took care of the eggs by protecting them from predators (including competing conspecific males), and fanned them with the pelvic fins, providing an adequate supply of oxygen-rich water. Eggs that were removed from the clutch often died from fungal or other infections, suggesting cleaning activities by the male. The male’s parental care extended for most of the free-living embryo stage (Chapter 3.2).

The diameter of the spherical yellow eggs is usually in the range of 2.8-3.4 mm, with little difference within a single clutch. A thin perivitelline space surrounds a yolk mass of approximately 2.4-2.9 mm in diameter, also varying between, but almost not within clutches. The egg scale or envelope is composed of several layers. The external layer is composed of soft, sticky material, making the eggs stick to each other and the wall or roof of the nest cavity (Fig. 12A-B). A hexagonal surface pattern suggests the presence of cells in this layer (Fig. 12C below). This layer is almost 40 µm thick, and covers the 5 µm thick zona radiata 30 PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY

(Fig. 12D). A zonation in a zona radiata externa (responsible for the production of the outer coat) and a zona radiata interna (responsible for the strength of the egg scale) (Riehl & Patzner, 1998) is not clear; numerous transverse fibres are present throughout this layer (Fig. 12E). On the inside, the ca. 0.5 µm thick oolemma membrane contains numerous pores, that continue as canals in the zona radiata (Fig. 12F).

3.1.4. DISCUSSION

Eggs of several loricariid species have been documented in literature. Egg number is generally low (e.g., 70-140 eggs in Sturisoma aureum (Riehl & Patzner, 1991). This low fecundity appears to be general in loricariids, with the notable exception of Rhinelepis aspera, for which Suzuki et al. (2000) counted an average of 47370 eggs, with a maximum of 181200. The latter species displays no parental care. The inverse relationship between fish fecundity and the degree of parental care has been documented repeatedly in fishes (e.g., von Ihering et al., 1928; Svärdson, 1948; Nikolsky, 1963 in Suzuki et al., 2000); it has also been observed among several loricariids examined by Suzuki et al. (2000). The zona radiata in teleosts is sometimes referred to as chorion or chorionic membrane. As this term is generally employed for membrane present in insect and amniote eggs, the term zona radiata is preferred for teleosts (Kunz, 2004). The zona radiata (especially the internal part) of the egg envelope of fishes is adapted to the specific environmental conditions in which the eggs are laid (Davenport et al., 1986). The thickness of the zona radiata in loricariids has been found to correlate to the degree of parental care. Rhinelepis aspera has many small eggs and an intermediate zona radiata thickness. Ancistrus cf. triradiatus, Hypostomus ternetzi (egg diameter 4.36 mm) and Megalancistrus aculeatus (egg diameter 4.29 mm) protect their eggs in a hole; the zona radiata is relatively thin. Eggs of species like Loricaria cataphracta (egg diameter 2.9-3.4 mm), Loricariichthys platymetopon (egg diameter 3.08 mm) and Loricariichthys sp. are carried on the ventral body side; the thick zona radiata appears to be an adaptation to the exposure to abrasion when the parent fish lies or swims on the bottom (Suzuki et al., 2000; Nakatani et al., 2001; Schmidt, 2001). The sticky outer substance of the egg scale of Ancistrus cf. triradiatus, displaying a hexagonal cell-like pattern, might correspond to the jelly coat or muco-follicular epithelium described in several siluriforms (Riehl & Patzner, 1998; Rizzo et al., 2002). A hexagonal surface pattern was also found in the external layer of Corydoras aeneus (Callichthyidae), where hexagonal-shaped protuberances are present (Huysentruyt & Adriaens, 2005a). Egg scale structure, including adhesive properties, is highly variable, complicating recognition and naming of layers (Laale, 1980). No furrows as seen in the loricariid Sturisoma aureum, PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY 31 or similar sperm guiding systems (Riehl & Patzner, 1991; Riehl, 1999), are present in A. cf. triradiatus.

32 PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY

∗ 3.2. EARLY LIFE HISTORY

Abstract — Early life history stages of the loricariid catfish Ancistrus cf. triradiatus are described, from pre-hatch embryos to juveniles. The descriptions, as well as metric characters, indicate that the free-swimming embryonic stage is followed directly by the juvenile stage, without a true larval stage or metamorphosis. Intense, but gradual ontogenetic head shape changes are present during the embryonic and free-living embryonic stages: the suckermouth gradually shifts from an almost rostral to a ventral position. The external and internal transformations related to this shape change are considered an adaptation to both the loricariid algae-scraping feeding mode and the need of suckermouth functioning from the moment of hatching, when a ventrally situated suckermouth would be disadvantageous, as a large yolk sac is present.

3.2.1. INTRODUCTION

The earliest free-living stages in fish development are poorly known and almost not studied, except for aquacultural and fisheries-related purposes. Nonetheless these crucial phases in the life history of fishes must also function as organisms; the relative success or failure of a species may be determined largely by events that affect its early developmental stages, rather than the more conspicuous adult stage (Orton, 1955; Lauder et al., 1989). The most critical period is often the moment of depletion of endogenous yolk material: the shift to exogenous feeding must occur swiftly (Surlemont & Vandewalle, 1991; Galis et al., 1994). Internal morphology at this moment can be considered a compromise between the availability of cranial structures and the functional demands that may have to be coped with (Adriaens et al., 2001). Growth reductions or arrests after yolk depletion have been observed in several teleost species; in case of food shortage, even shrinkage may occur (Blaxter, 1969; Surlemont & Vandewalle, 1991; Oozeki et al., 1992; Yada & Furukawa, 1999; Gisbert et al., 2002). A large endogenous supply of yolk nutrients is a significant aid in surviving the period during which feeding structures are still developing. It creates or prolonges a learning period: mouth opening and prey capture can develop and improve before they become essential for survival (Coughlin, 1994; Hunt von Herbing et al., 1996a; Adriaens et al., 2001). It can also enable the definitive adult phenotype to develop directly, avoiding an intermediate larval stage and the cost of metamorphosis (Orton, 1953; Balon, 1986). Loricariidae, or suckermouth armoured catfishes, have a ventrally placed suckermouth, and it has been reported that newly hatched free-living embryos are able to attach immediately onto submerged substrates with their tiny suckermouth (Riehl & Patzner, 1991 on Sturisoma

∗ Slightly modified from: Geerinckx T., Verhaegen Y. & Adriaens D. Ontogenetic allometries and shape changes in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes), related to suckermouth attachment and yolk sac size. Submitted to the Journal of Fish Biology. PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY 33 aureum). My similar observations of Ancistrus cf. triradiatus embryos attaching to the nest cavity roof directly after hatching raised the question on how they overcome the spatial problem of a ventral suckermouth combined with a voluminous yolk sac on the ventral side of the body. Eggs of most loricariid catfishes are large, and clutch sizes are small (Mazzoni & Caramaschi, 1997; Suzuki et al., 2000). Parental care has been observed in many species (Lowe, 1964; Vaz-Ferreira & Señorans, 1971; Lopez-Rojas & Machado-Allison, 1975; Burgess, 1989; Sabaj et al., 1999; Suzuki et al., 2000; Schmidt, 2001). In loricariid species where hatchlings leave the shelter of the nest immediately (Riehl & Patzner, 1991; Suzuki et al., 2000), suction might be even more important, as many loricariids live in fast-flowing rivers.

3.2.2. BRIEF MATERIAL AND METHODS

For breeding of Ancistrus cf. triradiatus, I refer to paragraph 2.2.1. A total of 132 specimens were examined, ranging from pre-hatch embryos to fully grown adults; a list of measurements is given in paragraph 2.2.4. Some specimens were cleared and stained (see paragraph 2.2.5 for details on procedures, and figure 15 and Table I for specimens).

3.2.3. RESULTS

Breeding Ancistrus cf. triradiatus in aquarium conditions resulted in seven egg clutches, all fathered by the same, dominant male. Egg number was usually between 50 and 90 eggs per clutch. The male’s parental care extended for most of the free-living embryo stage (see below). Embryos removed from the egg scale are curved around the spherical yolk mass. The snout is very short; one continuous finfold develops (Fig. 13A-B). Live observation of 5.3 mm embryos reveals tail and body undulations, and up and down movements of the hyoid region. The use of diluted milk (as dye) indicates that no unidirectional water flow through the mouth cavity is present yet. Such efficient gill ventilation is present in 6.2 mm embryos, that probably have developed their oral valve and/or branchiostegal membranes. Shortly before hatching the upper and lower lip surfaces increase, and the maxillary barbel appears (Fig. 13C). The dorsal and ventral finfold, especially those parts that will be resorbed during later ontogeny, are well vascularized. Newly hatched, free-living embryos [or eleutherembryos (Balon, 1975)] have a standard length of 6.3 to 7.1 mm, usually about 6.8 mm. Of the unpaired fins, the dorsal and caudal fins are well recognizable. The eyes are pigmented, and the snout is elongating (Fig. 13D). 34 PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY

The yolk sac is ovally shaped. The embryos are able to attach to a substrate with their sucker- like mouth, while water inflow occurs via small furrows near the maxillary barbels, as in adults. Body and tail movements assist the attachment in this early stage, albeit not all the time. In both live observations and the study of serial sections, there is no evidence of adhesive glands, which have been described in several other fishes (e.g., Allis, 1889; Watanabe, 1994); the papillose lip surface might however aid in attachment. During the first three days after hatching, the yolk sac is still very large, and is gradually flattening dorsoventrally: its length increases but both width and depth decrease slowly (Fig. 13D-G; Table V). The anal fin also develops now, while the adipose fin, supported by one spiny ray, differentiates only after complete resorption of the yolk sac (Fig. 13H-I). First body pigmentation appears after one to two days after hatching. An adult-like appearance is present at the moment the last yolk is consumed: the free-living embryo stage is immediately followed by the juvenile stage. A real larval stage (Balon, 1975; 1999) is absent. Intestinal content is observed from three to four days after hatching, which is one to two days before yolk depletion. The day on which the nest is left varied among clutches, but was usually the fourth or fifth after hatching.

TABLE VI. Growth coefficients (slope and R²) of metric variables in 132 specimens of Ancistrus cf. triradiatus. See paragraph 2.2.4 for abbreviations.

Embryo Free-living embryo Juvenile and adult (n=37) (n=40) (n=55) Slope R² Slope R² Slope R² HL 1.2849 0.7331 1.3150 0.9387 0.9644 0.9900 TrL 0.8720 0.6145 0.8718 0.8047 1.1495 0.9825 TaL 1.3142 0.8905 1.4802 0.9604 0.9187 0.9909 SnL 2.2018 0.7544 1.9304 0.7778 1.0652 0.9842 TL 1.1379 0.9835 1.3047 0.9899 0.9808 0.9982 PDL 1.1376 0.8920 1.3856 0.9702 1.0083 0.9962 PPcL 1.2089 0.5973 1.3867 0.8811 0.8999 0.9657 PPvL 0.9284 0.7025 1.0152 0.9533 1.0346 0.9951 HW 1.6923 0.7582 0.9955 0.9279 1.0319 0.9941 CPD 1.1939 0.6169 1.5286 0.8280 0.9602 0.9832 YSL 0.2902 0.2609 -1.4474 0.5507 - - YSW -0.4861 0.3060 -0.8955 0.5552 - - YSD -0.3620 0.3527 -4.3956 0.6610 - -

I compared growth patterns of Ancistrus cf. triradiatus in three different periods: (1) before hatching (embryo), (2) between hatching and yolk sac depletion (free-living embryo), and (3) after yolk sac depletion (juvenile and adult). A comparison of growth in the head, trunk and tail regions indicates an accelerated growth in the head and tail regions during the first two periods, especially in the free-living embryo stage (Fig. 14A-C; Table VI). Growth coefficients of caudal peduncle depth and snout length point in the same direction (Table VI). The fastest allometric growth is observed in the snout, which grows exceptionally fast during PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY 35 embryonic and free-living embryonic stages (slopes 2.2 and 1.9 respectively; Fig. 14D; Table VI). During these periods, the tip of the snout also grows ventrally. Examination of the A. cf. triradiatus specimens revealed that this transformation also includes a shift of the angle of the upper lip, and the whole suckermouth disc, relative to the body axis. The lower lip, also part of this disc, is highly mobile and flexible, but the upper lip allowed for reliable measurements in the examined specimens (Figs 9, 14E). The result is a mouth suction disc able to attach to a substrate at all times during free-living ontogeny, without spatial hindrance of the yolk sac. Figure 13 illustrates the angle of the suckermouth disc (in relation to the body axis) gradually decreasing while the yolk sac diminishes. The disc can fully contact the substrate at all times. The gradual external changes reflect intense internal skeletal transformations. A previous study of the ontogeny of the chondrocranium of Ancistrus cf. triradiatus showed that the tip of the ethmoid cartilage grows very fast, but also ventrally (Chapter 4.1). Figure 15, showing both the cartilaginous and bony elements of the head skeleton, indicates how the base of the maxillary bone, as well as the maxillary cartilage, shift ventrally as well. These are the skeletal elements of the short maxillary barbel that support the suckermouth disc. The premaxilla, or upper jaw, is also shifted ventrally (see also Chapter 4.2).

3.2.4. DISCUSSION

Descriptions of hatchlings of other loricariid species indicate a common standard length of 6 to 8 mm, and the general presence of a large yolk sac (except for Loricaria cataphracta, which only has a little amount of yolk when hatching) (de Azevedo, 1938 in Page et al. (1993); Lopez-Rojas & Machado-Allison, 1975; Machado-Allison & Lopez-Rojas, 1975; Riehl & Patzner, 1991; Page et al., 1993; Nakatani et al., 2001; Schmidt, 2001). The direct transition of a free-swimming embryo to the juvenile stage in these species supports the idea that a real larval stage is absent in loricariids in general. Teleostean embryonic (and larval) stages are usually characterized by a high degree of allometric growth patterns (Fuiman, 1983; Osse et al., 1997; van Snik et al., 1997). In juvenile and adult stages, all growth coefficients approach one (near-isometric growth). For Ancistrus cf. triradiatus, this is shown in Table VI. Similar findings have been made in other teleosts, and it has been generally accepted that this accelerated growth is related to rapid development of essential feeding structures (head region), and the need of improved swimming performance (tail region) (Fuiman, 1983; Osse et al., 1997; van Snik et al., 1997). Some minor allometric growth persists in the juvenile to adult period, as also observed by Strauss (1995) in the loricariids Loricariichthys maculatus and Pseudohemiodon laticeps. This is not uncommon in fishes (Fuiman, 1983). 36 PART 3 — EGG CHARACTERISTICS AND EARLY LIFE HISTORY

The fast positive allometric growth of the snout (even compared to the head length) during the early free-swimming period, as seen in Ancistrus cf. triradiatus, has also been observed in the loricariids Chaetostoma stannii, Loricaria cataphracta, Loricariichthys maculatus, L. platymetopon and Pseudohemiodon laticeps (Machado-Allison & Lopez-Rojas, 1975; Page et al., 1993; Strauss, 1995; Schmidt, 2001). In the cases where head length was measured as well, it proved to be far less allometric than the snout length. I hypothesize that the fast allometric growth of the snout and the remarkable lip transformation are related to (1) the advantage or need of suckermouth attachment to substrates as soon as the embryo leaves the egg, and (2) the early development of a ventrally oriented suckermouth, advantageous for substrate scraping and suction, and typical for all loricariids. Given the large volume of the yolk sac in the earliest free-living stages, it is clear that the embryos benefit from a rostroventrally instead of ventrally oriented mouth. After resorption of the yolk, the exogenously feeding juveniles and adults have a ventrally oriented suckermouth, which can be considered an adaptation to the feeding on algae and other material attached to submerged substrates, while the fish can also attach firmly to these substrates, an advantage in fast-flowing rivers, where many loricariids thrive. A long and ventrally directed upper snout brings the upper jaw in a position anterior to the lower jaw, so that both jaws can assist in the scraping of the substrate (which is situated ventrally to the fish). The profile of a substrate-attached fish is also hydrodynamically advantageous if its snout region is pointed ventrally (and thus is closely appressed to the substrate): this reduces current drag when a fish is positioned against substrates in fast flowing rivers. Loricariids use both upper and lower jaws to scrape food from the substrate. As the premaxillae articulate with the mesethmoid bone of the neurocranial rostrum, feeding surely benefits from a ventrally oriented rostrum. It seems well possible that the ventral mesethmoid disc is an additional adaptation to further lowering the position of the premaxillae. It also increases their mobility (Schaefer & Lauder, 1986; Chapter 6.1). This mesethmoid disc is synapomorphic for loricariids and astroblepids, that both display a sucker-like mouth (Schaefer, 1990). This ontogenetic shift in shape changes thus reflects the needs during the various developmental stages, while allowing the formation of the highly specialized loricariid head configuration. Moreover, this shift occurs without the need of a larval stage followed by a drastic and possibly very energy-consuming remodulation stage or real metamorphosis, as seen in, e.g., flatfishes (Brewster, 1987; Wagemans et al., 1998).

PART 4

ONTOGENY OF THE SKULL

PART 4 — ONTOGENY OF THE SKULL 37

∗ 4.1. ONTOGENY OF THE CHONDROCRANIUM

Abstract — The chondrocranium of the suckermouth armoured catfish Ancistrus cf. triradiatus is studied. Its development is decribed based on specimens ranging from small pre- hatching stages with no cartilage visible, to larger post-hatching stages where the chondrocranium is reducing. Cleared and stained specimens, as well as serial sections revealed a cartilaginous skeleton with many features common for Siluriformes, yet several aspects of A. cf. triradiatus are not seen as such in other catfishes, or to a lesser extent. The skull is platybasic, but the acrochordal cartilage is very small and variably present, leaving the notochord protruding into the hypophyseal fenestra in the earlier stages. The ethmoid region is slender, with a rudimentary solum nasi. A lateral commissure and myodomes are present. The larger posterior myodome is roofed by a prootic bridge. The maxillary barbel is supported by a conspicuous cartilaginous rod from early pre-hatching stages. The ceratohyal has four prominent lateral processes. Infrapharyngobranchials I-II do not develop. During ontogeny, the skull lengthens, with an elongated ethmoid, pointing ventrally, and a long and bar-shaped hyosymplectic- pterygoquadrate plate. Meckel’s cartilages point medially instead of rostrally.

4.1.1. INTRODUCTION

The ontogeny of fishes, and other vertebrates merits attention for various reasons. First, description of ontogeny and ontogenetic transformations is essential for understanding the pattern behind body plan formations. Second, this knowledge provides information that can be used in reconstructing phylogenies. Third, attention must be given to the fact that an organism must be functional at each moment, including young, growing, ever-changing, and thus ‘temporary’ stages (Galis, 1993; Galis et al., 1994). Organisms can hardly be understood by considering only their adult forms, and study of their early ontogeny may be more revealing and is therefore very important (Balon, 1986). An interesting case, of which very little is known at the moment, is the ontogeny and growth in the catfish family Loricariidae, or suckermouth armoured catfishes. With more than 700 species (Ferraris et al., 2003; C. Ferraris, pers. commun.), this extremely diverse South American family is the largest within the Siluriformes and is renowned for its remarkable niche occupation, i.e., the scraping and sucking of algae and other food types off various substrates. Within the superfamily Loricarioidea the loricariids developed a highly specialized feeding apparatus, with a ventral suctorial mouth, tilted lower jaws and new muscle configurations that greatly increase jaw mobility as most eye-catching adaptations (Alexander, 1965; Schaefer & Lauder, 1986). A number of studies have focussed on the

∗ Slightly modified from: Geerinckx T., Brunain M. & Adriaens D., 2005. Development of the chondrocranium in the suckermouth armored catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Journal of Morphology, 266: 331-355. 38 PART 4 — ONTOGENY OF THE SKULL group, clarifying many aspects of the adult osteology and myology of the Loricariidae (a.o. Howes, 1983a; Schaefer, 1987, 1988; Schaefer & Lauder, 1986). Many questions concerning loricariid morphology are still unresolved. Virtually nothing is known about their ontogeny. One aspect was studied by Carter & Beadle (1931), who confirmed the development and function of the stomach as a respiratory organ in Liposarcus anisitsi. A critical question is whether a family with such aberrant adult head morphology shows the general siluriform tendencies during early development. Given the atypically siluriform adult morphology of Loricariidae, coupled to a peculiar, not completely understood feeding and repiratory behavior, one could question of how this affects early life stages. In addition, hatchlings appear to be able to attach to the substrate immediately, using their suckermouth, as noted by Riehl & Patzner (1991) in the loricariid Sturisoma aureum. A first step in answering questions concerning ontogeny and function in Loricariidae, and hence the ontogeny of function, is a proper knowledge of the changing morphology during ontogeny. This chapter deals with the development and growth of the chondrocranium in a representative loricariid species, the ‘bristlemouth’ suckermouth armoured catfish Ancistrus cf. triradiatus. The chondrocranium of several of the approximately 34 siluriform families has already been described. Accounts of one or more stages in the development of the chondrocranium are published for Ariidae (Ariopsis felis; Bamford, 1948; Arius jella; Srinivasachar, 1958a), Bagridae (Mystus vittatus, Rita sp.; Srinivasachar 1957a), Callichthyidae (Callichthys callichthys; Hoedeman, 1960b; Hoplosternum littorale; Ballantyne, 1930), Clariidae (Clarias gariepinus; Vandewalle et al., 1985; Surlemont et al., 1989; Surlemont & Vandewalle, 1991; Adriaens & Verraes, 1994, 1997c; Heterobranchus longifilis Valenciennes; Vandewalle et al., 1997), Claroteidae (Chrysichthys auratus, Vandewalle et al., 1999), Heteropneustidae (Heteropneustes fossilis; Srinivasachar, 1958b, 1959), Ictaluridae (Ameiurus nebulosus; Kindred, 1919), Pangasiidae (Pangasius pangasius; Srinivasachar, 1957b), Plotosidae (Plotosus canius; Srinivasachar 1958a), Schilbeidae (Ailia coila, Silonia silondia; Srinivasachar 1957b), and the suspensorium of Trichomycteridae (Arratia, 1990). Recent papers have shed light on generalities and trends, as well as the diversity in catfish chondrocrania (Arratia, 1992; Adriaens & Verraes, 1997c; Vandewalle et al., 1999; Adriaens & Vandewalle, 2003). The current study of the chondrocranium of a species of the family Loricariidae adds a rather aberrant type of siluriform to this list, and forms the basis of current work on the ontogeny of other structures in loricariids.

PART 4 — ONTOGENY OF THE SKULL 39

4.1.2. BRIEF MATERIAL AND METHODS

Twenty specimens were used for in toto clearing and staining (see paragraph 2.2.5). A 3D- reconstruction was made from serial sections of the 5.2 mm stage (details in paragraphs 2.2.7 and 2.2.8). Eight specimens have been described in detail, with reference to other specimens, or serial sections, wherever relevant. Serial sections of a Clarias gariepinus specimen have been used for comparison of the orbitonasal region.

4.1.3. RESULTS

The chondrocranium of Ancistrus cf. triradiatus is composed of cell-rich hyaline cartilage (Benjamin, 1990). Both appositional growth (proliferation of chondroblasts at the outer edge of cartilage) and interstitial growth (division of pre-existing, medially located chondrocytes, and subsequent addition of matrix) are observed during development. Matrix-rich hyaline cartilage is only found in the anterior cartilaginous head of the autopalatine bone in juveniles and adults, and not in the embryonic chondrocranium.

4.8 MM SL — 2 DAYS POST-FERTILIZATION

Serial sections show no evidence of cartilage or chondroblast differentiation in this stage.

5.2 MM SL — 3 DAYS POST-FERTILIZATION (FIG. 16)

NEUROCRANIUM Serial sectioning reveals the presence of a few cartilaginous structures. The anterior parts of the parachordal cartilages have formed, and in front of these the trabecular bars are well visible and continuous with the parachordal cartilages. Chondroblast differentiation at both sides of the tip of the notochord constitutes the onset of the acrochordal cartilage (Fig. 17A- B). The trabecular bars are wide apart and slightly curved, typical for platybasic teleosts, leaving a broad hypophyseal fissure. They do not touch rostrally yet. Except for the notochord, no supporting structures unite both halves of the young neurocranium. Differentiating chondroblasts are seen where the anterior otic cartilage will form.

SPLANCHNOCRANIUM The equally well stained hyoid bar is already present. The hyosymplectic-pterygoquadrate plate is less developed, but also visible to some extent. In this stage, no cartilage is seen at the future location of the interhyal. 40 PART 4 — ONTOGENY OF THE SKULL

5.6 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 18)

NEUROCRANIUM Most parts of the skull floor are now at least partly formed, supporting the developing brain and separating it from the underlying structures. The parachordal cartilages, bordering the notochord, and the collateral basiotic laminae, more anteriorly, are indistinguishably fused. The curved trabecular bars become broader rostrally, where they will soon form the solum nasi; they end at the ethmoid plate. In this stage it is impossible to distinguish the trabecular bars from the polar cartilages, as there is as yet no sign of a fissure for the arteria carotis interna yet; but, as deduced from the later stages, and by analogy with the observations of Adriaens & Verraes (1997c) and others, the posterior part probably corresponds to the polar cartilage. It is this part that connects with the basiotic lamina. The rudimentary acrochordal cartilage only covers the tip of the notochord dorsally, so that the notochord protrudes slightly into the hypophyseal fenestra. From posterior to anterior, the elements bordering the hypophyseal fenestra are: the tip of the notochord and the acrochordal cartilage, the plate-like basiotic lamina, the polar cartilages, the trabecular bars, and the ethmoid cartilage. The parachordal cartilages are connected with the otic capsule by means of the anterior basicapsular commissure at the level of the anterior otic cartilage. The posterior otic cartilage is continuous with the anterior one, and only distinguishable from it because it already carries a medial process that later will give rise to the basivestibular and posterior basicapsular commissures (see below). It also is less stained, indicating that it might have developed later than the anterior otic cartilage. The occipital pilae arise from the caudal ends of the parachordal cartilages and contact the posterior otic cartilages. The metotic fenestra, a large opening bordered by the parachordal cartilage medially, the anterior basicapsular commissure rostrally, the otic capsule laterally and the occipital pila caudally, accommodates the glossopharyngeal (IX) and vagal (X) nerves (as seen in serial sections of the 5.2 and 6.1 mm specimens). The lateral part of the otic capsule becomes closed now, except for a lateroventral opening in the capsule floor. The taenia marginalis starts to grow at the rostral end of the anterior otic cartilage. Near its origin a small foramen is present in the anterior otic cartilage. A part of the otic branch of the facial nerve, innervating the sensory canal, is seen passing through it in serial sections of the 6.1 mm and all later stages.

SPLANCHNOCRANIUM A short maxillary barbel cartilage is present at the base of the rudimentary maxillary barbel. Meckel’s cartilage has arisen, and bears a conspicuous coronoid process, which points dorsorostrally. The hyosymplectic-pterygoquadrate plate is continuous with the interhyal and PART 4 — ONTOGENY OF THE SKULL 41 the ceratohyal-hypohyal bar, and, albeit very weakly, with Meckel’s cartilage (this matrix- poor articular cartilage connection is only seen in serial sections). The hyosymplectic part has a foramen for the hyomandibular trunk of the facial nerve. Both hypohyals are continuous at the midline, whereas Meckel’s cartilages are not. No signs of the branchial basket are visible yet.

6.0 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 19)

NEUROCRANIUM The notochord still protrudes slightly into the hypophyseal fenestra (as in the previous stage, the acrochordal cartilage only covers the dorsal side of the tip of the notochord). The metotic fenestra is slightly constricted by a lateral projection of the parachordal cartilage and a broad medial process of the posterior otic cartilage. Serial sections of the 6.1 mm SL specimen show that this broad process encloses the glossopharyngeal nerve, thus proving its double nature, i.e., the combined onset of the basivestibular and posterior basicapsular commissures (see next stage for details on these commissures). The solum nasi can now be discerned as an anterior differentiation of the trabecular bars. The orbitonasal process grows upward on this solum nasi, towards the fully grown taenia marginalis. The latter branches, almost at the level of the orbitonasal process, in a medial extension, being the onset of the epiphysial bridge, and a short stub extending anteriorly. This minute stub could be called the (rudimentary) taenia marginalis ‘anterior,’ as opposed to the taenia marginalis ‘posterior,’ caudal to the epiphysial bridge. It will, however, branch near its origin in the next stage, reducing the taenia marginalis anterior almost completely. The sphenoid fissure is situated between the trabecular bar and the taenia marginalis. The postotic process is formed at the posterior end of the posterior otic cartilage, where it borders the occipital pilae. The occipital pilae form the occipital arch, from which the tectum posterius is developing. Like the epiphysial bridge, it is not yet continuous at the midline.

SPLANCHNOCRANIUM The palatine is visible. The posterior part of this cartilaginous element arises first, articulating with the solum nasi of the neurocranium. The maxillary barbel cartilage has lengthened. It consists of a row of flattened chondrocytes with little matrix between them, but surrounded by a thick layer of more darkly stained matrix (Fig. 17C). The pterygoquadrate- hyosymplectic, which articulates with the neurocranium at the level of the anterior otic cartilage, is still continuous with the interhyal and the ceratohyal-hypohyal bar, which bears a conspicuous ventral process. The pterygoquadrate-hyosymplectic now bears a rudimentary pterygoid processs. Meckel’s cartilages are fusing medially. The exact location of the boundary between ceratohyal and hypohyal elements cannot be made until the onset of 42 PART 4 — ONTOGENY OF THE SKULL ossification, as there is no clear hyoid artery incision in the chondrocranium of Ancistrus cf. triradiatus. The position of this incision can be used in distinguishing both elements in these early stages (Adriaens & Verraes, 1997c). The future position of the ossa hypohyale and ceratohyale (not shown) is used to distinguish both cartilage elements. Both hypohyals are still merged, and continuous with a medial bar comprising the first and second basibranchial. These elements are all fused from the beginning, and will only later separate (see later stages). No other basibranchials are present in this specimen. This bar, however, proves to be longer in the serial sections of the 6.1 mm specimen, up to the level of the third branchial arch, so most probably it includes the third basibranchial and thus corresponds to the anterior copula. Ceratobranchials I-IV and hypobranchials I-II are present. Of these elements, ceratobranchials I-III are more intensely stained with alcian blue, so probably arise first during ontogeny. Serial sections of a 6.1 mm stage suggest that corresponding cerato- and hypobranchials I-II arise as one unit. Epibranchials I-III are also present.

6.8 MM SL — 5 DAYS POST-FERTILIZATION (FIG. 20)

NEUROCRANIUM The skull floor has now become more solidly chondrified, with a broadened solum nasi, and the onset of anterior lengthening of the ethmoid cartilage (this lengthening will go on during further ontogeny). A small precerebral process is present on the tip of this ethmoid plate. This structure starts as two separate projections (6.0 mm stage), soon fusing, but keeping two distinct tips (6.8 mm and 7.4 mm stages). This is also corroborated by serial sections of the 7.0 mm SL specimen. In the skull roof, the anterior tip of the taenia marginalis develops further, with the rudimentary epiphysial bridge still growing (though still not touching medially), and a bipartite stub at its anterior end. This stub (referred to in the previous stage as the rudimentary taenia marginalis anterior) branches into a small mediorostral sphenoseptal commissure, growing in the direction of the precerebral process, and a lateroventral spheno-ethmoidal commissure, which grows towards a dorsal projection of the skull floor, the orbitonasal process. When contact is established, a compound transverse plate is formed, the orbitonasal lamina. A reinforcement in the corner between the spheno-ethmoidal commissure and the taenia marginalis fuses with a more caudal projection of the skull floor, forming the preorbital base, and leaving a small foramen for the ophthalmic branch of the trigeminal nerve. Both latter connections are established somewhere between the 6.8 mm 7.4 mm SL stages. Examination of serial sections of a 7.0 mm stage reveals that the orbitonasal lamina is fully formed and the preorbital base nearly so. There is no apparent acrochordal cartilage in this stage, leaving both sides of the skull floor separated in this region. PART 4 — ONTOGENY OF THE SKULL 43

The metotic fenestra has been divided into several small fenestrae. The two medial processes of the posterior otic cartilage (as seen in serial sections of the 6.1 mm stage) have connected to the lateral extension of the parachordal cartilage, forming the basivestibular commissure and the posterior basicapsular commissure. The posterior basicapsular fenestra, between these two commissures, accommodates the glossopharyngeal nerve (n. IX). The anterior basicapsular fenestra, between the anterior basicapsular and the basivestibular commissure, will shrink and disappear almost completely later during ontogeny. More caudally two posterior, obliquely oriented foramina are situated between the posterior basicapsular commissure and the occipital pila. These also are remnants of the larger metotic fenestra; they are separated by a thin strut of cartilage. The medial one will soon disappear; the lateral one stays throughout ontogeny, and accommodates the vagal nerve (n. X). Serial sections of a 6.1 mm stage prove that no nerve or blood vessel passes through the lateral opening in the otic capsule floor, lateral to the anterior basicapsular fenestra and anterior of the lateral semicircular septum. It seems to be closed by a membrane. Later it will form the recess for the utriculus of the inner ear. The lateral semicircular septum (dotted lines in Fig. 22) connects the floor and the roof of the otic capsule, and is surrounded by the horizontal semicircular canal. Anteriorly, only observed at the left side, a small blastema arises from the otic capsule (Fig. 20D). This is the prootic process, described by Swinnerton (1902), Bertmar (1959), and Daget (1964) as the onset of the lateral commissure (see Discussion for details). Dorsally, the otic capsule has two large fenestrae, not observed in other siluriform chondrocrania. One is situated more or less between the anterior and posterior otic cartilages (which now can no longer be distinguished), the other in the second half of the posterior otic cartilage, close to the postotic process. The names anterior and posterior otic fenestra are proposed for these structures. The tectum posterius is complete, both parts having fused medially, and closes the foramen magnum. At the dorsomedial margin of the otic capsules, anterior to the tectum posterius, small extensions can be seen that might correspond to a rudimentary tectum synoticum (see Discussion). As proved by later stages, however, they do not grow significantly.

SPLANCHNOCRANIUM The pterygoid process, only a short projection in the 6.0 mm stage, now further develops on the anterior edge of the pterygoquadrate-hyosymplectic complex, which remains bar- shaped on lateral view throughout the development. The ceratohyal part of the hyoid bar now bears four distinct processes near its lateral end: a small one oriented rostrally; one oriented dorsally, behind the interhyal connection; one oriented caudally; and a very large one 44 PART 4 — ONTOGENY OF THE SKULL oriented ventrally, pointing in the direction where the branchiostegal rays will develop (and articulate). Hypobranchials III-IV and epibranchial IV are added to the branchial basket. All hypobranchials are continuous with the corresponding ceratohyals. Basibranchials I to III, composing the first copula, are present and confluent with the hyoid bar.

7.4 MM SL — 6 DAYS POST-FERTILIZATION (FIG. 21)

NEUROCRANIUM In this stage all major components of the cartilaginous skull have formed. Remarkably, and opposed to the situation in previous stages, the acrochordal cartilage is well developed in this stage, also covering the rostral and ventral sides of the tip of the notochord. The sphenotic fenestra is now well demarcated. The epiphysial bridge is completed, so now prepineal and postpineal fontanelles can be discerned. The former is still continuous with the foramen filae olfactoriae, as the sphenoseptal commissures and the forked precerebral process still do not touch. The orbitonasal lamina grows laterally, forming a prominent transverse sheet. Ventral to the foramen of a branch of the orbitonasal vein, the larger orbitonasal foramen (for the orbitonasal artery) is now separated from the foramen filae olfactoriae. In the orbitonasal lamina a rostrocaudal foramen is now clearly seen, accommodating the superficial ophthalmic branch of the trigeminal nerve. The prootic process of the otic capsule has formed the lateral commissure on the right side, but is still growing on the left side (see also Fig. 17D). It grows from the rostroventral edge of the anterior otic cartilage to the rostral end of the polar cartilage, thus dividing the sphenoid fenestra into a large anterior fenestra and a small posterior fenestra (Fig. 17E). The taenia marginalis develops a postorbital process, including the foramen for the otic branch of the facial nerve. In this and in the next stages, the asymmetrical rudiments of the tectum synoticum sometimes demarcate a small foramen where the lateral accessory branch of the facial nerve passes. Caudal reinforcement of the skull starts with the fusion of the tectum posterius and the paired cartilaginous precursors of the neural arch of the fifth and/or sixth vertebra (see Discussion).

SPLANCHNOCRANIUM Hypo- and ceratobranchials I-II become separated; III-IV will remain continuous until ossification. A fifth pair of ceratobranchials is present (this is the only element of the fifth branchial arch to appear). As for the basibranchials, two cartilaginous structures are present: the first one consists of basibranchials I-III, and is still weakly connected to the hyoid bar; the second one consists of basibranchials IV-V. These two compound elements correspond to the PART 4 — ONTOGENY OF THE SKULL 45 anterior and posterior copula respectively. A small uncinate process develops on the third epibranchial.

8.0 MM SL — 7 DAYS POST-FERTILIZATION (FIG. 22)

NEUROCRANIUM The notochord in the cranium has now shrunken to half its postcranial diameter. The prepineal fontanelle and the foramina filae olfactoriae are now completely separated by the fusion of the sphenoseptal commissures and the (double) precerebral process. A transverse reinforcement starts to grow between both tips of the precerebral process, forming a precerebral lamina. The acrochordal cartilage is seen only underneath the rostral tip of the notochord. The lateral commissure is complete on both sides. Slightly more caudally, another small blastema appears on the rostroventral edge of the anterior otic cartilage. It is also visible in the following stages, but not at both sides. It never connects to the skull floor. The anterior part of the skull is lengthening more, and the ethmoid plate develops a ventral protuberance at the rostral tip. The anterior basicapsular fenestra shrinks and splits off a small caudal fenestra, which will disappear later during ontogeny. The tectum posterius grows stronger, broadening in an anterior but mostly a posterior direction, so that the dorsal connection between both otic capsules is reinforced.

SPLANCHNOCRANIUM As the snout region of the neurocranium lengthens, the pterygoquadrate-hyosymplectic becomes more elongate as well. The hyosymplectic bears a conspicuous opercular process. The retroarticular process of Meckel’s cartilage is very small, only visible as a small stub caudolateral to the articulation with the quadrate. The thin connection of articular cartilage between Meckel’s catrilage and the quadrate is no longer seen in serial sections of the 8.0 mm SL specimen. The center of the hyoid bar is only slightly stained by alcian blue, indicating that the hypohyals are becoming separated. The anterior copula of the branchial basket is shrinking, as basibranchial I is becoming reduced and basibranchial III becomes separated. Two infrapharyngobranchials have appeared; their location confirms that they are infrapharyngobranchials III and IV. Just behind the medial ends of the hypohyals a double, dumbbell-shaped cartilaginous nucleus is present, which will later become part of the bony parurohyal. In serial sections of the 8.0 mm stage it is seen that it is continuous with the first basibranchial, which is not well seen in the stained specimen (and further reduces in the next stages) (Fig. 17F).

46 PART 4 — ONTOGENY OF THE SKULL

8.9 MM SL — 8 DAYS POST-FERTILIZATION (FIG. 23)

NEUROCRANIUM No major transformations occur in the cartilaginous neurocranium during this stage. The rostro-caudal elongation of the snout region proceeds, as does the reinforcement of the occipital region: the tectum posterius becomes more and more extended posteriorly. The prepineal fontanelle becomes smaller, as the precerebral lamina extends backward. The outline of the hypophyseal fenestra changes: a medial fissure appears between the trabecular bar and the polar cartilage, accommodating the internal carotid artery. The appearance of this fissure seems to be the result of allometric growth of the trabecular bars and the polar cartilages: they simply broaden everywhere except at the site of the fissure. The lateral end of the orbitonasal lamina grows slightly rostrally, around the nasal sac, while the articular facet of the solum nasi for the palatine becomes ever more prominent. Serial sections of the 8.0, 10.2 and 12.4 mm stages allow a reconstruction of the main nerve paths in the sphenoid region (Fig. 24). The olfactory nerve exits via its separate foramen. The sphenoid fenestra is penetrated by the optic, oculomotor, trochlear and abducens nerves, as well as by the main part of the trigeminal and facial nerves. The hyomandibular trunk and opercular branch of the facial nerve exit posterior to the lateral commissure [as do a vein and an artery, probably the orbital artery (de Beer, 1927)], and the otic branch rises and leaves the skull via the postpineal fontanelle, close to the taenia marginalis. One division of the otic branch pierces this taenia at the level of the postorbital process. Two branches of the trigeminal nerve pass through the orbitonasal lamina; one part (unclear homology) passes through a groove at the ventral side of the lamina (but goes through a ventral foramen in the right side of the 10.2 mm stage); the other (superficial ophthalmic branch) always pierces the dorsal part of the lamina. Two other foramina in this region are not penetrated by any nerves: the orbitonasal foramen accommodates the orbitonasal artery, and a more dorsal foramen accommodates a branch of the orbitonasal vein.

SPLANCHNOCRANIUM The medial connection between the hypohyal parts of the hyoid bar is now completely invisible in the stained specimen: the bar is no longer continuous. In serial sections of a 10.2 mm specimen, however, it is still visible as a frail and thin rostral sheet. The connection between Meckel’s cartilages has disappeared. Their coronoid processes, however, are becoming more substantial. The first basibranchial seems to have been reduced completely.

PART 4 — ONTOGENY OF THE SKULL 47

9.9 MM SL — 10 DAYS POST-FERTILIZATION (FIG. 25)

NEUROCRANIUM There is little shape difference with the previous stage. The tip of the notochord becomes squeezed between the parachordal cartilages. The anterior basicapsular fenestra has disappeared. Due to the lengthening of the skull and the fully grown tectum posterius, the ratio of the chondrocranial skull length to skull height is now 4, compared to 2.9 in the 6.0 mm stage. In general, the chondrocranium is now slowly being replaced by the osteocranium.

SPLANCHNOCRANIUM Just below the anteroventral end of the palatine a small submaxillary cartilage has appeared. This is also visible in serial sections of the 8.0 mm specimen. The cartilaginous nucleus of the parurohyal is no longer stained by alcian blue, but can still be seen on sections of the 10.2 and 12.4 mm stages. The second copula and the central shafts of the epi- and ceratobranchials are also no longer stained.

4.1.4. DISCUSSION

Compared to other siluriforms in which the chondrocranium has been studied and of which data of the pre-hatching period and of the first appearance of the chondrocranium are available, the cartilaginous cephalic skeleton of Ancistrus cf. triradiatus is already remarkably well developed at hatching. A comparable state of development has been observed in the non-siluriform three-spined stickleback Gasterosteus aculeatus and the brown trout Salmo trutta fario (Swinnerton, 1902; de Beer, 1927). But even compared to these two species, A. cf. tririadiatus has a more developed chondrocranium at the moment of hatching, even though it has a much shorter pre-hatching period. Obviously there is a tendency that species hatching very early lack chondrocranium elements at hatching. In Heterobranchus longifilis, Clarias gariepinus and Chrysichthys auratus, both African catfishes, no cartilaginous structures are present at hatching, which occurs about one day after fertilization (Vandewalle et al., 1997; Adriaens et al., 1997a; Vandewalle et al., 1999). It would be interesting to elaborate on the state of development of the cranium at key moments (hatching, complete resorption of yolk sac) in different species, but it is difficult to obtain the needed amount of data for more species. As in most siluriforms of which data are available, in A. cf. triradiatus the first elements of the neurocranium and the splanchnocranium appear more or less simultaneously.

48 PART 4 — ONTOGENY OF THE SKULL

NEUROCRANIUM

SKULL FLOOR The first structures to arise in the chondrocranial skull of Ancistrus cf. triradiatus are the parachordal cartilages and the trabecular bars. As in other siluriforms, the skull is platybasic, in contrast to the derived tropibasic skull type in most other teleosts (variation in the degree of trabecular fusion does exist) (Swinnerton, 1902; Bhargava, 1958; Verraes, 1974a; Wagemans et al., 1998). In some siluriforms the ethmoid plate may be broad, and can consequently be incorrectly considered as a trabecula communis (Srinivasachar, 1958a). The platybasic skull type has been linked to the reduced eye size that is typical of catfishes (Verraes, 1974b; Adriaens & Verraes, 1997b). In all examined siluriforms, including Ancistrus cf. triradiatus, each trabecular bar and the collateral parachordal cartilage arise as one part. In teleosts this isn’t a general rule (Swinnerton, 1902; de Beer, 1927; Vandewalle et al., 1992). The notochord becomes more or less surrounded by the basal plate, which develops from the fusion of the parachordal cartilages. In silurifoms this plate usually starts as a small acrochordal cartilage, covering the dorsal, ventral and/or rostral side of the tip of the notochord to a various extent. In Ancistrus cf. triradiatus, the acrochordal cartilage, which herein can be considered the most rostral part of the basal plate, is variably present in the various stages examined in this study. The presence of cartilaginous tissue dorsal, ventral or rostral to the tip of the notochord was determined in the cleared and stained specimens: 5.6 mm: dorsal; 6.0 mm: dorsal; 6.8 mm: nothing; 7.4 mm: dorsal, ventral and rostral; 8.0 mm: dorsal; 8.9 mm: dorsal and rostral; 9.9 mm: dorsal and rostral. The absence of cartilage above or below the notochord might be due to insufficient alcian blue staining; the cartilage there is usually only one or two cell layers thick. In the specimens that underwent serial sectioning the presence of this cartilage also proved to be highly variable, but when present, was always quite visible: 5.2 mm: nothing; 6.1 mm: dorsal and almost ventral; 7.0 mm: dorsal and ventral; 8.0 mm: dorsal; 10.2 mm: dorsal and rostral, 12.4: dorsal and rostral. In other siluriforms the acrochordal cartilage has been reported to consist of a hypochordal or an epichordal bridge, or a combination, also covering the rostral tip of the notochord (Kindred, 1919; Bamford, 1948; Srinivasachar 1957a, b; Adriaens & Verraes, 1997c). The notochord actually protrudes into the hypophyseal fenestra only in the earliest stages of Ancistrus cf. triradiatus, a situation also seen in Ariopsis felis, Arius jella and Callichthys callichthys (Bamford, 1948; Srinivasachar, 1958a; Hoedeman, 1960b), but not in Clarias gariepinus (Adriaens & Verraes, 1997c). Two hypotheses could explain the rostral position of the acrochordal cartilage in the later stages: the tip of the notochord degenerates early [as stated by Goodrich (1958)], or the acrochordal cartilage extends rostrally during PART 4 — ONTOGENY OF THE SKULL 49 development. In the sea trout Salmo trutta trutta de Beer (1937) saw the formation of the prootic bridge out of a membrane situated rostral and dorsal to the notochord tip, thus at the position of the dorsally situated part of the acrochordal cartilage, or epichordal bridge, in A. cf. triradiatus. Here the ontogenetic series suggests that the basiotic laminae of both sides add to the acrochordal cartilage, thus narrowing the end of the hypophyseal fenestra and giving rise to the epichordal or prootic bridge, as seen in the 9.9 stage (Fig. 25). The trabecular bars in teleosts usually undergo transformations for the passage of the paired internal carotid artery, which is situated caudally in the hypophyseal fenestra, rostromedial to the polar cartilage. In several siluriforms the artery moves to a more lateral position and cartilage resorption affects the inner side of the bars so as to accommodate it [e.g. Clarias gariepinus (Adriaens & Verraes, 1997c)]. In Chrysichthys auratus and the non- siluriform Barbus barbus the bars even reduce completely at the level of this artery (Vandewalle et al., 1999; Vandewalle et al., 1992); in Scophthalmus maximus the trabecula communis goes through the same complete reduction (Wagemans et al., 1998). No evidence of cartilage reduction is present in Ancistrus cf. triradiatus. Although the outline of the hypophyseal fenestra does change, and a medial fissure appears, the appearance of this fissure is the result of allometric growth of the trabecular bars: the bars just broaden everywhere except there, and the hypophyseal fenestra becomes narrower. No cartilage resorption is seen in the serial sections. In the brown bullhead Ameiurus nebulosus and Ariopsis felis the bars seem to become narrower. Whether they completely reduce is not clear (Kindred, 1919; Bamford, 1948). Srinivasachar (1957b) reported the artery in a small foramen in the trabecular bar in the gangetic ailia Ailia coila. Remarkably, in Hoplosternum littorale and Callichthys callichthys a constriction of the hypophyseal fenestra is present, anterior of where the trabecular fissure would be expected (Ballantyne, 1930; Hoedeman, 1960b). No information was given, however, on the position of the internal carotid artery.

At the level of the nasal sacs in siluriforms, each trabecular bar often forms a broad solum nasi. However, in Ancistrus cf. triradiatus it fails to grow significantly after the 6.8 mm stage, leaving the nasal sacs without a real floor, as is also the case in Arius jella and Plotosus canius (Srinivasachar, 1958a). Srinivasachar also noticed already that there is a lot of variation in both the ventral and lateral support of the nasal sacs (the latter due to a variably developed rostral extension of the orbitonasal lamina, almost absent in A. cf. triradiatus). The ethmoid plate is an unpaired, horizontal plate originating from, and uniting the tips of the trabecular bars. Swinnerton (1902) distinguished two separate primordia of the ethmoid plate at the tip of each young trabecular bar in the non-siluriform Gasterosteus aculeatus. In Ancistrus cf. triradiatus it is V-shaped anteriorly; more posteriorly, it is flat, as in most catfishes. In Ameiurus nebulosus, much of it is also V-shaped (Kindred, 1919). The ethmoid 50 PART 4 — ONTOGENY OF THE SKULL plate of A. cf. triradiatus is rather narrow, with a long, rostral extension. This extension is unique in catfish chondrocrania described thus far, and is related to the specialized jaws: the upper jaws of free-living embryonic, juvenile and adult Loricariidae are situated well in front of the lower jaws, the latter being turned backward. Hence the supporting structures of the upper jaws are relatively elongated. There are no ethmoid cornua [pre-ethmoid cornua of Adriaens & Verraes (1997c)] at both sides of the tip. There are, though, two more caudal processes at the rostral end of the solum nasi that might be homologous to the ethmoid cornua of other siluriforms, although the vicinity of the articular facet for the palatine contradicts this hypothesis.

SKULL ROOF A major component of the skull roof in Ancistrus cf. triradiatus is the taenia marginalis [alisphenoid cartilage of Kindred (1919); anterior process or supraorbital bar of Ballantyne (1930); orbital cartilage of Srinivasachar (1957a, b, 1958a) and Hoedeman (1960b)]. As is a generality in siluriforms it originates from the anterior end of the otic capsule, and not as a separate element, as can be observed in many other teleosts (de Beer, 1927; Adriaens & Verraes, 1997c). The absence of a real taenia marginalis anterior, in front of the epiphysial bridge, as observed in Ancistrus cf. triradiatus, also conforms to a trend in siluriforms [a short taenia marginalis anterior persists in Arius jella and Plotosus canius, which both have fairly long and narrow chondrocrania (Srinivasachar, 1958a)]. It is present in many other teleosts [e.g. Hepsetus odoe (Characiformes; Bertmar, 1959)]. Also as typical for siluriforms (Srinivasachar, 1957a), the taenia marginalis posterior (part behind the bridge) does not become discontinuous. A well-developed taenia tecti medialis posterior is not present in siluriform chondrocrania. In Ancistrus cf. triradiatus, the shape of the epiphysial bridge at the midline varies, and in the 8.0 mm stage a small posterior curvature may be seen. This has also been detected in Rita sp. and Clarias gariepinus (Srinivasachar, 1957a; Adriaens & Verraes, 1997c), but in these catfishes it is a small rudiment compared to the situation in various non-siluriform skulls [an extreme example is Heterotis niloticus, with four separate fontanelles (Daget & d’Aubenton, 1957)].

The posterior part of the skull roof in Ancistrus cf. triradiatus consists of no more than a tectum posterius originating from the occipital pilae, that rise from the rear part of the parachordal cartilages. The closure of this bridge-like structure around the time of hatching is the first dorsal fortification of the cartilaginous skull, and corresponds to the formation of the foramen magnum. A tectum synoticum, formed by a fusion of the posterior otic cartilages, is PART 4 — ONTOGENY OF THE SKULL 51 absent in A. cf. triradiatus, as in C. callichthys (Hoedeman, 1960b) and Clarias gariepinus (Adriaens & Verraes, 1997c). Kindred (1919) and Srinivasachar (1957a, b, 1958a) mention a ‘practically reduced’ tectum synoticum, indistinguishably fused with the tectum posterius. They provide no data from early embryonic stages, which could help distinguish the origin of both parts. The occipital pilae are those parts situated behind the vagal nerve foramen in the skull floor, but more dorsally the difference is more difficult to see when no early stages are available. In A. cf. triradiatus the posterior otic fenestra is situated in the skull roof, anterior to the occipital pila. The small medial outgrowths of the otic capsule noticed in most stages described herein (after hatching) may, however, correspond to rudiments of the tectum synoticum. Similar projections were noticed by Bamford (1948) in Ariopsis felis, who also considered them to represent this tectum synoticum. There, a longitudinal groove is present at each side along the posterior end of the otic capsule, accommodating the lateral accessory branch of the facial nerve. This branch goes from the ganglionic mass of the facial nerve to the dorsal body musculature, exiting the skull before the tectum posterius, and lying on top of the postotic process. In various stages of A. cf. triradiatus this nerve penetrates the rudimentary tectum synoticum, or passes through a small slit (e.g., Fig. 20A). The fusion of the tectum posterius with elements of the first vertebrae in ostariophysans, as well as the ontogeny of the Weberian apparatus, is still a problematic topic, although many aspects have been resolved (Fink & Fink, 1981; Radermaker et al., 1989; Coburn & Futey, 1996). The ontogeny of the Weberian apparatus and the complex vertebra, however, will not be discussed here. Among catfishes, this fusion seems variable, or, at least, difficult to interpret: Kindred (1919) noticed a close contact between the tectum posterius and the third neural arch in Ameiurus nebulosus; Bamford (1948) mentioned the role of the third and fourth supradorsals of either side fusing into one mass of cartilage, including the third radial, in Arius jella. In Ancistrus cf. triradiatus, the anteriormost basidorsals seem to fuse with the corresponding supradorsals (Fig. 21E). These are not seen as separate cartilages in early stages. The next basidorsal and basiventral correspond to the first vertebra developing (large) ribs (pers. ob.), which Regan (1911b) and later authors named the sixth vertebra. This suggests that the anteriormost basidorsals might be part of the fifth vertebra.

SKULL WALL The skull wall in the ethmoid and orbital regions in Ancistrus cf. triradiatus, as in other teleosts, consists of vertical commissures connecting the ethmoid plate and trabecular bars with the taeniae marginales. The origin of these commissures can be single (growing from one of the above structures) or double (a dorsal and a ventral part growing towards each other). The anteriormost of these commissures has two possible points of origin: in Heteropneustes longifilis a broad transverse process, the precerebral lamina, rises from the 52 PART 4 — ONTOGENY OF THE SKULL anterior edge of the ethmoid cartilage, forks and grows towards the anterior ends of both taeniae (Vandewalle et al., 1997). In Clarias gariepinus most of it originates from the taeniae, where a sphenoseptal commissure emerges rostrally (Adriaens & Verraes, 1997c), and connects with the small precerebral lamina. The result looks much the same in both cases, and the broad lamina seems to be correlated to the broad ethmoid plate [as in Callichthys callichthys too (Hoedeman, 1960b)]. In A. cf. triradiatus both points of origin contribute equally. Moreover, its ethmoid cartilage is narrow, and the precerebral process doesn’t form a real lamina, but forks from the start (7.4 mm stage; Fig. 21). Later (10.0 mm stage) an oblique sheet (also called the precerebral lamina) fills the anterior end of what has become the prepineal fontanelle, as in the silond catfish Silonia silondia, the yellowtail catfish Pangasius pangasius and Rita sp., and in the striped dwarf catfish Mystus vittatus, where it has become so large that it has been called the tectum or the tegmen cranii (Srinivasachar, 1957a, b). In the latter species, and in Ailia coila and Arius jella, a posterior mediosagittal extension of the precerebral process, the internasal septum, separates (the anterior parts of) both nasal sacs. The precerebral lamina and the internasal septum can be considered homologous (Daget, 1964), and sometimes appear to grow very allometrically. Rita sp. of 12 mm TL has no septum at all (Srinivasachar, 1957a), while adult Rita rita (Hamilton) (formerly R. buchanani) has an unmistakable cartilaginous internasal septum (Bhimachar, 1933). An internasal septum is absent in A. cf. triradiatus. It is fairly common in tropibasic skulls (de Beer, 1927).

The next vertical commissure is the orbitonasal lamina [preorbital process or ectethmoid cartilage of Ballantyne (1930); orbitonasal lamina sensu largo of Adriaens & Verraes (1997c)], a transverse sheet composed of a ventrolateral outgrowth of the taenia marginalis, the spheno-ethmoidal commissure, and a dorsal process of the solum nasi, the orbitonasal process [orbitonasal lamina sensu stricto of Adriaens & Verraes (1997c)]. The term orbitonasal process is introduced herein to avoid confusion. The compound nature of the orbitonasal lamina, as seen in Ancistrus cf. triradiatus, has been confirmed by Adriaens & Verraes (1997c) in Clarias gariepinus as well. The lamina often carries a laterorostral process that protects the nasal sacs laterally; in A. cf. triradiatus this is rudimentary. Hoedeman (1960b) mistakenly called the first rudiments of the orbitonasal lamina the sphenoseptal commissure (see above). In several siluriforms this is the first pre-otic vertical commissure to develop (Ballantyne, 1930; Adriaens & Verraes, 1997c; Vandewalle et al., 1997). However, in A. cf. triradiatus another, more medial commissure, the preorbital base [preoptic root of Srinivasachar (1957b); lamina preorbitalis of Vandewalle et al. (1999)] appears almost simultaneously. In most siluriforms it is formed well after the orbitonasal lamina, but serial sections of the 7.0 mm stage show their almost synchronized formation. PART 4 — ONTOGENY OF THE SKULL 53

The preorbital base too, consists of a dorsal part, originating from the taenia marginalis, and a ventral part, rising from the trabecular bar. Fenestrae in this region are variably present in siluriforms, and have received various names, causing some terminology confusion. In Ancistrus cf. triradiatus, as in all siluriforms, the most rostral of these fenestrae is the foramen for the fila olfactoria (olfactory foramen or foramen I), innervating the nasal organ. When the orbitonasal lamina, bordering it posteriorly, is situated more rostrally, as in Mystus vittatus and Arius jella, the foramen reduces to a very small opening, directed more rostrocaudally (Srinivasachar, 1957a, 1958a). An orbitonasal foramen [orbital foramen of Kindred (1919); preoptic fontanelle or foramen of Srinivasachar (1957b)] is most often present between the orbitonasal lamina and the preorbital base in catfishes. In Ancistrus cf. triradiatus it becomes smaller during ontogeny as the preorbital base enlarges, and is pierced by the orbitonasal artery. Depending on the size of the preorbital base and its position relative to the orbitonasal lamina, the direction of the orbitonasal foramen may be rostrocaudal or mediolateral (and sometimes becoming very large), leading to misinterpretations (Adriaens & Verraes, 1997c). Figure 26 shows the different orientations of the foramen, due to the size of the preorbital base. The foramen seems to be completely absent in Bagridae (Srinivasachar, 1957a). Two more foramina are present in this region in Ancistrus cf. triradiatus: dorsal to the orbitonasal foramen a small foramen is seen in the preorbital base, containing a branch of the orbitonasal vein, as observed in serial sections of 7.0 mm and later stages, and seen by Bamford (1948) in Ariopsis felis as well. Another foramen pierces the orbitonasal lamina rostrocaudally, accommodating a part of the superficial ophthalmic branch of the trigeminal nerve, which innervates the skin of the dorsal snout region (not to be confused with the identically termed branch of the facial nerve, which runs caudal to it and innervates the supraorbital lateral line organs). A second part of this branch, innervating the skin lateral to the naris, passes below the lateral side of the lamina, through a small ventral slit, or through a foramen (in the right side of the 10.2 mm stage): this varies between examined specimens. The foramen for the superficial ophthalmic branch is generally featured in siluriforms, except for Rita sp. (Srinivasachar, 1957a). In Srinivasachar’s paper a ventral groove in the lamina carries the so-called profundus branch of the same nerve. Still according to that author, in Mystus vittatus one part of the nerve runs through a groove at the dorsal side of the lamina; another part runs through a foramen. In Arius jella, Srinivasachar (1958a) mentions the course of the superficial ophthalmic and profundus branches through two distinct dorsal foramina in the orbitonasal lamina. The identity of this ‘profundus’ branch should be investigated, as it is normally characterized by a path ventral to the eye musculature and its nerves, and is absent in the black bullhead Ameiurus melas, and in most bony fishes (Workman, 1900). 54 PART 4 — ONTOGENY OF THE SKULL

The optic, oculomotor, trochlear, trigeminal, abducens and facial nerves exit the skull via the sphenoid fenestra in all siluriforms, as is typical in teleosts. The only exception known so far is Ailia coila, in which Srinivasachar (1957b) noted a separate foramen for the oculomotor nerve in the preorbital base. As can be seen in all post-hatching stages, Ancistrus cf. triradiatus shows a very peculiar feature in having a vertical structure identical to the lateral commissure as observed in several fishes, though not in other siluriforms (de Beer, 1937) (Fig. 17D). Lateral to the trigemino-facialis chamber it originates as a prootic process from the anterior otic capsule, connecting with the anterior end of the polar cartilage. Only the hyomandibular trunk and opercular branch of the facial nerve (immediately giving rise to the hyomandibular and opercular branches), an artery (possibly the orbital artery) and a vein exit behind it (Fig. 17E). The lateral commissure in Gasterosteus aculeatus and Hepsetus odoe (Swinnerton, 1902; Bertmar, 1959) is formed in exactly the same way. In Salmo trutta fario it is formed from two sides: a postpalatine process originates from the basiotic lamina and connects to the prootic process (de Beer, 1927). The absence of a lateral commissure was previously considered typical in catfishes (de Beer, 1937; Daget, 1964), but is obviously present in A. cf. triradiatus. A second, small blastema posterior to the prootic process (as seen in the older stages described herein) is variably present, and does not seem to correspond to any other structure described in teleosts. Serial sections reveal it as a very thin, almost membranous projection.

In general, catfishes are believed to lack true myodomes accommodating the eye muscles as seen in most other teleosts (de Beer, 1937). This might be due to the lesser eye sizes, and consequently smaller extrinsic eye muscles in catfishes (Arratia, 2003). Nonetheless, Ancistrus cf. triradiatus possesses a posterior myodome resembling very well the configuration as described by de Beer (1937) in Salmo trutta trutta (Fig. 17G). The external rectus muscle enters the myodome laterally, and penetrates deepest into it. The internal rectus muscle also penetrates into it, and inserts on the developing parasphenoid bone. The inferior rectus muscle enters the braincase, but not the canal formed by the parasphenoid bone and the prootic bridge, and attaches on the basiotic lamina. Meanwhile, the superior rectus muscle inserts on this lamina anterior to the passage of the other three muscles into the braincase. The internal rectus muscle lies medial to the external one, while in S. trutta trutta it lies underneath it. This myodome can also be seen in adult A. cf. triradiatus. A smaller, anterior myodome is present too, housing the obliquus eye muscles: both superior and inferior obliquus muscles enter the braincase through the orbitonasal foramen and attach on the solum nasi. McMurrich (1884:297) observed a rudimentary, “almost aborted” posterior myodome-like structure in Ameiurus nebulosus. Kindred (1919) saw no evidence of this in 10 and 32 mm PART 4 — ONTOGENY OF THE SKULL 55 stages of the same species: the rectus muscles insert on the lateral surface of the trabecula in the posterior part of the orbit. Similarly, Srinivasachar (1957b) mentioned the absence of a posterior myodome in 8 and 18 mm stages of (a.o.) Silonia silondia, while Bhimachar (1933) observed a small myodome in adults of the same species (then named Silandia gangetica). In developmental studies on catfishes, the only mention of a small anterior myodome as in Ancistrus cf. triradiatus is in Rita sp. by Srinivasachar (1957a). Loricariidae have relatively large eyes, needed in clear waters where these fishes often occur, especially since loricariids lack most barbels (only the maxillary barbels are more or less developed). Adriaens & Verraes (1997b) found that on average, after Callichthyidae, Pimelodidae and Schilbeidae, Loricariidae have the largest eyes among siluriforms. Schaefer (1997) and Aquino (1998) described myodomes in adult hypoptopomatin loricariids.

In the auditory capsule of teleosts, the anterior otic cartilage generally develops first, very soon giving rise to the posterior one (de Beer, 1927; Goodrich, 1958; Daget, 1964; Adriaens & Verraes, 1997c). In Ancistrus cf. triradiatus this sequence is also suggested by the presence of chondroblasts in the 5.2 mm stage where the anterior otic cartilage will form. In the 5.6 mm stage the anterior part is clearly more developed, being much better stained by alcian blue. The initial fusion of the posterior otic cartilage with the skull floor differs from the sequence noticed in Salmo trutta fario and Clarias gariepinus (de Beer; 1927; Adriaens & Verraes, 1997c), since it first contacts the bases of the occipital pilae (5.6 mm), and only later touches the parachordal plate directly (6.8 mm). As in other teleosts, no nerve or other structure passes through the anterior basicapsular foramen (Daget, 1964). The foramen is absent in some catfishes, like Ameiurus nebulosus (Kindred, 1919) and Silonia silondia (Srinivasachar, 1957b). In the latter species, and in Pangasius pangasius and Ailia coila both nerves exit through the same opening (Srinivasachar, 1957b), which might point to the absence of the posterior basicapsular commissure. In Ancistrus cf. triradiatus the posterior cerebral vein does not exit the skull via the foramen for the vagal nerve, which, as Goodrich (1958) postulates, is the case in most fishes. In A. cf. triradiatus it passes laterally through the foramen magnum. At closure, the anterior basicapsular fenestra in Ancistrus cf. triradiatus becomes filled only by a very thin cartilage layer, forming the recess for the utriculus of the inner ear (as shown in Fig. 23D). The recess for the sacculus is situated caudomedial of the posterior basicapsular foramen. The recess for the lagena is situated medial of the foramen for the vagal nerve. These last two recesses or grooves are generally present in siluriforms (Srinivasachar, 1958a). 56 PART 4 — ONTOGENY OF THE SKULL

Also often present (but not in all catfishes) are swellings or prominences in the otic capsule indicating the path of the internal semicircular canals (Srinivasachar, 1957b); these are visible in Ancistrus cf. triradiatus. In addition, a vertical pillar of cartilage, the lateral semicircular septum, is situated at the inner side of the horizontal semicircular canal (Fig. 22F). The anterior and posterior semicircular septa, present in some catfishes (Srinivasachar, 1957b, 1958a), are rudimentary, as the lateral so-called cavum labyrinthi opens widely into the medial cavum cranii, exactly as in Ameiurus nebulosus (Kindred, 1919). The foramen for the otic branch of the facial nerve is variably present in Siluriformes: it is seen in the anterior otic cartilage in Silonia silondia and Pangasius pangasius (Srinivasachar, 1957b), in the postorbital process of the taenia marginalis near its origin in the otic capsule in Ameiurus nebulosus, Ailia coila, Arius jella, Callichthys callichthys and Clarias gariepinus (Srinivasachar, 1957a, b, 1958a; Hoedeman, 1960b; Adriaens & Verraes, 1997c), and is absent in Rita sp., Mystus vittatus, Plotosus canius and the stinging catfish Heteropneustes fossilis (Srinivasachar, 1957a, 1958a, 1959). In C. gariepinus the foramen is formed as the taenia marginalis becomes broader and encloses the medially situated otic branch (Adriaens & Verraes, 1997c). In Ancistrus cf. triradiatus the branch splits, and only the lateral part goes through the postorbital process.

Even in early pre-hatching stages of Ancistrus cf. triradiatus a maxillary barbel cartilage is present. This maxillary barbel of Loricariidae connects and supports the upper and lower lips that make up the typical suckermouth. In literature no evidence is found of such an early presence of the cartilage in other catfishes (in which the maxillary barbel is always present). The early presence is thought to be related to the fact that young A. cf. triradiatus can suck themselves onto a substrate immediately after hatching, something requiring a well- developed suckermouth. Mandibular barbels are absent in Loricariidae; hence supporting cartilages are not present. They are, however, fairly common in other catfishes, and can often be seen in the cartilaginous skeleton. Hoplosternum littorale and Callichthys callichthys, both Callichthyidae and thus Loricarioidea, have a pair of tiny cartilage rods, attached to the lateral borders of the ethmoid plate and supporting the maxillary barbels (Ballantyne, 1930; Hoedeman, 1960b). Ballantyne (1930), Bamford (1948), Srinivasachar (1957b, 1958b) and Hoedeman (1960b) have emphasized the development of cartilages supporting the mandibular barbels in H. littorale, Ariopsis felis, Ailia coila, Clarias gariepinus and C. callichthys.

PART 4 — ONTOGENY OF THE SKULL 57

SPLANCHNOCRANIUM

In Ancistrus cf. triradiatus the palatine of the premandibular arch arises independently from the pterygoid process of the pterygoquadrate (palatal bar of the mandibular arch), and stays a separate element, a situation considered a synapomorphy among siluriforms (Arratia, 1992; Vandewalle et al., 1999). An exception occurs in Arius jella (Srinivasachar, 1958a) where both elements fuse; in Callichthys callichthys the elements almost touch (Hoedeman, 1960b). In A. cf. triradiatus the palatine elongates during ontogeny, and will articulate with the maxillary and the premaxillary bones, situated far rostrally. Below the anterior tip of the palatine a small separate cartilage develops at around 8.0 mm SL (as seen in serial sections); it is not stained by alcian blue until the 10.0 mm stage. Adriaens & Verraes (1997c) described this submaxillary cartilage in Clarias gariepinus, where it appears together with the bony maxilla, and facilitates the palatine-maxillary mechanism. The head of the maxilla in Ancistrus cf. triradiatus starts to ossify at 8.0 mm, so a similar function is assumed here.

Meckel’s cartilages [mandibular cartilages of Ballantyne (1930)] arise independently in Ancistrus cf. triradiatus, then fuse, and then disconnect (when the bony elements of the lower jaw develop). This sequence is not seen in other siluriforms: the cartilages never fuse in Ariopsis felis, Silonia silondia, Pangasius pangasius, Mystus vittatus, Rita sp., Callichthys callichthys and Heteropneustes longifilis (Bamford, 1948; Srinivasachar, 1957a, b; Hoedeman, 1960b; Vandewalle, 1997); they most often touch, though, or are joined by connective tissue. In Ameiurus nebulosus, Arius jella, Plotosus canius, Clarias gariepinus and Chrysichthys auratus they are fused from the start, and only later separate (Kindred, 1919; Srinivasachar, 1958a; Adriaens & Verraes, 1997c; Vandewalle, 1999). In the latter two species, they are fused to the pterygoquadrate-hyosymplectic plate as well. The position of Meckel’s cartilages in A. cf. triradiatus is notably different from the situation in other catfishes, as they point medially instead of rostrally. This is coupled to the caudomedial position of the ventrally oriented dentary bones in juvenile and adult specimens. This position of the lower jaw is an adaptation for the scraping of algae from substrates, and for efficiently using the typically loricariid suckermouth while breathing. The suckermouth has to be functional at hatching, a requirement that is thought to have had a large impact on the evolution of this part of the chondrocranium. The medial fusion of Meckel’s cartilages supports the lower jaws and the lower lip as soon as they start to move, up to the moment that ossification occurs.

Compared to other siluriforms, the body of the suspensorium of Ancistrus cf. triradiatus is a rather narrow and straight bar, elongating during ontogeny, as well as becoming tilted, with 58 PART 4 — ONTOGENY OF THE SKULL the pterygoquadrate acquiring a more rostral position. The suspensorium of catfishes has been a point of interest, as its development and fusion of different parts is remarkable among teleosts (Arratia, 1990, 1992). At least partial fusion of the suspensorium, i.e. of the hyosymplectic with the quadrate, may be a synapomorphy of siluriforms (Arratia, 1992; Vandewalle et al., 1999). The most extensive fusion is seen in Clarias gariepinus, where the young splanchnocranium, including Meckel’s cartilages, hyoid bars and the first copula, consists of one part, fused with the neurocranium at the level of the hyosymplectic (Surlemont et al., 1989; Adriaens & Verraes, 1997c). The fusion with the neurocranium is also present in certain Loricarioidea, like Trichomycteridae (Arratia, 1990), and, even more intensely, the callichthyid Callichthys callichthys (Hoedeman, 1960b). A weakly chondrified zone connects the suspensorium to the neurocranium in A. cf. triradiatus, but this is not visible in in toto stained preparations. In A. cf. triradiatus the foramen for the hyomandibular branch of the facial nerve is present in the center of the hyosymplectic from the start; in many other siluriforms the nerve runs anterior to the hyosymplectic, and then becomes captured by the growing hyosymplectic (Kindred, 1919; Srinivasachar, 1957b; Adriaens & Verraes, 1997c), or sometimes stays ahead of it, in a groove or not (Srinivasachar, 1957a, b, 1958a), or starts posterior of the hyosymplectic, later ending up in a foramen as well (Bamford, 1948). The interhyal in Ancistrus cf. triradiatus [stylohyal of de Beer (1927)] connects the suspensorium with the ceratohyal, and is a strip of weakly chondrified cartilage tissue, being more broadly fused to the ceratohyal than with the hyosymplectic (Fig. 17H). The interhyal is strongly fused with the hyosymplectic in catfishes as Silonia silondia, Pangasius pangasius, Ailia coila, Heteropneustes fossilis and Clarias gariepinus (Srinivasachar, 1957b, 1959). It is soon seen as a separate element in Mystus vittatus, Rita sp., Arius jella and Plotosus canius (Srinivasachar, 1957a, 1958a). The four processes of the ceratohyal in Ancistrus cf. triradiatus are thought to have various functions. The large ventral process, common among siluriforms, albeit usually smaller, articulates with the branchiostegal rays. The other three are not known from other siluriforms: the dorsal process grows to the hyosymplectic, and might well reinforce the more anterior hyoid-interhyal-hyosymplectic articulation (e.g., Fig. 23E); the anterior and posterior processes broaden the lateral end of the hyoid bar, where it will later grow into the vicinity of the preopercular bone, attached to the ossified suspensorium. The true nature of the parurohyal dumbbell-like nucleus caudoventral to the hypohyals (Fig. 17F) is not clear; ensuing study of its link with the first basibranchial and its role in the formation of the parurohyal of Ancistrus cf. triradiatus during ossification will help clarify the possible homology of this part with elements of the parurohyal described by Arratia & Schultze (1990) (see Chapter 4.2).

PART 4 — ONTOGENY OF THE SKULL 59

As all catfishes, Ancistrus cf. triradiatus has no separate basihyal (Arratia & Schultze, 1990; Adriaens & Verraes, 1997c). The hypohyals fuse with the first basibranchials, which, like the other branchial elements, arise in a rostrocaudal sequence, as is the case in all siluriforms, but not in teleosts generally (Vandewalle et al., 1997). The basibranchials in Ancistrus cf. triradiatus are organized in two copula’s: the anterior copula starts continuous with the hyoid bar, and consists of basibranchials I-III. Later, it becomes separated from the hyoid bar; the first basibranchial disappears, while the other two will ossify independently. The posterior copula arises separately, and comprises basibranchials IV and V. This grouping is seen in several other siluriforms as well, e.g., Ariopsis felis, Clarias gariepinus and Heterobranchus longifilis (Srinivasachar, 1958a; Adriaens & Verraes, 1997c; Vandewalle et al., 1997). Vandewalle et al. (1997) mistakenly called the two copula’s basibranchial one and two. In Silonia silondia and Pangasius pangasius the posterior copula is homologous with the fifth basibranchial only (Srinivasachar, 1957b).

The overall appearance of the other branchial elements varies among catfishes. An example is the degree of fusion of the hypobranchials and the ceratobranchials of the third arch: in Clarias gariepinus the hypobranchials become separated from the corresponding ceratobranchials (Adriaens & Verraes, 1997c), in Ancistrus cf. triradiatus they remain fused with the ceratobranchials, and in Plotosus canius both hypobranchials are fused with each other in the midline (Srinivasachar, 1958a). This might be related to the presence and size of the copulas in this region; in Pangasius pangasius, where the fourth basibranchial seems to be rudimentary or absent, hypobranchials IV are also fused in the midline (Srinivasachar, 1957b).

Infrapharyngobranchials I and II are not present in Ancistrus cf. triradiatus. Among catfishes, the number of infrapharyngobranchials may range between two and four (Adriaens & Verraes, 1997c). Diplomystidae have four separately ossifying infrapharyngobranchials (Arratia, 1987). In Silonia silondia and Arius jella the posterior two become fused; in Plotosus canius only the first one remains separate (Srinivasachar, 1957b, 1958a). In Clarias gariepinus the first two are fused, while the last two stay apart (Adriaens & Verraes, 1997c). In A. cf. triradiatus the first two infrapharyngobranchials are completely lost, complying to a reductional trend Adriaens & Verraes (1997c) observed among catfishes.

60 PART 4 — ONTOGENY OF THE SKULL

∗ 4.2. ONTOGENY OF THE OSTEOCRANIUM

Abstract — The development of the osteocranium of the suckermouth armored catfish Ancistrus cf. triradiatus is described based on specimens ranging from pre-hatching stages to juvenile stages where the osteocranium is more or less fully formed. The first bony elements that arise are the opercle, jaws, and lateralmost branchiostegal rays, as well as the basioccipital and parasphenoid in the skull floor. The supracleithrum and the membranous and perichondral pterotic components form one large, double-layered skull bone during ontogeny, without clear evidence of the involvement of a supratemporal. The Baudelot’s ligament ossifies from two sides, i.e., from the basioccipital medially and the supracleithrum laterally. The lower jaw consists of a dentary, mentomeckelium and angulo-articular that all soon fuse. The parurohyal, formed by the fusion of a ventral sesamoid bone and a dorsal cartilage element associated with the first basibranchial, is pierced by a vene, unlike in some other siluriforms. The interhyal cartilage disappears during ontogeny; medially of it a small sesamoid bone appears in a ligament. The largest, canal bearing cheek plate is not homologous to the interopercle. The results of the present research, with emphasis on bone formations and homologies, are compared with studies on related catfishes.

4.2.1. INTRODUCTION

Development of structures and the early life history of fishes are closely related. Early life history stages must also function as organisms, so the study of ontogeny is an obvious necessity, if one wants to understand the integration of, e.g., the feeding apparatus (Lauder et al., 1989). The vital importance of events occurring during early development is easily overlooked (Orton, 1955). The current study on the loricariid Ancistrus cf. triradiatus describes and discusses the development of the osteocranium in detail, in continuation of work on the chondrocranium of the species (see Chapter 4.1). Loricariids are well known for their remarkable niche occupation, i.e., the scraping and sucking of algae and other food items off submerged substrates. Of some loricariid species only the adult skeletal morphology has been examined by Alexander (1965), Schaefer (1987, 1988, 1997), Schaefer & Lauder (1986) and others. Compared to the development of the siluriform chondrocranium, the development of the osteocranium has received remarkably less interest. Relevant publications on the development of the bony skull in catfishes discuss Ariidae (Bamford, 1948; Tilney & Hecht, 1993), Callichthyidae (Hoedeman, 1960b, c), Clariidae (Surlemont & Vandewalle, 1991; Vandewalle et al., 1997; Adriaens et al., 1997; Adriaens & Verraes, 1998), Claroteidae (Vandewalle et al., 1995), Ictaluridae (Kindred, 1919), Siluridae (Kobayakawa, 1992) and the

∗ Slightly modified from: Geerinckx T., Brunain M. & Adriaens D., 2007. Development of the osteocranium in the suckermouth armored catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Journal of Morphology, in press. PART 4 — ONTOGENY OF THE SKULL 61 suspensorium of Diplomystidae, Trichomycteridae and some other families (Arratia, 1990; 1992). The ontogeny of the Weberian apparatus in Clarias gariepinus and Corydoras paleatus has been examined by Radermaker et al. (1989) and Coburn & Grubach (1998) respectively. Homology of the anterior vertebrae has been treated by Hoffmann and Britz (2006). The Weberian apparatus and complex vertebra are only briefly discussed in this dissertation. Taking into consideration their complexity, a thorough study of these structures would merit a separate paper. Loricariids are an exceptionally interesting fish taxon, as they are able to respire while attaching to a substrate with their suckermouth. The added possibility of feeding in their typical manner (see above) is even more unusual. Knowledge of the ontogenetic origin of the osteocranium in loricariids is lacking. The development of many bones (e.g., jaws) is, however, highly interesting, as they are crucially modified key structures in the sucking and scraping device that has evolved in this large and diverse catfish group. Some of these key structures might well have had an influence on the diversity within the large loricariid family. Next to the description of several ontogenetic stages, I discuss the observations in the light of possible homologies or non-homologies with structures present in related taxa, in order to try to shed some light on the problematic identifications of skeletal elements in loricariids. Finally, some more functional-morphological considerations are made.

4.2.2. BRIEF MATERIAL AND METHODS

Twenty-seven specimens were used for in toto clearing and staining (see paragraph 2.2.5). Ten specimens have been described in detail, with reference to other specimens, or serial sections (see paragraph 2.2.7 for procedure), wherever relevant.

4.2.3. RESULTS

No endochondral bone is encountered in Ancistrus cf. triradiatus up to the 20.7 mm SL stage. The skull is composed of perichondral and membrane bones (including dermal bones), as well as a few sesamoid bones and bones of compound origin.

5.6 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 27)

Only one pair of bony elements is present in this specimen (in serial sections of the younger 4.8 and 5.2 mm specimens only a cartilaginous skeleton is visible). Near the posterior end of the cartilaginous suspensorium, at the distal tip of the opercular process, the 62 PART 4 — ONTOGENY OF THE SKULL opercles have appeared as small bony splints. No other bony structures are seen. At the future location of each premaxilla, however, four tooth primordia can be recognized.

6.0 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 28)

NEUROCRANIUM The lateralmost protuberances of the otic capsule bear a small odontode (or denticle, see note in paragraph 1.2.3), and another small odontode is visible on the right postotic process. Odontodes are extra-oral teeth, part of the body armour of several fish taxa (Bhatti, 1938; Reif, 1982). Serial sections of a 6.1 mm specimen show precursor tissue of the future supporting bones, but no ossification.

SPLANCHNOCRANIUM Thin premaxillae have appeared, supporting four to five teeth. The maxillae arise as thin bony splints supporting the maxillary barbel cartilages, and develop articulation facets for the growing, but still short palatine (or autopalatine) cartilages (pars autopalatina of Arratia, 1990). The cleared and stained 6.0 mm specimen has no dentaries, nor dentary teeth; however, serial sections of the 6.1 mm specimen show the presence of four short conical teeth at the location where each dentary is to be expected. Both the premaxillary and the dentary teeth are still embedded in soft tissue in the 6.1 and 7.0 mm specimens. Two thin branchiostegal rays attach to the paired ventrocaudal process of the hyoid bar. The position of their insertion suggests that they are the lateralmost rays III-IV. The opercles have grown somewhat, and a pair of odontodes is present in the skin covering them.

7.4 MM SL — 6 DAYS POST-FERTILIZATION (FIG. 29)

NEUROCRANIUM On the taeniae marginales the dermal frontals have started developing. They are accompanied by two odontodes in the skin. Odontodes can now also be seen on both postotic processes of the otic capsule. A first indication of the parasphenoid is seen in the posterior half of the hypophyseal fenestra on the serial sections in the 7.0 mm stage, although no bone is visible in the stained 7.4 mm specimen. Also in these sections, the notochordal sheath, where the notochord enters the skull, is slightly ossifying, giving the onset of the basioccipital (not shown on figure 29B). In the 7.4 mm specimen, two odontodes are present on the skin near the lateral protuberance of the cartilaginous otic capsule. Small perichondral ossifications of the otic capsule are seen underneath these last odontodes, representing the first anlage of the autopterotics. Serial sections (7.0 mm specimen) show that initially PART 4 — ONTOGENY OF THE SKULL 63 odontodes are not in direct contact with the dermal frontals or perichondral autopterotics, which are situated deeper in the skin.

SPLANCHNOCRANIUM The premaxillae have grown broader, providing space for about six teeth. The maxillae have also expanded, now touching the cartilaginous palatine cartilage bars. The dentaries are visible as curved bony plates. Seven to eight slightly curved conical teeth attach on the posterior half of these, pointing ventrally. In serial sections of a 7.0 mm specimen thin perichondral ossifications can be recognized on the hyoid bar and suspensorium, representing the onset of the anterior ceratohyals and hyomandibulars; these are not yet visible in the stained 7.4 mm specimen. The first signs of the parurohyal have appeared in the serial sections of the 7.0 mm specimen: a small plate-like sesamoid ossification is present in the tendon of each half of the sternohyoideus muscle (thus only urohyal parts, or urohyal bones, are present at the moment). The branchiostegal rays III and IV have elongated; branchiostegal rays II have also appeared.

8.0 MM SL — 7 DAYS POST-FERTILIZATION (FIG. 30)

NEUROCRANIUM The frontals are now more substantial, covering most of the taeniae marginales dorsally. Serial sections of an 8.0 mm specimen show that the membranodermal components, covering the taeniae, are already present, while the neurodermal components still have to arise, as well as the supraorbital canals which they will surround. Between the otic fenestrae another pair of odontodes has arisen. A thin sheet of perichondral bone partially covers each half of the otic capsule roof: the supraoccipital thus originates as a paired structure. This remarkable feature is confirmed by the serial sections of the 8.0 mm specimen. At this moment, as well as in later stages, no evidence is seen of separately developing parietal bones. At no moment during ontogeny is there a full separation between the perichondral and membranous parts of the parieto-supraoccipital (I use this name for the bone, but refer to the discussion for a more thorough treatise of this bone). The ossification center of the basioccipital is present as a bony sheath around the cranial part of the notochord, now continuing as a perichondral ossification of the neurocranium floor next to it. The transverse processes of the complex vertebra have appeared; they form the ventral floor of the swimbladder capsules (the complex vertebra and Weberian apparatus are not discussed further here). The parasphenoid is now visible as a narrow band of dermal bone in the lateral and posterior perimeter of the hypophyseal fenestra. Small lateral notches are left in the bone; together with fissures in the trabecular bars they form the foramina for the paired internal 64 PART 4 — ONTOGENY OF THE SKULL carotic arteries (Fig. 30A-B). In the serial sections, but not in the stained 8.0 mm specimen, initiation of perichondral ossification of the prootics is observed. The lateral protuberances of the otic capsule are now well covered by the paired autopterotics. The supracleithra of the pectoral girdle have appeared just posterior to the otic capsule floor, and are closely associated with the dorsal articular processes of the cleithra, as could best be seen in serial sections of the 8.0 mm stage. The sections also show that the supracleithra appear separate from the autopterotics. The first, separate ossification of the supracleithra could be recognized in serial sections of the 7.0 mm specimen (Fig. 31A).

SPLANCHNOCRANIUM The premaxillae have become higher, with the eight erected teeth inserting on the anterior margin. The maxillae have developed complete articular facets for the palatine cartilages. The dentaries now attach firmly to Meckel’s cartilages, which have also started to ossify medially, forming the mentomeckelian bones (Fig. 31B; not visible in figure 30). The dentaries are best developed posteriorly, and each one supports about ten conical teeth. An angulo-articular bone is not present at this time. Separate from the anterior ceratohyal, the hypohyal and the posterior ceratohyal are visible on the serial sections of the 8.0 mm stage (Ancistrus cf. triradiatus only develops a ventral hypohyal). The paired sesamoid component of the parurohyal (i.e., the urohyal bone) is growing, and a dumbbell-shaped cartilage nucleus has separated from the hypohyal region of the hyoid bar. This nucleus contacts the rostral end of the anterior copula, which can be considered the first basibranchial. The medialmost branchiostegal rays I have appeared, articulating with the medial ends of the broad caudoventral processes of the hyoid bar. The opercles are triangular elements, each composed of a horizontal rod that bears an odontode almost halfway, and a ventral plate extending in the direction of the outer branchiostegal ray.

9.1 MM SL — 8 DAYS POST-FERTILIZATION (FIG. 32)

NEUROCRANIUM At the tip of the snout, the hypoethmoid is present as a perichondral ossification of the ventral face of the ethmoid cartilage. The frontals and the now unpaired parieto- supraoccipital have started to cover parts of the postpineal fontanelle. The anterior tips of the frontals have reached the prepineal fontanelle. The autopterotics have grown significantly along the lateral and posterior walls of the otic capsule. The dermopterotics appear as posterior projections of the posterior vertical walls of the autopterotics. They form the onset of the roof of the swim bladder capsules. The supracleithra are now fused to the autopterotics and dermopterotics.

PART 4 — ONTOGENY OF THE SKULL 65

SPLANCHNOCRANIUM Premaxillae and dentaries now form basket-like structures. Their ventral edges are complete at the tooth-bearing side, but not yet at the other, mouthward side (posterior in premaxillae, anterior in dentaries). The posterior ends of the autopalatine bones have appeared. The tip of the pterygoid process of each suspensorium has started to ossify perichondrally, and a small membranous bony sheet forms around it. These perichondral and membranous elements constitute the anlage of each metapterygoid. From the rostral end of this bone a ligament stretches towards the ventralmost aspect of the posterior autopalatine ossification. A partial perichondral ossification of each hyomandibular is faintly visible, with membranous extensions caudally (leaving an opening for the path of the opercular branch of the facial nerve) and lateroventrally, in the direction of the preopercle. The long and slender membranodermal preopercles have appeared along the central part of the suspensoria (the neurodermal gutter-like part is not yet present in this stage). Posterior to the path of the inferior jugular vein the paired urohyal ossifications of the sternohyoideus tendon have fused. The dumbbell-shaped cartilage nucleus is in contact with the tendon anterior to the path of the vein. In this stage the parurohyal bone thus still consists of a sesamoid (‘urohyal’) component only. In the branchial basket paired, bony plates have developed, lying against each fourth infrapharyngobranchial. These upper pharyngeal toothplates (or ‘jaws’) already bear two pointed teeth each at this stage (Fig. 32C). Serial sections of a 10.2 mm specimen demonstrate, however, that these rudimentary teeth are still covered by the epidermal pharyngeal tissue, so they can not yet be functional.

9.8 MM SL — 10 DAYS POST-FERTILIZATION (FIG. 33)

NEUROCRANIUM Some additional bony structures have appeared, while the skull bones already present have become enlarged so as to form a more rigid support of the braincase. The frontals have elongated and now connect the otic capsules and the sphenoseptal commissures, and have overgrown a large part of the epiphysial bridge as well. The parieto-supraoccipital consists of a large plate, membranous as well as perichondral, and makes up a bony tectum between both otic capsules. Four bones provide support of the skull floor in the midline. As the notochord is relatively smaller, when compared to the 9.1 mm specimen, the basioccipital becomes the most important supporting element in the posterior skull floor. It can be discerned from the notochordal ossification of the complex vertebra. The parasphenoid now fills the whole hypophyseal fenestra, and has become rhomboid-shaped, as the medial cartilage in front of the fenestra has reduced. Only the two foramina for the internal carotic arteries remain. More 66 PART 4 — ONTOGENY OF THE SKULL anteriorly, cartilage reduction has freed space for the membranous prevomeral (or vomeral13) bone, suturing with the parasphenoid. Around the ethmoid cartilage the various elements of the mesethmoid are seen. The supraethmoid and the hypoethmoid are dorsal and ventral perichondral bones respectively, connected only at the tip. The hypoethmoid has a medioventral process, which is the onset of the typical loricariid mesethmoid disc (Fig. 33C). Most of this process lacks a cartilaginous precursor (Fig. 31C). On the dorsal side of the ethmoid cartilage a dermal sheet, the dermethmoid, develops. It covers the posterior part of the supraethmoid. Serial sections of a 10.2 mm specimen show that these three bony elements later fuse to form a tube-like mesethmoid bone around the ethmoid cartilage. The lateral end of the orbitonasal laminae have ossified perichondrally. Small membranous extensions bear one odontode each. Thus the lateral ethmoids are composed of perichondral and membranous elements. The nasal bones, containing the rostral part of the supraorbital canals, have arisen on top of the nasal sacs. In serial sections of the 10.2 mm specimen only the canal-supporting neurodermal element is already present. The neurodermal parts of the frontals have also developed, being well visible in the anterior halves of these bones. It is not sure whether the neurodermal parts of the frontals have arisen separately, or from ossification centra of the membranodermal parts. Only serial sections of a specimen between 8.0 and 10.2 mm SL could show this. The perichondral prootics are seen in the skull floor. In the serial sections of the 10.2 mm specimen most of the perichondral orbitosphenoids, pterosphenoids and autosphenotics are present and relatively well developed, but they are not seen on the stained 9.8 mm specimen. The compound pterotic bone complex has grown extensively; it consists of three main spatial elements: a largely perichondral casing covering the lateral and posterior walls of the cartilaginous otic capsule, a dorsal dermal extension forming the roof of the swimbladder capsule, and a ventral membranous extension forming part of this capsule’s floor. The original components of this compound bone can not be clearly distinguished anymore. From the ventral extension, at the location where the supracleithrum could be identified in the 9.1 mm specimen, the Baudelot’s ligament runs medially towards the posterior ventral surface of the basioccipital. It starts to ossify laterally, from the compound pterotic.

SPLANCHNOCRANIUM The ventrocaudal processes of the autopalatine bones, on which the extensor tentaculi muscles and the autopalatine-metapterygoid ligaments insert, are well developed. Both the upper and lower jaws have fully developed their tooth-bearing baskets. Thus the premaxillae

13 I use the name prevomer throughout this thesis. Argumentation exists for both the homology and the non- homology of this bone and the vomeral bone in Mammalia. If they are truly homologous, the name prevomer can be replaced by vomer in teleosts as well. PART 4 — ONTOGENY OF THE SKULL 67 are complete, while the coronoid processes of the dentaries are not yet completely developed. The latter bones have formed ventrolateral flanges towards the angulo-articulars. The ossification of the latter bones has started at the articulation facets for the suspensoria, and extend rostrally. All three elements of both suspensoria are now present. The membranous sheets of the metapterygoids have grown extensively dorsally and ventrally, giving the bones a triangular outline. The quadrates and hyomandibulars are now both present, and the canal- bearing preopercles have broadened. The preopercles surround the preopercular canals. The largely perichondral anterior ceratohyals have developed membranous sheets growing dorsally from the anterior edge of the bones, while the posterior ceratohyals still lack any membranous extension. The interhyals have partly reduced (not visible on Fig. 33), separating the suspensoria and the hyoid bar. Various ligaments have developed along the suspensorial-hyoid connection. In a ligament at the inner side of the rudimentary interhyal cartilages, sesamoid bones appear (Fig. 31D). The urohyal has further developed. The paired sesamoid bones in the sternohyoideus tendon are now fused anterior and posterior to the inferior jugular vein. Anteriorly both tips of the sesamoid bone almost touch the developing hypohyals posteroventrally; posteriorly the bone reaches up to a quarter of the sternohyoideus length. The cartilage nucleus still shows no sign of ossification. As seen in the serial sections of the 10.2 mm specimen, the remainder of the partly reduced first basibranchial is still continuous with this cartilage nucleus. The branchial basket is still completely cartilaginous, except for the upper pharyngeal jaws and newly formed anterior processes originating from the first ceratobranchials.

10.8 MM SL — 18 DAYS POST-FERTILIZATION (FIG. 34)

NEUROCRANIUM Between the 9.8 and 10.8 mm stages extensive ossification of the skull has taken place. All neurocranial bones are now present. In the skull roof the various bones have grown closer and often touch each other already. The frontals have completely overgrown the epiphysial bridge and now separate the anterior from the posterior fontanelle (in the chondrocranium these two fenestrae are referred to as the prepineal and postpineal fontanelle). From this stage on, the supraorbital canals and the anterior part of the otic canals are well visible in the stained specimens, running through the nasals, frontals and dermosphenotics. The orbitosphenoids are now well developed, and cover the preorbital bases and the anterior halves of the trabecular bars. In rostral view they are L-shaped, with broad horizontal parts, reaching towards the parasphenoid, and narrower vertical parts forming the anterior margins of the orbits, and covering the anterior parts of the taeniae marginales. The foramina for the ophthalmic branches of the trigeminal nerves are enclosed by them. The pterosphenoids have covered the main parts of the taeniae marginales. The sphenotic 68 PART 4 — ONTOGENY OF THE SKULL fenestrae are surrounded by the orbitosphenoids, pterosphenoids, and prootics. The prootics have now covered the basiotic laminae, the posterior parts of the trabecular bars and a part of the anterior basicapsular commissures. The lateral commissures of the chondrocranium are ossified as well, providing the prootics with an anterior foramen. The parieto-supraoccipital has grown to a massive, U-shaped compound bone, including a small posterior process. This bone, as well as the sphenotics and pterotics, take part in the closing of the two otic fenestrae that characterize the chondrocranial skull of Ancistrus cf. triradiatus (Chapter 4.1). The lateral ethmoids, grown extensively since the 9.8 mm stage, touch the frontals, but not yet the mesethmoid. Only posteriorly the hypoethmoid and supraethmoid parts of the mesethmoid are still unconnected. The ventral mesethmoid disc has grown to the level of the premaxillae, where a mesethmoid-premaxillary cartilage is present, as already shown by serial sections of the 10.2 mm specimen. In the skull floor the basioccipital has started to form deep sutures with the parasphenoid. The exoccipitals have appeared, and have foramina for the glossopharyngeal and vagal nerves. All paired skull floor bones are still separated by broad cartilage zones; only the unpaired medial bones interdigitate with each other, forming the main axial supportive structure of the skull. The epioccipitals, developing as perichondral ossifications of the posterior skull wall, can be recognized in the 10.2 mm serial sections. Two ventral processes project laterally from the posterior half of the basioccipital. These are the medial ossifications of Baudelot’s ligaments (analogous to the lateral, supracleithral parts of the ligaments, the medial ossifications spread from the basioccipital bone). From the ventral layer of the compound pterotics lateral trabecles have begun to grow dorsally. Thus the lateral walls of the airbladder capsules start to form. Some neuromasts of the infraorbital canals have appeared and invaginated; the first bony encapsulation of these canals occurs by the fifth infraorbitals, as shown by serial sections of the 10.2 mm specimen.

SPLANCHNOCRANIUM The upper and lower jaws now bear 12-15 teeth each. The posterior attachment facets for the retractor premaxillae muscles on the premaxillae are now well developed. The dermal dentaries and perichondral mentomeckelia have fused. This is already the case in the (serially sectioned) 10.2 mm specimen. The angulo-articular bones have covered the lateral halves of Meckel’s cartilages, leaving only small posterior regions unossified, which will remain as such to the adult stage. The angulo-articulars acquire rostrodorsal membranous sheets that provide space for the developing adductor mandibulae muscles (which also insert on the coronoid processes of the dentaries). More than half of the palatine cartilages is now enclosed by the autopalatine bones. The metapterygoids have developed lateral ridges. PART 4 — ONTOGENY OF THE SKULL 69

The cartilaginous hyoid bar is now almost completely replaced by bone. Both the anterior and posterior ceratohyals have developed anterior membranous extensions. These extensions are orientated slightly dorsally, so that the hyoid bar actually consists of a horizontal perichondral plane and an almost vertical membranous plane. In the anterior ceratohyals large notches are left for the passage of the arteries supplying the lateral parts of the musculus hyohyoideus inferior, originating from the hypobranchial artery (Fig. 34C). The central regions of all epibranchials and ceratobranchials have begun to ossify perichondrally. First signs of these ossifications are visible in serial sections of the 10.2 mm specimen. The second basibranchial and the first hypobranchials are also starting to ossify. Strips of membrane bone have developed against the fifth ceratobranchials. These lower pharyngeal tooth plates are not (yet) continuous to the ceratobranchials, and already bear two teeth each. Three to four teeth can be counted on the upper pharyngeal toothplates (or jaws).

11.7 MM SL — 30 DAYS POST-FERTILIZATION (FIG. 35)

NEUROCRANIUM The outline of the original chondrocranium is not easy to make out anymore, as most of it has now been covered by perichondral bone. Also, the dermal bony elements have overlain or hidden the cartilaginous skeleton. The otic fenestrae of the otic capsule roof are fully closed by the parieto-supraoccipital, sphenotics and pterotics, and the anterior and posterior fontanelles have severely shrunk as the parieto-supraoccipital and the frontals have expanded. The mesethmoid is now completely tube-like, with the hypoethmoid and supraethmoid parts fully connected at both left and right sides. The mesethmoid now reaches the frontals, but not yet covers them. The lateral ethmoids now enclose the nasal sacs on three sides, and touch each other below the dermethmoid roof. They have contacted the prevomer ventrally. The medial floor of the ethmoid cartilage reduces where the prevomer covers it ventrally. The three bones bordering the sphenoid fenestrae, i.e., the orbitosphenoids, pterosphenoids and prootics, restrict the fenestrae by the formation of membranous bony sheets at the perimeter. The compound pterotics have developed spectacularly, expanding the dorsal and ventral layers, as well as connecting them laterally by means of trabecles carrying numerous odontodes (the only connection between the dorsal and ventral layers thus far was rostrally, at the level of the otic pilae). These trabecles leave various small foramina. The connections with the transverse processes of the complex vertebra are reinforced by means of fine sutures dorsally and ventrally. The basioccipital, exoccipitals, epioccipitals and medial parts of the pterotics are tightly connected to these transverse processes. The basioccipital and the parasphenoid are deeply sutured. In the skin below the eye, three small canal bones are now present: infraorbitals IV to VI.

70 PART 4 — ONTOGENY OF THE SKULL

SPLANCHNOCRANIUM As the jaw bones had already attained their more or less final shape in the 10.8 mm specimen, the only significant difference now is an increase in size, and a closer contact including suturing between dento-mentomeckelia and angulo-articulars. The autopalatine bones are now completely formed; the cartilaginous rostral articulation heads for the maxillae remain cartilaginous during further ontogeny. The original outline of the cartilaginous suspensoria is not visible anymore, as the proportion of membrane bone has increased and the cartilage has become reduced, except for the symplectic cartilages and the articulation heads for the neurocranium and the opercle (the articular cartilage at the facet for the lower jaw is minute). The hyomandibulars have started suturing weakly with the pterotics dorsocaudally. The lateral ridges of the metapterygoids have expanded into large sheets. As the cartilaginous interhyal connections between the suspensoria and the hyoid bar have been lost, the ligamentous connections grow stronger. The final shape of the sesamoid bones medial to the former interhyal locations is cylindrical. Lateral to the quadrates, at the rostral tip of the preopercles, the preopercular canals now turn ventrally. Thin neurodermal ossification are present around them. These are the first signs of the largest of two cheek plates that will develop in this region (Fig. 31E). Membranodermal elements are not yet visible in this specimen. The branchiostegal rays have grown and are now all flattened (the lateral ones more than the medial ones). Both anterior tips of the sesamoid urohyal bone almost touch the developing hypohyals posteroventrally; posteriorly the bone reaches up to a third of the length of the sternohyoideus muscle. The hypohyals acquire a depression near the anterior (par)urohyal tips. The middle shaft region of the ceratobranchials and epibranchials have further ossified, leaving the heads (and growth regions) still cartilaginous. The fifth ceratobranchials and lower pharyngeal jaws have fused.

14.4 MM SL — 45 DAYS POST-FERTILIZATION (FIG. 36)

NEUROCRANIUM Both fontanelles are now completely closed. The posterior growth of the parieto- supraoccipital also has elongated the skull (in the earliest stages the posterior end of the skull was demarcated by the tectum posterius). The major skull roof bones have also started suturing. The dermal lateral processes of the sphenotics, enclosing the posterior portions of the infraorbital canals, have grown. On top of the lateral connections of the lateral ethmoids and the frontals thin odontode-bearing dermal layers have appeared, the prefrontal plates. Beneath the skull floor, the Baudelot’s ligaments are almost completely ossified: only a thin region of each ligament, between the medial ossification from the basioccipital and the lateral ossification from the ventral ridge of the supracleithrum (compound pterotic) remains PART 4 — ONTOGENY OF THE SKULL 71 ligamentous at this moment. Most infraorbital bones (usually six on each side in Ancistrus cf. triradiatus, but sometimes only five) are present, or at least an odontode can be seen indicating the future location of the bone (the supporting bone underneath is often more difficult to see than the odontode itself). Minute dermal platelets carrying odontodes now appear in the cheek region and behind the skull as well.

SPLANCHNOCRANIUM The suspensoria have attained their approximate final shape by now: the metapterygoids have made contact with the collateral quadrate and hyomandibular, suturing with both. They are ligamentously connected to the lateral ethmoids (dorsally) and the autopalatine bones (rostrally). The membranous dorsal parts of the hyomandibulars have become completed and form rounded, thin sheets, supporting the eyes. The hyomandibulars articulate with the neurocranium at the level of the contact between the collateral prootic, pterotic and sphenotic. Just anterior to this articulation the membranous parts of the hyomandibulars form two processes fitting into the serrate lateral edges of the prootics. These serrations have developed together with these hyomandibular processes. The hyomandibular-pterotic sutures have expanded to most of the contact zone between both bones, which is the final adult configuration, though the sutures will become stronger during further development. The preopercles approach the quadrates and the hyomandibulars, though are not yet fused to any of these bones. Between the preopercles and the most rostral margin of the pterotics the preopercular canals are now enclosed by the paired suprapreopercles, dermal canal-bones bearing odontodes just like the infraorbitals. Anterior processes are formed on both opercles and point ventromedially. Near these processes, sturdy bony elements are developing, bearing large spiny odontodes, the so-called cheek spines. The dermal cheek plates lateral to the quadrates bearing the end of the preopercular canals, have expanded a little, as membranodermal bony sheets supporting a few odontodes are added to the neurodermal gutter-like components. The adult shape of the parurohyal is more or less reached. The cartilage nucleus dorsal to the sesamoid urohyal bone has condensed, and on its ventral side bone formation connects it to this sesamoid part (Fig. 31F-H). The bone now has a compound nature. Both parurohyal tips fit in holes of the hypohyal bones. It appears that the centers of the hypohyal cartilages have reduced so that the depressions, mentioned in the previous stage, now pierce the bones and have become holes. The bony first ceratobranchials and their spongiose anterior processes have fused; in earlier stages the processes were only loosely connected to the first ceratobranchial bones via the cartilage at the medial ends of the bones. The uncinate processes of the epibranchials that have arisen since the bones started to develop are now 72 PART 4 — ONTOGENY OF THE SKULL well visible; the largest are borne by the third epibranchials. The infrapharyngobranchials III and IV show the first signs of perichondral ossification.

20.7 MM SL — 96 DAYS POST-FERTILIZATION (FIG. 37), AND FURTHER DEVELOPMENT

NEUROCRANIUM As all neurocranial elements had already attained their approximate final shape at the 14.4 mm stage, this and later stages are mainly characterized by growth and reinforcement of the skull by means of further suturing of dermal bones, closer synchondral contacts between the perichondral bones, and heavier ossification of the various elements. As the skull floor bones have also thickened the bony recesses for the paired maculae can be easily seen: the utriculus is borne by a recess in the prootic, while the more posteriorly situated sacculus is enclosed by the basioccipital, and the lagena by both the basioccipital and exoccipital. The Baudelot’s ligaments are completely ossified, providing a direct bony connection between the basioccipitals and compound pterotics. The ligaments form continuous transverse ridges, though sutures can be seen in the positions where the sesamoid ossifications from the basioccipital and supracleithral sides touch each other. No teeth develop on the ventral surface of the prevomer. The prefrontal plates have grown, and will become almost rectangular when specimens reach maturity. Of the major skull bones, the compound pterotics expand most during further growth, broadening the posterior part of the skull. As seen in the 20.7 mm specimen, the lateral ethmoids have not yet closed completely anteriorly, leaving the anterior part of the nasal sacs less supported. In the 33.5 mm specimen the closure is established. The infraorbital bones have grown and almost touch each other. The suprapreopercles and the preopercles have almost touched as well. So now the infraorbital and preopercular canals are more or less completely enclosed in bone. The number of platelets in the snout region has increased. They can be divided into prenasal plates (on top of the mesethmoid) and lateral plates (between the preopercular and infraorbital canals). Lateral to the hyomandibular and the pterotic bones platelets will also soon appear (the first are observed in the 25.0 mm stage; most are present at 31.0 mm SL). A complete coverage of similar, overlapping plates appears under the skin of the rest of the body, leaving only the belly unprotected. In the 31.0 mm specimen thin sclerotic bones have appeared, that support the eyeballs anteriorly and posteriorly.

SPLANCHNOCRANIUM The upper and lower jaws now carry about 30 teeth each, a number that will have doubled when the animals reach adulthood. The preopercles and the suspensoria have started to fuse. In later stages it is very difficult to differentiate between these bones. A short stretch of the PART 4 — ONTOGENY OF THE SKULL 73 preopercles lies directly under the skin, and now carries a few odontodes. During further development, the preopercles broaden somewhat, further overgrowing the quadrates and hyomandibulars. The anterior processes of each opercle has grown, and the articulation of the bone with the hyomandibular is reinforced by the presence of a serration on the opercle posterior to the articulation. A second, paired cheek plate has developed dorsal to the first one. It consists of an odontode-supporting plate only, thus not bearing a canal. The cheek spines are now well visible. The rostral processes of the parurohyal pierce the hypohyals; the horizontal sesamoid sheet continues to grow until the posterior margin is more or less rounded in the 25.0 mm stage, and reaches to almost half of the sternohyoideus length. The dorsal region of the parurohyal cartilage nucleus degenerates, and the ventral region further ossifies, although part of it stays cartilaginous even in adults. The very small third basibranchial is ossified in specimens of over 50 mm SL, while the posterior copula, consisting of the fourth and fifth, fused basibranchials, remains cartilaginous. The second hypobranchials remain cartilaginous as well, while the third to fifth hypobranchials stay fused to the cartilage heads of the corresponding ceratobranchials. The anterior processes of the first ceratobranchials have broadened and grown to the same length as the ceratobranchials. The tips of the cerato- and epibranchials don’t ossify. Also, the infrapharyngobranchials maintain their cartilaginous articular caps.

4.2.4. DISCUSSION

Ossification in Ancistrus cf. triradiatus starts as early as the fourth day after fertilization. At 5.6 mm SL the first dermal bone can be recognized, i.e., the opercle. This is only one day after the formation of the first chondrocranial elements (Chapter 4.1), or one day before hatching occurs. The early appearance of the opercle, followed by the formation of the first branchiostegal rays and most dentulous bones, is a general trend in siluriforms and other teleosts (Weisel, 1967; McElman & Balon, 1980; Tilney & Hecht, 1993; Vandewalle et al., 1995, 1997; Adriaens & Verraes, 1998). Generally, the onset of splanchnocranial ossification is earlier than the first appearance of neurocranial elements (e.g., de Beer, 1937; Bamford, 1948; Surlemont & Vandewalle, 1991; Tilney & Hecht, 1993; Vandewalle et al., 1994, 1995; Adriaens & Verraes, 1998). Within the neurocranium, dorsal and ventral elements appear more or less simultaneously in A. cf. triradiatus. The ontogeny of some skull bones of both neurocranium and splanchnocranium of A. cf. triradiatus is briefly treated, while a few merit a more thorough discussion.

74 PART 4 — ONTOGENY OF THE SKULL

NEUROCRANIUM

The most important bone in the ethmoid region of Ancistrus cf. triradiatus, the mesethmoid, has a complex development. The dorsal and ventral perichondral components can be identified as the supra- and hypoethmoid bones. They are connected over the rostral tip of the ethmoid cartilage. A dermal element, called dermethmoid or rostral by Patterson (1975), soon overgrows the supraethmoid and stretches further posteriorly. Although these parts can be discerned based on their perichondral or dermal nature, as wel as their location, they are actually connected more or less from the moment they arise. The ethmoid cartilage shows a small ventral protrusion, but most of the ventral disc of the mesethmoid does not form perichondrally around the ethmoid cartilage, but is an extension of the hypoethmoid. It is mostly membranous, although cartilage can be seen in serial sections (Fig. 31C). Its development is related to the formation of ligaments to the maxillae, premaxillae and mesethmoid-premaxillary cartilage (Chapter 4.1). The nasal sac is bordered on all sides by the lateral ethmoid only, as in the loricariids Hypostomus plecostomus (Schaefer, 1987) and Pterygoplichthys (Howes, 1983a) but unlike in Hypoptopoma, where the nasal sac is free anteriorly (Howes, 1983a).

The membranodermal component of the frontal appears first, with the neurodermal (canal- bearing) component ossifying from it (or perhaps separately) somewhat later in ontogeny. First both parts can be well distinguished, but later they fuse more intimately. This is often seen in other teleostean canal bones as well (Daget, 1964). Exceptions in Ancistrus cf. triradiatus are the infraorbital bones, nasal and canal bearing cheek plates (see below), where the first ossification occurs around the canal. The anterior part of the adult frontal is relatively narrow, while the posterior part is broad enough to reach the orbit [as in Otocinclus (Schaefer, 1997) and Farlowella (Retzer & Page, 1996) but unlike the situation in Hypostomus plecostomus (Schaefer, 1987)].

In many siluriforms and other teleosts the skull floor bones appear earlier than the roof bones; usually the parasphenoid is the first bone to arise, more or less together with the basioccipital ossification around the notochord (Kobayakawa, 1992; Vandewalle et al., 1995, 1997; Adriaens & Verraes, 1998). Hoedeman (1960c) noted a slightly earlier development of some roof bones, i.e., the frontals and pterotics. When the parasphenoid ossifies, it is a U- shaped dermal sheet in the perimeter of the hypophyseal fenestra, also seen in Chrysichthys auratus (Vandewalle et al., 1995) and Clarias gariepinus (Adriaens & Verraes, 1998). In various teleosts the early appearance of the parasphenoid has been linked to the necessity of protecting the braincase from the physical particularities of food passing in the buccal cavity, from the moment of the transition from endogenous to exogenous feeding (Verraes, 1974a, PART 4 — ONTOGENY OF THE SKULL 75

Vandewalle et al., 1997, 1999; Adriaens & Verraes, 1998; Wagemans et al., 1998; Gluckmann et al., 1999). In Ancistrus cf. triradiatus though, the bone appears at 7.4 mm SL or 6 days after fertilization. Exogenous feeding starts around 9 mm or 8 to 9 days after fertilization. It could be hypothesized that parasphenoid ossification is related to the respiration movements and buccal pressure differences during respiration. Mechanical loading might be an important factor in inducing ossification (Mabee & Trendle, 1996; Adriaens & Verraes, 1998).

In siluriforms the posterior skull roof is composed of one large bone, known as the compound parieto-supraoccipital in siluriforms. Argumentation for the developmental fusion of paired parietal and supraoccipital ossification centra is given by Bamford (1948), Arratia & Menu-Marque (1981, 1984) and Fink & Fink (1996). In some cases (Callichthyidae, Clariidae) no developmental evidence was found (Hoedeman, 1960c; Adriaens & Verraes, 1998). In the ontogeny of Ancistrus cf. triradiatus separate parietals are never present. The parieto-supraoccipital arises as a paired, mostly perichondral bone. The development of the bone is somewhat complicated by the early presence of odontodes on the postotic processes, and posterior to the location where the possible parietals could be expected. There, membranous bone is soon added to the perichondral layer. The fact that a cleared and stained 8.0 mm SL specimen, as well as serial sections of another 8.0 mm SL specimen show one paired ossification anlage for the parieto-supraoccipital, is unusual. This might be correlated to the presence of odontodes on the skull roof, possibly hastening ossification in the region of the bone that will support it. I can, however, only speculate on the exact cause of this ossification pattern. I apply the name parieto-supraoccipital, not supraoccipital alone, but whether the parietals are fused to the supraoccipitals, or missing, can, however, not be concluded unambiguously from these data. The ossification centra might be close to each other, masking a possible double origin. The ending of the parietal branch of the supraorbital canal between the frontal and the sphenotic could be interpreted as an argument that the parietal is absent, but the branch might just have been reduced as well. It is absent in various catfishes (Arratia & Huaquín, 1995). Arratia & Huaquín (1995) regarded the absence of a parietal branch of the supraorbital canal as a synapomorphic condition for loricarioids. According to their definition this branch commonly runs from the frontal into (or above) the parieto-supraoccipital bone in catfishes, or does not reach it (thus might have been reduced). Schaefer (1987) reported this branch in the sphenotic in Hypostomus plecostomus, and in the frontal in Otocinclus (Schaefer, 1997), where he, rather confusingly called it ‘posterior’ branch, posterior to a ‘parietal’ (= epiphysial) branch. Whether this branch has disappeared in all loricarioid taxa except loricariids, or has secondarily reappeared in this family, remains to be verified. 76 PART 4 — ONTOGENY OF THE SKULL

According to Arratia (2003) a separate epioccipital is missing in Nematogenys and most trichomycterids, and also in Hypostomus and in scoloplacids. It has, however, been noted in other loricariids and in trichomycterids (Schaefer, 1987, 1997; Schaefer & Aquino, 2000). In Ancistrus cf. triradiatus it appears as a perichondral ossification of the middle part of the occipital pila. I favour the name epioccipital, and not epiotic (Schaefer, 1997; Arratia, 2003), following the argumentation of Patterson (1975: 425).

In siluriforms the identification of the bone usually termed posttemporo-supracleithrum, in the posterolateral corner of the skull, is not easy. Arratia & Gayet (1995) stated that there is no developmental evidence that the bone termed posttemporo-supracleithrum in siluroids is a compound element. It could be the posttemporal or supracleithrum alone, or the result of the early fusion of both elements during the earliest moments of ossification. Adriaens et al. (1997) suggested that in Clarias gariepinus the cleithral notch and the attachment of the Baudelot’s or transscapular ligament on it indicate it is at least composed of the supracleithrum, while the presence of an anteroventral process connecting the posterior element to the pterotic and a dorsal oblique process attaching it to the epioccipital might indicate that the posttemporal is also part of it (Adriaens & Verraes, 1998). In Callichthyidae, Scoloplacidae, Astroblepidae and Loricariidae the (posttemporo)- supracleithrum is fused to the pterotic (Regan, 1911b; Arratia, 2003). The detection of the separate bones in this ‘compound pterotic’ in loricariids has been further complicated. Analogous to the argumentation of Adriaens et al. (1997), Aquino & Schaefer (2002) considered the cleithral articular notch and the Baudelot’s ligament on the compound pterotic as indirect evidence of the incorporation of the supracleithrum into the ventral aspect of the pterotic. As Lundberg (1975) did, they concluded that there is no real evidence indicating the incorporation of the posttemporal as well, but that doesn’t mean this component is absent in (all) loricariids. Aquino & Schaefer (2002) also referred to Coburn & Grubach (1998), who suggested that loss or fusion of these elements may correlate to the loss of the first two occipital vertebral segments which they observed in the development of Corydoras paleatus (Callichthyidae) (see below for an account on the anterior vertebrae of Ancistrus cf. triradiatus). In A. cf. triradiatus the supracleithrum could be observed as a separate ossification, before fusing to the ventral layer of the pterotic (Fig. 31A). The cleithral dorsal process is closely associated to the supracleithrum, and the articular notch will form on this part of the ‘compound pterotic’ bone. The Baudelot’s ligament attaches and ossifies from this point as well (in case the posttemporal would be incorporated too, it would lie between the pterotic and the supracleithrum). To me, the name ‘compound pterotic’ seems most appropriate for the complex. The posttemporal is most probably never present in A. cf. triradiatus. PART 4 — ONTOGENY OF THE SKULL 77

The double-layered nature of the compound pterotic is considered a derived characteristic for loricariids by Aquino & Schaefer (2002), who added that the dorsal layer of the bone in loricariids is not homologous to that in other catfishes, as it does not bear the postotic canal. Also, this dorsal layer would appear prior to the ventral layer, which includes the neurodermal component enclosing the canal. This ossification sequence of the dermopterotic is not seen as such in Ancistrus cf. triradiatus. The ventral layer develops slightly earlier, and is grown to almost its full extent when the dorsal, strictly membranodermal layer starts to reach over the swimbladder capsule. Both are continuous with each other and the autopterotic at the occipital pila from the beginning, but only late in development they are connected by trabecles laterally. The absence of a pterotic branch of the postotic canal was considered a synapomorphy of loricarioids by Arratia & Huaquín (1995), who noted its absence in trichomycterids and nematogenyids. In a paper discussing the pterotic branch homology, however, Schaefer & Aquino (2000) could identify this branch in all loricarioids except scoloplacids and astroblepids. It is present in Ancistrus cf. triradiatus as well.

The ossification pattern of the Baudelot’s ligament in Ancistrus cf. triradiatus is interesting. As in other catfishes, it stretches from the ventral face of the supracleithrum towards the basioccipital, thereby forming a transverse ridge on the posteroventral skull floor. Two ossification centers are present in A. cf. triradiatus, one from the attachment point on the basioccipital, and one from the supracleithrum. In adults the boundary between both parts can still be seen, at the level of the basioccipital-pterotic contact. The boundary is also seen in Hypostomus plecostomus and Otocinclus vittatus (Schaefer, 1987, 1997). Not much is known about the ossification sequence of the ligament in other siluriforms (Fink & Fink, 1996), but in the ictalurid Trogloglanis pattersoni only one large ossification seems to arise from the supracleithrum (Lundberg, 1982).

The identification of the complex vertebra in loricariids is problematic, as a reduction in number of the anterior vertebral centra appears to have occurred. In literature on loricariid morphology, there appears to be a general consensus on the identification of the sixth vertebra, which carries a pair of large ribs connecting the vertebral column with the lateral dermal plates posterior to the head (Alexander, 1964; Chardon, 1968; Schaefer, 1987, 1997). The earlier assumption of Bridge & Haddon (1893) that it might be the fifth, was based on an unclear account of Reissner (1859). The sixth centrum is immovably sutured to the fifth, and its neural spine sutures to the posterior process of the parieto-supraoccipital. Chardon (1968) distinguished the fifth centrum from the first four, the centra of which must have become reduced significantly. In a developmental study of the Weberian apparatus in the callichthyid 78 PART 4 — ONTOGENY OF THE SKULL

Corydoras paleatus, Coburn & Grubach (1998) concluded that the first two vertebrae are missing, and the third and fourth lack basidorsals and basiventrals. The situation in Ancistrus cf. triradiatus might be similar. The fourth and fifth centrum originate as one long vertebral centrum. Their paraphophyses form the bony encapsulation of the swimbladder. The basidorsals of the fifth vertebra reach towards the cartilaginous tectum posterius. The resulting complex of vertebrae (up to the fifth) has a length of twice that of vertebra six or seven. A recent paper by Hoffmann & Britz (2006) discusses the homology of the anterior vertebral centra among otophysans. Contrary to the previous view, it hypothesizes that it is the fourth (not the fifth) basidorsal which contacts the tectum synoticum, and thus the fifth (not the sixth) which bears the large ribs and touches the parieto-supraoccipital process. It is the fourth vertebra that forms the os suspensorium, a feature of the fourth vertebra in otophysans (Hoffmann & Britz, 2006: 327).

Six, or rarely five infraorbital bones are present in Ancistrus cf. triradiatus. Usually six are found in hypostomine loricariids, though only five are found in Hypoptopomatinae (Schaefer, 1997). In one specimen of A. cf. triradiatus, there is no canal in the second infraorbital. Such a disjunct canal is slightly reminiscent to the situation in certain Trichomycteridae, where most of the infraorbital canal is lost, except for the part in the first infraorbital (Arratia & Huaquín, 1995). The shape and late ossification of the first bone of the infraorbital series, as well as the absence of an antorbital branch of the infraorbital canal, suggest the bone corresponds to infraorbital I and not to the antorbital. It doesn’t match the criteria Arratia & Huaquín (1995) used for the identification of this bone in Diplomystidae. The prefrontal plate might be homologous to the supraorbital-like tendon bone Arratia (1987) described in Diplomystidae. Howes (1983a: 332) noticed the resemblance in position between the loricariid prefrontal plate with the supraorbital bone of some non-siluriform taxa, but they are most probably not homologous (Fink & Fink, 1981; Howes, 1983a). The ossification of the infraorbital, nasal and suprapreopercle and the canal-bearing cheek plate differs from the situation in other canal bones. First, a neurodermal tube of bone arises around the sensory canal; only later the membranodermal component is formed against and on top of it. Adriaens & Verraes (1998) described the same phenomenon in the infraorbital, nasal and suprapreopercular bones in Clarias gariepinus. The number of prenasal and lateral plates or scutes varies in different specimens; particularly the prenasal plates are variable in both number and shape. Schaefer (1997) mentioned a more rigid pattern of these plates in Otocinclus. Sclerotic bones are present in larger specimens, although Fink & Fink (1996) regarded their absence as synapomorphic for Siluriphysi. Similar bones were found though in young PART 4 — ONTOGENY OF THE SKULL 79

Callichthys sp. by Arratia (1987). As in Salmo, they are situated anterior and posterior to the eyeball, while in Gasterosteus they have been described dorsal and ventral to it (Rojo, 1991).

SPLANCHNOCRANIUM

During ontogeny, the tooth-bearing dentary of the lower jaw in Ancistrus cf. triradiatus fuses to the mentomeckelium (Fig. 31B) and angulo-articular bones. In the adult stage only a rudiment of Meckel’s cartilage persists. As in other loricariids, as well as astroblepids, callichthyids and most trichomycterids, a coronomeckelian bone is absent (Mo, 1991; de Pinna, 1993).

The double posterior process of the autopalatine bone acts as a double insertion point for the extensor tentaculi muscle subdivisions. In the basal siluriform Diplomystes and †Hypsidoris a similar but even larger, single posterior extension is present, posterior to the articular facet with the lateral ethmoid (Arratia, 1987; Grande, 1987). During early ontogeny, no sign is found of a palatine splint or sesamoid bone as seen in Otocinclus (Schaefer, 1997); it can, however, be seen in adult Ancistrus cf. triradiatus (Chapter 6.1). Schaefer (1997) considered it a dermal or sesamoid ossification, variably present in loricariids. The view of Howes & Teugels (1989), suggesting the presence of dermal ento- and ectopterygoids next to the perichondral metapterygoid in some catfishes, is opposed to the hypothesis supported by Alexander (1965), Gosline (1975), Arratia (1990, 1992), Fink & Fink (1996), and Adriaens & Verraes (1998), who reported only the metapterygoid to be present, as a perichondral ossification of the chondrocranial pterygoid process, and having membranous outgrowths. The latter interpretation is followed in this dissertation. The interpretation of Hoedeman (1960b), with the metapterygoid being part of the hyomandibular ossification, is incorrect. Ectopterygoids are only found in some individuals within the Diplomystidae (Arratia, 1992). Sesamoid ento- or ectopterygoids are also lacking in Ancistrus cf. triradiatus [they are present in several catfish families (Arratia, 1992; Kobayakawa, 1992; Diogo et al., 2001)]. The hyomandibular articulation with the sphenotic, prootic and pterotic bones has also been observed in some trichomycterids (Arratia, 2003), but is uncommon in siluriforms (where usually only one or two of these bones are involved). In adult individuals of Ancistrus cf. triradiatus the hyomandibular trunk enters the hyomandibular bone at the medial side and leaves it at its lateroventral margin, medial to the preopercle, whereas it leaves the bone at the lateral side in Hypostomus and Otocinclus (Schaefer, 1987, 1997).

Schaefer (1988) elaborated on the identity of the largest, canal-bearing cheek plate, present in many loricariids. He concluded that it is not homologous with the interopercle of 80 PART 4 — ONTOGENY OF THE SKULL most other catfishes, as no other teleosts possess a canal in the interopercle, and this canal communicates directly with the preopercular canal terminus (his exit 5) in primitive siluroids. In the loricariid genus Delturus a true interopercle might be present, although the homology issue remains problematic (Armbruster, 2004). No ontogenetic stage of Ancistrus cf. triradiatus shows any sign of the interoperculo-mandibular ligament, which is assumed to be lacking in most loricariids as well as in astroblepids (Schaefer, 1988; Armbruster, 2004). Development of the cheek plate in A. cf. triradiatus starts with a neurodermal gutter-like bone surrounding the canal at 11.7 mm SL, followed by the addition of a small odontode- bearing membranodermal component at 14.4 mm SL. Schaefer (1988) observed an opposite sequence in Sturisoma sp.: the odontode-bearing part arises before a canal is observed in the bone. In adults of Ancistrus cf. triradiatus the suprapreopercle is fused to the sixth infraorbital bone (Chapter 6.1). This could not (yet) be observed in any of the examined embryonic and juvenile specimens. The infraorbital and preopercular canals, however, remain separated. In Otocinclus both canals sometimes share a pore between the sphenotic, pterotic and posterior (fifth) infraorbital bone (Schaefer, 1997; Chapter 6.3).

The cartilaginous interhyal connects the chondrocranial hyoid arch with the hyosymplectic cartilage. In Ancistrus cf. triradiatus it is lost during ontogeny. The final articulation between the hyoid bar and the suspensorium is assisted by a series of ligaments (Chapter 6.1). The loss of the interhyal is also seen in Clarias gariepinus (Nawar, 1954; Adriaens & Verraes, 1994). A cylinder-shaped sesamoid bone arising in a ligament at the medial side of the original interhyal location (Fig. 31D) might well be unique for loricariids. It is present in Hypostomus plecostomus and Otocinclus vittatus [though interpreted as an interhyal by Schaefer (1987, 1997)]. It is hypothesized to act as a support, strengthening the articulation, which may well be needed to resist the strong forces exterted by the suction used by loricariids to keep the body attached to substrates, often in fast flowing water. The absence of the interhyal is shared by loricariids and scoloplacids (Bailey & Baskin, 1976). The branchiostegal rays articulate with the ventrocaudal process of the hyoid bar, which is a large and cartilaginous extension of the hyoid bar at the level of the joint between the anterior and posterior ceratohyal. Arratia (1987) saw a similar situation in Loricarichthys sp., and a different situation in Callichthys callichthys, where three separate cartilage elements connect the four branchiostegal rays with the hyoid arch, while the rays articulate with the ceratohyals directly in other catfishes including diplomystids.

The minuscule cartilage nucleus present in front of the infrapharyngobranchial III in adult specimens (Chapter 6.1) has not been found in any of the studied developmental stages, and PART 4 — ONTOGENY OF THE SKULL 81 must, therefore, develop later than the 12.4 mm stage (its apparent absence in the older in toto stained specimens (14.4-25.0 mm) might be due to very weak alcian blue staining). Alexander (1965: 136) stated that the loricariid Hypostomus plecostomus has no pharyngeal teeth, as “it does not require them.” This stands against the observation by Schaefer (1987), who counted numerous teeth on both the upper and lower pharyngeal jaw in the same species, exactly as in Ancistrus cf. triradiatus. The lower pharyngeal jaws arise independently as supporting plates for the pharyngeal teeth, and coalesce secondarily with the fifth ceratobranchials that appear at the same moment, as in other siluriforms (McMurrich, 1884; Vandewalle et al., 1999). The development of the upper pharyngeal jaws has started much earlier than that of the lower, a sequence also observed in other siluriforms (Adriaens & Verraes, 1998).

In Ancistrus cf. triradiatus the parurohyal is pierced by one blood vessel, the inferior jugular vein (Fig. 31H). This vein receives blood from vessels draining the hyohyoideus inferior, and intermandibularis anterior and posterior muscles14, before ascending through the center of the parurohyal. It then receives several more veins from the sternohyoideus, before running to the sinus venosus, crossing the ventral aorta at the right side. This vein has also been reported by Nawar (1955b) in Clarias gariepinus, though not piercing the parurohyal. Instead, in C. gariepinus a direct branch of the ventral aorta descends through it, before sending branches into the hyohyoideus inferior (Adriaens & Verraes, 1998; pers. ob.). In A. cf. triradiatus, the arteries irrigating the hyohyoideus inferior branch off from the aorta ventralis and run above the parurohyal, entering the muscle more laterally. These strikingly different configurations also differ from the situation in Nematogenys, Trichomycterus and Noturus, where it is the hypobranchial artery that pierces the parurohyal (Arratia & Schultze, 1990). In Clarias gariepinus the first basibranchial is absent (Nawar, 1954). In their developmental study Adriaens & Verraes (1998) concluded it is most likely incorporated in the parurohyal. The present study corroborates this hypothesis: serial sections of subsequent stages show that the first basibranchial splits off from the next basibranchials, and becomes reduced. It remains, however, continuous to the dumbbell-shaped cartilage nucleus of the parurohyal. It is difficult to say whether it disappears completely or forms the dorsalmost part of the medial dorsal ridge of the parurohyal anterior to the foramen for the inferior jugular vein. I refer to chapter 5.2 for a further treatise of the parurohyal bone in relation to the sternohyoideus muscle.

14 The intermandibularis posterior is often called protractor hyoidei. In loricariids, however, this name is erroneous (see Chapter 5.2). 82 PART 4 — ONTOGENY OF THE SKULL

PART 5

ONTOGENY OF THE CRANIAL MUSCULATURE

PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 83

5.1. ONTOGENY OF THE JAW AND MAXILLARY ∗ BARBEL MUSCULATURE

Abstract — The neotropical loricarioid catfishes include six families, the most speciose of which are the Callichthyidae and the Loricariidae. Loricariidae (suckermouth catfishes) have a highly specialized head Bauplan, including an exceptionally high number of muscles derived from the adductor mandibulae complex and the adductor arcus palatini. Terminology of these muscles varies among the literature, and no data exist on their ontogenetic origin. Thus far, hypotheses of identity or homologies have been based on adult morphology only. A detailed examination of the ontogeny of both a callichthyid and a loricariid representative now reveals the identity of the jaw and maxillary barbel musculature, and supports new hypotheses concerning homologies. The adductor mandibulae muscle itself is homologous to the A1-OST and A3’ of basal catfishes, and the A3’ has given rise to the loricariid retractor veli as well. The A2 and A3” have resulted in the retractor tentaculi of Callichthyidae and the retractor premaxillae of Loricariidae. Thus, these two muscles are homologous. In Loricariidae, the extensor tentaculi consists of two separate muscles inserting on the autopalatine, and evidence is given on the evolutionary origin of the loricariid levator tentaculi (previously known as retractor tentaculi) from the extensor tentaculi, and not the adductor mandibulae complex.

5.1.1. INTRODUCTION

Research on teleostean ontogeny has generally involved studies concerning external morphology and skeletal development. Only a few authors have investigated the myological transformations (e.g., Jarvik, 1980; Surlemont et al., 1989; Surlemont & Vandewalle, 1991; Adriaens & Verraes, 1996, 1997a, d, e). Data on the early ontogeny of the musculature might, however, yield important findings in cases where origins or homologies of muscles are unknown or questioned. Also, transformations of musculo-skeletal systems are key factors in changes in the efficiency of functional units during ontogeny, and, hence, survival of early life history stages of fishes. The knowledge about the musculature serving the jaws and the maxillary barbels in the ostariophysan teleosts suffers from the variety in number and properties of the muscles, although Takahasi (1925), Winterbottom (1974), Gosline (1989), Adriaens & Verraes (1996), Diogo & Chardon (2000a), Wu & Shen (2004) and others have published important contributions towards the understanding of the origin and homologies of these muscles. For the loricarioid catfishes the situation is even more complex, due to the increasing number of subdivisions with shifted insertions, added to the often highly mobile and differently oriented

∗ Slightly modified from: Geerinckx T., Huysentruyt F. & Adriaens D. Ontogeny of the jaw and maxillary barbel musculature in the armoured catfish families Loricariidae and Callichthyidae (Loricarioidea, Siluriformes), with a discussion on muscle homologies. Submitted to the Zoological Journal of the Linnean Society. 84 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE jaw bones (Howes, 1983a; Schaefer & Lauder, 1986, 1996; Chapter 6). As a consequence, the muscular nomenclature is almost as diverse as the musculature itself, and names are usually appointed on the basis of assumed muscle function, which may or may not be coupled to a homology. This chapter focusses on the adductor mandibulae complex, originating from the dorsal adductor part of the mandibular muscle plate, and the extensor tentaculi muscles, originating from the adductor arcus palatini, the anteriormost dorsal part of the hyoid muscle plate (Jarvik, 1980). The retractor tentaculi of most siluriforms is considered a derivative of the adductor mandibulae complex (e.g., Eaton, 1948; Howes, 1983b; Adriaens & Verraes, 1996). The ontogeny of the intermandibularis muscles, also inserting on the lower jaw, is beyond the scope of this publication and are dealt with in chapter 5.2. The ontogeny of the jaw and maxillary barbel musculature of a specialized loricariid was examined, and compared to that of a more basal callichthyid. Loricariid head musculature has been described by Howes (1983a), Schaefer & Lauder (1986), Schaefer (1997) (Fig. 38A). In most Loricariidae, the adductor mandibulae part that acts directly on the lower jaw [muscle b of Howes (1983a)] originates on the quadrate, hyomandibula and preopercle, and inserts on the angulo-articular and dentary bones of the lower jaw. The retractor premaxillae [muscle c of Howes (1983a)] originates on the hyomandibula, and inserts tendinously on the posterior aspect of the premaxillae. The retractor veli [muscle d or retractor palatini of Howes (1983a)] originates on the metapterygoid and sends its tendon into the oral valve. The ‘retractor tentaculi’ [muscle a of Howes (1983a)] has its origin on the anteroventral face of the lateral ethmoid and runs to the maxillary bone. The extensor tentaculi consists of a lateral and a medial part that are completely separated [e and f of Howes (1983a)]. The extensor tentaculi pars lateralis runs from a canal-like groove, formed by the lateral ethmoid and the metapterygoid, to the ventrolateral autopalatine process [this groove is not canal-like in those loricariids that lack a well-developed lateral metapterygoid ridge (Howes, 1983a; Armbruster, 2004)]. The extensor tentaculi pars medialis connects the ventromedial autopalatine process with the ventral skull surface and is considered the antagonist of the retractor premaxillae (Alexander, 1965; Howes, 1983a; see also Chapter 6.1). The most relevant accounts on callichthyid musculature have been published by Howes (1983a) and Schaefer & Lauder (1986) (Fig. 38B). The autopalatine-maxillary mechanism is synapomorphic for catfishes, but some significant structural diversity has been noticed within the group (Gosline, 1975; Ghiot, 1978; Ghiot et al., 1984; Adriaens & Verraes, 1997a; Diogo & Vandewalle, 2003a). The mechanism in Loricariidae and Callichthyidae is that of the rotating type, and is characterized by a neurocranial articulation at the posterior end of the autopalatine (Huysentruyt & Adriaens, 2005b; Chapter 6). PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 85

5.1.2. BRIEF MATERIAL AND METHODS

2 µm thick serial sections have been made of embryos of Ancistrus cf. triradiatus and Corydoras aeneus (see Tables I and II, and paragraph 2.2.7). 5 µm thick serial sections of (sub)adults of both species, and some other loricariids (Table II) were studied as well. See paragraph 2.2.8 for details on the 3D-reconstructions of the 6.1 and 8.0 mm SL A. cf. triradiatus specimens.

5.1.3. RESULTS

In this paper, ‘adductor mandibulae’ refers to the cheek musle division(s) inserting on the lower jaw, whereas ‘adductor mandibulae complex’ refers to the whole set of muscle subdivisions originating from the A1-OST, A2 and A3 parts (there are no A1 and Aω, see below). In the description of Ancistrus cf. triradiatus, the loricariid muscle known as the ‘retractor tentaculi’ (see Introduction) is here referred to as levator tentaculi, because of the hypothesis of a non-homology with the retractor tentaculi in Corydoras aeneus and other siluriforms. Argumentation concerning homologies, as well as the use of the nomenclature, is given in the Discussion.

ONTOGENY OF ANCISTRUS CF. TRIRADIATUS

6.1 MM SL — 4 DAYS POST-FERTILIZATION In this early embryonic specimen, more or less one day before hatching, most of the cheek musculature is already visible, albeit in a rudimentary state. Observations of living embryos (removed from the egg scale) indicate that only hyoid movements occur in the head. The chondrocranium is only partially formed, but the trabecular bar and main splanchnocranial elements are present (Chapter 4.1). Notice that in figure 39 (and 41) the branchial basket and the developing teeth are not shown. The adductor mandibulae complex consists of four recognizable divisions. The outermost muscle, the external part of the adductor mandibulae, originates near the anterolateral margin of the hyosymplectic part of the cartilaginous suspensorium, close to the palatoquadrate part. It runs dorsolateral to the palatoquadrate, and ends near, but not on, the dorsolateral aspect of Meckel’s cartilage. The second, medioventral muscle, the internal part of the adductor mandibulae, originates anterior to the anterodorsal margin of the palatoquadrate. The origin of this muscle is confluent with the external adductor mandibulae part (Fig. 40C). It almost reaches the posterodorsal face of Meckel’s cartilage, slightly medial to the basis of the coronoid process (Fig. 39). 86 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

The third muscle, the retractor premaxillae, is situated mediodorsally to the external adductor mandibulae part, their posterior fibres merging together (Fig. 40C). The fibres of the retractor premaxillae pass the coronoid process of Meckel’s cartilage dorsally, and end just anterior to it (thus having no insertion). On the medioventral aspect of the anteriormost part of the muscle, a developing aponeurosis is seen (Fig. 40A), without any contact to a recognizable structure whatsoever. The premaxilla is only recognizable by the presence of a few tooth germs, borne by a thin sheet, which is barely ossified. The retractor veli, the fourth, medialmost muscle of the adductor mandibulae complex, originates close to the anterodorsal margin of the palatoquadrate, more or less where the pterygoid process will later develop (Fig. 40B). Except anteriorly, the fibres are continuous with those of the internal adductor mandibulae part, and thus, indirectly, with the external adductor mandibulae and the retractor premaxillae (Fig. 40C). The anterior end of the retractor veli is situated posterodorsally to Meckel’s cartilage. Its fibres diverge medially from the rest of the complex, and are directed towards the tissue bordering the oral cavity, at the point where the oral valve is developing. The extensor tentaculi has not yet developed by this time. The posterior end of the palatine cartilage, to which the muscle inserts in the older specimens, is just beginning to develop. The levator tentaculi, however, has already developed, well apart from the adductor mandibulae complex. It originates close to the trabecular bar of the skull floor, near the still rudimentary orbitonasal lamina. It follows a slightly lateral course, passing just lateral to the palatine cartilage, and ending without a clear insertion between the palatine and maxillary cartilages (Fig. 40A). Only the first anlage (no calcification) of the maxilla is seen, anterior to the rostral end of the muscle. The path of some of the main nerve branches emerging from the infraorbital trunk of the trigeminofacial complex can already be followed. The mandibular branch of the trigeminal nerve separates from the trigemino-facial complex outside the skull, and crosses the retractor premaxillae from medial to lateral (Fig. 40B), subsequently entering the adductor mandibulae complex between the retractor premaxillae and the external adductor mandibulae part. It passes over Meckel’s cartilage just medial to the coronoid process, thus anterior to the rostral end of the internal adductor mandibulae part. The anastomosed maxillary and buccal branches run in a rostral direction, lateroventral of the trabecular bar. They are not adjacent to the adductor mandibulae complex, but lie just laterally to the levator tentaculi (Fig. 40A). In this stage it is not clear where both branches separate. At the level of the anterior end of the levator tentaculi the fibres diverge, but they are not sufficiently developed to allow further examination of their courses. For the same reason, no muscle innervation can be observed at this stage.

PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 87

8.0 MM SL — 7 DAYS POST-FERTILIZATION In this specimen still few elements are ossifying, those in the cheek region being the dento-mentomeckelium (the dentary barely fused to the mentomeckelium), the premaxilla and the maxilla. Cheek muscles that were already present have enlarged substantially. The extensor tentaculi is present as well, having most probably appeared somewhere between the 6.1 and 7.0 mm stages. The external adductor mandibulae part inserts on the lateral portion of Meckel’s cartilage and on its coronoid process, which points rostrodorsally. In the posterior half of the muscle, the fibres still can not be separated from those of the other muscles of the adductor mandibulae complex (Fig. 42B). The origin of the external adductor mandibulae part is on the hyosymplectic part of the suspensorium, up to the insertion place of the levator arcus palatini, which is situated posteromedial to it. As dense connective tissue connects the muscle to both its insertion surfaces, it can be assumed that the muscle could now well be functional. The internal adductor mandibulae part now has both its insertions as well. It originates on the lateral aspect of the anterior part of the hyosymplectic cartilage, while an equal portion of the muscle originates on the dorsal tissue sheet extending between this suspensorial part and the now completely developed pterygoid process. It can be supposed that, as long as this sheet hasn’t ossified into the dermal part of the metapterygoid bone, these ventralmost fibres probably are not functional. The muscle appears to have developed an insertion on Meckel’s cartilage (Fig. 42A). The retractor premaxillae is substantially larger compared to the earlier specimens, and has become more separated from the external adductor mandibulae part. In transsection, the muscle is flattened posteriorly, but more or less round for most of its length. Its fibres originate on the hyosymplectic, medial to those of the external adductor mandibulae part, and anterior to anteromedial to the levator arcus palatini. The muscle has considerably lengthened, even compared to the other cheek muscles, and now reaches anterior to the lower jaw. Its tendon, developed from the ventromedial aponeurosis, is continuous with the connective tissue sheet that stretches between the developing maxilla and premaxilla, and the anterior tip of the palatine cartilage. Considering the location of this connective tissue sheet, it corresponds to the primordial ligament as defined by Gosline (1986:707). Thus no direct tendinous contact exists between the retractor premaxillae and the maxilla or the premaxilla. The retractor veli is still continuous with the internal adductor mandibulae part for half of its length, with its origin at the basis of the pterygoid process. It can now be clearly observed that anteriorly the fibres run up to the lateral edge of the oral valve, contacting the dense connective tissue that continues inside this valve (Fig. 42A-B). The extensor tentaculi pars medialis originates near the ventral side of the trabecular bar, at the level of the preorbital basis. It is a horizontally flattened muscle, ending at the 88 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE posteriomedial face of the palatine cartilage. It is improbable that the muscle is functional it this moment, as there seems to be no contact between the developing tendon and the palatine cartilage. Caudally, the muscle fibres touch those of the adductor arcus palatini, although there is no continuity between both muscles. The extensor tentaculi pars lateralis is more adjacent to the levator tentaculi (see below) than to the pars medialis. The origin of the extensor tentaculi pars lateralis is near the ventral side of the trabecular bar as well, at the level of the orbitonasal foramen, but most fibres originate more laterally, ventral to the orbitonasal lamina. The pars lateralis runs further rostrally compared to the pars medialis, towards the posterolateral edge of the back of the palatine cartilage. It does not seem, however, to reach it, nor does it provide any sign of a tendinous insertion. There is no contact between the fibres of both retractor tentaculi parts. The levator tentaculi very much resembles the extensor tentaculi pars lateralis: it lies directly laterally to it, is equally flattened in a vertical plane, and also originates near the ventral side of the orbitonasal lamina. Far posteriorly, the fibres are in contact, but do not merge. The levator tentaculi, however, continues further rostrally along the lateral aspect of the palatine cartilage. The ventral fibres are now directed somewhat laterally, so that, at the level of the anterior palatine end, the muscle is horizontally instead of vertically flattened (Fig. 41). As such, the ventral fibres insert most distally on the maxilla; the dorsal ones most proximally, close to the palatine articulation. The levator tentaculi and both extensor tentaculi muscles are already present in the younger 7.0 mm specimen as well, however, less developed, and with fibres also close to each other but not confluent (Figs 40D, 42A). The maxillary and buccal nerve branches are still anastomosed. After separating from the mandibular branch dorsal to the pterygoid process basis, they continue as one strand, mediodorsal to the adductor mandibulae complex, along the lateroventral edge of the levator tentaculi. Two minor branchlets leave the bundle to innervate the levator tentaculi. On arrival on the dorsal surface of the retractor premaxillae, the mandibular branch splits off a medial branch that enters the adductor mandibulae complex dorsomedially, and then continues laterally before descending between the retractor premaxillae and the external adductor mandibulae part. It runs dorsally on the internal adductor mandibulae part where this muscle inserts on the lower jaw, and then runs over Meckel’s cartilage before innervating the dentary region and the muscles ventral to it (Fig. 41A-C).

12.4 MM SL — 43 DAYS POST-FERTILIZATION As most skeletal structures of the skull start to ossify at a standard length of 9 to 11 mm, important changes in shape are seen by the time a standard length of 12.4 mm is reached (Chapter 4.2). During the onset of this major ossification period, the upper snout region goes PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 89 through a phase of elongation, while the lower jaw acquires its more or less final shape. These transformations are reflected in the muscular development. The orientation of the suspensorium has become more horizontal. The longitudinal axis of the lower jaw is directed medially instead of rostrally (as is the case in most teleosts), and the direction of the muscles inserting onto the lower jaw is perpendicular to it. Therefore, an important change during the ontogeny of the jaw musculature in Ancistrus cf. triradiatus is the broadening of the external and internal parts of the adductor mandibulae, with the external part lying more dorsal to the internal part, when compared with earlier stages (Fig. 42c). Both these adductor mandibulae parts also become broader. Ossification of the lower jaw has significantly progressed by now, with the final insertion site of the external adductor mandibulae part now being the dorsal aspect of the angulo- articular and the lateral face of the dentary coronoid process. The internal adductor mandibulae part now attaches to the dorsocaudal aspect of the dento-mentomeckelian bone, as well as in the Meckelian fossa, and the medialmost part of the angulo-articular. Near its origin on the suspensorium, the lateralmost fibres of the internal part are still fused to the external part, but lie separate along most of their length. The external part originates on the hyomandibula and the dorsal aspect of the preopercle, that has developed by now. The internal part originates on the quadrate and the metapterygoid; only a few fibres insert on the hyomandibula. This already corresponds to the adult configuration. In the internal part, two layers can be discerned, intimately connected at their anterior, tendinous insertion, but well separable posteriorly (Fig. 43A). The upper snout elongation affects those muscles inserting on it, i.e., the retractor premaxillae and the levator tentaculi. The retractor premaxillae has further grown anteriorly, and now reaches the posterior face of the premaxilla, and can thus be considered functional. Anteriorly, the ventral aponeurosis of the retractor premaxillae has largely lost its contact with the connective tissue part attached to the maxilla; in adults, there is no connection whatsoever. An interesting feature in the 12.4 mm specimen concerns the path of the mandibular nerve branch relative to the retractor premaxillae: in one side of the specimen the nerve branch runs dorsally and laterally to the muscle, as described in the earlier stages, but in the other side it enters and exits the muscle dorsally, locally separating a small group of fibres from the main muscle mass for a short distance (Fig. 42D). The same phenomenon is observed in one side of the 10.2 mm specimen, although here it occurs more laterally (Fig. 42E). The retractor veli is now clearly discernible from the internal adductor mandibulae part, and has its origin on the metapterygoid, on and just posterior to the basis of the ossified pterygoid process. 90 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

The extensor tentaculi pars medialis originates on the ventral faces of the lateral ethmoid and the orbitosphenoid. In larger specimens attachment is also observed on the lateral side of the ventral parasphenoid ridge. The muscle inserts tendinously on the ventromedial process of the autopalatine. The metapterygoid has now developed a lateral apolamella, which serves as additional insertion space for the extensor tentaculi pars lateralis in larger specimens, but not yet in the 12.4 mm specimen. A round aponeurosis lies in the centre of the muscle, and is continuous with the short tendon attaching to the ventrolateral process of the autopalatine. Both the medial and the lateral part of the extensor tentaculi can be considered functional in this stage. The lateral migration of the levator tentaculi is the most prominent event occurring between this and the former stage (Fig. 42C). The orbitonasal skull region is broadening; the muscle origin is now on the lateral half of the rostroventral aspect of the lateral ethmoid, a bone that expands rostrally and laterally to accommodate and support the olfactory organ. Also the maxilla grows allometrically, becoming considerably longer during the early ossification phase (Chapter 4.2). Thus both points of attachment of the muscle are shifting laterally, with the logical consequence that the muscle migrates laterally as well. As a result, the levator tentaculi lies no longer directly against the palatine cartilage, which is now becoming replaced by the autopalatine bone. The anterior end of the muscle has become more flattened, its width increasing as the maxilla elongates. Except for the irregular path of the mandibular nerve branch, there is no significant difference between the position of the nerves in the cheek region in this specimen and in the adults (see below).

ONTOGENY OF CORYDORAS AENEUS

4.9 MM SL — 6 DAYS POST-FERTILIZATION The adductor mandibulae complex consists of three subdivisions that are continuous for most of their length: two dorsal and one ventral one. It is only little differentiated. Most of it originates on the hyosymplectic part of the suspensorium, lateral and anterior to the insertion of the levator arcus palatini. Some fibres originate on the palatoquadrate part. The ventral division inserts musculously on Meckel’s cartilage, and thus is considered as the true adductor mandibulae muscle (Fig. 44B). At this moment, the separation into an internal and an external part is not clear, except anteriorly (Fig. 44B). A dorsolateral and a dorsomedial division pass the lower jaw dorsally, uniting on a ventrally placed aponeurosis that is continuous with the primordial ligament. More caudally these two dorsal divisions are separated by the mandibular branch of the trigeminal nerve. Posterior to the lower jaw, the dorsolateral division is unrecognisibly fused to the adductor mandibulae muscle. PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 91

The extensor tentaculi is a single muscle that originates on the trabecular bar near the orbitonasal lamina, as well as on the medial floor of this lamina. It runs towards the laterocaudal end of the palatine cartilage (Fig. 44A). The (bucco)maxillary branch of the trigeminal nerve runs between the extensor tentaculi and the dorsomedial division of the adductor mandibulae complex (Fig. 44A).

9.3 MM SL — 16 DAYS POST-FERTILIZATION The dorsomedial division, medial to the mandibular nerve branch, is now relatively well distinguishable from the remainder of the adductor mandibulae complex (Fig. 44D), except near both insertions, i.e., posteriorly, on the hyomandibula, and anteriorly, near its tendon, where it is joined by the dorsolateral division. This dorsolateral division, lateral to the mandibular nerve branch, still can’t be separated from the adductor mandibulae inserting on the lower jaw, except for the fact that its fibres extend more anteriorly, fusing with the dorsomedial division (Fig. 44C). The dorsomedial and dorsolateral divisions attach to a ventral aponeurosis that inserts on the primordial ligament. As such, this compound muscle [retractor tentaculi of Howes (1983a)] is penetrated posteriorly by the mandibular nerve branch. In the adductor mandibulae muscle, an external and internal part can now be distinguished. The mandibular branch runs over the internal part before entering the lower jaw, as it does in Ancistrus cf. triradiatus. The extensor tentaculi now inserts on a ventrolateral process on the posterior tip of the autopalatine. A remarkable observation concerns the neurocranial origin of this muscle. At the level of the nasal organ the muscle becomes significantly broader (Fig. 44C), and the posterior portion of the muscle is split in two parts: a medial one originating on the trabecular bar and the preorbital base, and a laterodorsal one originating almost completely laterally on the ventral aspect of the orbitonasal lamina. This posterior separation is even more pronounced in adult specimens (Fig. 45B) (pers. ob.).

MUSCLE INNERVATIONS AND NERVE POSITIONS IN ANCISTRUS CF. TRIRADIATUS

In (sub)adult Ancistrus cf. triradiatus and Corydoras aeneus the main branches of the infraorbital trunk of the trigeminofacial root are complexly intertwined, but can be identified for most of their course in the serial sections of the 12.4 and 33.5 mm specimens (A. cf. triradiatus) and the 9.3 and 39.0 mm specimens (C. aeneus).

In Ancistrus cf. triradiatus, the three main branches of the infraorbital trunk are the maxillary and mandibular branches of the trigeminal nerve and the buccal branch of the facial nerve. The buccal branch separates from the trunk, but re-enters it at the level of the posterior eye margin. Here, the trunk breaks up in two large parts. The ventral one is the mandibular 92 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE branch, and the dorsal one is the anastomosed maxillary and buccal branches, the latter sending off small branches laterally (to pit organs and the infraorbital canal). Some of the buccal branchlets stay anastomosed to the maxillary branch (all of these minor buccal branches are omitted in the figures). The ventralmost fibres of the mandibular branch descend into the dorsomedial portion of the adductor mandibulae complex, between the retractor premaxillae and the retractor veli. The fibres divide into two portions, the dorsal one innervating the retractor premaxillae, the ventral one serving the internal and external parts of the adductor mandibulae, and the retractor veli (visible, but not indicated on figures 42D and 43B). The main part of the mandibular branch runs ventrally, between the retractor premaxillae (medial) and the external adductor mandibulae part (lateral). In one side of the 33.5 mm specimen, the mandibular branch even enters the retractor premaxillae, splitting the muscle into a large medial part and a somewhat smaller lateral part (Fig. 43B). Apart from this, the two parts can not be distinguished from each other, neither posterior nor anterior to the nerve course. This is even more aberrant than the separation of only a few fibres of the retractor premaxillae in the 10.2 and 12.4 mm specimens (see above). The separation of only a few fibres is also observed in the other side of the 33.5 mm specimen. The mandibular branch then runs ventral to the retractor premaxillae and dorsal to the internal adductor mandibulae part (Fig. 43A). Close to the insertion of the latter muscle it splits into a part serving the dentary region, and a part running to the musculature between the lower jaw and the hyoid. Analogous to the papers of Vetter (1878), Juge (1899) and Atoda (1936) I name the branch that innervates the dentary itself the external mandibular branch (although in Ancistrus cf. triradiatus it is situated most medially, as a result of the medially rotated jaw); the internal mandibular branch is the one that sends branches into the intermandibular and protractor hyoideus muscles. A comparison with Pterygoplichthys lituratus (Hypostominae), Otocinclus vestitus (Hypoptopomatinae) and Farlowella acus (Loricariinae) yields that the mandibular branch in these specimens always runs between the external adductor mandibulae part and the retractor premaxillae, not separating fibres of the latter muscle. The maxillary branch runs dorsolaterally to the retractor premaxillae, and continues along the lateroventral edge of the levator tentaculi, sending one portion to the premaxillary region in the snout, and a larger portion into the maxillary barbel. The buccal branch sends off several small and larger twigs along its course in the cheek. Some fibres stay anastomosed to a minor bundle of the maxillary branch, ending in and near the premaxilla, a situation also described in Parasilurus asotus (Atoda, 1936). Both parts of the extensor tentaculi are innervated by separate portions of the branch that serves the adductor arcus palatini as well, and originates from the hyomandibular trunk of the trigeminofacial root (Fig. 43B). The levator tentaculi is innervated by a branch of thick fibres originating from the anastomosed maxillary-buccal branch (Figs 42D, 43B). They can be PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 93 traced up to the level of the posterior eye margin, where they seem to lose themselves in two small groups absorbed by the trigeminal part of the nerve complex formed by fibres of the mandibular and maxillary branches. Except for the mandibular branch of the trigeminal nerve (see above), no significant differences were observed in the nerve positions and innervations in the loricariid Pterygoplichthys lituratus, Otocinclus vestitus and Farlowella acus.

MUSCLE INNERVATIONS AND NERVE POSITIONS IN CORYDORAS AENEUS

Near their origin, it is hard to discriminate between the buccal and the mandibular and maxillary branches, as several parts of the buccal branch anastomose to the other branches. Along the whole of the length of the adductor mandibulae complex, the well developed maxillary branch lies mediodorsal to this complex, and lateroventral to the extensor tentaculi, most of its fibres finally entering the maxillary barbel. The adductor mandibulae complex is innervated by a ventral twig of the mandibular branch, which has separated from the maxillary branch immediately after its emergence from the skull. The mandibular branch then enters the complex separating the dorsolateral and dorsomedial divisions (thus splitting the retractor tentaculi in two parts; Fig. 45B). It then continues, rostrally separating the largest external part of the adductor mandibulae and the smaller internal part (Fig. 45A). It then sends off its branchlets to the dentary region and the musculature in the ventral region of the lower jaw. The extensor tentaculi is innervated by the same branch that serves the adductor arcus palatini, and that originates from the hyomandibular trunk of the trigeminofacial root.

5.1.4. DISCUSSION

THE ADDUCTOR MANDIBULAE COMPLEX

Vetter’s (1878) terminology of the adductor mandibulae sections can been applied to siluriforms, though has been replaced by more recent treminologies (see below). Vetter’s A2- section typically lies ventrolaterally in teleosts, and inserts on the dorsal face of the lower jaw, usually including the dentary coronoid process, and often also the angulo-articular and the Meckelian fossa (Vetter, 1878; Winterbottom, 1974). The more medial A3-section usually inserts on the medial face of the dentary and/or the Meckelian fossa (Vetter, 1878; Winterbottom, 1974). In ostariophysans it always inserts (at least) on the Meckelian fossa (Gosline, 1989; Wu & Shen, 2004). Both sections have been found in siluriforms, and can be discerned as follows (in at least some taxa; Takahasi, 1925): the external A2-section is fused to an outer portion of the internal A3-section, thus giving rise to the combined-A2A3’ 94 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE section, that is separated posteriorly from the inner-A3 portion (or A3”) by the insertion of the levator arcus palatini. Takahasi’s identification of an A3’-part fused to the A2 is based on the relation of the muscle to the Aω-section in those cypriniforms and siluriforms examined by him. His configuration and nomenclature have been followed by, e.g., Gosline (1989) and Adriaens & Verraes (1996). In some cases the mandibular branch of the trigeminal nerve passes between the A2- and A3-sections, but this is not a general situation (Edgeworth, 1935; Winterbottom, 1974; Gosline, 1989; see below). The A2-section of Vetter (1878) has been named ‘external division’ (Gosline, 1989), A1-OST (Diogo & Chardon, 2000a) and A2α (Wu & Shen, 2004). Vetter’s A3-section most probably corresponds to the ‘internal division’ of Gosline (1989), although the latter author incompletely described the inner divisions in the basal catfish Diplomystes, compared to the work of Diogo & Chardon (2000a) and Diogo (2005). Wu & Shen (2004) synonymized their A2β to the A2 of Diogo & Chardon (2000a) and the internal section of Gosline (1989). Alexander (1965) erroneously presumed the presence of an A1-section in the adductor mandibulae part that inserts on the lower jaw in both callichthyids and loricariids; his erroneous terminology was followed by Howes (1983a), Schaefer & Lauder (1986) and Schaefer (1997). Of the various recent terminologies, Wu & Shen’s (2004) uses derivatives of the terms A2 and A3 pointing to the various subdivisions of the adductor mandibulae complex, and avoids any reference to the term A1. An A1 section as observed in acantomorphs is absent in siluriforms (Adriaens & Verraes, 1996; Diogo & Chardon, 2000a; Diogo, 2005). Still, it is more appropriate to follow the terminology of Diogo & Chardon (2000a), Diogo (2005) and Diogo et al. (2006), given the thoroughness of their comparative work on siluriforms. Note that their A2 is not homologous to the A2 of Vetter (1878). Given the limited systematic range of taxa studied by me, I am not in a position to condemn or reject any of these proposed nomenclatures.

In Diplomystidae, the most basal catfish family, Diogo & Chardon (2000a) discern a large, lateral A1-OST with the A2 lying mediodorsal to it. The former muscle inserts on the coronomeckelian and angulo-articular, the latter on the coronomeckelian alone. Concerning the deeper adductor parts Diogo & Chardon (2000a) and Diogo (2005) provide more detail than Gosline (1989). They discriminate a dorsal and a ventral part of the A3’ that lie medioventral to the other divisions and insert on the posterior part of the coronomeckelian and the angulo-articular, respectively. Diplomystes sp. also possesses an Aω stretching between the A2 tendon and the inner face of the dentary. Although several bundles are present, the adductor mandibulae complex of Diplomystidae is relatively undifferentiated at the level of its anterior insertions, and it lacks a direct connection to the primordial ligament PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 95

(Arratia, 1987; Gosline, 1989; Diogo & Chardon, 2000a). The course of the mandibular branch of the trigeminal nerve is lateral to all divisions (Gosline, 1989). Within the loricarioid lineage, Nematogenyidae have been studied by Howes (1983b), Diogo (2005), and Diogo et al. (2006). The lateral A1-OST and A2 are present, as well as an A3”-disivion, situated medially to the uppermost fibres of the A2. The medial A3’-section consists of a dorsal and a ventral part, as in Diplomystidae. These authors found that Nematogenys inermis lacks a retractor tentaculi. While Diogo (2005) states that there is no relation between the A3” and the primordial ligament, Howes (1983b:12) mentions an anterior tissue sheet connected to the upper, inner division (A3”): “Stemming from the antero-medial surface of this muscle, and extending across the dorsal surface of the mandible is a thick sheet of connective tissue. This sheet bifurcates, the upper strand attaching to the posterior face of the maxilla, the lower to the distal portion of the maxilla where it forms a sheath around the maxillary barbel.” In the adductor mandibulae complex of Trichomycterus rivulatus (Trichomycteridae), Howes (1983a) mentions the presence of an outer portion inserting on the lower jaw, and a tendon connecting (part of?) it to connective tissue running to the maxilla. An additional medial section (A3’?) inserts onto the inner aspects of the dentary. The precise identification of these muscles remains unclear. The association between any of the adductor mandibulae bundles and the maxilla or the primordial ligament is contradicted by Diogo (2005), who investigated some other Trichomycterus species and Hatcheria macraei. A retractor tentaculi muscle was not found in the trichomycterids examined by Schaefer & Lauder (1986).

In the callichthyid Corydoras aeneus an external and an internal adductor mandibulae part can be distinguished. Based on its ventrolateral position and insertion on the outer portion of the lower jaw, the external adductor mandibulae part can be considered homologous to the A1-OST. Based on its ventromedial position and insertion on the posteromedial face of the lower jaw, the internal adductor mandibulae part can most probably be identified as the A3’, as it takes the place of the A3’ of Diplomystes sp. and Nematogenys inermis, and the additional medial section of Trichomycterus rivulatus. In C. aeneus, the A3’ is not divided in a dorsal and a ventral part. The ontogeny of the adductor mandibulae complex suggests a double origin of the retractor tentaculi in Corydoras aeneus: the dorsolateral division is partially confluent with the external adductor mandibulae part or A1-OST which inserts on the lower jaw. I identify this division as the A2 of Diogo (2005). The dorsomedial division, originating medially and anterior to the other division, then corresponds to the A3” of Nematogenyidae. Diogo (2005) also stated that the callichthyid retractor tentaculi results from the A3”, without, however, an 96 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE involvement of the A2. The mandibular branch of the trigeminal nerve, running between the A2 and the A3”, is discussed below. Some authors have hypothesized an A1 origin of the siluriform retractor tentaculi (Takahasi, 1925; Edgeworth, 1935). The most supported idea, however, states that the retractor tentaculi originates from an inner division of the adductor mandibulae complex (McMurrich, 1884; Lubosch, 1938; Eaton, 1948; Alexander, 1965; Howes, 1983b; Adriaens & Verraes, 1996, 1997a; Diogo, 2005). A retractor tentaculi muscle is present in several siluriform families like Clariidae, Ictaluridae, Cranoglanididae and Bagridae, and is believed to originate from the A3” (Nawar, 1955a; Adriaens & Verraes, 1996; Diogo, 2005). Both Howes (1983b) and Diogo (2005) support the idea that the muscle has been derived independently in several siluriform lineages, via a connection with the primordial ligament. The absence of a retractor tentaculi in the basal loricarioid families Nematogenyidae and Trichomycteridae, as well as in Diplomystidae, suggests that the retractor tentaculi has independently evolved within the loricarioid lineage. Clarias gariepinus is the only siluriform of which the ontogeny of the musculature has been studied so far (e.g., Adriaens & Verraes, 1996). A re-evaluation of their data reveals that their retractor tentaculi might not be completely homologous to the retractor tentaculi of Corydoras aeneus (i.e., having originated from more ventral fibres of the complex). In both species, the retractor tentaculi is a medial derivative of the adductor mandibulae complex that might have evolved independently (see above). The ontogenetic evidence on C. gariepinus suggests that the composition of the adductor mandibulae complex might well correspond to the general siluriform pattern of Diogo & Chardon (2000a) and Diogo (2005), which is corroborated by the results of the current dissertation. A lateral A1-OST (A2A3’β of Adriaens

& Verraes, 1996) and A2 (A2A3’ α) can be distinguished. With respect to the medial adductor part, a comparison with the configuration in Diplomystes sp., Nematogenys inermis and

Ancistrus cf. triradiatus suggest that their superficial and deeper A3” parts are homologous to the dorsal and ventral part of the A3’ of the former species (also in C. gariepinus, one part lies more dorsal to the other). As the more medial retractor tentaculi during ontogeny shows no affinities with the A2, but with the A3’, I hypothesize that it must be derived from this A3’ (and thus is not homologous to the A3” of loricarioids). Thus, the retractor tentaculi might have different evolutionary origins in siluriforms. In Callichthyidae, the retractor tentaculi provides a biomechanical coupling between the maxilla, the suspensorium and the lower jaw via a tendinous insertion on the primordial ligament (Howes, 1983a; Schaefer & Lauder, 1986). Moreover, a novel connection has also been established between the primordial ligament and the premaxilla. This bone has become highly mobile as the tight connection to the neurocranium has been loosened: a functional PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 97 coupling between the retractor tentaculi and the premaxilla has thus been created (Schaefer & Lauder, 1986).

The portion of the loricariid adductor mandibulae complex that effectively inserts on the lower jaw, as observed in Ancistrus cf. triradiatus, consists of two divisions. In order to compare these two divisions to divisions in Corydoras aeneus and other siluriforms, the fact that the loricariid lower jaw has rotated medially has to be taken into account when considering differences in jaw muscle topographies. Both loricariid lower jaws point towards each other. As a spatial consequence of this rotation, the external adductor mandibulae part would come to lie dorsal to the internal adductor part, in order to maintain their insertions on the lower jaw. I hypothesize the homology of the external adductor mandibulae muscle in Ancistrus cf. triradiatus and Corydoras aeneus, and, hence, the A1-OST. This hypothesis is supported by the similar anterior and posterior insertions, i.e., the hyomandibula and preopercle, and the dorsolateral aspect of the angulo-articular and the lateral face of the dentary coronoid process. Also, the muscle lies well lateral to the mandibular nerve branch in both A. cf. triradiatus and C. aeneus. The internal part of the adductor mandibulae of A. cf. triradiatus and C. aeneus is considered homologous with the A3’ of Diplomystes sp. and the other above-mentioned siluriforms, based on the origin on the suspensorium and the insertion on the dorsomedial aspect of the lower jaw, including the Meckelian fossa. Additionally, in both species the mandibular nerve branch lies on the anterior tendon of this internal adductor part before passing down the Meckelian fossa. The dorsal and ventral portions of the internal part, observed in the 12.4 mm and all larger specimens of A. cf. triradiatus (Fig. 43A), might correspond to the dorsal and ventral parts of the A3’ in Diplomystes sp. (Diogo & Chardon, 2000a; Diogo, 2005) and Nematogenys inermis (Diogo et al., 2006). The loricariid retractor veli has been hypothesized to be a medial derivative of the adductor mandibulae complex (Howes, 1983a; Schaefer & Lauder, 1986). The muscle was first described as the ‘muscle of oral valve’ by Gradwell (1971b). It was called retractor palatini by Howes (1983a), and is featured as such in the papers of Schaefer & Lauder (1996), Schaefer (1997), Diogo & Vandewalle (2003b) and Diogo (2005). A non- homologous retractor palatini was described earlier in balistoids, as an anterior portion of the adductor arcus palatini, inserting on the autopalatine (Lubosch, 1929; Hofer, 1938). Based on this, and on functional considerations (i.e., the fact that it retracts the oral valve and not the autopalatine; see also Gradwell, 1971b), I propose the new name retractor veli for this loricariid neoformation (see also Chapter 6.1). This retractor veli even differentiates into two separate muscles in Otocinclus vestitus (Chapter 6.3). Considering that: (1) the muscle is absent in Callichthyidae and other non-loricariid loricarioids; (2) the innervating nerve twig 98 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE separates from the branch innervating the internal adductor mandibulae part (A3’), and (3) the latter muscle and the retractor veli are continuous for most of their length during early ontogeny of Ancistrus cf. triradiatus, it seems most plausible it is a medial derivative of this adductor part. I hypothetize the homology of the loricariid retractor premaxillae (Ancistrus cf. triradiatus) with the medial half (A3”) of, or even the whole (A2+A3”), callichthyid retractor tentaculi (Corydoras aeneus). Several arguments support this hypothesis. (1) Both muscles are connected to the premaxilla. The callichthyid retractor tentaculi has a connection to the premaxilla via a dorsal extension of the primordial ligament (Schaefer & Lauder, 1986; pers. ob.). The loricariid retractor premaxillae is directly attached to the premaxilla. (2) The anterior tendon continues as an aponeurosis in the ventral to ventromedial aspect of the retractor tentaculi or premaxillae muscle in C. aeneus and A. cf. triradiatus, respectively. (3) In both species the muscle arises as the dorsalmost part of the adductor mandibulae complex and is situated lateral to the pterygoid process of the cartilaginous suspensorium, reaching beyond the coronoid process of Meckel’s cartilage. In early ontogeny the lateral fibres of the muscle originate on the hyosymplectic and are continuous with those of the external adductor mandibulae part; the medial fibres originate anteromedially to the other parts on the hyosymplectic. (4) Innervation of both muscles is similar: the first twig to separate from the branch that innervates the whole adductor mandibulae complex, is the one serving the retractor tentaculi (C. aeneus) or the retractor premaxillae (A. cf. triradiatus). The main branch innervating the complex originates from the mandibular branch, and is present in other teleosts as well: siluriform examples are given by Atoda (1936), Mithel (1964a), Winterbottom (1974) and Adriaens & Verraes (1996). (5) The last argument relates to the relative position of the retractor tentaculi or premaxillae, and the buccomaxillary nerve branch. In embryonic and adult C. aeneus, and in embryonic A. cf. triradiatus, the position of this anastomosed branch is dorsomedial to the muscle. In adult A. cf. triradiatus this nerve cord migrates laterally, along with the levator tentaculi (see below).

Opposed to the buccomaxillary nerve branch, that invariably runs outside the adductor mandibulae complex, I do not want to overestimate the use of the course of the mandibular branch of the trigeminal nerve to prove homologies of adductor muscle parts. The nerve crosses or enters the adductor mandibulae complex differently in different taxa. It has proved its usefulness in several studies (e.g., Wu & Shen, 2004; Diogo, 2005), but sometimes relative positions are too variable to be reliable (Edgeworth, 1935; Mithel, 1964a; Gosline, 1989). The paths of nerves can be modified according to topographical changes of the innervated muscle bundles. This might affect the validity of the use of the course of the mandibular nerve branch for homology assessments within the adductor mandibulae complex PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 99

(Nakae & Sasaki, 2004). As mentioned above, while the mandibular nerve branch in Ancistrus cf. triradiatus passes lateral to the retractor premaxillae in some cases, it runs through the muscle in others (e.g., Fig. 42D, E). Even individual left/right variation is observed. In all examined specimens of Corydoras aeneus the mandibular nerve branch runs between both halves of the retractor tentaculi. Thus, it is impossible to use the mandibular nerve branch to unambiguously identify adductor mandibulae sections in Ancistrus cf. triradiatus. Also the posterior position of the levator arcus palatini in A. cf. triradiatus, and its absence in several other loricariids (Diogo, 2005) inhibits any sure identification. The position of the insertion of this suspensorial muscle has been used to discrimate between adductor parts in other siluriforms (Takahasi, 1925; Adriaens & Verraes, 1996). In A. cf. triradiatus, the retractor premaxillae thus consists of the A2+A3”, or, perhaps, of the A3” alone. The presence of a retractor premaxillae and an (indirectly) antagonistic muscle originated from the medial fibres of the extensor tentaculi in loricariids (see Alexander, 1965; Chapter 6.1) can be linked to a novel protrusion mechanism that evolved in the loricarioid lineage. The antagonistic muscle acts on the premaxilla via the autopalatine [extensor tentaculi pars medialis (Alexander, 1965; Howes, 1983a; Chapter 6.1)]. An ascending process on the premaxilla is typical for those teleosts with a protractile upper jaw (Eaton, 1935). Such a process is present on the premaxilla in Loricariidae, though absent in most siluriforms. The protrusion mechanism of loriicarids is, however, different from those mechanisms described by Motta (1984), and even largely decoupled from the movements of the maxilla (Alexander, 1965; Chapter 6.1). In siluriforms, an Aω-section can be present (e.g., Takahasi, 1925; Gosline, 1989; Wu & Shen, 2004; Diogo, 2005) or absent (e.g., Gosline, 1989; Adriaens & Verraes, 1996; Diogo, 2005). Based on the definitions given by Vetter (1878) and Winterbottom (1974) it can be stated that there is no sign of an Aω in early or adult stages of both Corydoras aeneus and Ancistrus cf. triradiatus.

Within the loricarioid clade, the myology of the two remaining families, Astroblepidae and Scoloplacidae, is less known. As a consequence, only a tentative and incomplete comparison of their jaw musculature can be made here. No detailed information exists about the nature of the A1-OST and A3’ of the adductor mandibulae portion inserting on the lower jaw in astroblepids. A retractor premaxillae is present (Schaefer & Lauder, 1996), with the lower fibres not entirely separated from the outer adductor complex (Howes, 1983a), indicating a composition possibly homologous to A2+A3” (cf. the loricariid retractor premaxillae and the callichthyid retractor tentaculi). Howes’ (1983a) statement that fibres of the retractor premaxillae would originate on the 100 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE lateral ethmoid seems inconsistent with his accompanying figure, and is opposed by other authors, who describe a suspensorial origin (Diogo, 2005; S. Schaefer, pers. commun.). Little is known of the scoloplacid cranial musculature. Schaefer (1990) described an adductor part inserting on the lower jaw (without reference to any possible subdivisions) and a retractor premaxillae with a long tendon inserting on the premaxilla. Defining homologies with A2- and A3-divisions of other loricarioids requires a more detailed examination of scoloplacid musculature. According to Diogo (2005), the origin of the retractor premaxillae is on the suspensorium. Both authors describe a ‘retractor tentaculi’ originating on the lateral ethmoid; this muscle will be discussed in the next section.

THE EXTENSOR TENTACULI AND LEVATOR TENTACULI

My proposed homology of the callichthyid retractor tentaculi and the loricariid retractor premaxillae implies a non-homology of the callichthyid retractor tentaculi and the loricariid ‘retractor tentaculi’ of Howes (1983a), Schaefer & Lauder (1986, 1996), Schaefer (1997), Diogo & Vandewalle (2003b), identical to the ‘adductor tentaculi’ of Alexander (1965) and ‘muscle a’ of Howes (1983a). Hence I propose the name levator tentaculi for this muscle, a name that, to my kwowing, is not taken for any other teleost muscle (Winterbottom, 1974; Diogo & Vandewalle, 2003b). Additionally, it aptly describes the motion of the maxilla and associated lateral lip tissue, which is elevated from the substrate (to which loricariids can attach with their sucker-like mouth). In the following discussion I will provide additional evidence for this hypothesis of non-homology. The extensor tentaculi is generally present in siluriforms, and can be a single or a subdivided muscle (Diogo & Vandewalle, 2003b). It is a single muscle in Diplomystidae and Nematogenyidae, with some fibres mixed with those of the adductor arcus palatini (Diogo, 2005), of which it is an anterior derivative (Takahasi, 1925; Alexander, 1965; Winterbottom, 1974; Gosline, 1975; Adriaens & Verraes, 1997a). Origin is on the neurocranium, usually (at least) on the ventral aspect of the lateral ethmoid, as in Trichomycteridae, Callichthyidae and Astroblepidae (Howes, 1983a; Schaefer & Lauder, 1986; pers. ob.). Whereaes the whole of the adductor mandibulae complex originates from the dorsal, masticatory part of the mandibular muscle plate and is innervated by a twig of the mandibular branch of the trigeminal nerve, the adductor arcus palatini and the extensor tentaculi are derived from the constrictor hyoideus dorsalis muscle plate, and are innervated by the palatine branch of the hyomandibular trunk of the facial nerve (Mithel, 1964a; Alexander, 1965; Singh, 1967; Jarvik, 1980). Innervation of the levator tentaculi of Ancistrus and other examined loricariids is discussed below. In Callichthyidae, Scoloplacidae and Astroblepidae Diogo (2005) noted the differentiation of the extensor tentaculi into two elongated ventral bundles attaching to the postero- PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 101 ventrolateral surface of the autopalatine, and an additional dorsal bundle essentially oriented dorsoventrally and attaching to the posterodorsal surface of this bone. Only one instead of two ventral bundles were found in Corydoras aeneus (Fig. 44D). The complete separation of the ventral bundle of the extensor tentaculi in two separate muscles, i.e., a pars medialis and a pars lateralis, is a character observed in loricariids only (Howes, 1983a; Diogo, 2005; this dissertation). Here, both parts, which appear to be separated from early ontogeny on, are innervated by twigs of the palatine branch of the hyomandibular trunk of the facial nerve.

The following is a summary of evidence from the above results, to support the non- homology of the loricariid levator tentaculi and any part of the adductor mandibulae complex, as well as the close affinity of the former muscle to the extensor tentaculi. First, the dorsolateral part of the extensor tentaculi of Corydoras aeneus originates on the same place as the levator tentaculi of Ancistrus cf. triradiatus, i.e., far laterally on the lateral ethmoid (Fig. 45B), or the orbitonasal lamina in early stages (Fig. 44D). Second, during early ontogeny the levator tentaculi of Ancistrus cf. triradiatus develops separately from the adductor mandibulae complex, but adjacent to the extensor tentaculi pars lateralis. This is clearly visible on figures 39, 40A-D, 41 and 42A. Third, the lateralmost fibres of the extensor tentaculi of Corydoras aeneus insert on the lateral aspect of the autopalatine (e.g. Fig. 44A), with part of the tendon even extending a bit further anteriorly. A shift of these muscle fibres via the connective tissue connecting the autopalatine to the dorsolateral soft palate and the maxilla seems more parsimonous than the shift of the origin of the callichthyid retractor tentaculi from the suspensorium to the lateral ethmoid and a de novo development of the loricariid retractor tentaculi from the adductor mandibulae complex. Fourth, the relative position of the extensor tentaculi (and levator tentaculi) to the anastomosed buccomaxillary nerve branch and the adductor mandibulae complex in Corydoras aeneus and (embryonic) Ancistrus cf. triradiatus is identical. In both species the buccomaxillary nerve branch is situated dorsomedial to the adductor mandibulae complex, but (ventro)lateral to the extensor tentaculi bundles and the levator tentaculi (present in A. cf. triradiatus only) (Figs 40A-D, 41, 42A-C, 43A, 44A-D, 45A-B). This evidence can be overlooked in adult A. cf. triradiatus, as during ontogeny, the levator tentaculi migrates and expands laterally (Figs 40A vs 42C), while the adductor mandibulae complex expands somewhat medially, as a spatial consequence of the more medial jaw position (see also Chapters 4.2 and 6.1). The anastomosed buccomaxillary nerve branch migrates laterally as well, still retaining its relative position to the levator tentaculi. The innervation of the loricariid levator tentaculi remains somewhat puzzling. Based on dissections, Howes (1983a) concluded that innervation occurs by a branch of the maxillary 102 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE branch of the trigeminal nerve, entering the muscle’s ventral surface. Serial sections showed me that a thin nerve branch indeed enters it at its ventrolateral side, immediately splitting in two small twigs innervating the muscle. The very large axons originate from the anastomosed buccomaxillary branch, and can be traced back some distance, to the level where also the mandibular branch forms part of the nerve. (The same innervation was observed in Pterygoplichthys lituratus, Otocinclus vestitus and Farlowella acus.) As the mandibular and buccomaxillary branches are undistinguishably connected here, it is unsure to which of these two branches the axons belong. The infraorbital trunk, and even the roots of the trigeminal (V) and facial (VII) nerves are intimately fused, as in other siluriforms and teleosts (Berkelbach, 1915). Comparison with the paths of the nerves in the cheek region in other siluriforms revealed that, first, separation, anastomosis and branching patterns of the mandibular, maxillary and buccal branches vary within the order, and, second, a possible homologue of the branch innervating the levator tentaculi could not be found in Silurus glanis, Parasilurus asotus, Mystus seenghala, Wallago attu, Bagarius bagarius, Clarias gariepinus and Corydoras aeneus (Juge, 1899; Atoda, 1936; Mithel, 1964a, b; pers. ob.). The fact that the levator tentaculi is not innervated by a the palatine branch of the hyomandibular trunk (as is the extensor tentaculi) does not support, nor oppose my hypothesis. An examination of the innervation pattern in the related Scoloplacidae and Astroblepidae could possibly resolve this question.

Based on the above mentioned argumentation and literature data, I subsequently hypothesize the homology of the ‘retractor tentaculi’ of Scoloplacidae and Astroblepidae (Howes, 1983a; Schaefer, 1990; Diogo, 2005) and the loricariid levator tentaculi. Both are broad muscles, originate on the neurocranium, and cover the retractor premaxillae dorsally. The position of this muscle in Scoloplacidae, as figured by Schaefer (1990:200) infers that it might well be an intermediate configuration between the dorsolateral extensor tentaculi division of Callichthyidae and the levator tentaculi of Loricariidae (‘retractor tentaculi’ of Astroblepidae). The shape of the muscle is an additional argument, as the muscle has extended only slightly beyond the posterior end of the autopalatine, and a long tendon crosses the retractor premaxillae dorsally and connects to the maxilla (Schaefer, 1990). In loricariids, the levator tentaculi might have slightly different orientations, and is sometimes more separated from the extensor tentaculi than in Ancistrus cf. triradiatus (Otocinclus vestitus, Hypoptopoma sp., Farlowella sp.) (Howes, 1983a; Chapters 6.1 and 6.2). In these cases the lateral ethmoid, and thus the origin of the muscle, have extended far laterally (O. vestitus, Hypoptopoma sp.), or the origin of the muscle has shifted rostrally (Farlowella sp.). PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 103

If the levator tentaculi has indeed originated from dorsolateral fibres of the extensor tentaculi, one could speculate on evolutionary key innovations that could have led towards the loricariid muscle configuration. A hypothetical ancestral configuration would probably have included a retractor tentaculi as in Corydoras aeneus, retracting the barbel, and perhaps already having an indirect effect on the premaxilla. The extensor tentaculi would have been broad, as in C. aeneus, but with lateral fibres inserting even more on the anterior part of the autopalatine and the tissue surrounding the autopalatine-maxillary articulation. Or contact with the maxilla could have been established by means of a long tendon, as in scoloplacids (see above). The tissue that connects the autopalatine to the dorsolateral soft palate in C. aeneus, and that is continuous with the autopalatine-maxillary articular tissue, could have generated such tendon. Without a detailed microscopic-histological examination of, especially, scoloplacid and astroblepid representatives, such hypotheses remain speculative. They are, however, justified, as the homology of the loricariid retractor premaxillae and the callichthyid retractor tentaculi implies an alternative, novel origin of the levator tentaculi within the loricarioids.

104 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

5.2. ONTOGENY OF THE INTERMANDIBULAR AND ∗ HYOID MUSCULATURE

Abstract — Loricariidae or suckermouth armoured catfishes are the most speciose catfish family, displaying morphological specializations towards the attachment onto substrates with their suckermouth, and the scraping of algae and other food items off these substrates. The intermandibular and hyoid musculature differs from the general siluriform situation. This detailed study on several developmental stages of a loricariid representative aims to provide insight in the ontogenetic origin of these muscles, as well as on their morphology and homology. Serial sections and 3D-reconstructions are used to visualize the early muscle configurations. The intermandibularis anterior muscle develops two parts, inserting on the lower jaw but also on the lower lip tissue. A similar differentiation into a dentary and a labial part occurs in the intermandibularis posterior (usually erroneously referred to as protractor hyoidei in loricariids). The protractor hyoidei has a compound nature in teleosts, but in loricariids no interhyoideus portion fuses to the posterior intermandibularis portion. Several arguments, including the absence of a myocomma and a double innervation, indicate the absence of an interhyoideus portion. A double innervation has been found in the hyohyoideus inferior. The posteriormost muscles in the hyoid region are relatively small during early ontogeny: the sternohyoideus halves fuse relatively late; the hyohyoidei adductores develop latest of all ventral head muscles. A remarkable shift in orientation characterizes the hyohyoideus abductor.

5.2.1. INTRODUCTION

The present chapter is part of the study on the ontogeny of hard and soft cranial structures of a highly specialized teleost taxon: the neotropical family Loricariidae or suckermouth armoured catfishes. These catfishes are special in having their lower lip folded back against the ventral side of the head, and the lower jaws rotated medially and ventrally, so that the teeth are pointed towards the substrate on which they feed (mostly algae and other encrusted matter and detritus) (Alexander, 1965; Schaefer & Lauder, 1986). These and other authors (e.g., Howes, 1983a) also elaborated on the high number of cranial muscles serving the highly mobile jaw elements. The specializations of the loricariid head, related to this feeding mode, but also their ability to suck onto substrates and maintain position even in torrential rivers, certainly have to involve modifications in the musculature of the ventral head region. It has already been shown that the so-called protractor hyoidei differs from the general catfish morphology, as it is connected to the modified lower jaw as well as the lower lip tissue (Schaefer & Lauder, 1986; see also Part 6). Differences in other hyoid muscles have been observed as well (e.g., a forked hyohyoideus inferior, and a transversely oriented

∗ Slightly modified from: Geerinckx T. & Adriaens D. Ontogeny of the intermandibular and hyoid musculature in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Submitted to Animal Biology. PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 105 hyohyoideus abductor connected to the sternohyoideus) (Howes, 1983a; Schaefer, 1997; Diogo, 2005; see also Part 6). The evolutionary origin and transformations of some of the loricariid cranial musculature have been hypothesized (Schaefer & Lauder, 1986, 1996; Chapter 5.1), while the ontogenetic transformations have not yet been studied, except those concerning the jaw and maxillary barbel muscles (Chapter 5.1). Studies on the ontogeny of cranial musculature of other taxa have, however, yielded important results concerning muscle homologies, muscle functions during early life history stages, and changes in muscle morphology and function (Otten, 1982; 1983; Surlemont et al., 1989; Surlemont & Vandewalle, 1991; Adriaens & Verraes, 1996, 1997a, d, e; Hunt von Herbing et al., 1996a, b; Schilling & Kimmel, 1997; Hernández et al., 2002). Here I present a detailed study of the muscles on the ventral aspect of the head of the bristlemouth catfish Ancistrus cf. triradiatus Eigenmann, a representative of the Loricariidae. The muscles found in this region are various parts of the intermandibular muscle, several hyoid muscles and the sternohyoideus. Developmentally, these muscles arise from three different muscle plates (Edgeworth, 1935; Miyake et al., 1992): the intermandibularis anterior and posterior arise from the ventral portion of the mandibular muscle plate, the hyohyoideus inferior, abductor and adductores belong to the ventral part of the hyoid muscle plate, and the sternohyoideus originates from the hypobranchial muscle plate. The protractor hyoidei, present in most teleosts, is composed of the intermandibularis posterior and an anterior division of the hyoid muscle plate, the interhyoideus. The muscles of the branchial arches are not dealt with in this dissertation. My major objectives are a detailed analysis of the ontogeny of the ventral head muscles, as well as providing hypotheses on the identity and homology of the intermandibularis and ‘protractor hyoidei’ muscle divisions, by comparing them with the same muscles or muscle divisions in non-loricariid catfishes. To do this, I examined not only the muscles and their insertions, but also the paths of and innervations by the relevant nerve branches. Finally, I add a few brief considerations on the functional-morphological aspects of some muscles. My discussion on the functionality of muscles is based on the anatomy only, as no biomechanical studies including EMG could be done on such small specimens. An excessive functional interpretation of the results is not appropriate here, not only because of this reason, but also because a comprehensive analysis should include the jaw, suspensorial and opercular musculature as well.

106 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

5.2.2. BRIEF MATERIAL AND METHODS

2 µm thick serial sections have been made of embryos of Ancistrus cf. triradiatus (see Table I and paragraph 2.2.7). 5 µm thick serial sections of a subadult were studied as well. See paragraph 2.2.8 for details on the 3D-reconstructions of the 6.1 and 8.0 mm SL A. cf. triradiatus specimens. 2 µm serial sections of 5.6 and 7.2 mm specimens of the clariid Clarias gariepinus from Adriaens and Verraes (1997d) were examined for a short comparison of the intermandibularis posterior innervation (see discussion).

5.2.3. RESULTS

6.1 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 46)

In this early embryonic stage, with a still incomplete chondrocranium, most muscles of the ventral head region are already present. The intermandibularis anterior is a broad transverse muscle sheet that broadens at both lateral ends. The caudoventral fibres, representing the pars labialis, run almost straight (in a transverse plane), ventral to the anterior margin of the hyoid bar. No insertion is observed on the lower lip tissue, which lies ventral to the muscle. The rostrodorsal fibres diverge somewhat rostrally on both sides, running in the direction of Meckel’s cartilage, without, however, reaching it. These fibres form the pars dentalis of the intermandibularis anterior (Fig. 46A). The intermandibularis posterior is a paired muscle, originating at the ventral face of the lateral part of the hyoid bar. It probably already attaches to the cartilage. Anteriorly, two parts can be distinguished that can’t be separated near the posterior origin. The intermandibularis posterior pars dentalis is a dorsal group of fibres that runs mediorostrally, and ends halfway between the hyoid bar and Meckel’s cartilage, lateral to the pars dentalis of the intermandibularis anterior. Another part of the muscle, corresponding to the intermandibularis posterior pars labialis, runs more ventrally, in the direction of the lower lip. Its anteriormost end almost meets the lateral end of the intermandibularis anterior pars labialis. There is no doubt, however, that there is no contact between fibres of both muscles. The hyohyoideus inferior is a single, broad muscle plate with a slightly V-shaped appearance. The fibres originate at the ventral side of the hyoid bar just medial to the intermandibularis posterior, and run posteromedially until both halves unite ventral to the branchial region. At this moment only the anterior copula of the branchial basket has developed, which is continuous with the hyoid bar. The sternohyoideus has arisen as a small, paired muscle, that lacks any insertion at this moment. The muscle stretches from the level of the anterior copula (lying ventrolateral to it) PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 107 almost to the level where the cleithrum of the pectoral girdle will develop. The paired muscle is still very small, and round in transsection. The hyohyoideus superior is an equally rudimentary muscle. The fibres start just lateral to the sternohyoideus, halfway the length of the latter muscle. They then run obliquely in the direction of the branchiostegal membrane. This membrane is still only partly formed, and only the lateralmost branchiostegal ray is developing at this moment (not shown on fig. 46). The muscle has not yet reached this branchiostegal ray. It has not yet divided into abductor and adductores parts.

8.0 MM SL — 7 DAYS POST-FERTILIZATION (FIG. 47)

At this moment the chondrocranium is more or less complete, and several bony elements are already present, e.g., the premaxillary and dentary parts supporting the teeth (Part 4). Both parts of the intermandibularis anterior are now well discernable, as the muscle is now clearly forked, and both parts contact each other only at the midline. The pars dentalis is becoming more C-shaped, as both lateral ends grow anteriorly. The absence of most of the dentary bone suggests this muscle is not fully functional at this moment. The pars labialis still is a transverse sheet, somewhat narrower medially, and lying caudoventral to the pars dentalis. It contacts the lower lip tissue not only at both lateral ends, but along the whole of its length. The intermandibularis posterior now originates on the ventral face of the cartilaginous ceratohyal. Both muscle parts have grown anteriorly. The pars dentalis has arrived at the level of the dentary, without any clear sign of insertion though. The ventral and more flattened pars labialis, still confluent with the pars dentalis caudally, now reaches to the lateral end of the intermandibularis anterior pars labialis. Neither in this stage, nor in any older stage, fibres have been found that are continuous between both muscles. The intermandibularis posterior pars labialis thus ends exactly at the place where the lower lip is folded backwards (the lower lip is folded back posteroventrally and forms the posterior half of the sucker that surrounds the loricariid mouth). From this stage on it is clear that the fibre diameter is significantly larger in the pars dentalis than in the pars labialis of both the intermandibulares anterior and posterior muscles. Since the 6.1 mm stage the ceratohyal has become notably broader laterally. Coupled to this, the hyohyoideus inferior has extended its insertion posteriorly, and has thus also broadened laterally. It is now the largest cranial muscle, being more voluminous even than the adductor mandibulae. Considerable growth has occurred in the sternohyoideus. Both halves still remain separate. Each half, however, has expanded substantially: anteriorly, insertion is on the dorsal surface of each urohyal half of the developing parurohyal; posteriorly, the broad muscle inserts 108 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE somewhat dorsally on the anterior edge of the cleithrum. As yet, there is no articulation between the urohyal halves and the hyoid bar, but a connection of the urohyal to the anterior copula ensures an indirect connection between the sternohyoideus and the hyoid bar (Chapter 4.2). Both the hyohyoideus inferior and the sternohyoideus could well be functional by now. All four branchiostegal rays have now developed, and the hyohyoideus superior now consists of two distinct parts, the medial hyohyoideus abductor and the more lateral hyohyoidei adductores. The hyohyoideus abductor reaches more or less up to the dorsal aspect of the first, medialmost ray, which is just appearing in this stage (Fig. 47A). The muscle still runs somewhat anteromedially, but the insertion on the ventral fascia of the sternohyoideus has shifted to the posterior part of the latter muscle (probably due to allometric growth of this muscle instead of to a real migration of the hyohyoideus abductor tendon). The hyohyoideus abductor is the only hyoid muscle that has a substantial tendon (Fig. 47B). Other muscle insertions, including the posterior hyohyoideus abductor insertion, are all primarily musculous. The hyohyoidei adductores have differentiated between the 6.1 and 7.0 mm stages, as shown by serial sections of these specimens. The plural name reflects the fact that several short muscle bands, running from one branchiostegal ray to another, together constitute the adductor of the branchiostegal membrane. Two of these parts are already observed in this stage: a first muscle, stretching between the opercle and the lateralmost, fourth branchiostegal ray, and a second, still smaller muscle, connecting the fourth to the third ray. The third, medialmost part, connecting the third to the second ray, has not yet appeared.

12.4 MM SL — 43 DAYS POST-FERTILIZATION (FIG. 48)

By the moment most of the osteocranium elements are present, muscle differentiation has more or less reached its completion. Only minor changes in relative size or orientation of intermandibular or hyoid muscles are observed between this and older stages. The teeth-bearing part of the dentary has the form of an oval basket, suspended to a lateral handle including the angulo-articular, part of the dentary, and the mentomeckelium that has fused dorsally to the dentary. The intermandibularis anterior pars dentalis inserts in a shallow fossa on the lateral aspect of the dentary, anteroventral to the fusion of the dentary to the angulo-articular. As such, the curved pars dentalis surrounds the teeth-bearing baskets of both lower jaws caudally. The pars labialis of the intermandibularis anterior still contacts the pars dentalis medially, but can be considered as a separate functional unit. Both parts of the intermandibularis posterior have completely separated: posteriorly some of the fibres touch, but they are not continuous. The pars dentalis is a compact muscle, originating posterolateral to the pars labialis on the posterior ceratohyal, and inserting on the lateral aspect of the dentary. The corresponding fossa in this bone is minute, lying just PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 109 anterior to that of the pars dentalis of the intermandibularis anterior. The pars labialis of the intermandibularis posterior is a thin muscle, and has formed several separate small bundles that look like diverging threads, all inserting on the lower lip (Fig. 48A). The hyohyoideus inferior expands as the hyoid (now mostly ossified) becomes broader. Due to its posterior expansion on the caudoventral process of the hyoid, the muscle almost reaches the bases of the branchiostegal rays. The left and right halves of the sternohyoideus have fused medially, now also ensuring a direct mechanical couple between the pectoral girdle and the hyoid: the parurohyal bone is well developed, and two anterior processes are firmly connected to, and articulate with, the hypohyals. The hyohyoideus abductor has come to lie in an almost transverse plane, although not yet as in adult specimens (Fig. 63A). It attaches tendinously to the sternohyoideus medially. The fibres primarily insert on the medial, first branchiostegal ray, and a few tendinous fibres stretch somewhat further, towards the second ray. Except for these few fibres, no muscle, but a short ligamentous band connects the first and second branchiostegal rays, both in this and adult specimens (Fig. 48B). In the 10.2 mm stage, the third part of the hyohyoidei adductores has appeared. In the 12.4 mm and later stages the three parts, connecting the opercle to the fourth ray, the fourth to the third ray, and the third to the second ray, continue to broaden as the branchiostegal rays elongate.

NERVE PATTERNS AND MUSCLE INNERVATIONS

Both parts of the intermandibularis anterior are innervated by the inferior mandibular branch of the trigeminal nerve, that originates from the infraorbital nerve trunk, passes over the lower jaw and runs back somewhat caudally, finally reaching the muscle. This nerve also enters the intermandibularis posterior, innervating both of its parts (Fig. 49A-C). It is the only innervation of these muscles; no twig of the hyoid branch of the facial nerve enters the caudal portion of the intermandibularis posterior. The hyohyoideus inferior receives a branch from the hyoid branch of the facial nerve (Fig. 49D), but also, remarkably, receives a thin branch of the inferior mandibular branch of the trigeminal nerve (Fig. 49C). This latter innervation was unambiguously observed in all examined specimens except the 6.1 mm specimen (where most nerves are still poorly visible). Hyoidei abductor and adductores are innervated by posterior branches of the hyoid branch of the facial nerve (hyomandibular trunk portion). A branch of the occipito-spinal nerve supplies each half of the sternohyoideus.

110 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

5.2.4. DISCUSSION

Ancistrus cf. triradiatus is the second siluriform species of which the ontogeny of the intermandibular and hyoid musculature is described. A few comparisons can be made with the African catfish Clarias gariepinus, that has been the focus of Surlemont et al. (1989), Surlemont & Vandewalle (1991) and Adriaens & Verraes (1997d). In A. cf. triradiatus all ventral head muscles except the hyohyoidei adductores are observed in the 6.1 mm SL specimen. Of these, the sternohyoideus is least developed, as its size relative to the muscle in juvenile and adult specimens is almost negligible. The other muscles are already more substantial, compared to their final sizes. Insertions are observed in the hyohyoideus inferior and intermandibularis posterior (posteriorly). In C. gariepinus, the intermandibularis anterior, protractor hyoidei and hyohyoideus inferior are present in the 4.7 mm TL specimen (Surlemont & Vandewalle, 1991), while in the 5.2 mm TL specimen the sternohyoideus is seen (Surlemont et al., 1989). Only in the 7.2 mm SL specimen the hyohyoidei abductor and adductores are observed (Adriaens & Verraes, 1997d). Notice that Surlemont and collaborators used total length and not standard length to denominate their stages.

INTERMANDIBULARIS ANTERIOR

Both intermandibularis anterior parts, i.e., the pars dentalis and the pars labialis, run transversely in all examined Ancistrus cf. triradiatus stages. They originate as one, forked muscle, and separate during further ontogeny. In adults, they are completely separated, even in the midline (Chapter 6.1). The identification of this muscle as intermandibularis anterior (the anteriormost, transversely oriented part of the ventral intermandibularis muscle complex in teleosts) corresponds to the definition introduced, or applied by Vetter (1878), Edgeworth (1935), Greenwood (1971), Anker (1974), Winterbottom (1974), Miyake et al. (1992), Schilling & Kimmel (1997) and Diogo & Vandewalle (2003). The differentiation of the intermandibularis anterior into dentary and labial bundles has not been described in related loricarioid families; detailed accounts on this muscle in loricariids are lacking (Howes, 1983a; Schaefer & Lauder, 1986; Schaefer, 1990). In the related callichthyids only a dentary part is present (Huysentruyt et al., submitted).

INTERMANDIBULARIS POSTERIOR

The muscle here named intermandibularis posterior is usually termed protractor hyoidei in loricariids and other siluriforms (in loricariids, e.g., Alexander, 1956; Howes, 1983a; Schaefer & Lauder, 1986; 1996; Schaefer, 1997). The term protractor hyoidei (often called geniohyoideus) has been generally used to identify the muscle stretching between the lower PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 111 jaw (dentary) and the lateral part of the hyoid bar [Greenwood (1971) and Winterbottom (1974) argumented why the use of the name geniohyoideus should be avoided in teleosts]. The name protractor hyoidei has been applied first by Holmqvist (1910), and concerns a muscle with an anterior part derived from the intermandibularis posterior and a posterior part derived from an anterior portion of the ventral hyoid muscle plate (interhyoideus; Edgeworth, 1935; Winterbottom, 1974). In zebrafish, the anterior interhyoideus portion of this hyoid muscle plate originates separate from the posterior hyohyoideus portion (Schilling & Kimmel, 1997). Here, and in several other teleosts, the fusion of the intermandibularis posterior and the interhyoideus, connected at the midline, results in an X-shaped protractor hyoidei (Hernández et al., 2002). The compound nature of the protractor hyoidei is always coupled to a double innervation. The anterior segment is innervated by the inferior mandibular nerve branch of the trigeminal nerve (V) [the mandibular branch of this nerve serves the whole mandibular muscle plate (Jarvik, 1980)]. Innervation of the posterior segment is by, at least, the hyoidean branch of the hyomandibular nerve trunk of the facial nerve (VII) (Dietz, 1914, Winterbottom, 1974). This nerve trunk innervates all muscles derived from the hyoid muscle plate (Jarvik, 1980). Usually a transverse myocomma is observed on the line separating both muscle parts (Greenwood, 1971; Winterbottom, 1974). The relative contribution of the interhyoideus part in the protractor hyoidei varies widely, from a very large part in, e.g., the osteoglossomorph Pantodon (Greenwood, 1971), to a small posterior portion in, e.g., the siluriform Ictalurus (Winterbottom, 1974). In Notopteridae and Mormyridae, as well as in some (unspecified) ‘siluroid ostariophysans’ the interhyoideus portion is absent (Greenwood, 1971: 49), so that there is, in fact, no compound protractor hyoidei, but only an intermandibularis posterior. The same morphology is present in Ancistrus cf. triradiatus. Two nomenclatural options remain. First, following the terminology of Greenwood (1971), the usage of the name protractor hyoidei could be banned for those ostariophysans with the muscle lacking an interhyoideus portion and the corresponding hyoidean nerve branch innervation. This, of course, would imply some revisions of taxa in which the muscle has been described without coverage of the innervation. Second, the name protractor hyoidei could be expanded to that muscle connecting the lower jaw and the hyoid arch, and consisting of the intermandibularis posterior and the anterior portion of the interhyoideus, or of the intermandibularis posterior alone. As far as is known, the presence of the latter muscle in the protractor hyoidei appears to be almost universal (but see the remark on Hypophthalmus below). An expansion of the usage of the name protractor hyoidei has been proposed by Winterbottom (1974), to avoid nomenclatural confusion originating from the names for the intermandibular part (intermandibularis posterior or intermandibularis II) (Holmqvist, 1911; Edgeworth, 1928; 1935). 112 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

A revision of descriptions among actinopterygians led me to the conclusion that the first option is preferable. The protractor hyoidei should be used only if evidence of both muscle parts is present. In most taxa that have been thoroughly examined, a compound protractor hyoidei is indeed present, as a transverse myocomma has been observed (e.g., Dutta, 1968; Kampf, 1961; Kirchhoff, 1958; Thiele, 1963), or both innervations have been observed (e.g., Vetter, 1878; Dietz, 1914; Edgeworth, 1935; Datta Munshi & Singh, 1967). In several siluriforms, however, in which the muscle has been described, only the inferior mandibular branch innervation has been found to be present [Juge, 1899 on Silurus glanis and Singh, 1967 on Clarias batrachus, Callichrous pabda (now valid as Ompok pabda), Eutropiichthys vacha, Rita rita and Wallago attu]. Thus in several siluriforms there is no compound protractor hyoidei, but an intermandibularis posterior only. An exceptional, opposite configuration appears to be present in the pimelodid catfish Hypophthalmus edentatus, in which Howes (1983b) only mentions innervation of the protractor hyoidei by branches of the ramus hyoideus VII. Whether the anterior innervation is present, or the muscle consists of the interhyoideus alone, should be verified, as Howes (1983b) didn’t explicitly mention the (aberrant) absence of the inferior mandibular branch innervation. Winterbottom (1974) found a myocomma in Diplomystes and Ictalurus. The situation in the latter species differs from the description by Ghiot et al. (1984), who did not mention this myocomma. Neither of these two authors elaborated on the innervation. The nature of the so- called protractor hyoidei in other siluriforms should therefore be verified in future myological studies. It appears that the presence of an intermandibularis posterior or a protractor hyoidei varies within the siluriform order. The diversity in muscle composition might very well be related to, or even explained by the development of a varying number of mandibular barbels in most catfish taxa (see below). Verification of the serial sections used by Adriaens & Verraes (1997d) revealed that the ‘protractor hyoidei’ in Clarias gariepinus actually lacks the hyoidean branch innervation and the transverse myocomma (pers. ob.). It can thus be concluded that also here, the muscle in fact corresponds to the intermandibularis posterior only. A comparative early embryological study using antibody labelling, as done by Hernández et al. (2002) on zebrafish, carried out on related siluriforms with and without a true protractor hyoidei, would be of most interest. I am unable to state any unambiguous homology between the bundles of the intermandibularis posterior in Ancistrus cf. triradiatus and the pars ventralis, pars dorsalis and pars ventralis of the intermandibularis posterior in Clarias gariepinus (Adriaens & Verraes, 1997d). Fusion of the pars ventralis and the pars lateralis of the latter species gives rise to four different fields of superficial fibres for the manipulation of the mandibular barbels. PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 113

Generally, in non-loricarioid siluriforms there is a secondary subdivision of the intermandibularis posterior/protractor hyoidei, with several bundles serving the bases of the mandibular barbels (Takahasi, 1925; Adriaens & Verraes, 1997d; Diogo & Chardon, 2000). This is not the case in loricarioids (most families of which lack these barbels) (Diogo, 2005). In the loricarioid callichthyids, that possess two pairs of mandibular barbels, these are not served by separate muscles bundles (Huysentruyt et al., submitted). Loricariids thus are among the most aberrant catfish groups: they have an expanded lower lip that is ‘folded back’ posteroventrally, and the taxon must most probably have evolved muscle fibres connecting it to the hyoid arch (the pars dentalis) independently from non-loricarioid catfishes having mandibular barbels and associated muscle bundles. It is highly probable that the pars labialis of the intermandibularis posterior is essential in moving the lower lip and thus the funtion of the suckermouth (see Chapter 6.1). Saxena & Chandy (1966) described a remarkably diverse protractor hyoidei in the cyprinids Garra, Crossocheilus and Psilorhynchus, fishes that also demonstrate a sucker-like mouth (both the double innervation and the myocomma are present in these species).

HYOHYOIDEUS INFERIOR

The posterior portion of the ventral hyoid muscle plate gives rise to the hyohyoideus inferior, hyohyoideus abductor and hyohyoidei adductores in most teleosts (a few taxa have an undifferentiated interhyoideus posterior) (Takahasi, 1925; Greenwood, 1971; Winterbottom, 1974). The hyohyoideus inferior, connecting both sides of the hyoid bar ventrally, is especially well developed in loricariids, and has been found to be somewhat forked laterally (Schaefer, 1997; Howes, 1983a; see also Part 6). Ontogenetically, the posterior part develops somewhat earlier than the anterior part, with insertion on the posterior ventral surface of the cartilaginous hyoid bar, as well as its cartilaginous ventrocaudal process to which the branchiostegal rays attach. The insertion of the anterior part is well rostral to the insertion of the intermandibularis posterior. The unexpected innervation in Ancistrus cf. triradiatus of the anteriormost fibres of the hyohyoideus inferior by a thin branch of the inferior mandibular branch of the trigeminal nerve, clearly shown in the serial sections, is highly unusual, as it is not known from other teleosts (Winterbottom, 1974). One could speculate on possible different contraction patterns, as well as on the homology of these anteriormost fibres. They do, however, originate together with the remainder of the hyohyoideus inferior, and no visible aponeurosis separates them from it.

114 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

HYOHYOIDEUS ABDUCTOR

In Ancistrus cf. triradiatus the fibre direction of the paired hyohyoideus abductor changes during ontogeny. The medial insertion shifts backwards, until the muscle lies completely transverse. It then stretches from the medialmost branchiostegal ray straight towards the midline, not reaching it, but inserting on the ventral fascia of the sternohyoideus, whose fibres run perpendicular to it. This transverse direction of the hyohyoideus abductor is intriguing. In most teleosts its medial insertion is more rostral, on the hyoid bar (Winterbottom, 1974). Thus the muscle direction usually is oblique, and the muscle reaches or crosses the midline anteriorly. It has been stated that the expected function of the muscle, i.e., the expansion of the branchiostegal membrane, is not possible when the fibres run transversely (manipulation of the muscle effectively closes the branchiostegal opening; Chapter 6.1). In teleosts without a distinct hyohyoideus abductor, a caudal shift of fibres of the hyohyoideus inferior (undifferentiated interhyoideus posterior?) has been described, inserting on the proximal region of the branchiostegal rays; also the intermandibularis posterior/protractor hyoidei may reach the rays (Winterbottom, 1974). As such, these muscles can generate expansion of the branchiostegal membrane. It might be possible that the insertion of the posterior fibres of the hyohyoideus inferior on the ventrocaudal process of the hyoid bar can generate such an expansion as well; this process is cartilaginous. Schaefer (1990) observed insertion of these posterior fibres on the bases of the branchiostegal rays in loricariids and some other loricarioids. It has to be mentioned that the loricariid branchiostegal membrane is very small compared to that of most other siluriforms, where the number of branchiostegal rays can be much higher (up to 20) (McAllister, 1968; Adriaens & Verraes, 1998). The long cartilaginous ventrocaudal process of the hyoid bar, observed in loricariids, is not present in other siluriforms, where the rays usually articulate with the ceratohyals directly (Arratia, 1987).

HYOHYOIDEI ADDUCTORES

The three strands of hyohyoidei adductores on each side arise in the same sequence as the branchiostegal rays: the medialmost strand (between the second and third ray) in Ancistrus cf. triradiatus develops latest. The strands are not continuous in any developmental stage nor in adults, as is sometimes the case in teleosts (Winterbottom, 1974). While both the abductor hyohyoideus and hyohyoidei adductores originate from the hyohyoideus superior, the muscle observed in the 6.1 mm specimen could well represent the abductor alone, as its position corresponds to the medial portion of the abductor (while the adductores will develop far more laterally). Also, the adductores strands develop from lateral to medial, suggesting that the PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 115 separation into the abductor and adductores parts might occur before the muscles can be recognized on toluidine blue stained serial sections.

STERNOHYOIDEUS

During development of Ancistrus cf. triradiatus, the sternohyoideus grows from a pair of narrow muscles to one broad muscle mass stretching from the dorsal aspect of the parurohyal to the edge of the whole horizontal limbs of the cleithra. This muscle is rather conserved among teleosts (Winterbottom, 1974). Its development is more or less identical to that of the muscle in Clarias gariepinus (Adriaens & Verraes, 1997d), and its relation to the sesamoid, ‘urohyal’ part of the parurohyal bone corroborates the thesis that this part develops as an ossification of the paired but fusing anterior tendons of the sternohyoideus muscle (Arratia & Schultze, 1990).

Finally, some brief, functional considerations may be added to this discussion. One could hypothesize that the contribution of the pressure pump, driven primarily by the hyoid movements, is relatively larger in Ancistrus cf. triradiatus than in other siluriforms (and some other bottom-living teleosts), where the suction pump is more important (Hughes, 1970; Adriaens & Verraes, 1997d). The relative importance of the suction pump system is correlated to the size and mobility of the opercle and branchiostegal membrane. The latter membrane of A. cf. triradiatus is exceptionally small for a siluriform (see above). Its relative size also substantially decreases during ontogeny; the size and extent of the branchiostegal rays in the 8.0 mm specimen (Fig. 47A) and the 33.5 mm specimen (Fig. 63A) reflects this. The opercle is also small and is almost not moving during normal respiration (Howes, 1983a; Chapter 6.1). In many bottom-living teleosts the branchiostegal membrane is very large (Hughes, 1970). In the siluriform Clarias gariepinus, the buccal pressure pump is believed to be most important in early life history stages (11.4 mm TL); the importance of the opercular suction pump increases in the 15.5 mm and 21.4 mm TL specimens (Vandewalle et al., 1985). In this species the number of branchiostegal rays (and relative size of the membrane) gradually increases up to 10 during ontogeny (Adriaens & Verraes, 1998). The shift in orientation of the hyohyoideus abductor in Ancistrus cf. triradiatus might indicate a functional shift of an abductor of the branchiostegal membrane (as in most teleosts) to an adductor (as hypothesized for adult A. cf. triradiatus; Chapter 6.1). The respiration and feeding mechanism of A. cf. triradiatus (and other loricariids), being well adapted to the need of forceful suction for attachment, however, most probably involves more modifications than those of the hyoid region alone. The roles of the lower jaws and the oral valve, as well as the associated musculature, hitherto is only hypothesized for adult specimens (Chapter 6.1). A more general study of all head muscle systems, especially in embryonic and juvenile 116 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE loricariids, will yield a more complete comprehension of the cranial movements and functional shifts during ontogeny.

PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 117

5.3. ONTOGENY OF THE SUSPENSORIAL AND ∗ OPERCULAR MUSCULATURE

Abstract — Several morphological features characterizing Loricariidae or suckermouth armoured catfishes are related to their ability to attach onto substrates with their suckermouth, and to scrape algae and other food items from these substrates. Suspensorial and opercular muscles are among those muscles usually involved in respiration (and feeding). In several loricariids including the genus Ancistrus, the opercular musculature is decoupled from the respiratory mechanisms. This study of the ontogeny of both suspensorial and opercular muscles aims to describe how the muscles develop and transform during early life history. The adductor arcus palatini is relatively large throughout the whole ontogeny, while the levator arcus palatini is minute. It develops in association with the dilatator operculi, which exhibits substantial growth only in the juvenile and adult stages. The levator and adductor operculi are connected during early ontogeny, and anterior fibres of the latter muscle differentiate into the adductor hyomandibulae, a muscle previously thought to be absent in loricariids. Some functional considerations are made concerning hypothetical free-living embryonic to adult kinetics of the hyoid-suspensorium mechanism.

5.3.1. INTRODUCTION

The suspensorial and opercular musculature is part of the dynamic apparatus involved in respiration and feeding in fishes. Muscle activity can result in (1) expansion and contraction of the orobranchial cavity, and (2) expansion and contraction of the opercular cavity. The first set of actions can be referred to as the pressure pump, as contraction of the orobranchial cavity forces water through the gills. Expansion of the opercular cavity sucks water from the former cavity via the gills; this opercular pump has been called the suction pump (Hughes, 1960, 1970). The anatomical basis of the first system is most complex, as orobranchial volume changes can be generated by movements of the jaw and hyoid bars (ventrally), the neurocranium (dorsally), and the suspensoria (laterally) (e.g., Schaeffer & Rosen, 1961; Elshoud, 1978; Lauder & Liem, 1980; Lauder, 1985; Muller, 1987; Aerts, 1991). Opercular musculature is also often involved in the mouth-opening mechanism (Elshoud, 1978; Aerts et al., 1987; Lauder & Liem, 1980). Studies on cranial ontogeny have suggested transformations in the structural basis for respiratory kinematics, tightly coupled to the developmental sequence of participating elements (especially muscles and skeletal structures) (Vandewalle et al., 1985; Surlemont et al., 1989; Surlemont & Vandewalle, 1991; Adriaens & Verraes, 1997e; Adriaens et al., 2001).

∗ Slightly modified from: Geerinckx T. & Adriaens D. Ontogeny of the suspensorial and opercular musculature in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Submitted to Zoomorphology. 118 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

In loricariid catfishes, the suspensorial and opercular musculature is known to present a great degree of variation in size, position and even mere presence (Howes, 1983a). Loricariidae or suckermouth armoured catfishes are a highly specialized, but diverse and speciose family, and typically have a ventrally placed suckermouth with an enlarged and posteriorly oriented lower lip and medioventrally rotated lower jaws (Alexander, 1965; Schaefer & Lauder, 1986; this dissertation). The generally well developed adductor arcus palatini ontogenetically arises from the anterior part of the constrictor dorsalis of the hyoid muscle plate; the posterior part of this constrictor dorsalis gives rise to the adductor operculi, of which the levator operculi and adductor hyomandibulae are believed to have derived (Schaeffer & Rosen, 1961; Winterbottom, 1974; Diogo & Vandewalle, 2003a). The levator arcus palatini and dilatator operculi originate from the constrictor dorsalis of the mandibular muscle plate (Winterbottom, 1974). In several loricariid genera, including Ancistrus, the opercle has probably lost its role in the respiratory mechanism, and has acquired a role in a defensive apparatus in which cheek spines (enlarged odontodes) can be erected (Howes, 1983a; see Part 8 for more details). In Ancistrus, some of the opercular muscles are greatly enlarged, whereas the presence of the small adductor hyomandibulae has generally been overlooked (see also Part 6). The role of the suspensorial muscles in the buccal pump system is interesting, as the posterior edge of the hyomandibula is sutured to the neurocranium in many loricariid genera (Armbruster, 2004), restricting the mobility of the suspensorium. The present chapter describes the ontogeny of all suspensorial and opercular muscles in the bristlenose catfish Ancistrus cf. triradiatus. The discussion on the functionality of muscles is based on the anatomy only, as no biomechanical studies including EMG could be done on such small specimens. In all catfishes, including loricariids, the anteriormost fibres of the adductor arcus palatini have differentiated into muscles serving the autopalatine-maxillary system (Singh, 1967; Gosline, 1975). Ontogeny of these muscles in loricariids is discussed in chapter 5.1.

5.3.2. BRIEF MATERIAL AND METHODS

PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 119

5.3.3. RESULTS

6.1 MM SL — 4 DAYS POST-FERTILIZATION (FIG. 50)

While in earlier stages (4.8 and 5.2 mm SL) no recognizable muscle tissue was present, most suspensorial and opercular muscles can be identified in the 6.1 mm specimen. Hatching occurs at a standard length of 6.8 mm on average; this specimen was still in the egg when anaesthesized. No ossification is observed on the neurocranium floor or the suspensorium. The adductor arcus palatini is already present between these two structures. The flat muscle clearly lacks any insertion, neither on the neurocranium nor on the suspensorium. The suspensorium is oriented in a vertical plane; a vertical line can be drawn along the neurocranium-hyosymplectic articulation, the interhyal and the hyoid arch (Fig. 50). The levator arcus palatini has developed as a small muscle against the lateral to rostral side of the hyosymplectic part of the cartilaginous suspensorium. Fibre direction is ventrally from the region lateral to the suspensorium-neurocranium articulation. No insertion is observed. Near its origin some muscle fibres are continuous with the fascia of the dilatator operculi. The latter muscle originates from the same region, but is somewhat larger and runs caudoventrally along the suspensorium, also lacking an insertion. No visible muscle fibres or tendon reaches the opercle, which has appeared as a tiny splint of bone near the opercular process of the suspensorium. The levator operculi and adductor operculi are in contact with each other near their ventral end. Both muscles originate on the ventral floor of the otic capsule of the neurocranium, the levator more rostrolateral than the adductor operculi. The levator operculi, being the smallest of both muscles, runs towards the dorsal aspect of the opercle, without effectively reaching it. The adductor operculi almost reaches it somewhat more medially, and is fused to the medialmost fibres of the levator.

8.0 MM SL — 7 DAYS POST-FERTILIZATION (FIG. 51)

In this specimen the adductor arcus palatini has substantially expanded and has become a broad muscle plate originating from the trabecular bar and inserting on the base of the pterygoid process and the medial edge of the hyosymplectic of the cartilaginous suspensorium (at this moment perichondral ossification is just starting to cover some parts of the chondrocranium, but not this region of the suspensorium). The caudalmost fibres have no insertion at all, as they end halfway between the trabecular bar and the suspensorium, the latter still lacking its dorsomedial membranous outgrowth (see 12.4 mm specimen). The muscle reaches the perichondrium of both cartilaginous neurocranium and suspensorium, so the muscle might well be functional by now. 120 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

The direction of the levator arcus palatini has slightly shifted: the muscle now runs rostroventrally instead of ventrally from the neurocranium-suspensorium articulation. A clear insertion on the neurocranium is not present, and the muscle is still in contact with the dilatator operculi. Serial sections actually show its origin to be on the fascia of the latter muscle, and, thus indirectly, on the neurocranium. It inserts on the rostrolateral edge of the hyosymplectic. The dilatator operculi originates on the lateral side of the anterior portion of the otic capsule, just dorsal to the articulation of the hyosymplectic to the neurocranium. Insertion is doubtful: a long tendon reaches towards the dorsal aspect of the opercle, but might or might not effectively attach to it. The orientation of the dilatator operculi has become more ventral, compared to the orientation in the 6.1 mm specimen. This is due to allometric growth of chondrocranium elements. The suspensorium has also elongated and its ventral part has shifted anteriorly (compare figures 50C and 51C). This marks a major transformation in the chondrocranium Bauplan: not only is the anterior region (including most of the suspensorium) elongating, the relative positions of the articulations of the suspensorium with the neurocranium, lower jaw and hyoid have changed. The levator operculi still is a narrow, vertical muscle band stretching from the lateralmost protuberance of the otic capsule in the direction of the dorsal aspect of the opercle, well posterior to the insertion of the dilatator operculi tendon. Medially, its fascia is still in contact with the fascia of the larger adductor operculi. This latter muscle, without any doubt the largest of the opercular muscles at this moment, has expanded and almost reaches the opercle only slightly more medially than the levator. Interestingly, some anterior fibres of the adductor operculi insert on the opercular process of the suspensorium. Unambiguous insertion of the adductor or levator operculi on the opercle is not yet observed. A rostrodorsal differentiation of the adductor operculi has resulted in a thin muscle band ending in connective tissue near the medial side of the hyosymplectic part of the suspensorium. This band thus runs completely rostrally. I identify this muscle as the adductor hyomandibulae (see Discussion for an account on the usage of this name in teleostean myology).

12.4 MM SL — 43 DAYS POST-FERTILIZATION

Between the 8.0 and 12.4 mm stages most cranial bones have appeared, including all perichondral bones. Most of the cartilage of the neurocranium and suspensorium is covered by a bone layer. The opercle has grown, having become thicker as well as deeper. Origins and insertions of the suspensorial and opercular muscles are on the osteoskeleton now. Little has changed in the shape or relative size of the adductor arcus palatini. Origin is on the orbitosphenoid and parasphenoid. The posterior fibres run in a lateral direction, towards a large, membranous medial extension of the hyomandibular bone. The more anterior fibres PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 121 run more rostrolaterally, reaching both the hyomandibula and the metapterygoid, the latter of which has grown around the cartilaginous pterygoid process. The levator arcus palatini now has its final origin, i.e. on the lateral extension of the sphenotic (dermosphenotic portion), that was not yet present in the 8.0 and 10.2 mm stages. It has lost the contact to the dilatator operculi. The latter muscle is now provided with a large aponeurosis. Muscle fibres diverge from it, both medially and (less) laterally. The medial fibres originate on the perichondral ossification of the otic capsule wall (autopterotic portion of pterotic bone), and the lateral fibres attach on a membranous tissue sheet covering the muscle and lying directly under the skin. Current ossification in this sheet is part of the dermopterotic portion of the pterotic bone. Both the levator and adductor operculi now insert on the opercle: the levator on the dorsal margin of the bone, posterior to the dilatator operculi insertion, and the adductor medial of this dorsal margin. Origin of the levator operculi is now also on the pterotic instead of on the cartilage of the otic capsule. The adductor operculi now originates rather posteriorly on the neurocranium floor, with a clear tendon that also contacts the adductor hyomandibulae muscle. Origin is on the ventral aspect of the pterotic bone, more or less where the supracleithral ossification of Baudelot’s ligament initiates. The supracleithrum is fused to the posteroventral aspect of the pterotic in loricariids (Chapter 4.2). The vertically flattened adductor hyomandibulae, originating on a tissue sheet that is continuous with the tendon of the adductor operculi, inserts rather loosely on connective tissue attaching to the inner side of the hyomandibula. As the relative growth of the dilatator and levator operculi has surpassed that of the adductor operculi, the dilatator is now the largest opercular muscle, and the levator and adductor are of similar size.

FURTHER ONTOGENY IN JUVENILE ANCISTRUS CF. TRIRADIATUS

The adductor and levator arcus palatini, as well as the adductor operculi, grow more or less isometrically during further growth of Ancistrus cf. triradiatus (see Figs 62B, 63c). The adductor hyomandibulae remains very small and slender, being almost unrecognizable in adult specimens (in fact, it is easily overlooked without relying on serial sections; Fig. 52). The levator operculi grows relatively large, becoming a pyramidal muscle extending in a cavity of the (compound) pterotic bone. The most drastic growth is observed in the dilatator operculi, allometrically increasing in size during juvenile and even adult growth. Especially in large males, the muscle is housed in a cranial cavity that reaches towards the midline, above the brain, but below the dorsal skull roof. Intense modifications in the skull roof are present. 122 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

MUSCLE INNERVATIONS

Serial sections clearly show the innervations of all suspensorial and opercular muscles. The adductor arcus palatini is innervated by a small branch of the hyomandibular trunk of the facial nerve. This branch, aptly named ramus adductor arcus palatini (Atoda, 1936), also supplies both parts of the extensor tentaculi (muscles being derived from the adductor arcus palatini, see Chapter 5.1). Further along the hyomandibular trunk the opercular branch separates from it, almost immediately splitting in two parts: one supplying the levator operculi, and another the adductor operculi as well as the adductor hyomandibulae. A nerve branch separates from the infraorbital trunk (consisting of bundles of both trigeminal and facial nerves) soon after it emerges from the skull. This branch [ramus levator arcus palatini of Atoda (1936)] splits into a lateral twig to the levator arcus palatini and a posterior twig to the dilatator operculi.

5.3.4. DISCUSSION

In Ancistrus cf. triradiatus most suspensorial and opercular muscles are present in the 6.1 mm SL specimen; only the adductor hyomandibulae has not yet differentiated. None of the present muscles has distinct insertions. It is not clear whether any muscle contractions in this stage could move the chondrocranial elements via the not yet differentiated connective tissue that separates the muscle tissue from the cartilage. The only cranial muscle of which a probable insertion is observed in this stage is the hyohyoideus inferior. Origin is on the hyoid bar, but no anterior insertion is present in the intermandibularis posterior (Chapter 5.2). At this moment, one day before hatching, a weak contraction and expansion of the orobranchial chamber can be observed. In the 8.0 mm specimen the adductor hyomandibulae has differentiated from the adductor operceli, but lacks an anterior insertion on the suspensorium. The other suspensorial muscles are now provided of both origin and insertion. Insertion of the opercular muscles, however, is still uncertain (dilatator operculi), absent (levator operculi) or absent except for a few fibres attaching to the suspensorium (adductor operculi). In the 12.4 mm specimen all final insertions are present. This sequence in muscle appearance and insertion was also observed in the clariid Clarias gariepinus, the only siluriform of which the muscular ontogeny has been studied into detail thus far: appearance (at 4.7 mm TL) and insertion (at 5.2 mm TL) of the adductor and levator arcus palatini occur first, while the opercular muscles are first seen at 5.2 mm TL (Surlemont et al., 1989; Surlemont & Vandewalle, 1991). The dilatator operculi probably attaches to the opercular process of the hyosymplectic in the 5.2 mm TL specimen (Surlemont & Vandewalle, 1991), and insertion on the opercle appears not yet completely PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 123 developed in a 7.2 mm SL specimen (Adriaens & Verraes, 1997e). Insertion of the levator operculi on the opercle is effective at 6.8 mm SL. At this stage, the adductor operculi is closely associated to the opercular process of the hyosymplectic (as in Ancistrus cf. triradiatus); an undubious insertion on the opercle is only observed at 46.8 mm SL (Adriaens & Verraes, 1997e). C. gariepinus lacks an adductor hyomandibulae.

In Ancistrus cf. triradiatus the adductor arcus palatini shows no important positional or size-related transformations during ontogeny. Its early position and insertion are already identical to the adult configuration. The levator arcus palatini is more influenced by changes in shape and position: its fibre direction changes from slightly caudoventrally (Fig. 50) to essentially rostrally (Fig. 62B). Opposed to its origin, its insertion is never associated to that of the dilatator operculi, thus contradicting Howes’ (1983a) description of adult Ancistrus sp. The insertion of the adductor and levator arcus palatini on the suspensorium suggests their role in the adduction and abduction of the wall of the orobranchial chamber, as in other siluriforms and most teleosts (Nawar, 1955a; Osse, 1969; Lauder & Liem, 1980; Adriaens & Verraes, 1997e). The function of the opercle and associated musculature in Ancistrus differs from the general teleostean situation, in which abduction and adduction of the opercle causes the opercular cavity to expand and constrict, respectively. The opercle and opercular muscles are involved in a mechanism of erectile cheek spines, used as a defensive apparatus. This altered morphology and function was noticed by Alexander (1965) and Howes (1983a), and is described into detail in part 8. The opercular muscles thus have acquired an altered function, although the adductor operculi still adducts the opercle, and the dilatator and levator operculi abduct it. The effects of the latter muscles are identical, as the long articular hinge allows opercular movement only in one rotational plane (compare left and right side on figure 63C; Part 8). It can be observed that the opercle does not move synchronously with the ventrally positioned branchiostegal membrane. This, and the fact that it is used in the defensive cheek- spine apparatus, and not respiration, infers that the opercular muscles appear to have lost their respiratory role (Howes, 1983a; Chapter 6.1 and Part 8). The biomechanical couple between the opercle and the lower jaw (Elshoud, 1978; Aerts & Verraes, 1984; Aerts et al., 1987; Westneat, 1990) is lost as well, as the interopercle and interoperculo-mandibular ligament are absent in many loricariids including Ancistrus (Armbruster, 2004; Chapter 6). The function of the adductor hyomandibulae is uncertain. The muscle remains very small during later ontogeny. In some papers (e.g., Howes, 1983a; Schaefer, 1997) the name ‘adductor hyomandibulae’ is erroneously used for the adductor arcus palatini. The true loricariid adductor hyomandibulae is an anterior derivation of the adductor operculi. It was not seen in loricariids by Howes (1983a) and Schaefer (1997), but I found it in 124 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

Pterygoplichthys lituratus (extremely small), Farlowella acus and Otocinclus vestitus. Among siluriforms, the muscle has also been found in Amblycipitidae, Ariidae, Bagridae, Callichthyidae, Claroteidae, Cranoglanididae, Plotosidae and Siluridae (Takahasi, 1925: ‘retractor hyomandibularis’; Diogo & Vandewalle, 2003a; F. Huysentruyt, pers. commun.). It is not present in the basal siluriform Diplomystidae and Nematogenyidae (Diogo & Vandewalle, 2003a; Diogo et al., 2006). It appears to have developed independently in several non-siluriform lineages as well (Winterbottom, 1974).

The relative volume of the suspensorial musculature during early ontogeny of Ancistrus cf. triradiatus (e.g., 8.0 mm specimen) is interesting: the adductor arcus palatini is very large, and the levator arcus palatini is very small, with a fibre direction (line of action) that is not ideal to abduct the suspensorium. Of the larger hyoid muscles, the hyohyoideus inferior is more substantial and has its insertions earlier than the sternohyoideus (Chapter 5.2). Thus, the musculature responsible for a compression of the orobranchial chamber at the level of the hyoid and suspensorium [hyohyoideus inferior, adductor arcus palatini, possibly (temporarily) adductor operculi] seems more substantial and functional earlier than the antagonistic musculature responsible for the orobranchial expansion (sternohyoideus, levator arcus palatini). I observed actual movements of the hyoid bar and suspensoria from about 6.0 mm SL on. The expansion phase might well profit from the elastic properties of the cartilaginous skeleton. The contribution of elasticity of cartilage has been suggested for Clarias gariepinus as well (Vandewalle et al., 1985; Surlemont et al., 1989). The zones of articular cartilage, containing hypertrophied cartilage cells with little matrix, are highlighted in figure 53. In other teleosts such cartilage tissue has been found in joints and symphyses subject to compression (Anker, 1989; Benjamin, 1990; Aerts, 1991). A hypothetical respiratory cycle would include a compression phase in which the hyohyoideus inferior and adductor arcus palatini would contract: the articular cartilage between the hyosymplectic and the neurocranium (1) and the interhyal cartilage (2) would bend; the hyoid symphysial cartilage (3) would bend and compress (especially posteriorly). In the expansion phase, the elasticity of the chondrocranium, as well as the action of the (still small) sternohyoideus and levator arcus palatini, could bring the hyoid bar and suspensoria back to their original position. Newly hatched specimens (free-living embryos) of Ancistrus cf. triradiatus immediately cling to vertical or even totally inclined surfaces in the nest cavity [parental care is common in Ancistrus (Burgess, 1989)]. Only the suckermouth is used for attachment, so suction is particularly important. Considering the relatively weakly developed sternohyoideus and levator arcus palatini during early ontogeny (compared to the antagonistic muscles), I assume it might well be impossible to maintain this lower pressure inside the orobranchial cavity PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE 125 without the cartilage elasticity. Apparantly, possible adhesive properties of the papillose surface of the only partly developed lips are not enough to maintain attachment, as is proven by experiments with an anaesthesizing substance (MS-222), that halts muscle activity. Suction fails immediately. During ontogeny, several major transformations occur that influence the mode of respiration: the proceeding development of the cranial muscles and ligaments, and their insertions, the ossification of the skeletal elements, and the lengthening of the skull, including the change in shape and position of the suspensorium (compare figures 50C and 51C). In the 6.1 mm specimen the articular points (1-3 in figure 53) are all in one vertical plane. During further ontogeny points 2-3 shift anteriorly. The same has been observed for Clarias gariepinus (Vandewalle et al., 1985: fig. 1). Compared to adult Ancistrus cf. triradiatus, the hyoid symphysis has shifted significantly more anteriorly in adult C. gariepinus. In the latter species, Vandewalle et al., (1985) correlated this shift to a reduction of the buccal cavity and a parallel enlargement of the opercular fold and branchiostegal membrane. In adult A. cf. triradiatus the smaller branchiostegal membrane and more posterior position of the (rotated) lower jaw and lip are probably related to the more posterior position of the hyoid arch. A functional consequence of the smaller branchiostegal membrane (and the fact that the opercle may not be involved in respiratory movements) is the absence of a forceful suction pump system in adult (and free-living embryonic) A. cf. triradiatus, compared to C. gariepinus and many other bottom-living fishes (Hughes, 1970; Adriaens & Verraes, 1997e). This is also reflected in the small size of the branchiostegal musculature (see Part 6). My observations corroborate the thesis that in bottom-living fishes with a dorsoventrally flattened skull (‘a broad head’) the hyoid movements contribute more to the volume changes of the orobranchial cavity, compared to fishes with narrower heads (Alexander, 1970; Gosline, 1973; Adriaens & Verraes, 1997b). The suspensorium of loricariids is relatively weakly movable, due to the tight connection of the metapterygoid to the lateral ethmoid (Alexander, 1965; Arratia, 1990), the dorsal interdigitation of the hyomandibula to the prootic, and (in many loricariids) the posterior suture of the hyomandibula to the compound pterotic (Howes, 1983a; Armbruster, 2004; see also Part 8). Observation of adults (and early life stages) clearly shows that suspensorial movements are considerably less than hyoid movements. It might well be possible that the depression and elevation of the hyoid bar are responsible for most of the suspensorial movements. Such interactions have been found in many teleosts (e.g., Alexander, 1969, 1970; Anker, 1974; Elshoud-Oldenhave & Osse, 1976; Elshoud, 1978; Muller, 1989a; Aerts, 1991; Hunt von Herbing et al., 1996). The ontogenetic replacement of the cartilaginous interhyal by a sesamoid bone in a more ligamentous hyoid- suspensorial connection (Chapter 4.2) may reflect a change in cranial kinetism [the significance of the interhyal has been stressed before (e.g., Anker, 1974; Lauder, 1980; 126 PART 5 — ONTOGENY OF THE CRANIAL MUSCULATURE

Adriaens & Verraes, 1994)]. The consequences of the weak mobility of the adult suspensorium have been discussed by Alexander (1965), and is also discussed in chapter 6.1 of this dissertation. The rigidity of the suspensorium might be important to resist the forces exerted by powerful suction when attaching to a substrate. It might also reinforce the suspension of the hyoid. A certain elasticity of the bony suspensorium itself has however been suggested (Alexander, 1965; Anker, 1974). The complex interaction between the suspensorium and the hyoid probably differs significantly between free-living embryonic and adult specimens. Changes in articulation points might be as important as changes in the connected elements: the increasing rigidity of the suspensorium-neurocranial contact, the loss of the interhyal and the formation of a sesamoid bone probably are key transformations in the change in respiratory kinetism that is expected to occur during ontogeny of Ancistrus cf. triradiatus. These changes, as well as the sequence of muscle and bone development and ligament formation and the positional changes of the suspensorium, are all adaptations to the specific functional requirements of fishes throughout their ontogeny (Liem, 1991; Adriaens et al., 2001). In loricariids, this will undoubtedly be dominated by the ontogenetic and evolutionary transformations towards a suckermouth with a scraping feeding mode.

PART 6

ADULT MORPHOLOGY

PART 6 — ADULT MORPHOLOGY 127

∗ 6.1. ANCISTRUS CF. TRIRADIATUS

Abstract — The neotropical loricariid catfishes are highly specialized for attaching to substrates, and can continue breathing and even scrape food from these surfaces while using the mouth for suction. A detailed study integrating bones, muscles and ligaments was performed on Ancistrus cf. triradiatus, using cleared and stained specimens, dissections and manipulations on fresh specimens, serial sections, and histological examination of key tissues. A limited kinematic study using high-speed video was performed as well. The suspensorium is a rather rigid structure; the hyoid is more movable and associated muscles are more substantial; it appears to be more important in the buccal pump system. The transverse orientation of the hyohyoideus abductor suggests it can’t open the branchiostegal membrane. This movement might be passive. Apart from divisions inserting on the lower and upper jaws, a medial adductor mandibulae division, the retractor veli, inserts on the oral valve. The levator tentaculi and the lateral part of the completely subdivided extensor tentaculi move the maxillary barbel, a structure that allows controlled inspiration preventing failing of the suction system. Rotational movements of the lower and upper jaws result in scraping the substrate. Antagonistic muscles for the adductor mandibulae divisions inserting on the lower and upper jaws might be a part of the intermandibularis posterior and the medial part of the extensor tentaculi. The lower jaws are most mobile, not being linked to the hyoid arch medially. A medial cartilage plug acts as a supporting and gliding device for the lower jaws.

6.1.1. INTRODUCTION

Many fishes have adopted a behaviour in which attachment to substrates proves to be advantageous. Three typical features of the body plan that seem to have an adaptive value, can be discerned (Hora, 1930): firstly, a depressed body shape reduces drag of torrential water when lying on a substrate. Paired fins that are closely appressed to the substrate can aid in maintainance of this close contact. Secondly, frictional devices may evolve, such as spines or odontodes on the ventral side of the body and fins, making it less likely an adhering fish is washed away. Thirdly, a suction apparatus can attach a fish more firmly, irrespective of the substrate inclination or the current direction. The latter two mechanisms are also of great value in non-flowing water systems, and (especially the third) allow the attachment to inclined substrates. Examples of the latter mechanism in recent teleosts are numerous. Thoracic discs providing both friction and suction, as well as paired fins with adhesive surfaces have been described in Sisoridae and Erethistidae (Hora, 1930; Saxena & Chandy, 1966; Tilak, 1976). A suction disc is formed by the pelvic fins in various families, such as Gobiidae (Hora, 1930), or by the ventral surface of the lower lip in the cyprinid Garra (Hora, 1930; Saxena &

∗ Slightly modified from: Geerinckx T., Brunain M., Herrel A., Aerts P. & Adriaens D., 2007. A head with a suckermouth: a functional-morphological study of the head of the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Belgian Journal of Zoology, in press. 128 PART 6 — ADULT MORPHOLOGY

Chandy, 1966). Gyrinocheilidae have a modified gill opening, providing both an inspiratory and an expiratory canal. Consequently, when attaching to a substrate, the mouth can continue to work as a sucking cup (Smith, 1945; Jayaram, 1950; Benjamin, 1986; Roberts & Kottelat, 1993; Nelson, 1994). A small ventral mouth with thick, papillated lips, as well as tentacles and minute spines on fin rays ensure attachment in certain Amphiliidae (Hora, 1930; Diogo et al., 2000). A ventral suckermouth is also well developed in members of the Mochokidae and Astroblepidae. The latter group can even use the suckermouth for terrestrial movements (Johnson, 1912). Also, in some fishes the coupling of a suckermouth to structural transformations in the jaws or the lips has allowed an evolutionary development of a substrate-scraping device. The single largest group in which the combination of oral suction and a scraping feeding apparatus is found, is the family of the South American Loricariidae, or suckermouth armoured catfishes. The family is characterized by remarkable features such as a bony armour, a ventrally oriented suckermouth, ventrally tilted lower jaws and new and unique muscle configurations increasing jaw mobility (Schaefer & Lauder, 1986). Not only this morphology, but also their status as the largest catfish family [more than 700 species according to a recent count (C. Ferraris, pers. commun.)] has drawn focus to the group. Among the most important papers discussing loricariid cranial morphology are Alexander (1965), Howes (1983a), Schaefer (1987, 1988), Schaefer & Lauder (1986, 1996) and Armbruster (2004). Accounts on postcranial morphology were given by Shelden (1937), Lundberg (1969) and Schaefer (1987, 1997), while Alexander (1964) and Chardon (1968) treated the loricariid Weberian apparatus. Cranial osteology and myology of the loricariid Otocinclus vittatus have been described by Schaefer (1997), with a strong systematic and phylogenetic interest. The aim of the present chapter is to describe in detail the cranial osteology and myology of one loricariid species, and to discuss those parts of the head involved in suction, respiration and feeding, with an emphasis on functionality of structures rather than their use as systematic characters. The anatomical results are coupled to a limited kinematic data set. The species chosen for this study is Ancistrus cf. triradiatus (see Chapter 2.1), a species from the northern drainages of Venezuela, as well as Colombia (Isbrücker, 1980; Page et al., 1993; Fisch-Muller, 2003).

6.1.2. BRIEF MATERIAL AND METHODS

Examined specimens: clearing and staining (paragraph 2.2.5): 4 (male: 88 mm SL, 90 mm SL; female: 44 mm SL; gender unknown: 36 mm SL); dissection (paragraph 2.2.6): 3 (male: 94 mm SL; female: 68 mm SL, 74 mm SL); serial sections (paragraph 2.2.7): 1 (gender PART 6 — ADULT MORPHOLOGY 129 unknown: 33.5 mm SL); histological study (paragraph 2.2.7): 1 (female: 71 mm SL). Some specimens of Farlowella acus, Otocinclus vestitus and Pterygoplichthys lituratus were studied for comparison (see Table II). For live observations, see paragraph 2.2.2. Filming of three specimens (in lateral, ventral and oblique ventrolateral view) was done with a Redlake Motionscope digital video camera at 200 frames per second (see paragraph 2.2.2). Only each eighth frame was analysed (thus 25 frames per second). Observations are qualitative; no markers or length measurements were included.

6.1.3. RESULTS

Except where noted, osteological terminology follows Schaefer (1987), and myological terminology follows Winterbottom (1974).

NEUROCRANIUM (FIGS 54, 55, 56, 57)

The anterior half of the long mesethmoid is almost cylindrical and gives the snout region rigidity. Anterolateral cornua are absent, and there is an expanded ventral disc projecting ventrally (Figs 55A, 57A-B). The large lateral process of the lateral ethmoid [antorbital process of Schaefer (1987)] has an articular facet for the autopalatine anteriorly, and an articular facet with a supporting ridge for the metapterygoid ventrally. The anterior part of the frontal is relatively narrow, while the posterior part is broader, reaching the orbit. The sphenotic has a prominent lateral process, enclosing the infraorbital canal. Posterior to the orbit, the sphenotic contributes to the articulation with the hyomandibula (Fig. 55B). In the skull floor the toothless prevomer is a narrow bone, without well developed lateral wings. It sutures deeply with the mesethmoid (anteriorly) and the parasphenoid (posteriorly), which sutures with the basioccipital as well. The parasphenoid forms a longitudinal protruding ridge on the ventral side of the neurocranium. It has a pair of small lateral wings. A major part of the skull wall and floor lateral to the parasphenoid is occupied by the orbitosphenoid. Together with the pterosphenoid and the prootic it forms the border of the sphenotic fenestra (Fig. 55B). The prootic contributes to the posterior skull floor and, to a lesser degree, to the neurocranium wall, where it bears almost half of the articulation facet for the hyomandibula. The basioccipital is fused posteriorly with the ossified Baudelot’s ligaments into a T- shaped bone (Fig. 56A). Laterally, a certain degree of fusion has occurred with the exoccipital. The ossified Baudelot’s ligaments are vertical bony ridges protruding ventrally 130 PART 6 — ADULT MORPHOLOGY from the skull floor. They extend towards, and suture with, ventral flanges of the compound pterotics, which are the continuation of this transverse ventral ridge. A substantial part of the posterior skull roof, skull floor and caudolateral wall of the brain case is taken by the pterotic. Thus it is a double-layered bone, providing ample insertion space for the opercular muscles. Where it contacts the sphenotic and prootic, it forms the caudal edge of the hyomandibular articulation. The pterotic is fused to the more ventrocaudal supracleithrum (or even posttemporo-supracleithrum) of the pectoral girdle. The true nature of this fused bone complex, as well as its relation to the ossified Baudelot’s ligament, has not been unambiguously resolved up till now (Lundberg, 1975; Fink & Fink, 1996); its development is discussed in chapter 4.2. In this dissertation it is referred to as compound pterotic. The epioccipital [epiotic of Schaefer (1997)] is a small element composing part of the caudal neurocranium wall (Fig. 55C). Finally, the parieto-supraoccipital lacks a pronounced posterior process. Posteroventrally, a V-shaped medial ridge is fused with the fused neural arch of the second and third vertebrae. Moreover, the posterior tip is fused with the neural spine of the sixth vertebra, as in Hypostomus plecostomus (Alexander, 1964) (Fig. 55C). There is no dorsomedial crest. Six infraorbital canal-bearing bones are present on either side. These bones, as well as the canal-bearing nasal, are part of the dermal plating of the snout and cheek region (Figs 54A, 55A). Between both infraorbital series, a total of 24 prenasal plates were counted in the specimen illustrated in the figures (this number varies somewhat among the examined specimens). Ventrolateral to the infraorbitals a group of lateral plates is present, also varying in number and size, even between both sides of the same specimen (nine at left, eight at right side of drawn specimen; Figs 54A, 55A). The paired, square-shaped prefrontal plate partly covers the lateral process of the lateral ethmoid, bordering the nostril and the orbit.

SPLANCHNOCRANIUM (FIGS 54, 55, 56, 57, 58, 59)

The paired autopalatine articulates with the anterior end of the lateral ethmoid. Ventral to the articulation facet (which is directed caudodorsally) two processes are present (Figs 56A, 57C-D). The tendons of the pars medialis and lateralis of the extensor tentaculi muscle insert on these medial and lateral processes respectively. Anteriorly a large cartilaginous head is present on the autopalatine, on which the bar-shaped maxilla articulates by means of a double medial process, providing two articular surfaces. The autopalatine splint or sesamoid bone, as seen in some other loricariids (Schaefer, 1987, 1997), is rudimentary in Ancistrus cf. triradiatus, fused to the anterolateral side of the autopalatine (Fig. 57C). The slightly curved maxilla is flattened (Fig. 57E-F), providing abundant insertion space for the levator tentaculi muscle posteriorly. The distal end is fused to the maxillary barbel cartilage, which supports a short barbel. The premaxillae are oval to rectangular basket-shaped and provide space for PART 6 — ADULT MORPHOLOGY 131 developing teeth rows (Fig. 58A-B). One row of functional teeth (number of teeth per jaw averaging 60 to 80 in adults) inserts on the inner rostral side. The dento-mentomeckelium and angulo-articular of which the lower jaw consists are strongly sutured rostrally; a long splint of the dento-mentomeckelium overlies the angulo- articular ventrally (Fig. 57C-D). Caudally, the small cartilaginous remnant of the Meckel’s cartilage connects both bones. There is no separate coronomeckelian bone. The dento- mentomeckelium resembles the premaxilla in being basket-shaped, and bearing functional and developing teeth rows. As in the premaxilla, the functional teeth (60 to 85 on average) point ventrally, as the whole lower jaw is twisted ventrally and medially. The coronoid process of the dento-mentomeckelium, and, more importantly, the high dorsal ridge of the dento-mentomeckelium and the angulo-articular serve as ample insertion surface for the adductor mandibulae muscle (Fig. 55A-B). Laterally, each angulo-articular articulates with the quadrate; medially, both dento-mentomeckelian bones are loosely embedded in soft tissue containing a cartilaginous plug (Figs 60D-E, 63B). This configuration allows both lower jaws to move independent from each other. The cartilage plug acts as a supporting device for the free medial end of the dento-mentomeckelian bones. The long, triangulate suspensorium is a very sturdy structure. The quadrate and the hyomandibula are tightly sutured, as well as reinforced by the preopercle that overlies their lateral surfaces, tightly coalesced to both bones (Figs 54B, 55B, 59). Dorsally the small symplectic cartilage is situated between both bones. The hyomandibula articulates with the neurocranium at the point where the prootic, sphenotic and compound pterotic meet. The combined hyomandibular crest for the adductor mandibulae and levator arcus palatini is very conspicuous (Fig. 54B). The restricted mobility of the suspensorium is largely caused by an additional, long and almost suture-like articulation between the posterior edge of the hyomandibula and the compound pterotic (Figs 55B, 56A). Both quadrate and hyomandibula are loosely sutured to the metapterygoid (Fig. 55B). Dorsally this latter bone articulates with the lateral ethmoid, the long articulation being most rigid in the posterior half. The body of the metapterygoid and a dorsolateral lamina extending towards the autopalatine, provide a groove-like housing for the pars lateralis of the extensor tentaculi muscle (Fig. 55B). Although the entire suspensorium is rather rigid, some bending along the joints and sutures allows a certain degree of movement towards the medial, while elasticity seems to return it to the resting position. The hyoid (Fig. 56B-C) consists of paired (ventral) hypohyals, and anterior and posterior ceratohyals [anterohyal and posterohyal of Schaefer (1987, 1997)]. The anterior ceratohyal has synchondral joints with the hypohyal and the posterior ceratohyal. The latter joint is reinforced by means of a suture between the anterior laminae of both bones. The posterior ceratohyal has a long hinge with the medial face of the hyomandibula: halfway along the 132 PART 6 — ADULT MORPHOLOGY hyomandibula it has a cartilaginous articulation; at the rostral end of the hyomandibula, near the symplectic cartilage, a second, more movable ligamentous connection is present, with ligaments from the posterior ceratohyal to the preopercle and to the symplectic cartilage too. In the latter ligaments a minute sesamoid bone is found (Fig. 56C). An interhyal is not present. The rostral margin of the anterior ceratohyal attaches to the quadrate and metapterygoid with still longer ligamentous fibres. The resulting articulation between the hyoid and the suspensorium is strongest posteriorly. The articular configuration restricts the mobility of the hyoid to a, still considerable, oblique dorsoventral movement, reducing or enlarging the oral cavity. The anterior and posterior ceratohyals share a large, cartilaginous ventrocaudal process, with which four branchiostegal rays articulate (Fig. 56B). The compound parurohyal bone connects the sternohyoideus muscle to the hyoid arch. It bears two rostral articular processes, each fitting into a cavity of one hypohyal (Fig. 56B-C). Strong ligaments keep these elements well connected. A mediodorsal ridge of the parurohyal increases the insertion surface for the sternohyoideus muscle. In Ancistrus cf. triradiatus only the second and third basibranchials are ossified (ossification of the third is inconspicuous and only present in the largest specimens). Basibranchial I can’t be distinguished, and basibranchials IV and V are fused and remain cartilaginous. Basibranchial II is connected to the parurohyal ligamentously. Hypobranchials I and II are separate from their corresponding ceratobranchials, whereas III and IV are fused to them. Hypobranchial V is reduced to such an extent that it can’t be discerned from ceratobranchial V. Only hypobranchial I is ossified. The ossified ceratobranchials I-IV are long and bar-shaped; V is flattened, has gill filaments on the anterior side only, and bears about 35 conical pharyngeal teeth. Ceratobranchial I bears an accessory process, as large as the bone itself (Fig. 56C). This process is loosely attached to the hyomandibula, and bears gill rakers on its posterior side. The process, present in many loricariids, is unique among siluriforms (Schaefer, 1987; Armbruster, 2004). All epibranchials except the fourth bear posterior uncinate processes, of which the third is the largest. Epibranchial I bears an additional small anterior process. Ossified infrapharyngobranchials III and IV are present, articulating with epibranchials II-IV, the prootic and the upper pharyngeal jaws, the latter being paired elongated dermal bones that bear about 25 conical teeth each. Infrapharyngobranchial III is a short bar, whereas IV is more square-formed. In front of infrapharyngobranchial III a minuscule cartilage nucleus can be seen (Fig. 56C). Whether this corresponds to infrapharyngobranchial I or II could not be determined. Of the opercular bones, only the opercle itself is a separate structure in Ancistrus cf. triradiatus. The suprapreopercle, a simple canal bone with a ventral flange, fuses with the sixth infraorbital in large specimens, resulting in an apparently ‘double’ canal bone (Fig. 55A). As mentioned above, the preopercle rigidly connects the quadrate and the PART 6 — ADULT MORPHOLOGY 133 hyomandibula in all examined specimens. It is an elongated, flattened canal bone overlying the lateral margin of these suspensorial bones, attaching them to the armoured skin. There is no interopercle, nor an interoperculo-mandibular ligament (there is, however, a lateral mandibulo-hyoid ligament). The opercle has a complex shape. It has a long joint with the suspensorium, consisting of a main articulation with a cartilaginous hyomandibular head and opercular socket, a rather stiff, bony articulation more posteriorly, and a tooth-like fortification of the joint more anteriorly (see also Part 8). Movements along this hinge have an effect on the ventral process of the opercle (Fig. 55A-B) that will push the large cheek spines, a set of very large odontodes, to a lateral, erect position. These spines insert on small bony platelets, which are embedded in ligamentous tissue and so articulate with the opercular process. Two cheek plates are present. The anteriormost cheek plate is the larger, and is situated more ventrally. It articulates with the spine-bearing platelets and the quadrate. The resulting cheek-spine mechanism is been described and discussed in part 8.

CRANIAL LATERAL-LINE CANALS (FIGS 54A, 55A)

The supraorbital canal runs from the nasal to the frontal. Pores are present before and after the nasal. The epiphyseal branch of the supraorbital canal splits off medially, running to a pore between both frontals, while the parietal branch ends in a pore posterior to the frontal bone. The main canal enters the sphenotic to immediately fork into two canals: the infraorbital and the otic. These two canals are situated in the deeper part of the sphenotic. The infraorbital canal follows a medial ridge down the lateral process of the sphenotic towards the six infraorbitals, with a pore between the infraorbitals and before the first one. The otic canal goes even deeper into the cranium, and almost touches the anterior and horizontal semicircular canals, separated only by a thin layer of bone. It then enters the ventral part of the compound pterotic, where it forks into the postotic canal and the preopercular canal. The preopercular canal goes through the suprapreopercle and the preopercle, with one opening halfway along the latter bone, and one near its rostral end. Additionally, it passes through the largest cheek plate. As in all Loricariidae (and Astroblepidae) there is no mandibular canal (Schaefer, 1987). In the more dorsal part of the compound pterotic, the postotic canal splits off a very narrow, almost invisible pterotic branch [posterolateral branch of Lundberg (1975) and Schaefer (1987)], ending in a single pore when leaving the compound pterotic laterally. The postotic canal then runs uncovered on the ventral floor of the compound pterotic, passing the swimbladder ventrolaterally. It then surfaces posterior to the compound pterotic, where it continues as the lateral line canal on the trunk. Serial sections confirm the uncovered, fleshy nature of the canal on the pterotic floor, below the swimbladder capsule (Bleckmann et al., 1991; Schaefer & Aquino, 2000).

134 PART 6 — ADULT MORPHOLOGY

JAW MUSCULATURE (FIGS 61, 62, 63)

Loricariid jaw musculature is highly complex. In addition, the homology between the adductor mandibulae subdivisions as described in loricariids and A1, A2 and A3 sections has been a matter of debate (Alexander, 1965; Winterbottom, 1974; Howes, 1983a; Schaefer, 1997; Diogo, 2005). Ideally, developmental studies, and a comparative examination of several loricarioid families should be done in order to clarify all possible homologies and de novo formations. At this moment it is most appropriate to describe muscle divisions according to their position, presumed function and (eventual) previous nomenclature in loricariids. In the jaw muscle complex of Ancistrus cf. triradiatus, different muscle divisions can be discerned. The main part of the complex is the adductor mandibulae sensu stricto [muscle b of Howes (1983a)]. Two main subdivisions can be discerned. The longest, external bundle originates on the preopercle and on the lateral surface of the hyomandibula, anterior to, and on the prominent lateral hyomandibular ridge, and inserts on the high dorsal ridge of the dento-mentomeckelium and the angulo-articular of the lower jaw. The shorter, more ventral and interior part of the muscle, hidden below the external one, consists of two flat, distinct bundles (Fig. 61B) that originate on the quadrate, the hyomandibula and the metapterygoid. They insert dorsocaudally on the meckelian cartilage and the body of the angulo-articular. Experimentally pulling the adductor mandibulae results in an adduction of the mandible, but also rotates it, so that the teeth row swings anteriorly. This rotation appears to be due to the rather dorsal insertion on the lower jaw of the muscle. Medial to the dorsal bundle of the adductor mandibulae lies the retractor premaxillae [muscle c of Howes (1983a)], which also originates on the hyomandibula. Manipulation of the muscle pivots the premaxilla around its dorsal articulation. The result is analogous to that of the combined adductor mandibulae bundles on the lower jaw: the teeth scrape on the substrate where the fish is lying on. The third part of the complex is the thin and strap-like retractor veli [muscle d of Howes (1983a); Figs 61B, 62C], originating posteriorly on the metapterygoid, and running medial to the retractor premaxillae. Almost a third of the muscle is composed of the thin aponeurosis, from which the fibres diverge in the oral valve (or velum). The presence of these collagen fibres running in both halves of the oral valve was shown by the Verhoeff-Van Gieson’s stain methods. The collagen fibres intermingle at the midline where they enclose a band of elastic tissue running rostrocaudally (Fig. 60B-C), connecting the valve with the cartilage-like tissue between both premaxillae. The soft connective tissue between the autopalatine and the metapterygoid is loosely attached to the aponeurosis, but pulling the muscle has no apparent effect other than retracting the valve backward. PART 6 — ADULT MORPHOLOGY 135

The intermandibularis anterior muscle consists of two separate divisions (Fig. 63A). The dorsalmost, here called pars dentalis, connects the posterior aspects of both lower jaws, inserting in a groove along the teeth-bearing dento-mentomeckelian bones. Medially, where the contralateral halves meet in a slim raphe, they attach firmly to the inner tissue of the lower lip. The effect of this muscle is not clear, but it appears to pull the mandibles together and rotate them backward, which causes the lower lip to purse towards the substrate. The pars labialis of the intermandibularis anterior is a separate division, attached to both lateral sides of the inner lower lip tissue ventral and caudal to the pars dentalis. Laterally, the fibres run into the connective tissue posterior to the basis of the maxillary barbel. Contrary to the pars dentalis, some muscle fibres do cross the midline, and there is no raphe. Manipulation of this muscle part leads to pursing of the lower lip. There is no muscular contact between the intermandibularis anterior and the subdivisions of the intermandibularis posterior. The intermandibularis posterior muscle consists of two completely separate subdivisions having different directions and insertions (Fig. 63A). Terminology of subdivisions in literature can be confusing, as the names often refer to the relative position of these subdivisions, but these can vary from taxon to taxon. Hence names referring to the insertions are given here. A separate, band-like subdivision is the intermandibularis posterior pars dentalis [part a of Howes (1983a); lateral division of Schaefer (1997)], connecting the ventral side of the posterior ceratohyal to the ventrocaudal side of the dento-mentomeckelium. Pulling the pars dentalis retracts and rotates the mandibles around their axis and brings the teeth rows in a position ready for scraping; the movement of the mandibles automatically retracts the lower lip somewhat. The hyoid stays almost motionless. Another subdivision is the pars labialis [part b of Howes (1983a); medial division of Schaefer (1997)]. It runs between the posterior ceratohyal and the lower lip tissue. It diverges in several small bundles before reaching the lip. The effect appears to be on the lower lip only: it is flattened and retracted dorsally. Again, no significant movement of the hyoid is observed.

HYOID MUSCULATURE (FIG. 63)

The largest ventral muscle is the unpaired hyohyoideus inferior (or hyohyoides inferior), lying ventral to the hyoid arch (Fig. 63B). It is narrower at the midline, where a medial raphe connects both halves. Fibre direction is from the medial raphe towards the lateral attachment on the ventral side of the anterior and posterior ceratohyals. As suggested by manipulation, any contraction elevates the hyoid, reducing the branchial cavity. An additional, indirect effect, probably caused mainly by the caudalmost fibres, appears to be on the proximal ends of the branchiostegal rays that articulate with the hyoid arch. As a result, the branchiostegal membrane is opened slightly. 136 PART 6 — ADULT MORPHOLOGY

The antagonist of the hyohyoideus inferior is the sternohyoideus muscle, that diverges from the dorsal surface of the parurohyal to the anterior edge of the cleithrum of the pectoral girdle. The muscle is much broader than long. As the pectoral girdle is an almost immobile structure in Ancistrus cf. triradiatus, the only effect of manipulating the sternohyoideus is the retraction and depression of the hyoid arch. Although some movement of the cartilage plug anterior to the hyoid arch is observed, no effect is seen on the mandibles. The almost cylindrical plug of cartilage is connected to the ventral face of the hypohyals, projecting rostrally, between both dento-mentomeckelia. The cartilage is elastic, while no collagen is seen except for a very thin perichondral layer. No collagen fibres attach to the dento- mentomeckelian bones (Fig. 63D-E). The two medialmost branchiostegal rays are kept close to each other by means of short ligamentous fibres. The second and the third, and the third and the fourth, respectively, are interconnected by short muscles, running from the lateral edge of the medial to the dorsal surface of the more lateral ray. From the ventral face of the fourth, lateralmost and broadest ray, an additional, narrowing muscle attaches to the medial aspect of the opercle. These three different muscles are collectively referred to as the hyohyoidei adductores (Figs 62C, 63B). They force the branchiostegal rays together and towards the opercle which results in an adduction of the branchiostegal membrane and, hence, the closure of the gill opening. The first, medial branchiostegal ray provides insertion for the hyohyoideus abductor, a flat and straight transverse muscle projecting towards the midline, which it does not reach (Fig. 63A). At four fifths of its length it continues as a thin aponeurosis and connects to the ventral fascia of the sternohyoideus, thus only forming indirect contact with its counterpart. Pulling one or both abductors unambiguously closes the branchiostegal membrane.

SUSPENSORIAL MUSCULATURE (FIGS 61, 62, 63)

The levator arcus palatini is a minute muscle running obliquely from the sphenotic (medial to the emergence of the preopercular canal) to the hyomandibula, where it inserts on the posterior side of the ridge that also forms the attachment site for the adductor mandibulae. Attachment is muscular and relatively broad on the hyomandibula and tendinous and slim on the sphenotic (Figs 61B, 62B). These observations contradict Howes’ (1983a) statement that the muscle joins the dilatator operculi in Ancistrus. The muscle is too small to allow manipulation, but probably can only slightly elevate the well-attached suspensorium. A prominent adductor arcus palatini connects the medial face of the suspensorium (hyomandibula and metapterygoid) with the base of the neurocranium. Origin is on the ventral ridge of the parasphenoid as well as a large surface of the orbitosphenoid. Manipulation of the muscle brings the suspensorium to a more medial position. PART 6 — ADULT MORPHOLOGY 137

In siluriforms the extensor tentaculi is a muscle derived from the adductor arcus palatini and the antagonist of the levator tentaculi (Winterbottom, 1974). In Ancistrus cf. triradiatus, as in other loricariids (Howes, 1983a; Schaefer, 1997), the extensor tentaculi is divided in two completely separate segments, which I refer to as the pars lateralis and pars dentalis (Fig. 63C). The extensor tentaculi pars lateralis [muscle e of Howes (1983a)] inserts on the ventrolateral process of the autopalatine. The muscle is circular in transsection and has an anterior aponeurosis continuing halfway inside the muscle itself. Fibres radiate from this aponeurosis. Insertion space for this ‘circularly pinnate’ muscle is provided by a bony canal formed by the ventral face of the lateral ethmoid and a groove formed by the lateral face and a lateral process of the metapterygoid. Pulling the muscle in fresh specimens results in a lateroventral swinging of the autopalatine and a corresponding ventral movement of the tip of the maxilla, as both are coupled through their articulation and the mesethmo-maxillary ligament. The maxillary barbel and the lateral parts of the lips are consequently pushed against the substrate. The second segment, the extensor tentaculi pars medialis [muscle f of Howes (1983a)], is a more flattened muscle connecting the longer ventromedial process of the autopalatine with the lateral ethmoid, the orbitosphenoid and the lateral side of the ventral parasphenoid ridge. The tendon inserting on the autopalatine continues as a ventral aponeurosis to which the slightly dorsally oriented fibres attach. The effect of pulling the muscle is a ventral movement of the autopalatine. The autopalatine-maxillary articulation pushes the caudal edge of the premaxilla downward, pivoting the premaxilla around its ligamentous articulation with the mesethmoid. Due to the configuration of the articulation, effect on the distal tip of the maxilla itself is negligible. The abduction of the premaxilla brings the teeth in a position ready for scraping. The extensor tentaculi pars medialis can thus be considered the antagonist of the retractor premaxillae. The dorsal levator tentaculi [muscle a of Howes (1983a); Figs 61A, 62A] overlies the adductor mandibulae musculature. It is a broad band-like muscle, originating from an anterior ridge on the ventral side of the lateral ethmoid. Two thirds of the muscle fibres run straight to the posterior face of the maxilla; the other third, comprising the lateroventral fibres, runs in a slightly more lateral direction, inserting closer to the distal tip of the bone. Contraction of the muscle appears to pull the tip of the maxilla in a caudodorsal direction. Due to the position of the bone and the maxillary barbel cartilage (where upper and lower lip meet) the lateral lip tissues are retracted towards the dorsal. The small and extremely thin adductor hyomandibulae originates from the ventral floor of the compound pterotic (together with the adductor operculi), and loosely inserts on connective tissue at the medial side of the hyomandibula (Fig. 63C). Given its size, it is not easy to manipulate it, nor to even speculate about its function. The name adductor 138 PART 6 — ADULT MORPHOLOGY hyomandibulae has been used for the adductor arcus palatini in loricariids (Howes, 1983a; Schaefer, 1997). This is erroneous, as both muscles are not homologous (Winterbottom, 1974; Diogo & Vandewalle, 2003a).

OPERCULAR MUSCULATURE (FIGS 61, 62, 63)

The largest opercular muscle is the dilatator operculi, which has its origin on a large surface on the posterior part of the neurocranium and inserts via a thick aponeurosis on the dorsal side of the opercle, lateral to the articulation with the hyomandibula (Figs 61B, 62C). Its different bundles originate mainly on the compound pterotic, parieto-supraoccipital, sphenotic and the posterior margin of the hyomandibula. Due to the configuration of the hyomandibula-opercular articulation, any force exerted via the aponeurosis rotates the opercle dorsally, erecting the long cheek spines. In large adults some bones of the skull roof and walls show expanded ventral or medial laminae separating the dilatator operculi from the braincase. Immediately posterior to the insertion of the dilatator operculi on the opercle, the short and sturdy aponeurosis of the levator operculi inserts. It has a broad origin on the compound pterotic. The effect of pulling this muscle is identical to that of the dilatator operculi. Experimental tension on both muscles indicates no role in the opening of the branchiostegal membrane. The antagonist of these muscles is the adductor operculi (Fig. 63C). It has its tendinous origin on the ventral transverse rim of the compound pterotic contacting the Baudelot’s ligament. It inserts muscularly on the medial side of the opercle just posterior to the insertion of the lateralmost division of the hyohyoidei adductores.

OTHER CRANIAL MUSCLES

The musculature of the branchial basket is not discussed here. See Schaefer (1997) for a description of the branchial muscles in Otocinclus and Fernandes et al. (1995) for a short account on the gill filament muscles in Hypostomus and Rhinelepis. Extrinsic eye musculature is very similar to Otocinclus as well (Schaefer, 1997), including the presence of a posterior myodome. The external and internal rectus muscles penetrate into it, while the inferior and superior rectus muscles do enter the skull, but not the myodome itself. In Ancistrus cf. triradiatus, the obliquus muscles originate from the anterior myodome, a minute cavity in the posterior lateral wall of the lateral ethmoid.

PART 6 — ADULT MORPHOLOGY 139

LIGAMENTS IN THE ROSTRAL REGION

The mesethmoid, lacking the lateral cornua typical for most siluriforms but bearing a ventral disc-like process, influences the mobility of the upper jaw. The mesethmoid- premaxillary cartilage or meniscus rests on the ventral edge of the mesethmoid disc like a small cap and forms two articulation cups for the premaxillae (Fig. 63C). It is V-shaped in transsection (Fig. 60A). Strong connective tissues, and ligaments in the rostral region ensure the relative position of these elements. The mesethmoid-cartilage ligaments connect the rostral tip of the mesethmoid with both posterodorsal sides of the mesethmoid-premaxillary cartilage, running along both sides of the mesethmoid disc (Fig. 61B). The lateral mesethmoid-premaxillary ligaments attach to the anterodorsal side of both premaxillae and the rostral end of the mesethmoid, thus running more or less horizontally (Fig. 61A). The medial mesethmoid-premaxillary ligaments run vertically from the anterior face of the premaxillae to the mesethmoid disc right above it (Fig. 61B). The paired mesethmoid- maxillary ligament, as mentioned above, connects the maxilla, close to its ventral head, to the mesethmoid, immediately behind the other ligament attachments (Figs 61A, 62A). The short interpremaxillary ligament is an unpaired ligament, running transversely behind the mesethmoid-premaxillary cartilage. It keeps both premaxillae closely apposed to each other, restricting their relative movement. The mesethmoid-autopalatine ligament is a broad ligament between the mesethmoid shaft and the autopalatine; the anteriormost fibres are longest, and also contact the autopalatine-maxillary articular tissue (Figs 61A, 62A). The paired rostromaxillary and ventral labial ligaments as seen by Schaefer (1997) in Otocinclus vittatus are not observed in Ancistrus cf. triradiatus.

KINEMATICS (FIGS 64, 65)

Respiratory cycles while sucking onto a vertical glass wall took approximately 250 to 400 ms (with a water temperature of 23° C). The fishes preferred to support themselves with their tail on the bottom of the aquarium. In a first phase the oral cavity expands by a depression of the hyoid (and lower jaw region), and a slight abduction of the suspensoria (Figs 64, 65A). During this expansion phase the oral valve is open, and the branchiostegal membrane is closed (Fig. 65A, E). At maximal expansion the oral valve is closing. Complete closure is observed only well after the onset of elevation of the hyoid and lower jaw region, and adduction of the suspensoria (Fig. 65F). Movements of these latter elements is synchronous. At maximal constriction of the oral cavity the oral valve is still closed; it bulges out ventrally, probably due to the water pressure inside the mouth. It starts to open again only when a new expansion phase starts. During constriction the branchiostegal membrane opens, allowing water to flow out (Fig. 65B, F). This water flow could be visualized by the use of diluted 140 PART 6 — ADULT MORPHOLOGY milk. At maximal constriction the branchiostegal membrane starts to close again (Fig. 65C). The same milk experiment also shows the inflow of water through the narrow openings created in the lateral lip tissue by an elevation of the maxillary barbel. The opening and closure of this inflow opening coincides with the same movements of the oral valve (Fig. 65D). Sometimes only one of both sides of the lips is opening. During the whole cycle, opercular movements are restricted to a very small in- and outward movement, synchronous with the suspensorial movements. Respiratory cycles accelerate during feeding (Fig. 64). In addition, extensive rotation of both lower and upper jaws occurs. During the constriction phase the mandibles not only depress, but also move posteriorly. An additional rotation of the jaws along their axis brings the teeth rows even further posteriorly. At the same time, the premaxillae protrude anteriorly by a rotation around their hinge to the mesethmoid disc (Fig. 65G). During the subsequent expansion phase the mandibles move and rotate so that the teeth row scrapes the glass surface in an anterior direction. An analogous action occurs in the premaxillae: the teeth row scrapes the surface in a posterior direction. At peak expansion the teeth of lower and upper jaws almost touch each other (Fig. 65H). The rough, papillose surface of the large lower lip moves posteriorly and anteriorly together with the lower jaw. Fishes often proceed somewhat anteriorly during each cycle, so that a fresh substrate can be fed on. Opening of the lateral lip tissue is less pronounced and difficult to observe during feeding. The extensive jaw movements might well allow a sufficient inflow of water. Often (but not always) body and tail movements accompany the feeding sequences.

6.1.4. DISCUSSION

Anatomical observations, aided by the manipulation of muscles in fresh specimens, yielded some remarkable morphological considerations. Together with the limited kinematic data set functional hypotheses could sometimes be formulated. Here I group the morphological structures and kinematic facts in six known functional mechanisms present in loricariids, providing a comprehensive overview of elements, movements, and possible functional-morphological links. The morphological basis of some typical loricariid structures or systems is also discussed and compared to non-loricariids.

THE BUCCAL PUMP SYSTEM

In Ancistrus cf. triradiatus, oral expansion has a double function: inflow of water and food, and the maintenance of the suction power when attaching to a substrate. In fishes with a depressed head, the hyoid depression generally has a greater share in the buccal pump system PART 6 — ADULT MORPHOLOGY 141 than the suspensorium abduction (Alexander, 1970; Gosline, 1973; Adriaens & Verraes, 1997b). The sternohyoideus and hyohyoideus inferior muscles, respectively depressing and elevating the hyoid arch in many fishes (e.g., Ballintijn & Hughes, 1965), are among the largest and most vascularized muscles in A. cf. triradiatus, indicating their importance in this system. The articulation between the hyoid arch and the suspensorium is long and strong, compared to several other teleosts (Anker, 1974; Aerts, 1991; Hunt von Herbing et al., 1996a). The large suspensorium is rigid (the metapterygoid is connected to the other bones with sutures), and anchored to the skull via the posterior hyomandibula-pterotic suture and the dorsal connection of both the metapterygoid and the hyomandibula to the neurocranium. It is also connected with the tough, armoured skin at the level of the preopercle, even more reducing its mobility. These connections can all be considered reinforcements of the hyoid- suspensorium system [some are characteristic for loricariids or the more apomorphic loricariid taxa (Schaefer,1990; Armbruster, 2004)], and might be adaptations to the high forces probably generated during suction while attaching to a substrate. The ossified Baudelot’s ligaments, vertical ridges between the basioccipital and the supracleithral part of the compound pterotic, form an ossified posterior wall of the branchial cavity, which Alexander (1965) regarded as an adaptation to withstand the negative pressure caused by the suction, hydrostatically isolating the orobranchial cavity from the intestinal cavity. The relative resilience of the suspensorium and the strengthening of the suspensorial- neurocranial connection could be interpreted in a similar way: a weaker and more movable suspensorium would be susceptible to collapsing when extreme suction is needed. Figures 61B and 63C clearly show that the masses of the two muscles inserting on the suspensorium are far from equal: the adductor arcus palatini is large, while the levator arcus palatini is almost rudimentary. It can be hypothesized that the relatively large adductor arcus palatini can cause a slight adduction of the suspensorium, and that elevation is mainly a result of the return of the suspensorium to its original position. My manipulations during dissection, and the absence of the levator arcus palatini in some loricariids (Alexander, 1965; Howes, 1983a) corroborate this hypothesis. During dissection, elastic properties of the skin and the ligamentous tissue in the hyomandibula-pterotic suture appeared to be the main cause of such passive elevation, rather than the bending of the bony suspensorium itself, which was suggested by Alexander (1965). This also explains the presence of a relatively robust adductor arcus palatini, needed to overcome the resilience of the suspensorium. Whether the buccal pump system is able to maintain a negative pressure in the oral cavity has since long been a matter of debate. Hora (1930) believed that the lips could not function as a sucker while respiration continued, since the inflowing water would cause the system to fail. Alexander (1965) demonstrated that respiration and suction can function simultaneously, and that both actions continue when the fish is pulled away from the substrate (a vertical 142 PART 6 — ADULT MORPHOLOGY aquarium glass). My results indicate that inflowing water was limited to a thin stream passing under the sucker immediately posterior to each maxillary barbel, a phenomenon also observed by Vandewalle et al. (1986) in Hypostomus punctatus.

THE BRANCHIOSTEGAL MEMBRANE

Water leaves the orobranchial cavity through the branchiostegal opening. It does not enter there, as initially supposed by Regan (1904), but is inhaled via the mouth. The functioning of the muscles associated with the branchiostegal membrane in Ancistrus cf. triradiatus can not easily be understood. As indicated by manipulation, contraction of the hyohyoidei adductores closes the branchiostegal membrane. Schaefer (1997) did not mention this muscle in his myological account on Otocinclus, but it is present in O. vestitus (Chapter 6.3). In non-loricariid siluriforms like Clarias gariepinus the paired hyohyoideus abductor opens the branchiostegal membrane, inserting on the medialmost branchiostegal ray and originating directly on the hyoid arch, or indirectly, via a medial aponeurosis, thus running more or less rostrocaudally (Adriaens & Verraes, 1997d). Remarkably, in Ancistrus cf. triradiatus the muscle runs completely transversely (Fig. 63A), so that experimental contraction closes the membrane instead of opening it, the muscle having become a functional adductor. The angle between the hyohyoideus abductor and the fibres of the sternohyoideus on which they attach is 90°; hence no interaction of functions is assumed. Also, due to its shortness, the sternohyoideus can’t induce a significant change in orientation of the hyohyoideus abductor. Two possible mechanisms might cause the abduction of the membrane in A. cf. triradiatus: it might have become a passive movement, induced by the high pressure in the branchial cavity caused by the contraction of the hyohyoideus inferior, or it might be initiated by the posterior part of the latter muscle, attaching to the posterior ceratohyal, near the articulation of the first branchiostegal ray. Manipulation suggested that movements of the hyoid have a slight effect on the position of the rays. Schaefer (1990) observed that in some loricariids, this part of the hyohyoideus inferior inserts on the branchiostegal rays themselves. My anatomical study and observations suggest the movements of the opercle in the respiratory cycle are negligible. It is moved very slightly by the movements of the suspensorium, with which it articulates. In Ancistrus cf. triradiatus, it has a prominent role in the erection of the large cheek spines (Howes, 1983a; Part 8).

MOVEMENTS OF MAXILLARY BARBELS AND LIPS

Initiation of suction requires a close contact of the upper and lower lips to the substrate. In the most plausible theoretical scenario the outer edges of the suction disc, formed by the PART 6 — ADULT MORPHOLOGY 143 fused lips, are pushed against the substrate. A subsequent creation of a negative pressure by expansion of the oral cavity (posterior to the valve) enables transferring water from the cavity anterior to the oral valve. A fish adopting a sucking position brings the less mobile upper lip against the substrate by literally swimming against it or pushing the head downward. Pulling both parts of the intermandibularis anterior muscle appears to result in a pursing of the lower lip. Antagonistic movement, i.e. retracting the lower lip towards the body, is most probably achieved by contraction of the intermandibularis posterior pars labialis, as suggested by manipulations (see also section on the lower jaws). The maxillae of most loricariids support only small maxillary barbels. It appears that their main function has become to mediate the movements of the lateral lip tissue in which they are embedded. The levator tentaculi muscle lifts this part of the lips from the substrate, as in many siluriforms (Winterbottom, 1974), allowing water to enter the oral cavity. Experiments with diluted milk showed that when the fish clings to a vertical substrate water often enters only one side of the mouth, and only this side of the lip is seen moving. This suggests the independent contraction of the left or right levator tentaculi. Vandewalle et al. (1986) hypothesized that the elevation of the lateral lip tissue is mostly caused by lateral movements of the lower jaws. I found, however, no anatomical link connecting both elements. The lateral motion of the lower jaws is negligible when compared to their dorsoventral movements during respiration. Also, when only one of both lip sides is elevated, no visible asymmetry in the movements of the lower jaws is present. The closing of the lip furrow by the action of the levator arcus palatini, suggested by the same authors, is highly improbable, not only because of the absence of an anatomical link, but also because the levator arcus palatini is absent in several loricariids (Howes, 1983a). My hypothesis, based on the anatomy of Ancistrus cf. triradiatus, that the extensor tentaculi pars lateralis may close the lip opening, should ideally be tested by electromyographical experiments. The extensor tentaculi [not subdivided in catfishes other than loricariids (Schaefer 1990; Diogo, 2005)] is responsible for a downward or forward movement of the maxillary barbel in most catfishes (Alexander, 1965; Gosline, 1975). The origin of the pars lateralis in the lateral ethmoid-metapterygoid groove is unique among siluriforms, and might have evolved several times within the loricariid family (Howes, 1983a). The autopalatine-maxillary mechanism of Ancistrus cf. triradiatus and other loricariids is comparable to the situation seen in the African mochokid Euchilichthys, also having a sucker-like mouth. Here too, the cartilaginous rostral tip of the autopalatine is situated on top of the premaxilla, the extensor tentaculi muscle is directed completely rostrocaudal, and the line between the lateral ethmoid-autopalatine joint and the muscle insertion on the autopalatine is almost vertical instead of horizontal (Gosline, 1975).

144 PART 6 — ADULT MORPHOLOGY

THE ORAL VALVE

The oral valve of loricariids attaches to the anterodorsal mouth roof, and, when closed, separates a small anterior cavity from the larger oral cavity. The muscle I name retractor veli has previously been called muscle d or retractor palatini (Howes, 1983a; Schaefer, 1997; Diogo & Vandewalle, 2003a). It is, however, not homologous to the retractor palatini of Lubosch (1929, in Winterbottom, 1974) and Hofer (1938), which is a derivative of the adductor arcus palatini in some perciform and tetraodontiform fishes. Contrary to Howes’ (1983a) claim, Alexander (1965) did not name it retractor premaxillae, but omitted it. The muscle is unique for loricariids (Schaefer, 1990). Its function has at best been vaguely described, based on dissection of preserved specimens. It would insert on a ‘complexly divided connective tissue sheet’ (Howes, 1983a:313), attached to the autopalatine, the premaxilla, and the lower jaw, and would pull the autopalatine ventrally in preserved specimens. However, my manipulation of the muscle in fresh specimens of Ancistrus cf. triradiatus revealed no ventral movement of the autopalatine. Gradwell (1971b) described a ‘muscle of oral valve’ in several loricariids: Hypostomus punctatus, Hemiancistrus annectens, Rineloricaria microlepidogaster, Ancistrus occidentalis and Otocinclus mariae. It would insert on the lateral sides of the oral valve, and originate ‘on the dorsal surface of the skull, anterior to the eyes and lateral to the nares’ (Gradwell, 1971b:837). His biomechanical analysis proved it to contract at the moment the valve closed. In spite of the questionable description of the muscle origin mentioned in the his text, his figures corroborate my assumption that it is the retractor veli. The retractor palatini of Howes (1983a) and the muscle of the oral valve of Gradwell (1971b) are presented as separate muscles in the myological review of Diogo & Vandewalle (2003a), but are identical. Apart from Ancistrus cf. triradiatus, I also observed this muscle in Pterygoplichthys lituratus, Farlowella acus and Otocinclus vestitus. In F. acus it inserts on the valve via two tendons, while in O. vestitus two separate retractor veli divisions are present, only connected posteriorly. The dorsal division inserts more anteriorly on the valve. In breathing specimens of A. cf. triradiatus it can be observed that the valve is closed at the moment the hyoid elevates (Fig. 64). The exact movements of the valve, both rostrocaudally and dorsoventrally, are difficult to explain, as retractor veli activity, elastic recoil and water pressure and flow probably have a combined effect.

THE UPPER JAWS

The high mobility of the premaxillae must be one of the most important evolutionary innovations encountered in the loricarioid lineage, and is a synapomorphy for callichthyids, scoloplacids, astroblepids and loricariids (Howes, 1983a; Schaefer, 1990; Diogo, 2005). PART 6 — ADULT MORPHOLOGY 145

Schaefer & Lauder (1986, 1996) listed this release of constraint as one of the decoupling events observed in the evolution leading to the loricariids. In most siluriforms the premaxillae are firmly attached to the lateral cornua of the mesethmoid; structural and functional changes of the premaxilla and maxilla have led to the inability to perform protrusion (Adriaens & Verraes, 1997b). However, the development of a small dorsal premaxillary process and the disappearance of the mesethmoid cornua have triggered an important shift in the relation between the neurocranium and the upper jaw, enabling a novel protrusion mechanism. My results suggest this mechanism is most probably mediated by the extensor tentaculi pars medialis and the retractor premaxillae in loricariids. The latter muscle, derived from the adductor mandibulae complex and constituting a key innovation in the loricarioid lineage (Schaefer & Lauder, 1986), has a direct insertion on the premaxilla in astroblepids and scoloplacids as well, and an indirect connection via a connective sheet in callichthyids (Schaefer, 1990; Schaefer & Lauder, 1986, 1996), which also have a dorsal premaxillary process (Huysentruyt & Adriaens, 2005b), and show some degree of premaxilla protrusion (Alexander, 1965). The vertical ventral mesethmoid disc, present in all loricariids, and, to a lesser degree, astroblepids (Schaefer, 1990), is important for the movements of the premaxillae. The mesethmoid-premaxillary cartilage, a meniscus forming a double articulation cup, keeps the premaxillae in place, aided by the series of ligaments that assist and direct their movements (Fig. 60A). Consequently, contraction of the retractor premaxillae swings the premaxilla teeth row caudally, thus scraping the substrate. The effect of the assumed antagonistic extensor tentaculi pars medialis is more complex: the muscle unambiguously pulls the rostral tip of the autopalatine ventrally. When manipulating the autopalatine in this direction, the autopalatine- maxilla joint pushes against the posterior part of the premaxilla. Both the premaxilla and the maxilla (the part ventral to the joint) are provided with a strong tissue cushion in the region where they touch (Fig. 60A). The ligamentous suspension of the premaxilla makes it rotate about a transverse horizontal axis instead of merely being pushed forward. As a result the premaxilla protrudes, and swings its teeth row rostrally. Dissection and manipulation of this apparatus showed no significant effect on the movement of the tip of the maxilla. Independent motion of each premaxilla seems to be limited by the short interpremaxillary ligament.

THE LOWER JAWS

The evolutionary rotation of the lower jaws to a medial position with the teeth pointing rostroventrally, as well as the loss of both the interoperculo-mandibular ligament and the medial connection between both dento-mentomeckelian bones, have laid open new possibilities concerning rotational mobility (Schaefer & Lauder, 1986). During normal 146 PART 6 — ADULT MORPHOLOGY respiration in Ancistrus cf. triradiatus a slight up-and-down motion of the lower jaws and the adjacent lower lip tissue is seen, probably anatomically and functionally coupled to the hyoid movements. Whether this movement is (partly) caused by the adductor mandibulae muscle remains to be verified. Only at feeding the lower jaws are seen rotating and scraping the substrate, more or less synchronously with the upper jaws. In the most probable hypothetical scenario the adductor mandibulae and intermandibularis posterior muscles act as antagonists. The scraping movement, in which the teeth are moved rostrally, would be achieved by a contraction of the adductor mandibulae (dissection and manipulation confirmed that the most dorsal part certainly has to be involved). The intermandibularis posterior pars dentalis may then perform the antagonistic movement of swinging the lower jaw and its teeth row back caudally. The intermandibularis posterior pars labialis can be considered the retractor of the lower lip. The cartilage plug, attached to the hyoid arch at the midline, and protruding into the space behind and between the lower jaws, is hypothesized to be a supporting device for the dento- mentomeckelian bones, preventing them from being merely pulled caudally. Their caudal motion is restricted, and the effect of contraction of the adductor mandibulae, inserting on the dorsocaudal aspect of the jaws, is partly transformed in a rotation around the longitudinal axis of the jaws. The consequence is that the teeth can scrape a larger surface. Previously, Schaefer & Lauder (1986, 1996) appointed a different function of the cartilage plug: it was suggested to act as a novel anatomical link between the hyoid and the lower jaws, unique to loricariids, allowing the sternohyoideus muscle to retract the lower jaws via the hyoid arch. This is contradicted by the present study on Ancistrus cf. triradiatus, as the plug attachment to the hyoid arch is relatively strong, while it is not strongly attached to the lower jaws (Figs 60D-E, 63B). The dento-mentomeckelian bones move and roll against the plug, but are not attached to it. Hence, any force exerted by the sternohyoideus retracts the hyoid arch, and, to a lesser extent, the attached cartilage plug, but has no significant effect on the mandibles. This was shown by manipulation of freshly killed and dissected specimens (tissue characteristics are strongly altered in preserved specimens). Another argument is histological: the cartilage plug connected to the hypohyals consists of elastic cartilage, ideal for a supporting, gliding device, but inappropriate for the efficient transmission of pulling forces in the caudal direction, a mechanism that would benefit more from a tendinous ligament. The mandibulo-hyoid ligament attaches to the lateralmost aspect of the angulo-articular, and its role in retraction of the medially pointed lower jaw seems very slim. The lower lip is moved rostrally and caudally together with the lower jaws. Unicellular keratinized projections or unculi form numerous brushes on thick epidermal papillae, that may aid during substrate scraping (Fig. 60D). These have been observed in other loricariids as well (Ono, 1980; Roberts, 1982). PART 6 — ADULT MORPHOLOGY 147

The remarkable habitus of loricariids evoked Gregory (1933:196) to state that “in these heavily armoured forms the siluroid skull attains its highest specialization.” Considering the results of the present dissertation, in addition to the works of Alexander (1965), Howes (1983a), Schaefer (1987, 1997), Schaefer & Lauder (1986; 1996) and others, it can be concluded that this has not been an idle statement. On the contrary, the loricariid head is one of the most impressive examples of structural diversification and refinement shaped by evolution.

148 PART 6 — ADULT MORPHOLOGY

6.2. FARLOWELLA ACUS

6.2.1. INTRODUCTION

Whereas the hypostomine genera Ancistrus, Pterygoplichthys and Panaque can be considered rather stocky and stoutly built loricariids, the genus Farlowella and other loricariines are characterized by a very long and slender body. They represent the other end of the wide spectrum of body shapes within the family, and are an ideal choice if one wants to explore the variety in loricariid internal head morphology. The subfamily Loricariinae is the sistergroup of the Hypostominae (Schaefer, 1987; Armbruster, 2004), and comprises 31 genera with about 209 species (Ferraris et al., 2003). Loricariinae all have an elongated body, a flattened caudal peduncle, and often extremely long dorsal and ventral procurrent caudal fin rays. The monophyly of the subfamily is well established (Schaefer, 1987), although less is known about the relationships within it. Typical for the subfamily is the presence of a more complete bony coverage of the head and the body. The 25 recognized species of Farlowella are found in a large area of South America. Retzer & Page (1996) formulated a hypothesis of relationships between them, based on external morphological features. Farlowella is remarkable as it has an elongated rostrum, sometimes almost as long as the rest of the head itself. These so-called stick catfishes hide themselves by means of camouflage; they spend most of the daytime attached to leafs, twigs or other substrates. The rostrum, as well as their pigmentation, may help in veiling the body shape: both morphologically and behaviourally they mimic sticks. A few other loricariine genera show the same trend, but to a far lesser extent. Farlowella acus Eigenmann & Eigenmann is native to the region of the Lago Valencia and the Río Torito in Venezuela. The largest reported standard length is 14.5 cm (Retzer & Page, 1996). Males have a broader rostrum, with enlarged ‘breeding’ odontodes projecting laterally.

6.2.2. BRIEF MATERIAL AND METHODS

For examined specimens, see Table II. For clearing and staining, dissections, and serial sectioning procedures, see paragraphs 2.2.5, 2.2.6 and 2.2.7, respectively.

PART 6 — ADULT MORPHOLOGY 149

6.2.3. RESULTS

OSTEOLOGY (FIGS 66, 67, 68, 69)

The mesethmoid has excessively elongated rostrally, and supports the long rostrum (Fig. 66). It is the only skull bone involved in the rostrum, and its remarkable length is due to an elongation of the anteriormost tip of the bone, as shown by the position of the ventral process, in the posterior half of the bone (Fig. 66D). The process is not as ‘disc-like’ as in Ancistrus cf. triradiatus and Otocinclus vestitus, and is situated between both premaxillae, as is the case in all loricariids. Only the dorsal surface of the posterior tip of the mesethmoid is visible externally (Fig. 67A); the remainder is covered by dermal platelets (see below). Like many of the skull bones, the lateral ethmoid is narrower than that in A. cf. triradiatus. The ventral ridge to which the metapterygoid attaches is remarkably high. The nasal sac is completely supported by the lateral ethmoid. The nasal bone is square-formed, and partially covers the mesethmoid; the supraorbital canal is situated in the lateral part of the bone. The frontal is small: it doesn’t reach halfway the nasal sac, and doesn’t participate in the bordering of the orbit (Fig. 67B). The position of the supraorbital canal and its epiphysial and parietal branches is similar to that in A. cf. triradiatus. Probably related to the smaller size of the frontal, the sphenotic has somewhat extended anteriorly, touching the prefrontal plate and bordering more of the orbit. The lateral wings of the prevomer are extremely reduced (Fig. 68B). The parasphenoid wings are short as well. Compared to A. cf. triradiatus, no significant difference is observed in the shape of the orbitosphenoid and pterosphenoid. The prootic lacks the anterior semicircular arch. The deep interdigitation between the parasphenoid and the basioccipital is situated relatively caudally, making the parasphenoid somewhat longer and the basioccipital somewhat shorter. No relevant size or shape differences (when compared to Ancistrus cf. triradiatus) can be seen in the exoccipital and epioccipital. Both the basioccipital and the supracleithral parts of the Baudelot’s ligament are present, and are deeper than in A. cf. triradiatus. The compound pterotic takes part in the encapsulation of the swimbladder, and has the typical double-layered nature as described for A. cf. triradiatus in chapter 6.1 and part 8. The dorsal sheet has extended considerably posterior to the epioccipital (and thus braincase). Anteriorly, around the brain, the dorsal and ventral layers are closer to each other than in A. cf. triradiatus (Chapter 6.1 and Part 8). The pterotic fenestrae are very small but numerous, and occupy the lateral half of the bone surface (not drawn on figures). The posterior process of the parieto-supraoccipital is straight instead of pointed. Analogous to the compound pterotic, this bone has extended very much posteriorly. The posteriormost part of the parieto-supraoccipital is covered by the dermal body plates. 150 PART 6 — ADULT MORPHOLOGY

Except for the compound pterotic, no skull bone shows the double-layered structure as seen in adult Ancistrus cf. triradiatus (Part 8). The complex vertebra and the associated transverse process are closely appressed to the skull. Dorsal contact of the transverse process and dorsal vertebral processes with the elongated compound pterotic and parieto- supraoccipital is as intense as in A. cf. triradiatus. The skull of Farlowella acus is completely covered by a well fitting armour of dermal plates (Figs 67A, 68A, 69A). There are six plate-like infraorbital bones. The first three are longer than the others, so that the infraorbital canal is very long when compared to the other examined loricariid species. Whereas the first infraorbital in the other loricariids is situated at the level of the lower jaw (e.g., Fig. 55A), it reaches far beyond the upper jaw in F. acus (Fig. 69A). Thus the bone takes part in the bony covering of the rostrum. The other dermal plates covering the head can be divided in various groups; these are generally more numerous than in Ancistrus cf. triradiatus. An average of seven prenasal plates are situated dorsally, covering most of the mesethmoid between the infraorbital series. The rostrum itself is covered by a high number of irregular rostral plates. On the (male) specimen illustrated in figures 66 to 69 a total of 75 rostral plates was counted. The number is somewhat lower in females, that have a narrower rostrum. The lateral odontodes on the male’s rostrum are enlarged. The paired prefrontal plate appears to be fused somewhat to the lateral ethmoid. Three oral plates border the mouth. Approximately 12 lateral plates are present between the oral and infraorbital plates, and around 23 ventral postoral plates are counted posterior to the mouth. There is no sharp distinction between these plates and the bony armour of the trunk region. A single large cheek plate is present at each side, between the third oral plate and the postoral plates, bordering the mouth laterocaudally. It bears the end of the preopercular canal (see below). All these dermal plates fit well together. About 20 loose platelets are present in the skin of the mouth, anterior to the upper lip. Between the mouth and the second and third oral plates a paired cavity is seen, penetrating amazingly deep in the head, and extending up to the level of the hyoid arch (Figs 70-71; see discussion for details). These cavities, opening lateral to the mouth, are situated between the suspensorium (and attached muscles) and the dermal armour. The name paroral cavity is introduced for these cavities (see discussion).

The autopalatine is slightly curved, with a convex medial side and a concave lateral side. Its two posterior processes are well developed. A small splint-like sesamoid bone is attached to the autopalatine anterolaterally, extending a bit along the rostral cartilage head of the autopalatine (Fig. 69B). The tip of the maxilla is slightly curved backward. The premaxilla is narrower relative to that of Ancistrus cf. triradiatus, and bears about 30 teeth. The same observation is made in the lower jaw; a comparable number of teeth is present on the dentary. PART 6 — ADULT MORPHOLOGY 151

The bony contact between the dento-mentomeckelium and the angulo-articular is weaker, while the remnant of Meckel’s cartilage is more substantial than in A. cf. triradiatus. In the suspensorium, the quadrate is slender and almost perfectly triangulate. The hyomandibula articulates with the neurocranium at the joint spread along the prootic, sphenotic and compound pterotic. Caudally it is not as broad as in Ancistrus cf. triradiatus, a feature that influences the position and orientation of the opercular articulation (see Part 8 for details on the opercle). The long suture between the hyomandibula and the compound pterotic, observed in A. cf. triradiatus, is short in Farlowella acus. The metapterygoid is not so deep; it articulates with a high ventral ridge of the lateral ethmoid, the deepness of which compensates for the less deep metapterygoid, so that ample insertion space for the lateral part of the extensor tentaculi muscle is still available (Fig. 71B). The lateral ridge of the metapterygoid is well developed, and is positioned relatively high on the bone (due to its lesser deepness). Thus the lateral ethmoid has a relatively great contribution to the so-called metapterygoid channel. The hyoid arch is very similar to that of Ancistrus cf. triradiatus, except for some subtle shape differences: the rostral portion of the anterior ceratohyal is somewhat larger. A sesamoid bone is present in the ligamentous joint between the hyoid arch and the suspensorium (not drawn on figures). An interhyal is absent. The parurohyal is fairly round, carrying a medial dorsal ridge and a foramen that accommodates the inferior jugular vene. Next to basibranchial II, also basibranchial III is ossified. A separate basibranchial I is absent, and IV and V are cartilaginous and fused (the so-called posterior copula). The first hypobranchial is separate and ossified; all the others are fused to the ceratobranchials. An anterior process is present on the first ceratobranchial, though slightly shorter than in Ancistrus cf. triradiatus (Fig. 69C). The first three epibranchials bear small uncinate processes. Only infrapharyngobranchials III and IV are present. The upper pharyngeal jaws are well developed, bearing more than 50 small pointed teeth. Also the lower pharyngeal jaws, borne by the fifth ceratobranchials, are large structures, bearing almost 40 teeth each. The opercle is short and little movable; it lacks an anterior process. The preopercle, well fused to the quadrate and the hyomandibula, bears a canal only in its posterior half. It is S- shaped, following the ventral margin of the suspensorium. The suprapreopercle is almost indinguishably fused to the sixth infraorbital (Fig. 69A).

CRANIAL LATERAL LINE CANALS (FIGS 67A, 68A, 69A)

The cranial lateral line has to the general loricariid configuration (Schaefer, 1987; Chapter 6.1). The otic canal lies deep below the surface, being unvisible externally in most of the sphenotic and compound pterotic. The parietal branch of the supraorbital canal and the pterotic branch of the otic canal are present. A few minor differences with Ancistrus cf. 152 PART 6 — ADULT MORPHOLOGY triradiatus (Schaefer, 1987; Chapter 6.1) are listed. The infraorbital canal extends further rostrally, well beyond the mouth opening and upper lip. The preopercular canal exits halfway the preopercle. The preopercle is completely covered by the lateral dermal plates and doesn’t reach the skin. The plates, however, leave a thin strip of skin in between them, exactly where the canal is situated within the preopercle. The rostral part of the preopercular canal is directed rostrally, passing through the long cheek plate.

MYOLOGY (FIGS 72, 73, 74)

All divisions of the adductor mandibulae complex, as found in Ancistrus cf. triradiatus, are present. Generally, they are narrower in Farlowella acus. The adductor mandibulae s.s. is very similar to the configuration in the previous species; the external part, the bulky dorsolateral bundle, is the longest and largest part, originating on the hyomandibula and inserting on the angulo-articular and the lateral aspect of the coronoid process. The flattened internal part originates on the quadrate and metapterygoid, and inserts on the meckelian cartilage and the posterior aspect of the coronoid process (this part is not visible on figure 74A, but can be seen on figure 73B). The mandibular branch of the trigeminal nerve lies on top of it, close to the insertion point of the muscle (more posteriorly the nerve branch separates the retractor premaxillae from the external part of the adductor mandibulae; Fig. 71A). The retractor premaxillae, the longest muscle of the adductor mandibulae complex, originates on the hyomandibula, and is closely associated with the external part of the adductor mandibulae (and partially hidden by it on figure 74A). The position and size of the retractor veli are similar as in A. cf. triradiatus (Fig. 74B). Serial sections, however, show the splitting of the muscle in two bundles anteriorly, just before it continues as a tendon in the oral valve. This splitting, albeit faint and not well visible in dissected specimens, is reminiscent of the situation in Otocinclus vestitus (see Chapter 6.3). The intermandibularis anterior pars dentalis is absent in Farlowella acus. The pars labialis is present and well developed, stretching between both halves of the lower lip, reaching far laterally, almost towards the contact zone between upper and lower lips (Fig. 73A). The fibres appear to touch the anteriormost fibres of the intermandibularis posterior pars labialis. They are, however, not continuous, and fibre direction in both muscles is almost perpendicular. The intermandibularis posterior is similarly constituted as in Ancistrus cf. triradiatus. The pars dentalis originates far posteriorly on the posterior ceratohyal, and inserts via a long tendon on the dentary part of the lower jaw, close to, but not on the teeth-bearing basket (as opposed to the insertion in A. cf. triradiatus). The pars labialis originates mainly on the anterior ceratohyal and diverges rostrally, forming about five separate bundles before inserting on the lower lip (Fig. 73A). PART 6 — ADULT MORPHOLOGY 153

The hyohyoideus inferior is slightly forked, the posterior half extending further laterally and running more obliquely (Fig. 73A). The sternohyoideus is more or less identical to the one in Ancistrus cf. triradiatus (Fig. 73A). Its ventral fascia is the origin of the medial tendon of the hyohyoideus abductor, which inserts on the medialmost branchiostegal ray, and, unlike in A. cf. triradiatus, narrows when approaching the midline (which it does not reach; Fig. 73A). A conspicuous feature in the composition of the hyohyoidei adductores in Farlowella acus is the origin of the anteriormost dorsal fibres on the ventromedial preopercle margin, in addition to that on the opercle more posteriorly (not shown on the figures). More ventrally, the adductores interconnect the branchiostegal rays. The levator arcus palatini is a short and tiny, weakly developed bunch of fibres stretching between the anteromedial edge of the sphenotic and the lateral aspect of the hyomandibula, close to the articulation with the neurocranium (Figs 72, 74A). The adductor arcus palatini diverges less than in Ancistrus cf. triradiatus, and almost exclusively inserts on the hyomandibula (Fig. 73B). Both parts of the extensor tentaculi are relatively smaller than in Ancistrus cf. triradiatus. The pars lateralis is identically shaped, round, with an aponeurosis in the centre of the muscle and continuing as a tendon inserting on the ventrolateral process of the autopalatine (Figs 73B, 74A). Origin is along the whole channel formed by the metapterygoid and lateral ethmoid (a relatively greater share is taken by the lateral ethmoid in Farlowella acus – see above). Related to the narrower skull, the pars medialis is less broad than in A. cf. triradiatus (Fig. 73B). Points of origin and insertion, however, don’t differ. A ventrally situated aponeurosis is present and transmits the force of the obliquely directed muscle fibres to the tendon attached to the ventromedial autopalatine process. The point of origin of the levator tentaculi on the lateral ethmoid is smaller, and the muscle is well separated from the adductor mandibulae complex (in Ancistrus cf. triradiatus the muscle bundles are so large they touch intimately). Rostrally it forks into two bundles, both inserting on the maxilla (Figs 72, 74A). The muscle is separated only by a thin skin layer from the paroral cavity. The dilatator operculi is a modest muscle in Farlowella acus, originating on the sphenotic and the anterior medial margin of the compound pterotic. The long tendon inserts on the dorsal aspect of the opercle, lateral to the opercle-hyomandibular articulation (Fig. 74A). The particularly underdeveloped levator operculi is a feeble strip of fibres connecting the caudodorsal aspect of the opercle to the ventral compound pterotic margin (Fig. 79AB). The myodome formed by the double-layered structure of the compound pterotic is very small when compared to that of Ancistrus cf. triradiatus (Part 8). The adductor operculi is also very small; it inserts on the medial face of the opercle and has its origin on the medial fascia of the protractor pectoralis muscle (Figs 73B, 74B). The adductor hyomandibulae is similarly sized, 154 PART 6 — ADULT MORPHOLOGY situated somewhat more dorsally, and has the same fibre direction. This muscle runs from the protractor pectoralis fascia as well, and ends in a tendon that appears to attach to, or near the medial aspect of the hyomandibula. This insertion is not easy to see, and thus very much resembles the situation in A. cf. triradiatus.

LIGAMENTS (FIG. 70)

As in all examined loricariids, the premaxillae are attached to, but separated from the mesethmoid, by the mesethmoid-premaxillary cartilage. The mesethmoid-cartilage ligaments and mesethmoid-maxillary ligaments are present. Of the ligaments connecting the premaxillae to the neurocranium, the medial mesethmoid-premaxillary ligaments are present and similar to the configuration in Ancistrus cf. triradiatus. The lateral mesethmoid- premaxillary ligaments are exceptionally long, connecting the premaxillae to the mesethmoid far rostrally, at the level of the first oral plates. More than in A. cf. triradiatus and Otocinclus vestitus the serial sections show a peculiar, aberrant histological nature of these long ligaments (Fig. 70B versus C). The short interpremaxillary ligament is found in Farlowella acus, as are the ventral labial ligaments. The ventral labial ligaments are not seen in A. cf. triradiatus, but have been found in Farlowella acus and in several Otocinclus species (Chapter 6.3). They connect the lateral upper lip tissue to the rostral end of the ventral mesethmoid process. There are no rostromaxillary ligaments. Instead, the space between the maxillary and the snout armour is taken by the opening of the paroral cavities.

6.2.4. DISCUSSION

In this chapter, as well as the subsequent (on Otocinclus vestitus), only some major anatomical differences with Ancistrus cf. triradiatus are discussed. Points relating to both Farlowella acus and O. vestitus are discussed in the last chapter.

The remarkable rostral extension of Farlowella acus appears to have little impact on the skull, except for the drastic elongation of the mesethmoid. The dermal plates in the ethmoid region have grown in number, and the first infraorbital reaches far beyond the upper jaw in F. acus, unlike in the other loricariids. I agree with Retzer & Page (1996) that the rostrum most probably functions in the stick-like mimicry, and that the broader rostrum in males, provided with large lateral odontodes, is a substantial sexual dimorphism. The two layers of the compound pterotic are closer to each other than in adult Ancistrus cf. triradiatus, but comparable to Pterygoplichthys lituratus, Otocinclus vestitus, and juvenile A. PART 6 — ADULT MORPHOLOGY 155 cf. triradiatus. This can most probably be related to the absence of opercular musculature invading the skull, and the absence of an erectile cheek-spine apparatus. The single large cheek plate might be considered homologous to the largest cheek plate in Ancistrus cf. triradiatus. Both carry the distal end of the preopercular canal. Given the interspecific variability of the dermal armour of the head, this cheek plate might however have developed separately in the evolution to both species. In that case they are, of course, not homologous. Detailed ontogenetic information of other species than A. cf. triradiatus is lacking, but is needed to resolve this issue. The absence of a canal-bearing cheek plate in Callichthyidae, Astroblepidae, and several basal loricariid species (Armbruster, 2004) suggests the possibly independent evolutionary development of such cheek plate(s) within the loricariid family. The term paroral cavity is introduced in this dissertation to describe the paired lateral and deep skin invagination penetrating under the skin and dermal armour, starting lateral to the mouth, and ending halfway the third infraorbital bone, just anterior to the eye. Among the studied loricariids, it has been found only in Farlowella acus and Sturisoma aureum, both loricariines. Its function is most probably related to the fact that the dermal armour in the head, including the mouth region, hindres the movements of the hyoid and suspensoria. The extent of the dermal armour is similar in the loricariine Rineloricaria parva, where such cavity, however, is not present. The paroral cavities separate the suspensoria from the lateral dermal armour, allowing movements of these elements. The jaw musculature, lying directly medial to the paroral cavities, probably benefits as well, as it has more space to expand. Thus the development of these paroral cavities might represent an adaptation to accommodate muscular expansion, as has been hypothesized as well for the suspensorial fenestrae in Gobiidae (Decleyre et al., 1991), and the dermatocranial fenestrae in diapsids and mammals. The consequence of the absence of the dentary part of the intermandibular muscle is not clear. The lower jaws appear to be more tightly connected to the lower lip tissue than in Ancistrus cf. triradiatus. Perhaps the labial part of the muscle has an indirect effect on the dentaries as well.

156 PART 6 — ADULT MORPHOLOGY

6.3. OTOCINCLUS VESTITUS

6.3.1. INTRODUCTION

The sistergroup of Loricariinae + Hypostominae is a third speciose subfamily: the Hypoptopomatinae, with 16 genera and about 79 species (Ferraris et al., 2003). It comprises some smaller species, compared to the previous taxa, often with discontinuous lateral lines (Schaefer, 1991). The hypoptopomatine genus Otocinclus was chosen for the present study of the cranial morphology, thus adding a small, almost miniaturized loricariid to the list. In most Otocinclus species a sexual dimorphism is observed: odontodes on the male caudal peduncle are modified into a contact organ, which probably assists in the positioning of the male urogenital region with that of the female prior to the release of the sperm (Aquino, 1994; 1997; Schaefer, 1997). Also, females are 10-20% larger than males (Schaefer, 1997; opposite as in Ancistrus cf. triradiatus). Compared to most loricariids, the eyes are relatively large. Otocinclus vestitus Cope is the type species of this genus, which has been found to be paraphyletic (Schaefer, 1991). The species can be distinguished from related species in the O. vestitus group by the absence of the lateral line canal in the caudalmost dermal plates of the trunk (Schaefer, 1997). The dermal armour covers the whole head except for the ventral hyoid region, and the odontodes are very large when scaled to Ancistrus cf. triradiatus and Farlowella acus. The odontodes on the tip of the snout are even larger. Odontode size, especially in small loricarioids, has been related to their protective function (Bhatti, 1938; Sire & Meunier, 1993; Sire & Huysseune, 1996). Otocinclus vestitus grows up to 4.5 cm SL (Schaefer, 1997). Compared to Ancistrus cf. triradiatus and Farlowella acus, it is a more active species, also during the day. It forages on leaves rather than rocks or wood; aquarium observations show it is less able to attach to overhanging substrates. It is known from geographically disjunct regions in the western and southwestern Amazon basin (Peru, Bolivia) and the lower Paraná basin (Paraguay) (Schaefer, 1997).

6.3.2. BRIEF MATERIAL AND METHODS

For examined specimens, see Table II. For clearing and staining, dissections, and serial sectioning procedures, see paragraphs 2.2.5, 2.2.6 and 2.2.7, respectively.

PART 6 — ADULT MORPHOLOGY 157

6.3.3. RESULTS

OSTEOLOGY (FIGS 75, 76, 77)

The relative size of the mesethmoid is comparable to that of Ancistrus cf. triradiatus, but is its anterior tip is broader. It is firmly connected to the rostral plate (see below) by means of various short ligaments. The ventral disc is well developed (Fig. 77B). Contrary to the previous species, Otocinclus vestitus has a nasal sac not completely enclosed by the lateral ethmoid. Only the posterior half is supported by the bone, the dermal part of which has barely extended rostrally from the ossified orbitonasal lamina. The nasal is short and pentagonal, and only posteriorly bears the supraorbital canal. The frontal is large and broad, bordering the orbit dorsally. The sphenotic has a lateral process bordering the orbit posteriorly, like in A. cf. triradiatus, but unlike the sphenotic in Farlowella acus. The prevomer has the common loricariid shape, with a single process anteriorly and posteriorly, suturing with the double anterior process of the parasphenoid. The latter bone has well developed lateral wings. The orbitosphenoids are joint together above the parasphenoid, and are fused to a certain degree. The pterosphenoid is pentagonal, very similar to the same bone in the other loricariids, touching the orbitosphenoid, frontal, sphenotic and prootic, and bordering the sphenotic fenestra dorsally. This fenestra, pierced by the trigeminofacial and optic nerves, is also bordered by the orbitosphenoid and prootic, a configuration that appears to be conserved among loricariids. The prootic bears an anterior arch, that is also found in Ancistrus cf. triradiatus. The basioccipital is broad; its interdigitation with the parasphenoid consists of only one long process, fitting between two posterior processes of the parasphenoid. This, like most other interdigitations, is less elaborate in Otocinclus vestitus than in the other examined loricariids. The exoccipital and epioccipital don’t show significant size or shape differences, when compared to the other species. The compound pterotic has a shape externally similar to that of A. cf. triradiatus. However, a few important differences can be noted. First, the pterotic fenestrae are exceptionally large, being separated by only thin bone tubercles (Figs 75, 77). Second, the two layers are very close to each other anteriorly, leaving only a narrow space between them. Only posteriorly, at the level of the swimbladder encapsulation, they are wide apart. The posterior process of the parieto-supraoccipital is long and pointed, thus differing a lot from the previous species. The contacts with the dorsal processes of the complex and sixth vertebrae are present. Contact between the transverse process of the complex vertebra and the compound pterotic is similar to that of the previous species. There are only five infraorbital bones (Fig. 77A). As in the other examined loricariids, the posterior three border the orbit; anterior to these, an infraorbital bone borders the nostril, but 158 PART 6 — ADULT MORPHOLOGY only one instead of two are found more rostrally. As in Farlowella acus, the bony armour of the head is tightly fitting. Only ventrally, the skin is more naked. Usually three to four prenasal plates are found on the snout. A single, heavy rostral plate covers the tip of the snout, and is well connected to the mesethmoid. It bears enlarged odontodes. A series of three thick plates, also bearing these large odontodes, lies laterally, bordering the mouth. In analogy to the plates in F. acus, they are termed oral plates. Schaefer (1997) called them postrostral plates II-IV. Between these, the rostral, prenasal and infraorbital series, another plate is present [postrostral plate I of Schaefer (1997) (see discussion)]. A small prefrontal plate covers the posterior portion of the lateral ethmoid. Two cheek plates are present on either side, the first of which bears the remnant of the preopercular canal.

The autopalatine is accompanied by a long splint-like sesamoid bone, that attaches to the anterior cartilaginous of the autopalatine. It reaches towards the nasal capsules and is associated with the lateral base of the nasal flap (Figs 75B, 76B, 77B, 78A). The maxilla is relatively short, supporting an even shorter maxillary barbel cartilage distally (the maxillary barbel of Otocinclus vestitus is very short). Both premaxilla and dento-mentomeckelium are short and less round than in the previous species. Each jaw bears 13 teeth on average. The bony contact between the dento-mentomeckelium and the angulo-articular is even weaker than in Farlowella acus, the remnant of Meckel’s cartilage being the most important contact between both parts of the lower jaw. The suspensorium is a more fragile structure as well. The quadrate and hyomandibula are very thin bones with various small fenestrae (not drawn on figures). Their contact is almost exclusively established by the symplectic cartilage, as the preopercle merely contacts and overlies the quadrate, but is not realy fused to it. Only posteriorly it is fused somewhat to the hyomandibula. As in Farlowella acus, it doesn’t reach the skin surface, but is hidden underneath the dermal armour. The hyomandibula articulates with the synchondral joint between the prootic, sphenotic and compound pterotic. The metapterygoid barely interdigitates with the rest of the suspensorium, only suturing well to the quadrate just anterior to the symplectic cartilage. It lacks a lateral ridge for attachment of the lateral part of the extensor tentaculi muscle. The elements of the hyoid arch are separated from each other by a relatively thick synchondral zone. The anterior sheet of the posterior ceratohyal, making bony contact with the anterior ceratohyal, is weakly developed, compared to the other loricariids. The caudal margin of the anterior ceratohyal bears a few irregular protrusions that increase the insertion space for the hyohyoideus inferior muscle. There is no interhyal bone (it was found in none of the examined loricariid species). The sesamoid bone in the ligament between the posterior PART 6 — ADULT MORPHOLOGY 159 ceratohyal and the hyomandibula is ovally shaped instead of being more elongate (not drawn on figures). The shape of the posterior edge of the parurohyal is slightly forked. Only the second basibranchial is ossified; the third is cartilaginous, as well as the fourth and fifth, which are fused. The first two pairs of hypobranchials are separate structures, not fused to the corresponding ceratobranchials; the first pair is ossified. The anterior process of the first ceratobranchial is slender but as long as the ceratobranchial itself. Contrary to the previous species, Otocinclus vestitus also has an anterior process on the first epibranchial as well (Fig. 77C). The process does not attach to or near the medial cartilage head of the first epibranchial, as is the case of the first ceratobranchial process in O. vestitus and the other examined species, but almost halfway on the bony epibranchial itself. The lateral ends of both processes reach towards each other and the medial side of the hyomandibula, functioning as support for an accessory gill. The fourth epibranchial is broader than in the other species. Shape and position of infrapharyngobranchials III and IV are comparable. The upper pharyngeal jaws are minute and bear only four to five pointed teeth. The lower pharyngeal jaws are even less developed: the fifth ceratobranchials are not enlarged, and bear about three teeth each. The opercular series of Otocinclus vestitus lacks a suprapreopercle. The opercle is well developed, but is overlapped by the second cheek plate anteriorly. It has no anterior process. The preopercle, as stated above, is slender, and barely fused to the suspensorium.

CRANIAL LATERAL LINE CANALS (FIGS 75A, 76A, 77A)

The lateral line system has undergone some reductions in Otocinclus vestitus. The preopercular canal has disappeared except for its base in the compound pterotic and a short stretch in the first, largest cheek plate. The infraorbital canal is shortened, related to the presence of only five infraorbital bones. The supraorbital canal exits halfway the nasal bone instead of at its rostral margin. When leaving the skull, the postotic canal is almost not embedded in the compound pterotic, as in the other species. It passes below the dorsal bone layer, and then enters the first body scute as the trunk lateral canal.

MYOLOGY (FIGS 79, 80, 81)

The bundles of the adductor mandibulae have the same origin and insertion points as the previously mentioned species. As the adductor mandibulae is relatively small, the retractor premaxillae doesn’t overlie it posteriorly, as it does in the other loricariids (Fig. 81A). Both muscles are extremely flattened posteriorly, as the large eye occupies much of the available space dorsolateral to the suspensorium (Fig. 78B). Two distinct bundles can be discerned in the retractor veli, both originating on the metapterygoid and inserting tendinously on the oral 160 PART 6 — ADULT MORPHOLOGY valve (Figs 78A, 81B). Anteriorly both bundles are widely separated, and the dorsal one extends somewhat further rostrally, its tendon entering the valve more anterodorsally. It also originates somewhat more posteriorly on the hyomandibula (Fig. 78B). Both the intermandibularis anterior pars dentalis and pars labialis are well developed in Otocinclus vestitus (Fig. 80A). The pars dentalis is similar to that in Ancistrus cf. triradiatus, connecting the lateral aspects of both lower jaws, and running posteroventrally from one side to the other. The pars labialis is relatively straight, not running very far laterally. The intermandibularis posterior pars dentalis inserts nearly on the teeth bearing, basket-like part of the dento-mentomeckelium (Fig. 80A), a configuration intermediate to that of A. cf. triradiatus (insertion on the basket) and Farlowella acus (insertion not on the basket). The intermandibularis posterior pars labialis is almost identical to the one in A. cf. triradiatus: a medial and lateral part can be distinguished, and five to six separate bundles insert on the lower lip. As in the previous species, this number of bundles varies intraspecifically. The configuration of the hyohyoideus inferior appears to be very conservative within the examined loricariid subfamilies; in Otocinclus vestitus it is slightly laterally forked as well (Fig. 80A). The presence of an anterior and posterior half was recognized as a synapomorphy for loricariids, astroblepids and scoloplacids by Schaefer (1990). I found, however, no true subdivision in two parts, not even laterally (and not in the other loricariids either). The sternohyoideus is well developed, but rather straight (instead of widely diverging posteriorly, as in Ancistrus cf. triradiatus; Fig. 80A). The hyohyoideus abductor is relatively broad and long, reaching its counterpart via an aponeurosis instead of inserting only on the sternohyoideus (Fig. 80A). Fibre direction, as in the other species, is transverse. The slender hyohyoidei adductores interconnect the branchiostegal rays and the medial side of the opercle. There is no contact to the preopercle. The number of adductor bundles (three) is the same in all species. The levator arcus palatini is remarkably well developed: it is short, but relatively thick (Fig. 81A). It connects the medial side of the sphenotic and the lateral face of the hyomandibula; the muscle broadens when approaching the hyomandibula. The muscle direction is transverse on serial sections, like in Farlowella acus, but unlike in Ancistrus cf. triradiatus, and appears to be more ideal for elevation of the suspensorium. The adductor arcus palatini is similar to that in A. cf. triradiatus, diverging broadly from the orbitosphenoid and parasphenoid towards the hyomandibula and the posterior end of the metapterygoid (Fig. 81B). The ‘canal’ for the extensor tentaculi pars lateralis is rudimentary, as the metapterygoid lacks a lateral ridge. However, in addition to the lateral face of the metapterygoid, a ventral groove-like cavity in the large lateral ethmoid offers ample insertion space (a situation reminiscent of the one in F. acus). Although it is relatively slim, the muscle doesn’t differ much from those in the other loricariids. The pars medialis is better developed, being similar PART 6 — ADULT MORPHOLOGY 161 to the general loricariid configuration (Fig. 80B). The levator tentaculi is well separated from the muscles of the adductor mandibulae complex. Origin is on the lateral ethmoid, lateral to the nasal capsule (Figs 79, 81A). This lateral position of the muscle was already noted by Schaefer (1997) (who called it ‘retractor tentaculi’, see Chapter 5.2). It is relatively unconsolidated when compared to the one in the other loricariids. The dilatator operculi is rudimentary, and, opposed to the situation in the other loricariids, connects the dorsal opercle margin not only to the medial side of the sphenotic, but also to the lateral aspect of the hyomandibula (Fig. 81A). The levator operculi is relatively larger than that of Farlowella acus, but negligible when compared to that of Ancistrus cf. triradiatus. It connects the opercle to the ventral compound pterotic margin. The adductor operculi and adductor hyomandibulae roughly have the same relative size and shape as in F. acus, also originating from the medial protractor pectoralis fascia (Figs 80B, 81B).

LIGAMENTS

The mesethmoid-premaxillary cartilage is present, as are the mesethmoid-cartilage ligaments. The position and orientation of the medial and lateral mesethmoid-premaxillary ligaments, as well as of the interpremaxillary and mesethmoid-maxillary ligaments, are similar to the situation in Ancistrus cf. triradiatus (without an elongated mesethmoid). The ventral labial ligaments are present but short. Rostromaxillary ligaments, not seen in any of the other loricariid species, connect the maxillary heads to the adjacent snout armour (the lateroposterior ends of the rostral plate). They were observed in Otocinclus vittatus as well (Schaefer, 1997).

6.3.4. DISCUSSION

Osteology and (incomplete) cranial myology of the related species Otocinclus vittatus has been described by Schaefer (1997). Aquino (1998) commented on the presence of eye muscle myodomes in Otocinclus and some other hypoptopomatines. Her and my results indicate that loricariids are an exception to the fact that siluriforms generally lack eye muscle myodomes (Alexander, 1975). Major differences with Ancistrus cf. triradiatus and Farlowella acus include the exceptionally large fenestrae in the compound pterotic (Figs 75, 76, 77). This might be functionally related to possible (but unknown) differences in the swimbladder and acoustic apparatus, but might equally be a consequence of the almost miniature size of the species. Reductions in ossification of elements is a rather general feature in miniaturized species. As far as is known, there might be only one miniature loricariid species, Microlepidogaster 162 PART 6 — ADULT MORPHOLOGY lophophanes, with a maximum standard length of 18 mm (Weitzman & Vari, 1988), and it is possible that larger specimens of this little known species exist. Otocinclus vestitus, however, shows several reductions in bones, as well as in the lateral line system. Some cranial reductions, when compared to the previous species, are mainly found in the splanchnocranium. Sutures are weaker (with few serrations), and relatively more cartilage is found in the cranial synchondral joints. The bones of the suspensorium are not tightly sutured, and the preopercle is not fixed to the quadrate, and only weakly to the hyomandibula. Less elaborate sutures and interdigitations are present throughout the skull of O. vestitus, suggesting miniaturization-related reductions. The lower jaw is less robust, containing relatively more cartilage connecting the dento-mentomeckelian and angulo- articular bones; teeth numbers are low [but this appears to be more correlated to mere scaling, and is the case in related species as well (Aquino, 1996)]. General papers on the anatomical-morphological effects of miniaturization are few; most data are to be found in individual descriptions of small-sized species (e.g., Friel, 1995; Soares-Porto et al., 1999). The structural consequences of scaling effects, however, are often quite dramatic. Novel arrangements of parts of the skull, brain, brain tissue, and sense organs may occur and proportional changes may have drastic effects on feeding biomechanics (Lauder et al., 1989). More has been done on the ecological consequences and (dis)advantages of small-sized species (Miller, 1979, 1996; Weitzman & Vari, 1988; Alexander, 1996). The opercular series of Otocinclus vestitus lacks a suprapreopercle. Related to this (and possibly to the near-miniaturized size of the species), the preopercular canal is disjunct: one part runs between the postotic canal and the lateral pterotic margin. It ends in a pore that is close to the posteriormost opening of the infraorbital canal. In O. vittatus, intraspecific variation was observed in the presence of two single pores, or one shared pore (Schaefer, 1997). A second part of the preopercular canal is present in the largest of two cheek plates. Analogous to the discussion on Farlowella acus, it is difficult to state a sure homology between the cheek plates of the examined loricariid species. Only five instead of six infraorbital bones are present. It is suggested here that the plate directly anterior to the first infraorbital might represent the canal-less, anamestic part of the actual first infraorbital. It has been termed postrostral plate I by Schaefer (1997). Aquarium observations of Otocinclus vestitus indicate that it is more active, but also less relying on its suckermouth in maintaining position on inclined substrates. The fishes usually lie on leaves or other substrates, supported by their pectoral fins. They are, however, capable of attaching to vertical (but not completely inclined) surfaces with their mouth. The opercle-hyomandibular articulation has an orientation similar to that in Farlowella acus and most other examined loricariids, except the Ancistrus species. This character is PART 6 — ADULT MORPHOLOGY 163 discussed in Part 8. The autopalatine sesamoid bone, small in Farlowella acus, and almost invisible in Ancistrus cf. triradiatus, is well developed in Otocinclus vestitus. This splint bone might have a function related to the generation of water flow in the nasal sac, being moved by the movements of the autopalatine (Schaefer, 1997). Serial sections show it to attach to the cartilaginous base of the nasal flap (Fig. 78A). Some important features in cranial myology of Otocinclus vestitus include the presence of two distinct retractor veli muscle bundles, that might result in an altered or more precise mobility of the oral valve. The laterally flattened jaw musculature can be related to the presence of relatively large eyes (Fig. 78B).

This limited appraisal of loricariid functional-morphological diversity has yielded one major conclusion: while many anatomical features can be found to be similar in the various examined species, subtle differences are found to exist. Differences observed between Ancistrus cf. triradiatus, Farlowella acus and Otocinclus vestitus include possibly important differences in size and shape of cranial elements, the presence or absence of elements like the labial intermandibular muscle part, or a second retractor veli muscle bundle. Considering the considerable diversity of the family, a lot of work still has to be done in order to assess the true morphological diversity of the Loricariidae, which will most probably provide more insight regarding the exceptional evolutionary success of the family. 164 PART 6 — ADULT MORPHOLOGY

PART 7

MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES 165

MORPHOLOGY AND ONTOGENY OF TEETH AND ∗ EPIDERMAL BRUSHES

Abstract — Loricariidae or suckermouth armoured catfishes are one of several aquatic taxa feeding on epilithic and epiphytic algae. Their upper and lower jaws bear exquisitely curved tooth, which usually are asymmetrically bicuspid. The enlarged lower lip carries papillae with keratinous unicellular epidermal brushes or unculi. Teeth, and probably unculi too, assist in scraping food off substrates. Their morphology, growth and replacement is examined and compared among several loricariid species, using cleared and stained specimens, serial sections and SEM. Apart from the general tooth form and crown shape, the anterior layer of soft tissue on the lower shaft region, present in several species, appears to be a specialization for enhancing the mobility of individual tooth when scraping on oneven surfaces. During early ontogeny, a transition from simple conical to mature tooth is observed. The first unculi appear together with the first teeth carrying a bicuspid crown, two days after the first exogenous feeding, but synchronous with the complete resorption of the yolk sac.

7.1. INTRODUCTION

Brush- and gouge-like devices are ideal for scraping food off substrates, and are most often found in aquatic organisms, particularly aquatic insect larvae (Arens, 1994), amphibian larvae (Orton, 1953; Wassersug & Yamahita, 2001), and fishes. In all these cases the diet primarily consists of adherent algae. Examples in teleosts are the rake-like denticles of the osmeriform ayu (Howes & Sanford, 1987; Uehara & Miyoshi, 1993), the tooth-like keratinous hooks of Gyrinocheilidae (Ono, 1980; Benjamin, 1986), spatulate teeth of certain Cichlidae (Vandervennet et al., 2006) and Mochokidae, and the scraping teeth of species of the Loricariidae or suckermouth armoured catfishes. The latter family exhibits the most exquisite and diverse teeth forms (e.g., Muller & Weber, 1992; Schaefer & Stewart, 1993; Delariva & Agostinho, 2001): the S- or Z-shaped recurved teeth are generally asymmetrically bicuspid, but, in some taxa, have one cusp only. Teeth of the related loricarioid scoloplacid and astroblepid families are usually symmetrically bifid (Schaefer, 1990), although shape variation exists (e.g., Cardona & Guerao, 1994). Teeth are absent in adults of the more basal callichthyids, while simple conical teeth have been found in small juveniles (Huysseune & Sire, 1997a). While many genera of the basal loricarioid trichomycterids have rather conical teeth, Henonemus has (unicuspid) recurved teeth, reminiscent of loricariid teeth (DoNascimiento & Provenzano, 2006).

∗ Slightly modified from: Geerinckx T., De Poorter J. & Adriaens D. Morphology and development of teeth and epidermal brushes in loricariid catfishes. Submitted to Zoology. 166 PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

Loricariidae are able to attach onto surfaces using a not well understood process allowing respiration during attachment, and scrape off algae and other food with their ventrally oriented upper and lower jaws. Ono (1980) and Roberts (1982) described unicellular keratinous lip projections or ‘unculi’ on the surface of the expanded lower lip of loricariids. These epidermal projections might serve as abrasive brushes or protective structures for the associated taste buds. Ono (1980) studied the internal microstructure of the loricariid unculi. Unicellular and multicellular keratinous structures are found in several teleostean orders, with functions often related to reproduction, protection, abrasion, adhesion and hydrodynamics (Branson, 1962; Wiley & Collette, 1970; Roberts, 1973; 1982; Arratia, 1987; Arratia & Huaquín, 1995; Chen & Arratia, 1996). In this chapter I describe the morphology of loricariid teeth and unculi, their growth, and their shape during early ontogeny. Results are then discussed in view of the possible role of both structures in feeding.

7.2. BRIEF MATERIAL AND METHODS

Teeth and lips of several loricariid species were examined, using the clearing and staining method described in paragraph 2.2.5, and SEM (paragraph 2.2.9): Ancistrus cf. triradiatus, Pterygoplichthys lituratus, Panaque nigrolineatus, Otocinclus vestitus, Rineloricaria parva, Farlowella acus and Sturisoma aureum (Tables I and II). For clearing and staining, and serial sectioning procedures, see paragraphs 2.2.5 and 2.2.7, respectively.

7.3. RESULTS

MORPHOLOGY AND GROWTH OF TEETH

Upper and lower jaws of loricariids are oriented so that the teeth point ventrally, touching the substrate to which the fish is attaching (Figs 5, 82A). Adult Ancistrus cf. triradiatus specimens carry one row of 40 to 67 emergent teeth per premaxilla (n=7, mean=55), and 58 to 79 teeth per dentary (n=7, mean=69). No distinct differences are noted in tooth shape for each jaw, except for those being on the lateral side being somewhat smaller. Teeth are Z- shaped, and are composed of a thick, curved base (which is covered by the jaw epithelium), a thin lower shaft, a thicker upper shaft and a curved bicuspid crown (Fig. 83C). The base is movably connected to the jaw bone. The base and the shaft form an angle of about 90°, the shaft and the crown form an angle of 90° as well. The small lateral cusp has an angle of 110 to 150° to the main cusp, and is about half as long.The anterior region of the lower shaft is PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES 167 not calcified, but is composed of soft material. A distal protuberance of this soft layer, seen only in A. cf. triradiatus, is stained blue with alcian blue and purple (metachromatic) with toluidine blue (Fig. 83C, I). Manipulating individual teeth of freshly killed A. cf. triradiatus specimens reveals that the lower shaft can actually bend up to 90° with respect to the tooth base (Fig. 84). Such considerable bending as shown in this figure (here caused by manipulation with tweezers) probably seldom occurs in natural circumstances, as the lip tissue, situated behind the teeth, hindres such excessive movement (e.g., fig. 85A). Manipulated bending in the opposite, anterior direction, invariably causes breaking after bending only about 10-20°. One row of teeth emerges from the jaw epithelium (Figs 82B-C, 83H). The jaws are essentially basket-shaped; the lower jaw also has a lateral ‘handle’ articulating with the quadrate and consisting of the angulo-articular and part of the dento-mentomeckelium (Chapters 4.2 and 6.1). Tooth germs are found deep inside the basket; growing teeth migrate within the basket. The individual tooth germs are not found in bony crypts; serial sections show that the bases of emergent teeth are connected to the bone via soft tissue, most probably containing a large amount of collagen. Thus the development is completely extraosseous, as in other siluriforms and many other teleosts (Trapani, 2001). In the premaxilla, there is a progression of less to more developed teeth from posterior to anterior (vice versa for dentary). An anteroposterior cross-section through the jaws (Fig. 82B-C) clearly shows this progression, but at the same time gives the false impression of the presence of numerous replacement tooth rows (up to 20). Serial sections, however, reveal the presence of only about four successive teeth in one tooth family: arrows on figure 83J show the replacement teeth originating from two distinct loci; an epithelial ‘track’ connects the subsequent replacement teeth. A tooth becomes erected after it has been fully formed (Fig. 82B-C): it penetrates the jaw epithelium, thereby rotating 40 to 80° (large arrowheads on figure 85A indicate emerging teeth). This possibly rather sudden tooth movement leaves trail-like scars in the weak jaw epithelium (small arrowheads on figures 83H and 85A). Once in use, the crown is worn and consequently shortened; figure 85B shows obvious signs of wear caused by scraping on substrates. Previous authors have described the diversity in number and shape of teeth in various loricariids (Muller & Weber, 1992; Delariva & Agostinho, 2001). I limit my description to the most remarkable shape observations. The teeth of Pterygoplichthys lituratus and Otocinclus vestitus resemble those of Ancistrus cf. triradiatus, but the soft layer in the lower shaft region is thinner, lacking the distal protuberance (Figs 83A, E, 85C-D). Long spatulate crowns with almost similarly sized cusps characterize Sturisoma aureum and Farlowella acus (Figs 83B, D, 85E-F). Panaque nigrolineatus has sturdy, unicuspid, spoon-shaped teeth (Figs 83G, 85G). Both unequally sized cusps of the teeth of Rineloricaria parva are pointed (Figs 168 PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

83F, 85H). Teeth in the latter two species are least curved and lack the soft layer. They appear to be rigid, while teeth of all other species can bend to a certain degree between the lower shaft and the base. While 15 to 25 apparent replacement tooth rows are seen at first sight in adult specimens of all specimens (Fig. 82B-C), only about four are present (as observed on serial sections of P. lituratus and Farlowella acus), as in A. cf. triradiatus. Analogous to this, the 7 to 8 apparent replacement rows seen in O. vestitus actually represent only two rows. One to two rows are probably present in P. nigrolineatus (5 apparent rows), and R. parva (4 apparent rows).

TOOTH SHAPE DURING EARLY ONTOGENY

Ancistrus cf. triradiatus hatches five to six days after fertilization. The yolk sac is depleted after an additional four to five days. Ingested food particles are found in the intestine from three to four days after hatching. The first teeth appear on the premaxilla and erupt before hatching, at four days after fertilization. The (five to nine) premaxillary teeth are conical, bearing no resemblance at all to the adult tooth shape (Fig. 86A). Analogous to the appearance of the skin odontodes, these first teeth are observed before the supporting bone materializes (Chapter 4.2). Six days after fertilization, half a day after hatching, similar teeth are still present on the premaxilla, and identical teeth have appeared on the dentary as well. One replacement tooth row is present (7 days PF; arrows on Fig. 87A). Serial sections show that, as in adults, these teeth are connected to the bone via soft (collagenous) tissue. At eight days new teeth possess a flattened, unicuspid tip (arrowheads on figure 86B); the curvature between the base and the shaft has now developed. Consequently, these new teeth are recurved backward (Fig. 86B), instead of slightly forward (conical teeth on figure 86A). A rudimentary bicuspid crown is present on some teeth at ten days (both cusps are indicated by arrowheads on figure 86C). The curvature between the shaft and the crown, added to the one between the base and the shaft, results in the first Z-like teeth. Only 14 days after fertilization bicuspid crowns are present (Fig. 86D); cartilage resorption has made place for already three to four replacement rows (arrows on figure 87B). This is already the maximum of tooth rows observed in Ancistrus cf. triradiatus. The number of teeth has risen to 13-17 teeth for each of the four jaw bones. Already some of the teeth appear to have a damaged crown. As opposed to the number of replacement rows, the number of teeth progressively increases during further ontogeny. Pharyngeal teeth are present in all examined species [as opposed to Alexander’s (1965) statement that they are absent]; they are cone-shaped. Their number more or less correlates with body size, without any further substantial difference between species. In Ancistrus cf. triradiatus, the first pharyngeal teeth appear at three days after hatching (Chapter 4.2), and about 30 teeth are present per single jaw quadrant. PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES 169

MORPHOLOGY AND GROWTH OF UNCULI

Unicellular epidermal brushes or unculi are found in a field on the top of labial papillae of the lower lip. In some species with unculi not all papillae bear unculi; unculi are rare on the upper lip. In those cases where taste buds are present on the papillae as well, the unculi are found rostrally to these. Counts of unculi on five randomly chosen papillae, and height measurements of ten randomly chosen unculi at the scanning electron microscope yielded the following ranges: Ancistrus cf. triradiatus: 40-80 unculi per papilla, sometimes covering the whole top of the papilla, height 7-15 µm, unculi slender with the tip curved rostrally (Fig. 88A). Pterygoplichthys lituratus: 25-50 unculi per papilla, height 12-16 µm, straigth with a slightly flattened tip (Fig. 88B). Panaque nigrolineatus: several hundreds of unculi per papilla, also covering the sides of the papilla, height 5-10 µm, broad with a flattened tip (Fig. 88C). Sturisoma aureum: 20-35 unculi per papilla, height 8-12 µm, nor broad nor slender, tip flattened. Farlowella acus: 40-70 unculi per papilla, height 12-17 µm, sometimes covering whole top of papilla, unculi slender, tip straight (not flattened) (Fig. 88D). Rineloricaria parva: no unculi present on papillae. Otocinclus vestitus: 40-70 unculi per papilla, height 2-4 µm, unculi broad and tip extremely flattened (Fig. 88E). Epidermal cell diameter is approximately 8 to 12 µm in all species. Taste buds are numerous on the papillae in Rineloricaria parva, rare in Farlowella acus, Otocinclus vestitus and Panaque nigrolineatus (arrowheads on figure 88C, E), and absent on many (but not all) papillae in the other species. Unculi appear to be replaced when parts of the upper epidermis layer of the lip are shed, as seen in Ancistrus cf. triradiatus (88F). Figure 88F, as well as figure 89, suggest that the shedding occurs more or less per papilla. In the specimen in figure 89, only three of the total amount of papillae were in an obvious process of shedding.

SHAPE OF UNCULI DURING EARLY ONTOGENY

Embryonic specimens of Ancistrus cf. triraditatus have rudimentary papillae one day after hatching (Fig. 90A). First sloughing of epidermis occurs at three days after hatching (Fig. 90B). The first unculi appear together with the first well developed tooth cusps, at five days after hatching (Fig. 90C-D). This is the moment of complete resorption of the yolk sac.

170 PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

7.4. DISCUSSION

As in several (but not all) other Ancistrus species, the number of dentary teeth in Ancistrus cf. triradiatus is equal to (or slightly higher than) the number of premaxillary teeth (Muller & Weber, 1992; Miquilarena et al., 1994; Fisch-Muller et al., 2001). No significant intraspecific differences between both jaws were noted in the other examined species. The morphology of teeth of A. cf. triradiatus appears to be the most interesting of all loricariids examined by me and other authors thus far (e.g., Schaefer, 1987; Muller & Weber, 1992; Schaefer & Stewart, 1993; Delariva & Agostinho, 2001; Armbruster, 2004). The teeth are characterized by a strong Z-shaped curvature, the differentiation of the shaft in a thin lower and a thick upper portion, and the presence of an anterior layer of soft tissue along the lower shaft. This layer has been found in all examined species except Panaque nigrolineatus and Rineloricaria parva. The thicker distal protuberance was found in A. cf. triradiatus only. The histological nature of the soft layer is unclear, but might well have a strain-resistant function: if the crown jolts along a rough substrate during scraping, the anterior region of the tooth is prone to strain, especially near the base of the lower shaft, where the bending occurs. A completely calcified tooth might easily break. Figure 83I suggests that the posterior lower shaft region (near the base) is far from completely calcified: it is almost not stained by the toluidine blue stain (compare tooth section ‘3’ to ‘1’ and ‘2’ on Figure 83I). This might represent an elaborate adaptation to the fact that the shaft of loricariid teeth encounters sideward (anterior) forces instead of axial forces during feeding. The unmineralized (collagenous) attachment of the tooth base, present in most ostarioclupeomorph teleosts (Fink, 1981) further increases the mobility of the individual tooth (with respect to the jaw bone). The link between tooth shape and diet has been inferred from several studies, with the general conclusion that slender teeth are appropriate for scraping smaller particles from surfaces, while larger, stronger and spatulate teeth are better for scraping coarser food items off hard surfaces (Delariva & Agostinho, 2001). This is best illustrated by the two extreme conditions. Robust teeth are present in taxa like Panaque and the Hypostomus cochliodon group (Schaefer & Stewart, 1993; Armbruster, 2004; this chapter), of which at least Panaque has been proven to be able to eat and digest wood (Nelson et al., 1999). On the other hand, sometimes complete absence of teeth is observed in some loricariines, a subfamily containing many detritus-feeders living on soft substrates (Salazar et al., 1982; Reis & Pereira, 2000; Rapp Py-Daniel, 2001). The lack of the soft tissue layer and the lesser mobility of the tooth base of Panaque nigrolineatus relative to the bone might reflect the need of more robust, better anchored and rigid (thus completely mineralized) teeth for scraping wood. The diet of the loricariine Rineloricaria parva is less known; aquarium specimens were commonly observed on sand and gravel, when compared to the other species (stones, wood pieces and PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES 171 plants were provided as well). Some Loricaria species are known to feed more on small animals and detritus on soft bottoms (as opposed to hard substrates) (Saul, 1975; Aranha et al., 1998; Reis & Pereira, 2000). More ecological data confirming this for R. parva are needed to substantiate the hypothesis that this species scrapes less on hard substrates than the other loricariids examined in this study. The lack of a soft layer in both P. nigrolineatus and R. parva also coincides with a low number of replacement tooth rows. The mineralized tooth portion of advanced actinopterygians doesn’t contain enamel (a purely epithelial product), but enameloid (to which ectomesenchymal tissue contributes) (Huysseune & Sire, 1998). However, a soft tooth portion, as found in the lower shaft of several loricariid species, has not been reported before, nor in any other teleost group. It is possible that the thin, ‘hard’ part of the lower shaft near the base is not as intensely mineralized as the remainder of the tooth. It is hard to find alternative explanations for the fact that this zone can bend without immediately breaking. A possible hypo-mineralization might be tested using microradiography.

No clear correlation, but a scala of different combinations is found between tooth and unculus shape in the examined species. Of the loricariines, Farlowella acus has slender teeth and unculi, and Sturisoma aureum has similar teeth but sturdier unculi. Unculi are absent in Rineloricaria parva, which has pointed tooth cusps and no soft anterior layer. Among the hypostomines, slender unculi co-occur with relatively slender teeth in Pterygoplichthys lituratus; unculi are somewhat thicker and rostrally inclined, and teeth are somewhat narrower in Ancistrus cf. triradiatus. Panaque nigrolineatus is characterized not only by the spoon-shaped teeth without anterior soft layer, but also by numerous short and flattened unculi. Unculi are even shorter and equally flattened, but less numerous in the hypoptopomatine Otocinclus vestitus, which has few and rather slender teeth. Unculi were not found in Otocinclus sp. by Ono (1980). The interspecific shape diversity, and the rostral inclination of the unculi on the lower lip in Ancistrus cf. triradiatus, corroborate the hypothesis that unculi in loricariids may serve as abrasive structures (Ono, 1980). Ono (1980) mentioned the presence of unculi on the upper lip of certain hypostomines; the upper lip is moved far less during scraping, and papillae are less numerous. She did not elaborate on the relative position of unculi and taste buds on the upper lip papillae. Keratinization, though rare in teleosts, has been found in several taxa; I refer to Das & Nag (2006) for an overview of such reports in teleosts and a histological examination in keratinized spines (~ unculi) in the cyprinid Garra gotyla gotyla. In the latter species, direction of the spines on the lower lip is opposite to the unculus direction in A. cf. triradiatus. This, and the distribution of the spines along the outer edge of the lip, 172 PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES surrounding a central callus part, suggest a function in adhesion rather than in feeding. In most loricariids, the unculi might well serve to both functions. It might be that the shedding of lip epidermis contributes to a ‘renewal’ of unculi. The shedding of the epidermis of single papillae on figures 88F and 89 proves that such shedding does occur in Ancistrus cf. triradiatus, although similar processes in teleosts are not well treated in literature. One could hypothesize that the abrasive function of the unculiferous papillae causes them to wear, a scenario in which shedding would be beneficial. Unculus structure might be an underestimated aspect of the adaptive radiation present in the loricariid family, next to shape, size and number of the teeth, morphology and orientation of the jaws, and the presence of labial filaments or fimbriae (Delariva & Agostinho, 2001; Rapp Py-Daniel, 2001; Armbruster, 2004), and deserves a closer look in ecological studies where several loricariid species are often found to live syntopically (Power, 1984; Buck & Sazima, 1995; Aranha et al., 1998; Delariva & Agostinho, 2001). As pointed out excellently by Zaret & Smith (1984), explanations for the small differences (and similarities) between ecologically important structures like teeth and unculi of similar or related species, are not easy and will have to rely on a significant amount of ecological information including present and past syntopy of the various species.

Generally the first tooth generations in teleosts consist of simple, conical teeth, irrespective of the adult tooth shape (Huysseune & Sire 1997b; Sire et al., 2002; Vandervennet et al., 2006). Even loricariids are no exception, as observed in Ancistrus cf. triradiatus. During further growth in the juvenile and adult phases, only a weak allometry in shaft and crown length has been observed in several Ancistrus and Hypostomus species (Muller & Weber, 1992). Schaefer & Stewart (1993) noted more pronounced shape transitions (from standard bicuspid to unicuspid and spoon-shaped) in juveniles of the Panaque dentex group. In the lower jaw, I did not observe cartilage resorption at the level of formation of individual tooth germs of the first generations, as observed in some cichlids by Huysseune (1990). This might be related to the relatively large distance between the tooth germs (and the dentary bone anlage) and Meckel’s cartilage in the 8.0 mm Ancistrus specimen. In the 10.2 mm specimen Meckel’s cartilage is completely resorbed at the level of the teeth, but I can not infer a direct relation with the tooth germs that now develop at the former location of the cartilage (Fig. 87B).

In conclusion, I consider the teeth and less known unculi of Loricariidae to be highly diverse tools, which most certainly are the result of an adaptive radiation. The unculi most probably have the same function as the teeth, i.e. scraping food off substrates, and as such, a PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES 173 comparable selective pressure can be expected. The morphology and diversity of teeth and unculi surely adds to the adaptations to the broad ecological niche occupied by this successful neotropical catfish family. 174 PART 7 — MORPHOLOGY AND ONTOGENY OF TEETH AND EPIDERMAL BRUSHES

PART 8

THE CHEEK-SPINE APPARATUS IN ANCISTRUS

PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS 175

THE CHEEK-SPINE APPARATUS IN ANCISTRUS ∗

Abstract — In the South American catfish family Loricariidae the opercle has been decoupled from the lower jaw, and has also lost its function in expiration. While many loricariid species have a small and slightly mobile opercle with reduced opercular musculature, within the hypostomine subfamily a novel opercular mechanism has developed that erects a tuft of enlarged odontodes anterior to the opercle. This defensive mechanism is examined in Ancistrus cf. triradiatus. The opercle has a prominent anterior process and the orientation of the reinforced articulation hinge to the hyomandibular bone has shifted. The opercular musculature is well developed, with a hypertrophied dilatator operculi, that extends deep inside the skull roof bones and towards the midline, over the brain, but below the superficial skull roof. Hence the frontal, sphenotic, parieto- supraoccipital and compound pterotic bones consist of a dorsal, superficial part and a deeper part separating the brain from the muscle: two functional skull roofs are thus formed. The impact on the path of the cranial sensory canals is substantial, moving canals away from the skull surface. Hypertrophy of cranial muscles is known from many teleosts, but the invasion of such large muscles into the skull, that is drastically modified and literally hollowed out, has never been described before. These cranial modifications are greater in males than in females, related to the territorial behaviour of the former, in which the erectile spines are used.

8.1. INTRODUCTION

Among the neotropical fish taxa, the suckermouth armoured catfishes or Loricariidae are remarkable for their atypical catfish morphology, specialized feeding apparatus and high species and shape diversity. Their armour, covering almost the whole body, is formed by numerous dermal bone plates carrying odontodes (or denticles, see note in paragraph 1.2.3), tooth-like structures composed of dentine, covered by a hypermineralized (enameloid) substance. These are firmly ankylosed to the bone or anchored by connective tissue fibres (Bhatti, 1938; Sire & Huysseune, 1996). This spiky external skeleton might well be an effective protection against predation. Odontodes are thought to have evolved independently in the loricarioid lineage (Bhatti, 1938; Reif, 1982), and are also found in e.g., callichthyids, scoloplacids, and the almost naked astroblepids, related loricarioid families (Schaefer et al., 1989; Sire & Huysseune, 1996; Schaefer & Buitrago-Suárez, 2002). In some loricariid genera the tuft of large, spiny odontodes anterior to the opercle can be erected. Alexander (1965) and Howes (1983a) mentioned the role of the dilatator operculi muscle in the erection of the odontodes. In Ancistrus, where this mechanism is most developed, both authors noted the impressing size of this muscle, reaching towards the dorsal midline of the skull, between the braincase and the dermal roof of the skull.

∗ Slightly modified from: Geerinckx T. & Adriaens D., 2006. The erectile cheek-spine apparatus in the bristlenose catfish Ancistrus (Loricariidae, Siluriformes), and its relation to the formation of a secondary skull roof. Zoology, 109: 287-299. 176 PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS

All loricariid odontodes, including the erectile ones, can be shed; marks on the dermal ossicles of Ancistrus cf. triradiatus and Chaetostoma sp. provide evidence of odontode loss or shedding (Howes, 1983a; pers. ob.). The fixed cheek odontodes of Loricaria uracantha develop in ca. 48 days and are shed after ca. 148 days (Moodie & Power, 1982). The number of erectile spines of the cheek-spine apparatus in Ancistrus and related genera is taxonomically informative (Eigenmann & Eigenmann, 1890; Muller, 1989b; Miquilarena et al., 1994; Ceas & Page, 1996; Fisch-Muller et al., 2001). Their movements, however, have not yet been studied. The present chapter aims to describe the morphology, function and use of the opercular odontode-erecting apparatus in Ancistrus cf. triradiatus Eigenmann, as well as discussing the seemingly indeterminantly growing opercular musculature, and its drastic impact on the spatial design of the skull roof bones.

8.2. BRIEF MATERIAL AND METHODS

Examined specimens of Ancistrus cf. triradiatus (with standard length): clearing and staining (paragraph 2.2.5): 4 (male: 101.9 mm, 95 mm; female: 76.6 mm, 69.6 mm); dissection (paragraph 2.2.4): 4 (male: 86 mm, 94 mm; female: 71 mm, 70 mm); manual sectioning (paragraph 2.2.7): 1 (male: 108 mm); 5 µm serial sections (paragraph 2.2.7): 1 (gender unknown: 33.5 mm). Some specimens of other loricariids were studied for comparison (see Table II): Ancistrus ranunculus (58 mm), Ancistrus dolichopterus (93 mm), Pterygoplichthys lituratus (63 mm; 150 mm), Farlowella acus (124 mm; 109 mm), Sturisoma aureum (83 mm; 85 mm), Rineloricaria parva (75 mm) and Otocinclus vestitus (23 mm; 24 mm). Live observations of the fast movements of the mechanism in A. cf. triradiatus were aided by the use of high-speed filming from different perspectives (see paragraph 2.2.2).

8.3. RESULTS

The following description is based on mature males, where the cheek-spine apparatus is best developed, and the associated structural implications for the neurocranium are most pronounced (see figure captions for length of specimens).

PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS 177

BONES OF THE CHEEK-SPINE APPARATUS

The cheek-spine apparatus consists of several bones in the lateral cheek region of the head (Fig. 91). The key element in the function of the cheek-spine apparatus is the opercle. Its articulation with the hyomandibular bone is reinforced by an anterior serration (two small ‘processes’), and a more weakly serrated, ligamentous connection between both bones posteriorly (Figs 92B, 93C-D). The result is one long, strong and more or less horizontal articular hinge around which the opercle pivots outward and inward (Fig. 93A-D). On the dorsal side insertion facets are present for the dilatator and levator operculi muscles, while the lateralmost hyohyoidei adductores bundle and the adductor operculi insert to the medioventral side (Fig. 92B-C). A very prominent anteroventral process articulates with the caudalmost ossicles carrying odontodes (Figs 92C, 93B). There are about 14 of these thick bony ossicles, most of which carry only one large odontode (Fig. 92D). To ease interpretation, and to avoid confusion with the larger cheek plates more anteriorly, I refer to these structures as cheek spines throughout the text. The odontodes are immobile relative to the ossicles to which they are anchored. There are four small and two larger cheek plates anterior to the cheek spines, that, like other dermal bones in loricariids, carry only minute odontodes (not shown on figures). The largest of these plates bears the distal portion of the preopercular canal (Fig. 92E). Its anterior edge articulates with the quadrate. All cheek plates and cheek spines are embedded in thick connective tissue, facilitating a chain of articulations (see below).

OPERCULAR MUSCULATURE

The dilatator operculi is the largest opercular muscle, and, in adult males, even the largest of all cranial muscles. In all male and female specimens four bundles are discernable (m-dil- op-I-IV on figures 94 to 96). The larger the specimen, the larger the relative size and area of origin of these bundles become. All bundles insert next to each other on the dorsal aspect on the opercle, lateral to the main articulation with the hyomandibula. Their origins, however, differ substantially. The first, anteriormost bundle is short, inserts on the opercle without an obvious tendon, and originates on the ventral edge of the suspensorium (this bone is not shown on figure 94). The second bundle is larger, and contains a distinct tendon that continues as an aponeurosis through most of the muscle length. It originates on the sphenotic and frontal bones. Both these bones are modified as such to accommodate the large bundle (see below). Medially, the second bundles of both sides of the head contact each other. A thick medial myocomma is present, which ossifies in the largest specimens. The third bundle is even more hypertrophied, and also contains a broad aponeurosis throughout most of its length. The tendon contacts the tendon of the second bundle when reaching the opercle. The 178 PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS bundle originates mostly from the compound pterotic and the parieto-supraoccipital, but also from the posterior part of the sphenotic, as well as from the dorsal process of the sixth vertebra. The compound pterotic and parieto-supraoccipital are significantly modified in large adults. Both bundles touch each other medially. The small fourth bundle lacks a real tendon, and originates on the caudal margin of the hyomandibula, below the passage of the third bundle. Manipulating any of the dilatator bundles results in an upward pivoting of the opercle along the hyomandibular articulation (Fig. 93D). The levator operculi muscle inserts on the dorsal aspect of the opercle, posterior to the dilatator operculi, and similarly diverges broadly into a cranial cavity. It consists of a single bundle with a plate-like aponeurosis throughout the muscle and originates broadly on the compound pterotic, without reaching the midline. Manipulation results in the same effect as the dilatator operculi: the opercle pivots upward (Fig. 93D). The adductor operculi muscle originates with a long, narrow tendon on the ventral surface of the compound pterotic. The muscle contacts the fascia of the protractor pectoralis muscle, the fibres of which run perpendicular to the direction of the adductor operculi. The muscle becomes somewhat broader before inserting on the ventromedial aspect of the opercle, posterior to the insertion of the lateralmost of the hyohyoidei adductores bundles (that originate on the fourth, lateralmost branchiostegal ray). Manipulation results in a downward rotation of the opercle, so that the muscle can be considered as the antagonist of the dilatator and levator operculi.

GROWTH OF THE DILATATOR OPERCULI AND THE FORMATION OF A SECONDARY SKULL ROOF

When comparing small and larger adults, a drastic change in the skull roof structure and the paths of the sensory canals is noted. This change does not seem to take place abruptly at maturation or at a certain standard length, but occurs progressively during subadult and adult growth. Until maturation, the major skull roof bones have a normal, more or less uni-layered architecture, as has been the case during the whole of the earlier development. Only the compound pterotic and, partly, the sphenotic arise as a ventral and a dorsal layer (Chapter 4.2). Figures 95 and 96 compare transsections of a subadult unsexed specimen and an adult male. As the relative size of the dilatator operculi increases, it invades the skull more deeply. In subadult specimens, sufficient insertion surface is present on the compound pterotic. The architecture of this bone is very complex (see figs 95C, 96C, 97A-B). The ventral layer carries the otic and postotic canals of the cranial lateral line system, and demarcates the brain laterally and posteriorly. The ventral layer contacts the dorsal surface layer laterally and caudally. During further growth, when the muscle becomes thicker, both layers are separated PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS 179 at their lateral contact zone with only two bony trabecles connecting both layers anteriorly. One of these carries the preopercular canal branch, bringing it to the surface in the dorsal bone layer (Fig. 96B). The pterotic branch of the postotic canal exits via the ventral layer, and the remainder of the postotic canal itself is seen uncovered on top of the ventral floor of the transverse process of the complex vertebra, lateroventral of the swim bladder. Further posteriorly, the postotic canal rises towards the dorsal layer of the compound pterotic, and then continues as the lateral line canal, passing through the body armour. The cavity (or ‘myodome’) in the compound pterotic, however, soon is not sufficient to house the dilatator operculi. In adults, modifications are seen on the sphenotic bone as well. When arising, the sphenotic consists of a superficial layer and a deeper oblique part. These two parts are still in close contact anteriorly and medially, and suture to the superficial and deeper layer of the compound pterotic respectively. In subadults the space between the two sphenotic layers increases, and the medial contact is soon lost (Fig. 95C). As the dilatator operculi grows further, it reaches between the surface layer and the deeper canal-bearing part, elongating and thinning out the rostral contact zone, so that only a thick bony trabecle remains between both diverging layers (Fig. 97D). The otic canal is found deep inside the skull, below the muscle. The infraorbital canal, which surfaces and exits the bone laterally, ascends through the bony trabecle (Fig. 96B). Both bone layers serve as points of insertion for the dilatator operculi in large specimens (Fig. 96B-C). Figure 98A shows the cavity in the sphenotic bone, here situated between the diverging otic and infraorbital canals.

As the muscle invades the skull more deeply, it grows between the brain and the dorsal skull roof. At the same time, the parieto-supraoccipital expands dorsoventrally. The result is a deepened, two-layered bone, hiding the brain deep below the dorsal skull surface. The bone is first split laterally, at the level of the dorsalmost part of the cartilaginous otic capsule remnant (Fig. 98B). The dorsal and ventral layers of the parieto-supraoccipital contact each other only at the midline, where a sagittal fragmented bony sheet is present (Fig. 99C). The dilatator operculi inserts on both layers, as well as on the sagittal sheet (Fig. 96C). Where this sheet is incomplete, and foramens are present (Fig. 99C), the left and right muscles touch each other. The last bone that is drastically modified is the frontal, to which most muscle fibres of the second dilatator operculi bundle reach (Fig. 96A-B). Only the posterior half of the bone is modified. The muscle enters the bone caudally, wedging in-between a ventral layer of the bone covering the braincase and a dorsal layer at the dorsal skull surface. The lateral portion of the frontal, connecting the ventral and dorsal layers and carrying the supraorbital canal of the cranial lateral line system, is deepened, and a vertical ridge is formed (Fig. 99A-B). This ridge is deepest posteriorly, where the supraorbital canal continues into the sphenotic. Still, the epiphysial and parietal branches of the supraorbital canal both exit dorsally in all 180 PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS specimens, i.e., passing through the surface layer. The epiphysial branch splits off the supraorbital canal where it still lies near the surface; the parietal branch, however, starts somewhat further posteriorly, where the main canal is already found deeper in the skull (Fig. 99B). Hence, in large specimens, the parietal branch is seen rising up back up to the surface layer of the frontal, where it exits at the frontal-sphenotic suture. An additional effect of the extreme hypertrophy of the dilatator operculi is the flattening of the anterior braincase itself. Whereas the pterosphenoid is a vertical bone in young specimens, it is squeezed and moved into an almost horizontal plane in large adults. Similarly, the plane of the epioccipital is shifted more horizontally (Fig. 94C). The tight contact between many of the skull bones makes some of the sutures hard to see in most specimens, e.g. between the prootic and the sphenotic. This tight contact might be promoted by the fact that a substantial part of several skull bones serves as muscle insertion, although similar tight contacts and even fusions in this skull region are seen in other loricarioids as well (Reis, 1998; Huysentruyt & Adriaens, 2005b). Thus, below the ‘primary’ skull roof, composing the dorsal roof of the skull, a ‘secondary’ skull roof is formed between the brain and the dilatator operculi muscles. The bones contributing to this extra roof are the frontal, sphenotic, parieto-supraoccipital, compound pterotic, pterosphenoid and epioccipital (the latter two only due to their altered orientation). The sensory canals are ‘forced’ deeper inside the skull, and exiting branches rise towards the skull surface via bone trabecles or vertical bone ridges.

FUNCTIONING AND USE OF THE CHEEK-SPINE APPARATUS

The cheek spines of Ancistrus cf. triradiatus can be erected and retracted within 100 milliseconds, or can stay erected for a longer period. The functioning of the apparatus can be understood by manipulating the dilatator or levator operculi muscles in freshly killed specimens. See figure 93 for an illustration of the functioning of the apparatus. Manipulation of both muscles (1 on figure 93D) suggests that they have the same function: rotating the opercle upward (2). The direction of this movement is related to the long, hinge-like articulation of the opercle to the hyomandibula (see above). The long anterior process of the opercle, pointing medioventrally at rest, is swung outward (3). The bases (ossicles) of the cheek spines lie in a concave plane when retracted, so that the spines are packed together (Fig. 93A-C). The opercular motion causes the cheek-spine bases to bulge laterally, directing the spines towards the possible danger or competitor. Also, the plane of the bases is now convex, so that the spines point in various directions (4) (see also figure 93B).The cheek- spine bases are anchored to the quadrate via the chain of cheek plates, limiting the outward movement, and thus causing them to lie in a convex plane: the divergence of the cheek spines when completely erected is caused by this limitation of outward movement (5). High-speed PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS 181 video images indeed show that the spines are diverged most only when they are completely erected. Pulling the adductor operculi results in retraction of the whole apparatus. The system appears to be functional from around the 21 mm SL stage, when all elements are present, and manipulation results in erection and retraction of the (still small) spines. The suturing of the hyomandibula to the compound pterotic is present in all loricariids examined by me, as well as in Hypostomus plecostomus (Schaefer, 1987), but is shorter and not as tight as in Ancistrus cf. triradiatus. The strong suture in A. cf. triradiatus can be functionally interpreted as a fortification preventing excessive outward movement of the hyomandibula when the opercular muscles pull the opercle outward. Also, some of the dilatator bundles originate on the hyomandibula instead of on the neurocranium, so that a rather immovable hyomandibula is advantageous to the function of these bundles. Among the examined species, the cheek-spine apparatus and the hypertrophied opercular musculature are also present in Ancistrus ranunculus and A. dolichopterus. The impact on the skull roof is similar, though slightly less pronounced: the two layers of the frontal are slightly less separated in adults of these species, and the medial contact zone between both layers of the parieto-supraoccipital is somewhat broader (the bone is less deep). Aquarium observations demonstrate the biological role of the cheek-spine apparatus in Ancistrus cf. triradiatus. In case of a limited food supply individuals might be willing or unwilling to share resources. In the latter case a sudden lateral bending of the body is performed so that the anterior body region is rammed towards the competitor. The erected pectoral fins are used as well. This behaviour has also been seen in A. ranunculus, Pterygoplichthys lituratus, Sturisoma aureum, Rineloricaria parva and Farlowella acus. The three examined Ancistrus species, however, often combine the ramming of the anterior body region with the erection of the cheek spines, and usually only at the side of the competitor. The fishes also erect the spines when threatened, or when two males compete for the same cavity (which provides protection and nesting opportunity to the winning male). The males face each other snout to snout, then position their head next to each other, and attempt to hit the opponent and push him away. Such territorial fights can continue for several minutes. Most likely, the cheek spines have a display function as well. The large, fleshy snout tentacles (Figs 5, 94A) might well have the same purpose of display. Sabaj et al. (1999) suggested the use of these tentacles as ‘larval’ mimicry to attract females. Surely, neither hypothesis excludes the other.

8.4. DISCUSSION

The hinge direction of the opercle-hyomandibular articulation has undergone an evolutionary shift. In the primitive catfish family Diplomystidae and the loricarioid 182 PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS

Nematogenyidae, it is almost vertical (Arratia, 1987, 1992). The same configuration is present in most teleosts, i.e., the opercular opening is directed posteriorly. In the loricarioid Trichomycteridae and Callichthyidae the direction has become more oblique, although not completely horizontal (Arratia, 1992; Huysentruyt & Adriaens, 2005b). In most Loricariidae the position of the opercular articulation with the hyomandibula is situated dorsally on the opercle, instead of anteriorly, a fact also noticed in other loricariids (Schaefer, 1987; Chapters 6.2 and 6.3). The hinge has become horizontal, so that the opercle swings not only outward and inward, but also upward and downward (Howes, 1983a; pers. ob.). Hence the hinge has shifted from functioning in a vertical plane, as in most fishes, to a horizontal plane. This situation is at its most extreme in the ancistrines, e.g., Ancistrus, with the strongly reinforced hinge orientated from dorsorostral to ventrocaudal (Fig. 93). This infers an evolutionary rotation of the opercle-hyomandibular articulation of more than 90°, compared to a diplomystid or trichomycterid-like ancestral configuration. The articulation plane is also reflected in the tendon direction of the dilatator and levator operculi (i.e., running dorsocaudally from the opercle, more or less perpendicular to the hinge; Fig. 93). I agree with Howes (1983a) that the position and mobility of the opercle implies that it takes little or (most probably) no part in opening or closing of the branchiostegal membrane, which is situated at the ventral, and not at the lateral side of the head. The absence of significant movement during normal respiration or feeding is supported by preliminary results from a kinematic study on Ancistrus cf. triradiatus and Pterygoplichthys lituratus (see Chapter 6.1 for A. cf. triradiatus). These and Howes’ observations corroborate the hypothesis that the opercle and the three opercular muscles have lost their role in the respiratory mechanism. The branchiostegal membrane slit, well ventral to the opercle, is the only exhalant opening; the membrane is operated by the hyohyoidei musculature. I define the cheek-spine apparatus as consisting of a series of several, loose cheek spines that articulate with the opercle, and can be erected by the movements of this bone. The presence of a large anteroventral opercular process is typical only for those loricariid taxa that possess a cheek-spine apparatus. It is absent in Farlowella acus, Sturisoma aureum, Rineloricaria parva, Otocinclus vestitus, and Pterygoplichthys lituratus. The latter species has a minuscule ‘opercular apparatus’, consisting merely of an evertible opercle (pers. ob.). The opercular musculature appears to be highly variable in both presence and size within the loricariid family. In his study of the loricariid cranial muscles, Howes (1983a) correlated the atrophy or disappearance of opercular muscles in some species to a small size and immobility of the opercle. Please refer to Howes (1983a: fig. 9) for an excellent illustration of the diversity in sizes and shapes of the opercular muscles in loricariids. The adductor operculi, which usually connects the opercle to the neurocranium floor in teleosts (Winterbottom, 1974), is seen inserting on the preopercle instead of the opercle in PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS 183

Pseudacanthicus, Stoniella and Panaque; it is weakly developed in Loricaria, Sturisoma and Farlowella, and is absent in Pseudohemiodon and Hemiodonichthys (Howes, 1983a; Chapter 6.2). The levator operculi, originating from the hyomandibula, sphenotic, pterotic or posttemporal in teleosts (Winterbottom, 1974), is weakly developed in Loricaria, Sturisoma and Farlowella, and absent in Pseudohemiodon and Hemiodonichthys (Howes, 1983a; Chapter 6.2). The dilatator operculi is found in all loricariids examined by Howes (1983a) and me, although it is small in Pseudohemiodon and Hemiodonichthys, where it is virtually reduced to a tendon running in a lateral hyomandibular groove (Howes, 1983a). Winterbottom (1974) mentions the sphenotic, frontal, pterotic and hyomandibula as possible sites of origin for the dilatator operculi in teleosts. Ancistrus is remarkable in having an extremely well developed dilatator operculi originating from all of these bones, as well as from the parieto-supraoccipital and the dorsal process of the sixth vertebra. The invasion of the skull roof bones by the opercular musculature appears to be unique among teleosts; it is more common for large cranial muscles (e.g., the adductor mandibulae) to overgrow the surface of the skull (e.g., Liem, 1980b; De Schepper et al., 2005). In Ancistrus, invasion starts at cartilaginous zones (taenia marginalis and otic capsule), and from there separates the upper and deeper layer of the bones, using both layers as points of insertion. In later stages, growth and remodulation of bone tissue itself, and not the ossification of growing cartilaginous tissue becomes dominant. This explains, for example, the progressive deepening of the medial part of the parieto-supraoccipital, without the presence of cartilage parts. The apparatus was found to be more developed in the male than in the female specimens. Related to this, the dilatator operculi size and the modifications of the skull bones are relatively larger in males. This corresponds well with the fact that the apparatus is used more by the males during territorial behaviour, and represents a sexual dimorphism that is mostly unrevealed externally. The female cheek spines never grow larger than those of similar-sized males that just reach maturity: this external difference appears to be better correlated with gross body size (but can still be called sexual dimorphism). The depth of the compound pterotic and the parieto-supraoccipital is substantially less in females. Also, only a negligible portion of the dorsal process of the sixth vertebra, and the caudalmost end of the frontal, serve as insertion for the dilatator operculi; hence, the latter bone is essentially unchanged in females.

Within the Loricariidae the erectile cheek-spine apparatus has evolved in the hypostomine Ancistrini (Isbrücker, 1980; Montoya-Burgos et al., 1997), and is one of the synapomorphies of this tribe (Schaefer, 1987). The more weakly developed ‘opercular apparatus’, as described above for Pterygoplichthys lituratus, is present in the tribe Pterygoplichthini (Armbruster, 184 PART 8 — THE CHEEK-SPINE APPARATUS IN ANCISTRUS

2004). Evertible cheek plates as in the cheek-spine apparatus of Ancistrus, but lacking hypertrophied odontodes, are present in some Pterygoplichthys species (Armbruster, 2004). The ancistrine cheek-spine apparatus could not have evolved as such without the decoupling of the opercular region from the jaw movements. The interoperculo-mandibular ligament has been lost in most loricariids and related astroblepids (Schaefer, 1988; Armbruster, 2004). In most loricariids the interopercle itself has been lost too (Schaefer, 1988; Armbruster, 2004; Chapters 4.2 and 6.1). The term interoperculum used by Howes (1983a:335), referring to the largest cheek plate, is erroneous. The two largest cheek plates, that have been mentioned in other loricariids as well (Schaefer, 1987, 1988, 1997; Chapters 6.2 and 6.3), might well have contributed to the evolutionary origin of the whole cheek-spine apparatus, as they provided a strong articular connection between the cheek spines and the suspensorium (quadrate), which I consider necessary for the apparatus to be used as an aggressive, powerful tool. The homology of this largest plate, carrying the distal portion of the preopercular canal, and the bone carrying the same canal in the other examined loricariids, seems plausible. A phylogenetic discussion on this bone and the possible true interopercle in the loricariid Delturus is given by Armbruster (2004). In conclusion, the evolution towards the ancistrine cheek-spine apparatus has been made possible by the evolutionary decoupling of the opercle from the lower jaw movements, the presence of odontodes on bony platelets in the skin, the possibility to develop a series of articulations between the opercle, the cheek spines, the cheek plates anterior to them, and the quadrate, and, last but not least, the possibility of the opercular musculature to expand substantially inside newly evolved skull cavities. A remarkable convergence to the loricariid cheek-spine apparatus is seen in a distantly related loricarioid family. Some trichomycterids (Branchioica, Pareiodon, Vandellia) also have an anteroventral opercular process and an enlarged dilatator operculi, growing on, not inside the skull (Howes, 1983a; pers. ob.). There are no separate erectile odontodes, but both the opercle and interopercle carry large odontodes and are alternately abducted, providing a means of attachment to the substrate of these detritivores (Eigenmann, 1918). Additionally, they function as anchors in the parasitic stegophiline and blood parasitic vandelliine trichomycterids (Kelley & Atz, 1964; Baskin et al., 1980; Moodie & Power, 1982; Breault, 1991; Spotte et al., 2001).

PART 9

GENERAL DISCUSSION

PART 9 — GENERAL DISCUSSION 185

In this general discussion on the results presented in this dissertation, the adult morphology of the examined loricariid species will be discussed first, with emphasis on Ancistrus cf. triradiatus. Some functional considerations are made (Chapter 9.1). Second, the evolution towards the loricariid morphology, and especially the suckermouth, the peculiar loricariid respiration and feeding modes are discussed, followed by a short appraisal of the link between evolution and the high diversity of the family (Chapter 9.2). Third, the ontogenetic results of this dissertation are linked to the first two parts, resulting in considerations on the ontogeny of function of the specific loricariid head configuration, and on homologies of structures (Chapter 9.3). Finally, a conclusive synopsis summarizes the major findings in an evolutionary and ecological context (Chapter 9.4).

9.1. MORPHOLOGY AND FUNCTION IN LORICARIIDAE

Morphological examination of Ancistrus cf. triradiatus and other loricariid species, combined with comparisons with other siluriforms and teleosts, as well as a literature review of relevant papers on loricariid morphology, allows a discussion on those morphological features that reflect, or, in fact, constitute the suckermouth apparatus in the loricariid head. This has already partially been done in part 6, especially in the discussion in chapter 6.1. Some preliminary answers were formulated to questions about the functional-morphological basis of the suckermouth, the various elements involved in the peculiar respiration in combination with substrate attachment, and the algae scraping apparatus. The execution of the basic functions of respiration and feeding might appear complicated in loricariids, when compared to most other teleosts. The attachment with a suckermouth, while breathing, has puzzled the first authors observing and describing the loricariid head (e.g., Regan, 1904; Hora, 1930), leading them to conclude that inspiration must occur via the branchiostegal slits, or that the fishes simply could not breathe and attach to a substrate at the same time. The simultaneous occurrence of oral inspiration and sucker attachment was discovered and demonstrated surprisingly late, by Alexander (1965). Subsequent studies on the function of the suckermouth are virtually non-existent, except for the short papers by Gradwell (1971b) and Vandewalle et al. (1986). The present dissertation has led to more insight in the elements that are involved in loricariid respiration. Ideally, a thorough kinematic analysis should further elucidate the role of each element. Any kinematic study, however, must be based on detailed knowledge of the anatomical basis of the biological 186 PART 9 — GENERAL DISCUSSION system under investigation. For that purpose, the detailed anatomical descriptions of loricariids in part 6 may be very useful. Intensely modified elements that are considered specializations or adaptations for respiration while attaching with the suckermouth, include the expanded upper and, most spectacularly, lower lip. The resulting mouth disc is the most conspicuous tool participating in the suction process. The epidermal brushes or unculi on the lip papillae might aid in attachment to rough or irregularly shaped substrates (Part 7). Mobility of, especially, the lower lip is substantial; both the intermandibularis anterior and posterior muscles have separate parts attaching to the lower lip tissue (Chapter 5.2). Inspiration via the small lateral lip furrow (Chapter 6.1) is controlled by movements of the short maxillary barbel, which is elevated by the levator tentaculi, a muscle unique to loricariids, and, perhaps, the related astroblepids and scoloplacids (Chapter 5.1). Its newly proposed name reflects its non- homology with the retractor tentaculi muscle of many siluriforms. As far as can be deduced from live observations and the study of anatomical structures, the hyoid and suspensorium movements, responsible for volume changes of the buccal cavity, are essentially the same as in other teleosts (Chapters 5.2 and 6.1; Alexander, 1969, 1970; Gosline, 1973; Anker, 1974; Elshoud-Oldenhave & Osse, 1976; Elshoud, 1978; Muller, 1989a; Aerts, 1991; Hunt von Herbing et al., 1996; Adriaens & Verraes, 1997b). The broad hyoid bar (in this thesis especially in the Ancistrus species) suggests that it has a relatively high contribution to volume changes, compared to the suspensoria (Chapter 5.2). The posterior and lateral walls of the orobranchial cavity appear to be very rigid in loricariids. Posteriorly, the immobile pectoral girdle, with substantial vertical medial flanges of the cleithrum, as well as the ventrally protruding Baudelot’s ligament of the neurocranium, form a rigid wall. Alexander (1965) already regarded this as an adaptation to withstand the negative pressure caused by the suction, hydrostatically isolating the orobranchial cavity from the intestinal cavity. Laterally, the metapterygoid and hyomandibula are well attached to the neurocranium (Chapters 4.2 and 6.1). In many loricariids a long suture is present between the hyomandibula and the compound pterotic (Armbruster, 2004). The very long articulation between the suspensorium and hyoid bar could be an additional adaptation to resist high negative forces while sucking. The implication of the loss of the interhyal cartilage and the formation a sesamoid bone in this articulation is still unclear. Two other structures, generally involved in controlling water flow in and out of the orobranchial cavity, are the paired branchiostegal membrane and the oral valve. An operculo- branchiostegal apparatus provides an efficient seal so that large negative pressures can be potentially generated in the orobranchial cavity (Drost & van den Boogaart, 1986b; Hunt von Herbing et al., 1996b). In loricariids, only a small slit is present, almost exclusively bordered and supported by the four branchiostegal rays, and not by the opercle. A small slit might be a PART 9 — GENERAL DISCUSSION 187 more efficiently closable seal than a longer one. The absence of actual participation (i.e., any significant movement) of the opercle in the movements of this slit (Howes, 1983a; Chapter 6.1) implies less to no contribution of the opercular musculature (Part 8), and, consequently, a relatively increased importance of the branchiostegal muscles. The reduction of opercular muscles in many loricariids points in this direction (Howes, 1983a; Chapter 6.1, Part 8). I have hypothesized in chapter 6.1 that the hyohyoideus abductor, having a direction that makes it impossible to open the branchiostegal membrane, may have acquired the same function as the hyohyoidei adductores, ensuring a firm branchiostegal seal that can possibly be closed faster or sooner than a more passive branchiostegal valve. The suckermouth attachment, generated by forceful suction, might benefit from this configuration. The opening of the slit might well have become passive, induced by the positive pressure as the orobranchial volume again decreases. Such speculations, however, cannot resolve this issue; kinematic data are needed. The differentiation of a subdivision of the adductor mandibulae complex, named retractor veli by me (Chapter 5.1, Part 6), into one or two muscle bundles mediating the oral valve (apparently closing it), might well be an additional adaptation to prevent backflow of water during this expiration phase (Part 6).

Several additional cranial elements differ from the head configuration in non-loricariid catfishes, and can be related to the scraping feeding mode of loricariids. Both the upper and lower jaws are modified as a specialization for the efficient scraping of algae (and other food items) from submerged substrates. Chapter 3.2 discusses the anteroventral extension of the upper snout region, bringing the upper jaws anterior to the lower jaws, and lowering them so that they reach the substrate. The ventrally directed mesethmoid, and especially its medioventral disc providing an even more ventral suspension point for the premaxillae, are beneficial for bringing the upper jaws more ventrally. The suction disc, formed by the upper and lower lips, also benefits from this snout morphology. The ventral mesethmoid disc is also found in astroblepids, which have a similar suckermouth (Schaefer, 1990). Even more spectacular is the ventral ánd medial rotation of the lower jaws. Unlike the upper jaws, that are tightly interconnected by the interpremaxillary ligament, the increased mobility of the lower jaws allows them to be moved more or less independently. No medial connection (symphysis) between them is present, nor is there any medial connection to the hyoid arch. The mandibulo-hyoid ligaments are short, and insert on the lateralmost aspect of the jaw, almost on the articulation with the quadrate. There is no interoperculo-mandibular ligament; neither is there an interopercle in Ancistrus cf. triradiatus and many other loricariids (Chapter 6.1; Armbruster, 2004). The upper jaws rotate around their suspension on the mesethmoid disc; the lower jaws rotate around the articulation point with the quadrate (and around their 188 PART 9 — GENERAL DISCUSSION axis), being supported by a medial cartilage plug, which is not connected to them (Chapter 6.1). This contradicts Schaefer & Lauder (1986), who described this plug as a novel link between the mandibular and hyoid arches, but is strongly supported by morphological, experimental and histological arguments (Chapter 6.1). This and other evolutionary skeletal decoupling events related to the high jaw mobility are discussed in Chapter 9.2. The presence of antagonistic musculature of the lower jaws (adductor mandibulae and intermandibularis posterior pars dentalis) is common in teleosts; the musculature acting on the upper jaws (retractor premaxillae and extensor tentaculi pars medialis) is a synapomorphy for loricariids (Chapters 5.1 and 6.1; see also Chapter 9.2). Especially for the lower jaws, bilateral asymmetry of muscle activity would allow a finer regulation of the scraping of irregular substrates; still, such asymmetry appears to be a rather common phenomenon in fishes (Liem, 1979). Kinematic analyses including electromyographic data are needed to investigate this in loricariids. The whole process of scraping food from substrates differs significantly from the suction- feeding dynamics in fishes feeding in the water column (e.g., Drost et al., 1988): the spatial organization of the low pressure when sucking against a substrate (instead of in the water column), and, consequently, water flow patterns into the mouth might be a rewarding research topic. A study of Nauwelaerts et al. (2006) on bamboo sharks feeding close to the sea floor showed that the fluid velocity field is strongly altered by the vicinity of a substrate. Also, Liem (1980a), on algae-feeding cichlids (that do not have a ventrally oriented mouth) recorded irregular and often non-cyclical jaw movements, apparently adjusted to the irregularities of the substrate. Live observations of feeding in various loricariid species suggest that this is the case in loricariids as well, with some differences observed between the movements of the left and right lower jaws. Figure 64 shows how the mouth closing phase while feeding becomes longer, relative to the (shorter) opening phase, most probably related to the resistance of the substrate. This, and the shortening of the gape cycle, have also been noted in a non-related but trophically similar organism: tadpoles (Larson & Reilly, 2003). A comparison between tadpoles and free-living embryos of Ancistrus cf. triradiatus raises the intriguing question whether the similar skeletal morphology is accompanied by similar (convergent) musculature and muscle activity patterns (Haas, 2001; Larson & Reilly, 2003). Except for the short papers by Gradwell (1971b) and Vandewalle et al. (1986), and chapter 6.1 of this dissertation, no attempt has been done to explore loricariid muscle activity or skeletal motion patterns. Thus, the biomechanical aspect of loricariid suckermouth attachment, respiration, and feeding, still remains an unexplored research field. It is, in fact, surprising, that the biomechanical processes in loricariid suckermouth attachment, respiration, and feeding, have escaped a thorough examination thus far. In the frame of the PART 9 — GENERAL DISCUSSION 189 project introduced in chapter 1.1, biomechanical data on Pterygoplichthys lituratus, including high-speed video (normal and X-ray) and electromyographical sequences, have been gathered by Anthony Herrel and me. At this moment, the analysis of these data is still to be completed.

190 PART 9 — GENERAL DISCUSSION

9.2. EVOLUTION OF THE SUCKERMOUTH APPARATUS

REAPPRAISAL OF THE PAPER BY SCHAEFER & LAUDER (1986) Phylogenetic affinities within the loricarioid clade are relatively well known, and are supported by both morphological and molecular studies (Baskin, 1973; Howes, 1983a; de Pinna, 1993; 1998; Diogo, 2005; Sullivan et al., 2006). Schaefer & Lauder (1986) proposed an evolutionary pattern that would have led to the loricariid head morphology with a suckermouth and a food scraping apparatus. A short summary of the most important structural innovations in the clade constituted by the Callichthyidae, Astroblepidae and Loricariidae, is given below. Due to insufficient material, Schaefer & Lauder (1986) did not examine Scoloplacidae (sister group of Astroblepidae + Loricariidae), but see the paper of Schaefer (1990) on Scoloplacidae. Their conclusions are merely summarized here; a reappraisal and discussion are given below.

For all three families (as opposed to Trichomycteridae and Nematogenyidae): 1) The presence of highly mobile premaxillae, thanks to changes in the joint with the mesethmoid. 2) A dorsal extension of a portion of the adductor mandibulae complex, reaching to a connective tissue sheet (primordial ligament of Chapter 5.1). This muscle division has been termed retractor tentaculi by Howes (1983a), Huysentruyt et al. (submitted) and me (Chapter 5.1, Part 6).

For Astroblepidae + Loricariidae: 3) Direct insertion of the adductor mandibulae division mentioned in (2) onto the premaxilla, resulting in an independent movement of the upper jaw. 4) Loss of the connection between the connective tissue sheet (primordial ligament) and the maxilla. 5) Novel acquisition of a muscle [referred to as muscle c by Howes (1983a)], with a direct insertion onto the premaxilla.* 6) Loss of the interoperculo-mandibular ligament, resulting in the loss of the opercular jaw depression mechanism. [Later it was found that not all loricariids have lost this ligament (Armbruster, 2004).] 7) Loss of the tight connection between both lower jaws, giving these elements more (independent) mobility. PART 9 — GENERAL DISCUSSION 191

8) Novel acquisition of a cartilage plug, situated medially between the mandibular and hyoid arches. 9) Subdivision of the intermandibularis posterior (‘geniohyoideus’) into two separate parts, inserting on the lower jaw and on the cartilage plug, respectively. 10) Bilateral bifurcation of the hyohyoideus inferior.

For Loricariidae: 11) Establishment of a new direct connection between the mandibular and hyoid arches via the cartilage plug of (8). 12) Novel acquisition of a muscle, termed retractor palatini [muscle d of Howes (1983a)], derived from the adductor mandibulae complex. It has an indirect insertion onto the premaxilla via a connective tissue sheet.*

* Due to misinterpretation, feature (5) and (12) were erroneously switched in the paper of Schaefer & Lauder (1986), but later this was adjusted by Schaefer (1990:201). It is, however, not clear then how the muscle in (5) relates to the muscle in (3) in Schaefer’s argumentation. As shown in chapter 5.1, they are the same (see also below).

COUPLINGS, DECOUPLINGS AND NEW STRUCTURES The evolutionary transformation resulting in the loricariid head configuration of both hard and soft parts includes coupling and decoupling events, and the establishment of new structures. One could discuss on where to draw the line between a decoupling event and the establishment of a new structure, as most novel structures arise from existing elements, but are often modified, expanded, or intensely remodulated. Examples of true new structures in loricariid evolution are the cartilage plug (see Part 6 and feature (8) above), and the sesamoid bone embedded in the articular ligament between the suspensorium and the posterior ceratohyal (Chapter 4.2). Here, cartilage and bone, respectively, have developed from other connective tissue types. Thus, three main categories of structural innovations exist: new or de novo structures, coupling events and decoupling events. The intrinsic value of a de novo structure needs no further explanation (but see footnote in paragraph 1.1.2). Coupling events combine or link two or more structures, giving rise to new functions that are possible thanks to these linkings. The fact that the coupled structures usually become unable to perform independent tasks, could be considered as a constraint, but this constraint does not weigh substantially against the advantage of the coupling: in a case where the disadvantage of the constraint is larger than the advantage of the coupling, the coupling would not be selected for, and would therefore not exist. Only in descendant taxa, encountering other environmental and thus 192 PART 9 — GENERAL DISCUSSION possibly, structural, needs, such coupling could impair adaptability. Decoupling events, on the other hand, of two plesiomorpically linked structures, can be considered a release of constraint and an onset of new possibilities of functions performed by one of the structures, while the other continues to perform the original function (Lauder et al., 1989). Decouplings have often been linked to an increased diversity (structural, and in number of species; see below).

THE UPPER JAW In the evolution leading towards the loricariids, Schaefer & Lauder (1986) found a remarkable number of key events that permitted a structural and functional diversification. The list given above contains some events related to the hyoid musculature, but most concern the jaw region. As already mentioned in chapter 5.1, I fully agree with the importance of premaxillary mobility and the mesethmoid connection via the ventral disc, allowing the scraping movements of the upper jaw. Its forward movement (rotation) can truly be termed protrusion, albeit in a way that is different from the four types of protrusion listed by Motta (1984). In the same chapter, I elaborated on the origin of the loricariid retractor premaxillae from the adductor mandibulae division termed ‘retractor tentaculi’ in callichthyids (feature 2 in the list above). This contradicts the ‘sudden’ novel acquisition of the retractor premaxillae as stated by Schaefer & Lauder (1986; feature 5), but represents a well argumented decoupling event (Chapter 5.1). The antagonistic muscle, responsible for the protruding anterior movement of the premaxillae, is the medial part of the fully subdivided extensor tentaculi (Chapter 5.1, Part 6). This complete subdivision is found in loricariids only; a rudimentary differentiation is seen in callichthyids, scoloplacids and astroblepids (Diogo, 2005). I regard this subdivision as an additional decoupling event, occuring gradually in the loricarioid lineage, and allowing an independent mobility of the maxilla and the premaxilla (via the lateral and medial parts of the extensor tentaculi, respectively). This decoupling event most surely benefitted from, or even coincided with the release of a physical constraint as the maxilla and premaxilla lost their tight connection via the primordial ligament (see also feature 4 above).

THE LOWER JAW Schaefer & Lauder (1986) list the loss of the interoperculo-mandibular ligament and the loss of the medial connection between both lower jaws as important events in the evolution towards the high mobility of the lower jaws (features 6 and 7). The medial cartilage plug situated between the mandibular and hyoid arches is presumed to function as a novel link between both arches in loricariids (feature 11). Based on experimental, morphological and histological arguments (Chapter 6.1), I definitively disagree with this latter hypothesis: I PART 9 — GENERAL DISCUSSION 193 found that the plug acts as a supporting and gliding device for the rotating lower jaws, without a firm attachment to them. One could only speculate about ancestral lower jaw configurations intermediate to the plesiomorphic one, as in Callichthyidae (jaws directed anteriorly), and the apomorphic one in Loricariidae (jaws directed medially). Such intermediate, yet selectively advantageous configurations could have been useful to loosen and collect food items on the bottom, especially when the lower lip became relatively mobile as well. It is not easy to imagine an intermediate configuration that perfectly matches any hypothetic intermediate trophic (or other) need. However, one should not overlook the fact that newly evolving patterns and configurations are not ‘perfectly adapted’ from the start, but need only to be adequate, or slightly advantageous, for the current functional needs, in order to provide a selective advantage (Gans, 1988, in Lauder et al., 1989). I consider the angle of the adductor mandibulae muscle in relation to the lower jaw direction a major player in the evolutionary shift of lower jaw movement, from an up-and- down rotation around the articulation with the quadrate (as in Callichthyidae), to an anteroposterior rotational movement around this articulation, and a rotation around the longitudinal axis of the lower jaw (as in Loricariidae). The relative positions of the articulation point, the teeth row, and the coronoid process in figures 37C and 55A-B, illustrate how this second rotation, present in Loricariidae (and possibly Astroblepidae) originates. The importance of the position of the coronoid process in evolutionary changes in jaw mobility has been noticed in other clades as well (Lauder, 1979); the muscle attachment and orientation highly influences its effect on the jaw bone movements. The subdivision of the intermandibularis posterior into dentary and labial parts could also have implied a key innovation, leading to a more mobile, as well as more ventrally oriented food grabbing or collecting tool. Regardless of how and when this subdivision occurred, I consider it an important decoupling event: the lower lip mobility in loricariids and astroblepids is mainly mediated by the labial parts of the intermandibularis anterior and posterior muscles, whereas the dentary parts of both muscles act upon the lower jaws (and possibly indirectly on the lower lip). A detailed functional-morphological study of Astroblepidae and Scoloplacidae could be excitingly informative on both the role of the lower jaw orientation and the intermandibularis posterior differentiation. The findings on intermandibularis posterior morphology in all loricariids examined in this study (Chapter 5.2, Part 6) contradict feature (9) of Schaefer & Lauder (1986), which states that the labial part (their ventral protractor hyoidei part) would insert on the cartilage plug, and not on the lower lip.

194 PART 9 — GENERAL DISCUSSION

HYOID MOVEMENTS Halecostomians possess two biomechanical pathways to lower the mandible: the hyoid coupling (present in all actinopterygians), and the opercular coupling (a synapomorphy of halecostomians) (Schaeffer & Rosen, 1961; Lauder, 1979, 1980). The lower jaws of the loricariids I examined slightly move up and down along with the hyoid during respiration: this confirms the observation of Muller & Osse (1984) that this primary pathway remains functional in most extant species, and serves to increase the volume of the buccal cavity. The only firm connection between the mandible and the hyoid is the mandibulo-hyoid ligament, inserting on the lower jaw just lateral of the articulation with the quadrate. The positional changes in the loricariid lower jaw configuration, and the altered movement of the lower jaw, might be related to the loss of the second, operculo-mandibular, pathway in many loricariids. The decoupling of the lower jaw and the opercular apparatus coincides with a decrease in size of the opercular muscles in many loricariids (Howes, 1983a; Parts 6 and 8), but has opened the possibility to develop the cheek-spine apparatus in Ancistrus and some other genera (Part 8). Taking into account the complete non-related functions of a jaw-lowering system and a defensive spine-erecting mechanism, I consider this the most remarkable (though not necessarily the most important) decoupling event in loricariid evolution.

THE ORAL VALVE MUSCULATURE The retractor veli [retractor palatini in feature (12) of Schaefer & Lauder (1986), but see Chapter 6.1] might have evolved after the establishment of a (loose) connection between the adductor mandibulae and the epithelium lining the oral cavity near the oral valve. This might have aided in a more efficient closure of the valve while sucking onto a substrate, preventing backflow leakage during expiration. Subsequent subdivision of one or two (Otocinclus) separate retractor veli muscles would then have increased the efficiency during later evolution.

EVOLUTION OF MUSCLE ACTIVITY PATTERNS It can be concluded that the number of morphological innovations in the lineage towards Loricariidae is exceptionally high. A still mostly unexplored aspect is the possibility of evolutionary transformations in muscle activity patterns (see also Chapter 9.1). No attempt has been done so far to explore this interesting aspect in view of loricariid feeding. No paper addresses the kinematics or underlying muscle activities with regard to the feeding movements. The preliminary kinematic analysis in chapter 6.1 should best be followed by a more thorough study. After all, both the morphology of the feeding apparatus and the patterns of muscle contraction and skeletal motion observed during feeding are the two major components of the feeding mechanism in teleosts (and other animals) (Wainwright, 1996). In PART 9 — GENERAL DISCUSSION 195 some case studies in other teleostean taxa, evolution of feeding mechanisms has been linked to primarily morphological changes (e.g., Werner, 1977; Wainwright & Lauder, 1992 in Wainwright, 1996), or to essentially relevant transformations in muscle activity patterns (Liem, 1980b; Lauder, 1983). Still, even in the absence of a substantial changes in muscle activity patterns, morphological changes alone could have permitted an important shift in breathing and feeding mechanisms. Mechanical models of hyoid motion, upper jaw protrusion, and lower jaw motion have been used as frameworks for interpreting the consequences of morphological variation among fish species for kinematic patterns during feeding (Wainwright et al., 2000). Considering the fact that especially the latter two points, i.e., upper and lower jaw morphology in Loricariidae, are so vastly different from more basal taxa (e.g., Diplomystidae, Trichomycteridae, Callichthyidae), one could imagine that morphological innovations alone could well have driven most of the transformations in the loricariid feeding mechanism.

EVOLUTIONARY ORDER OF SUCKERMOUTH ATTACHMENT VERSUS SCRAPING FEEDING MODE An intriguing question regarding evolutionary patterns in the Loricarioidea has been posed by Adriaens (2003): would the suckermouth attachment mechanism have evolved first, providing the opportunity for the scraping feeding mode to develop, or vice versa? The first possibility could have occurred if suckermouth attachment was selectively advantageous to enter and successfully colonize fast-running rivers, where other species could barely maintain viable populations. An argument supporting the first hypothesis is the fact that many fish species inhabiting turbulent environments such as rivers or shorelines display attachment modes, albeit often developed from other parts than the mouth (e.g., pelvic fins; Hora, 1930; Saxena & Chandy, 1966; Tilak, 1976). The second possibility, i.e., the scraping feeding mode originated first and led to a ventrally oriented mouth, after which the suction attachment abilities developed, is supported by the fact that a suckermouth appears not to be necessary to live in fast flowing rivers: Fernández & Aquino (1993) found loricariids, as well as other taxa that lack a suckermouth or other attachment modes, in the same turbulent rivers. Of course, the advantage could be small, though still important enough to act as a selective power on those individuals, populations or species possessing some kind of attachment mode. The ability to feed efficiently on epilithic and epiphytic food items such as algae might appear more advantageous than the ability to attach onto a stone or submerged tree branch, but can not be used as a very strong argument to support the second hypothesis. Nor can the amazing array of small details in the feeding apparatus (jaw size and orientation, and tooth and unculus numbers and properties) be applied to unambiguously state that the altered feeding mode is (or has been) most important in loricariid evolution and early radiation. 196 PART 9 — GENERAL DISCUSSION

LORICARIID DIVERSITY Another interesting topic is the exceptionally high species number of Loricariidae (paragraph 1.2.4). One of the basic factors allowing or inducing the wide radiation of the family might be the significant number of evolutionary decoupling events that took place. Decouplings can (but do not have to) result in increased structural or functional diversity in the system in descendant taxa. Examples of decoupling as a process permitting diversity are provided by feeding systems of fishes, salamanders and mammals (Lauder et al., 1989). The process is considered important in the evolutionary radiation of the African cichlids (Liem, 1991). Schaefer & Lauder (1996) tested and confirmed the correlation between decoupling events and morphological diversity in loricarioid taxa. Diversity of loricariids (and other neotropical freshwater taxa) might have been additionally boosted by intense neotropical river drainage pattern changes during the history of the South American continent (Lundberg et al., 1998; Lundberg, 2001). One could argue that highly specialized taxa (e.g., in terms of their feeding apparatus) would be restricted in their range of structural possibilities, and, hence, would be expected to be low in species diversity when compared to taxa displaying a more general morphology and feeding behaviour. Specialization could make taxa more vulnerable in environments that are not very stable throughout long periods. Loricariidae prove that specialization in the head Bauplan can, however, still allow a substantial amount of diversification. I assume that the success of the family is at least partly due to the fact that the loricariid head morphology has not led to a limited and specialized trophic niche occupation, but has opened a vast niche, that could not be exploited so efficiently by taxa without the ventrally placed jaws and suckermouth. The ‘specialized’ morphology still allows to feed on a wide variety of items, associated with underwater substrates (Angelescu & Gneri, 1949; Saul, 1975; Angermeier & Karr, 1983; Power, 1984; Schaefer & Stewart, 1993; Buck & Sazima, 1995; Fugi et al., 1996; Grosman et al., 1996; Aranha et al., 1998; Nelson et al., 1999; Delariva & Agostinho, 2001; Nelson, 2002). Of these food items, algae, and especially detritus, are an abundant resource in many tropical river systems (Bowen, 1984; Power, 1984). Also, several very similar loricariid species can be found to live syntopically (Power, 1984; Jégu et al., 2004). These species often differ in (micro)habitat use, something that can be reflected in differences in body shape (e.g., Armbruster et al., 2001). A large diet overlap, however, is often found, as in similar species groups of other fish families as well (Lowe- McConnell, 1969; McKaye & Marsh, 1983; Mol, 1995). While often trophically relevant differences are found in small details in, e.g., the teeth (Zaret & Smith, 1984; Part 7), it has been found that similar-looking, closely related species often feed on the same food items most of the time, but switch to a narrower food source when food becomes scarce (McKaye & Marsh, 1983). Also, often other factors than food limitation, like predation, may be most PART 9 — GENERAL DISCUSSION 197 important in limiting population size of the different species, thus excluding severe food competition (Mol, 1995). Also, two species are unlikely to share a niche all the time (at all times of the year or throughout life), or in all sections of the habitat. Small oscillations in selection pressures probably have a vital importance in permitting coexistence, operating in different ways on the various stages of the life cycles of co-habiting species (Lowe- McConnell, 1969).

198 PART 9 — GENERAL DISCUSSION

9.3. CONSIDERATIONS REGARDING LORICARIID ONTOGENY

Chapter 3.2 discusses the advantages of a large endogenous yolk supply for ontogenetic trajectories: it can enable the direct development of the definitive adult phenotype, leaving out the larval stage (Orton, 1953; Balon, 1986; see also paragraph 2.3.2). The absence of the larval stage is effectively observed in Ancistrus cf. triradiatus. A substantial yolk reserve also postpones the critical moment of the onset of exogenous feeding to a point where more anatomical structures have had the opportunity to develop. From the moment of hatching, the free-living embryo, especially if it doesn’t live floating pelagically, may be hindred in its movements by the bulky burden of the yolk sac. Thus, this early life strategy has its disadvantage, which, in A. cf. triradiatus, is coped with by means of a slow rostral to ventral shift of the mouth. Balon (1984) considers the transition to exogenous feeding, rather than hatching, to be the decisive threshold of ultimate survival value. Related to this, yolk contributes considerably to individual survival chances, as it allows some kind of learning period in which coordination of viscerocranial structures can increase, and thus prepare for obligate exogenous feeding, while relying on the yolk sac and avoiding starvation in the transition phase (Chapter 3.2; Hunt von Herbing et al., 1996a).

The success in terms of survival during early life history stages can be linked to evolutionary success; the relevance of regarding the earliest free-living stages as equally important for natural selection as the adult morphology and behaviour, has already been recognized by Darwin (1859), and has been repeatedly pleaded for ever since (e.g., Orton, 1955; Lauder et al., 1989; Wake & Roth, 1989; Galis et al., 1994; Adriaens et al., 2001). The fact that ontogenetic stages must be functional at all times, while allowing the adult phenotype to develop, has drawn attention to the expected, and confirmed, correlation between the anatomical structures and behavioural characteristics, and the functional needs related to the environment and size of the individuals (e.g., Liem, 1991; Galis et al., 1994; Kohno et al., 1996; Doi et al., 1997; Osse et al., 1997; Adriaens et al., 2001). Therefore, the ontogenetic trajectory in Ancistrus cf. triradiatus can be expected to display some features or transformation sequences that can possibly be interpreted as such. The gradual mouth displacement from rostral to ventral is obviously the most conspicuous feature. Some other embryonic and free-living embryonic traits that are dealt with in chapter 3.2, have been observed in many other fishes, and appear to be general teleostan features: PART 9 — GENERAL DISCUSSION 199 especially the allometric growth of the head and tail regions, and the presence of the larval finfold (Fuiman, 1983; Strauss, 1995; Osse et al., 1997; van Snik et al., 1997). In Ancistrus cf. triradiatus, development of some anatomical structures can be linked to the mouth displacement (see Chapter 3.2). Not only suckermouth attachment is important for free-living embryos of A. cf. triradiatus: actual respiration is observed from before hatching, and occurs together with attachment from the moment of hatching. In chapters 4.1 and 4.2 the early development of the maxillary cartilages and bones is described; these elements are responsible for the creation of lateral lip furrows allowing inspiration during attachment. They are relatively large in the earliest stages (e.g., fig. 29), when compared to juvenile and adult proportions (e.g., fig. 37). It is difficult to infer early muscle functionality from visible insertions or muscle fibre development; clear insertions of the muscles acting upon the barbel are actually only seen unambiguously at 8.0 mm SL, and not yet at 6.1 mm SL (just before hatching; Chapter 5.1). The first observations of barbel and lip movements is at 6.3 mm SL; Chapter 3.2). The last branchiostegal ray (appearing between 7.4 and 8.0 mm) develops after the first observation of true, unidirectional respiration (6.3 mm), but the membrane could well close efficiently with less than four rays in these young (and small) specimens. The hyohyoidei adductores are not observed in the 6.1 mm stage, but are present in the 8.0 mm stage; the hyohyoidei abductor is present (but without insertion) in the 6.1 mm stage (Chapter 5.2). While opening of this membrane might occur passively, the closing (especially when sucking onto a substrate) clearly benefits from a muscular aid. The importance of the opercular or branchiostegal seal for mouth suction (whether for respiration, attachment, or suction feeding) has been stressed for early life stages of other teleosts as well (Drost & van den Boogaart, 1986b). The synergetic effect of the hyohyoidei adductores and abductor in loricariids has been discussed earlier (Chapters 6.1 and 9.1). Another transformation that can be related to respiration has been discussed in chapter 5.3: the angle between the suspensorium and the neurocranium. The ventral portion of the (still largely cartilaginous) suspensorium shifts forward, bringing the hyoid bar somewhat more anteriorly (compare figs 16, 18 and 25). This is not unique for Ancistrus cf. triradiatus, as it has been noted in Clarias gariepinus as well (Vandewalle et al., 1985). In A. cf. triradiatus, the anterior shift of the hyoid bar is even less than in C. gariepinus, what can probably be related to the relatively posterior position of the lower jaw (which is also rotated medially). As opposed to C. gariepinus (Vandewalle et al., 1985), the small size of the branchiostegal membrane (and musculature) in A. cf. triradiatus throughout ontogeny has led me to the conclusion that in the latter species, no significant ontogenetic swith in the importance of the buccal pressure pump and the operculo-branchiostegal suction pump occurs: the latter is clearly less developed. 200 PART 9 — GENERAL DISCUSSION

Some changes in the hyoid-suspensorial apparatus have been assumed, at the level of articulation points (Chapter 5.3): during ontogeny, the continuous cartilage of the chondrocranium containing pliable zones (Fig. 53) is replaced by bony elements with definitive, cartilaginous articulations. In the 6.1 mm stage the hyoid, suspensoria and neurocranium are all continuous. During ontogeny the articulation between the hyoid and the suspensorium changes from a point articulation (at the interhyal connection; e.g., figs 19 and 53) to a long articulation with a ligamentously embedded sesamoid bone. The functional implications of these changes are difficult to assess. Their timing and sequence might be adjusted to certain functional needs, or they might be a mere consequence of the fact that it takes some time to develop definitive structures like bones and ligaments.

The first exogenous feeding is observed much later than the first respiration movements (Chapter 3.2). Intestinal content was observed two to three days before yolk sac depletion (thus around 8.7 mm SL). Actual feeding movements (i.e., including jaw movements) could well start later, as detritus might have simply been swallowed after unintentionally being sucked into the mouth during suckermouth attachment. Serial sections have shown that gill rakers, providing an efficient sieve for feeding, are only rudimentary in the 8.0 mm stage, but are well developed in the 10.2 mm and later stages, suggesting that the complete development of the gill rakers conincides wit the depletion of the yolk sac. Teeth with a biscuspid crown, as well as unculi, are present only at the moment of complete yolk resorption (Part 7). As such, it appears that movements enabling the most rudimentary form of feeding (i.e. suction, primarily serving respiration) are effective before the complete development of the devices needed for efficient food processing. There is no recognizable time lap between visible insertions of muscles related to respiration only, and insertions of those muscles that can be presumed to be functional during feeding only (e.g., retractor premaxillae; Chapters 5.1, 5.2 and 5.3). Still, it is clear that true respiration (from the moment of hatching) starts well before feeding in Ancistrus cf. triradiatus. This observation is completely opposite to what Hunt von Herbing (2001) observed in cod larvae. She concluded that for cod, Gadus morhua, and perhaps most fish larvae, initial viscerocranial structures appear to be specialized for feeding. Once fish larvae have survived the critical transition from endogenous to exogenous feeding, the transition to branchial respiration can proceed. Two factors might explain this great difference between G. morhua and A. cf. triradiatus. The first is the large supply of yolk in the latter species, allowing it to grow to a larger size without the need of feeding. Due to this size, muscular branchial respiration will sooner be essential, as cutaneous respiration is only sufficient in the smallest stages. The second factor might be the difference between a cod larva floating and swimming freely in the ocean, and the loricariid embryo, sucking onto the PART 9 — GENERAL DISCUSSION 201 roof of the nest cavity. Oxygen availability probably is much higher for the pelagic cod larva than for the loricariid embryo, which is handicaped in terms of water inflow and oxygen availability, especially in its branchial cavity.

The ‘luxury’ of the extended period of endogenous yolk feeding in Ancistrus cf. triradiatus is accompanied by the absence of an ontogenetic diet shift, or even an observed (thus substantial) shift in feeding mechanics. Young fish of many species display a shift of ram-feeding during the larval stage, followed by the definitive suction-feeding in juvenile and adult stages (e.g., Coughlin, 1994; Cook, 1996). Ontogenetic diet shifts are a general feature, related to the mouth gape, and the sequence of the development of structures involved in food capturing and processing (Werner & Gilliam, 1984; Cook, 1996; Hunt von Herbing, 1996a; Kohno et al., 1996; Doi et al., 1997). Food items in those loricariids that mainly feed on algae and detritus, remain more or less the same size, and never approach the size of the mouth gape, not even in the smallest stages. This is clear from the observations I did on A. cf. triradiatus. Moreover, Mérigoux & Ponton (1998) studied the trophic ecology of several neotropical freshwater fish taxa, and found that most taxa showed clear ontogenetic diet shifts, while the herbivorous and detritivorous Ancistrus cf. hoplogenys belonged to the same diet group from early life stages to older juveniles. When present, ontogenetic diet shifts are often associated with or caused by shifts in habitat (Werner & Gilliam, 1984), but in most freshwater fishes the early life history stages and the adults occupy more similar environments (Orton, 1955). Some spatial segregation might occur (e.g., river banks or flooded forest areas versus deeper river beds), but this is never as substantial as in, especially, marine fishes (nurseries in mangroves and other coastal habitats).

Throughout this dissertation, argumentations are given for the homology of anatomical structures in Ancistrus cf. triradiatus with structures in other teleosts (especially siluriforms). Ontogenetic arguments can be extremely useful when studying evolution. Without the study of ontogeny, evolutionary transformations are based on the presence or absence of structures, and possible similarities, only. In the case of loricarioid evolution, the paper of Schaefer & Lauder (1986) has been complemented, or often adjusted, by several facts obtained from ontogenetic examination in the present dissertation (see the previous Chapter 9.2). The ontogeny criterion is a good addition to the outgroup criterion, as it (theoretically) allows the direct observation of transformations that reflect evolutionary novelties or other changes. Direct observations of ontogenetic transformations in one single organism are usually impossible in large animals (as opposed to, for example, nematodes), because of the technical difficulties of studying internal anatomy of an organism through time (Kluge & Strauss, 202 PART 9 — GENERAL DISCUSSION

1985). The study of a high number of ontogenetic stages, though, approaches the value of the direct observation. I will not repeat all assumed or hypothesized homologies (or non-homologies) featuring in the previous chapters. Some major findings are the non-homology of the interhyal cartilage and the (sesamoid) bone between the posterior ceratohyal and the hyomandibula, the non- homology of the largest cheek plate (Ancistrus cf. triradiatus) and the interopercle (non- loricariid siluriforms), the homology of the retractor premaxillae (A. cf. triradiatus) and the retractor tentaculi (Corydoras aeneus), the non-homology of the levator tentaculi (A. cf. triradiatus) and the retractor tentaculi (C. aeneus), and the homology of the adductor hyomandibulae in A. cf. triradiatus and some other siluriform families.

Some ontogenetic transformations also appear to reflect the evolutionary origin of complex shapes or configurations in Ancistrus cf. triradiatus. The first phases in the gradual shift in the upper snout region (Chapter 3.2) are reminiscent to the plesiomorphic state of an almost rostral mouth in loricarioids and other basal siluriforms; the apomorphic loricariid ventral position of the mouth is acquired during ontogeny. Juvenile, intermediate configurations of the double skull roof (Part 8) might be very similar to the adult configuration of the skull roof in evolutionary predecessors of the extant species, as the selective advantage of increasingly larger opercular muscles to operate the cheek-spine apparatus probably paved the way to the further evolution of the dramatic skull Bauplan as observed today. Many of the above mentioned key innovations that lead to the loricariid morphology involve structural changes that can be classified as heterotopies [in the definition of Zelditch et al. (2000)], while evidence of heterochronic processes are less conspicuous. A future comparison between the ontogenetic trajectories of Ancistrus cf. triradiatus and the callichthyid Corydoras aeneus (see Chapter 1.1) may provide more insight in the nature of some of the more important loricariid key innovations.

PART 9 — GENERAL DISCUSSION 203

9.4. SYNOPSIS

The cranial morphology of loricariids is remarkable among siluriforms. At the onset of this dissertation (paragraph 1.1.2, Aims) it has been hypothesized that many loricariid characteristics can be related to the peculiar ecological niche which loricariids occupy (i.e., scraping food from submerged substrates, while being able to attach to these substrates, and respire while being attached with the suckermouth). After the current functional- morphological examination, the comparison to related taxa (thus evolutionary morphology), and the detailed treatise of the ontogenetic origin of the cranial structures, it has been become clear that several loricariid characteristics can indeed truly be considered specializations for this niche. Transformations that have occurred in the lineage towards the loricariid family include intense modifications in the skull architecture, as well as decouplings of structures, that enhance the mobility of cranial elements and explore the vast possibilities of remodulation of the versatile vertebrate head. Drastic as some of these transformations might appear, they still enabled (and enable) all life history stages to survive and prosper, and a wide array of diversity to develop within the family. In the early life stages of Ancistrus cf. triradiatus, suckermouth attachment (and respiration) is functional immediately after hatching, thus largely preceding the onset of actual feeding. This situation, opposed to the situation in many other teleosts, can be related to the parameters of the ecological niche of the free-living embryos. The large yolk supply appears to be of paramount importance, as it allows a postponed onset of exogenous feeding. This, on its turn, might well be related to the absence of a (possibly disadvantageous) metamorphosis or a larval stage. Even more important, however, is the high versatility of the fish skull configuration and its ontogeny, molded and transformed by evolution’s primary tool: natural selection. Reminiscent of the evolutionary transformation of the head, the upper snout region elongates during early ontogeny, and grows ventrally, bringing the upper lip and jaw downwards. The upper lip is joined by the expanded lower lip to form the suckermouth, so typical for the loricariids. The upper jaw, released from its plesiomorphic tight constraint, now freely articulates with a medioventral vertical disc of the mesethmoid bone. The lower jaw, positioned posterior to the upper jaw and lacking a medial symphysis, or any tight connection to the hyoid, has evolutionary rotated both medially and ventrally. The result of these transformations is that the subtly Z-shaped teeth of both jaws can now scrape food from underwater substrates. Already during early ontogeny, Meckel’s cartilages are situated rather posteriorly, and the suspensorium elongates less anteriorly than in catfishes with a ‘normal’ anteriorly oriented lower jaw. During ontogeny, the early point articulation between the suspensorium and the hyoid, formed by the interhyal cartilage, is replaced by a strong, long 204 PART 9 — GENERAL DISCUSSION and hinge-like ligamentous articulation, including a sesamoid bone of which the probable function or effect remains somewhat puzzling. The connection of the suspensorium to the neurocranium is even more reinforced by sutures and ligaments. This fortification of the lateral wall of the orobranchial cavity, and the rigid posterior wall formed by the cleithral flanges and the deep Baudelot’s ligaments, can be considered probable adaptations to the high negative forces that might well be generated during forceful suction onto substrates in the fast-flowing rivers, where many loricariids thrive. While the basic respiration pattern is probably essentially similar to that in other teleosts, the design of the ventral head region has enabled the new, apomorphic function of suckermouth attachment. Some of the most striking internal evolutionary transformations have been decouplings that occurred in the muscular basis of the respiratory and feeding movements. The high mobility of the upper jaw is operated by a distinct, medial portion of the extensor tentaculi and its antagonist, the retractor premaxillae. This latter muscle is a subdivision of the jaw muscle complex; it is homologous to the retractor tentaculi of many other siluriforms, and unique in its present form in loricariids and their closest relatives, the astroblepids. Identification and homology of these muscles was revealed by examination of the early ontogenetic stages of the loricariid Ancistrus cf. triradiatus, and comparison with the callichthyid Corydoras aeneus. No decoupling, but mere shifts in orientation of structures have led to the scraping action of the lower jaw. Due to the altered position and orientation of the lower jaw, the effect of the action of the adductor mandibulae and the dentary part of the intermandibularis posterior is a double rotation of the jaw (around the lateral articulation point and around its axis). The somewhat shifted relative positions of the internal and external adductor mandibulae portions have hampered, but not inhibited the recognition of homologies between these and other adductor mandibulae derivatives in loricariids, and the adductor mandibulae subdivisions found in basal and other catfishes. In this case, ontogeny has proved to be a truly invaluable tool. The lateral portion of the extensor tentaculi performs the same function as in other siluriforms, i.e., protraction of the maxillary barbel. However, in loricariids this barbel has acquired a different task: allowing controlled inspiration without failure of the suckermouth attachment, instead of mere probing for food. Thus, indirectly, the biological function of this muscle has changed. Its usual antagonist in siluriforms, the retractor tentaculi, also has experienced a functional shift (and has become the retractor premaxillae), but a new muscle has functionally taken its place: the levator tentaculi. Ontogeny, and a comparison to related taxa, have revealed that this typical loricariid muscle was derived from the muscle complex including the adductor arcus palatini and extensor tentaculi. The striking non-homology with any of the jaw muscle subdivisions, not anticipated by earlier studies, is unambiguously PART 9 — GENERAL DISCUSSION 205 shown in the current work. The incorporation of the maxillary barbel (and the associated, fused lips) in the respiratory process has not been described for other taxa. Moreover, the labial parts of the intermandibularis anterior and posterior muscles are considered to be crucial elements in the suckermouth function, providing the lower lip, dotted with unculiferous papillae, with a high, probably partly independent mobility. The positional shift of the hyohyoideus abductor during ontogeny is striking. The fact that both the hyohyoidei adductor and abductor muscles close the small branchiostegal membrane, and the fact that a unique muscle (retractor veli) operates the oral valve in loricariids, can be related to the suckermouth attachment apparatus. The opercular bone, not participating in expiratory movements, has in some loricariids become the central player in a totally different functional entity: the erectile cheek-spine apparatus. Related to this apparatus, some cranial muscles expand drastically during late ontogeny. This extreme muscle hypertrophy literally invades and remodulates the braincase itself. The findings presented in this doctoral dissertation show that the arising of the loricariid morphological design, including all mentioned evolutionary and developmental transformations, surely ranks as one of those subtle but genuine masterpieces created by evolution. 206 PART 9 — GENERAL DISCUSSION

PART 10

SUMMARY & SAMENVATTING

PART 10 — SUMMARY 207

10.1. SUMMARY

This doctoral dissertation describes and discusses the early life history, ontogeny and functional morphology of Ancistrus cf. triradiatus, a representative species of the neotropical suckermouth armoured catfish family Loricariidae. Several aspects concerning the adult morphology have been examined in other loricariid species as well. This summary lists the most important results and conclusions of this study, thus reflecting the aims stated in paragraph 1.1.3. First, the introductory part and the material and methods section are summarized.

Part 1 — An overview is given of the general frame of this study. Several aspects (ontogeny, functional morphology, biomechanics, evolutionary considerations and convergent characters) of the catfish taxa Loricarioidea (South America) and Mochokidae (Africa) are examined in this larger frame. In both these taxa, independent evolution towards an algae scraping feeding apparatus has developed, including a suckermouth and ventrally oriented lower jaws. Some problems regarding ontogeny and functional morphology, to be treated in this thesis, are explained. The aims of this part of the project are listed (see paragraph 1.1.3), introducing the structure of the dissertation. The family Loricariidae is introduced: its systematical position is described, including an overview of higher taxa, and of the related loricarioid families. An overview is given of the current knowledge of the systematics within the family, and of some general aspects of the family, including general morphology, reproduction, and feeding ecology.

Part 2 — A list is provided of the specimens of all species that were used in this thesis. Argumentation is given for the choice of these specific species. All methods that were applied in the course of the study are explained comprehensively. These include the culture of Ancistrus cf. triradiatus, live observations and high-speed filming, preparation of specimens for study, metric characters, in toto clearing and staining methods, dissections, serial sections, 3D-reconstructions and scanning electron microscopy.

Part 3 — The eggs and early life history stages of Ancistrus cf. triradiatus are described. The eggs have an adhesive shell, and contain a large amount of yolk. Embryos hatch after about five to six days; the yolk sac of the free-living embryos is resorbed in another six days. There is no true larval stage nor a metamorphosis. The head and tail regions grow positively allometrically. This is most pronounced in the snout length. Intense, but gradual ontogenetic head shape changes are present during the embryonic and free-living embryonic stages: the suckermouth gradually shifts from an almost rostral to a ventral position. The external shape 208 PART 10 — SUMMARY change reflects equally intense internal transformations. The resulting habitus transformation can be considered as an adaptation to both the loricariid algae-scraping feeding mode and the need of suckermouth functioning from the moment of hatching, when a ventrally situated suckermouth would be disadvantageous, as a large yolk sac is present ventrally.

Part 4 — The skeletal development of both the early chondrocranium and later and definitive osteocranium is decribed. The chondrocranial skull is platybasic from the start, representing a general siluriform trait. Compared to other siluriform species of which the early development is known, the ethmoid region is slender, with a rudimentary solum nasi. During ontogeny, especially this ethmoid region lengthens, and the ethmoid tip grows ventrally. This is related to the rostroventral growth of the snout, bringing the upper jaw region anterior to the lower jaw. The ventral suspensorium tip (palato-quadrate) shifts anteriorly, but less pronounced than in other catfishes. Meckel’s cartilages point medially instead of rostrally (which is the case in most teleosts). A lateral commissure and eye muscle myodomes (usually absent in catfishes) develop. The first elements of the osteocranium that arise are the opercle, lateralmost branchiostegal ray, both upper and lower jaws, and the medial skull floor bones. The supracleithrum and the dermal and perichondral pterotic components form one large, double- layered skull bone during ontogeny, without clear evidence of the involvement of a supratemporal. The Baudelot’s ligament ossifies from two sides, i.e., from the basioccipital medially and from the supracleithrum laterally. The parurohyal, formed by the fusion of a ventral sesamoid bone and a dorsal cartilage element associated with the first basibranchial, is pierced by a vene, unlike in some other siluriforms. The interhyal cartilage disappears during ontogeny; medially of it a small sesamoid bone appears in a ligament. The largest, canal bearing cheek plate is not homologous to the interopercle.

Part 5 — The ontogeny of all cranial muscles except the branchial and eye muscles is decribed. Ontogenetic evidence supports homology or non-homology of some muscles in loricariids and non-loricariids. In the upper snout region, an exceptionally high number of muscles is present, derived from the adductor mandibulae complex and the adductor arcus palatini. Terminology of these muscles varies among the literature, and no data exist on their ontogenetic origin. For these jaw and maxillary barbel muscles, the ontogeny of the callichthyid Corydoras aeneus is examined as well, resulting in some new hypotheses concerning muscle homologies. The adductor mandibulae muscle s.s. (inserting on the lower jaw) is homologous to the A1-OST and A3’ of basal catfishes, and the A3’ has given rise to the loricariid retractor veli as well. The A2 and A3” have resulted in the retractor tentaculi of Callichthyidae and the retractor premaxillae of Loricariidae. These two muscles are PART 10 — SUMMARY 209 homologous. In Loricariidae, the extensor tentaculi consists of two separate muscles inserting on the autopalatine. Evidence is given on the evolutionary origin of the loricariid levator tentaculi (previously known as retractor tentaculi) from the extensor tentaculi, and not the adductor mandibulae complex, as was previously thought. The intermandibular and hyoid musculature differs from the general siluriform situation. Several modifications can be related to the presence of the suckermouth (especially the expanded lower lip), and the altered mobility of the lower jaws. The intermandibularis anterior muscle differentiates into two parts, inserting on the lower jaw but also on the lower lip tissue. A similar differentiation into a dentary and a labial part occurs in the intermandibularis posterior. The latter muscle is joined by the interhyoideus to form a compound protractor hyoidei in many teleosts, but not in loricariids. Several arguments, including the absence of both a myocomma and a double innervation, indicate the absence of an interhyoideus portion. The term protractor hyoidei is thus erroneous for the muscle in loricariids; the name intermandibularis posterior must here be used. A double innervation has been found in the hyohyoideus inferior. The posteriormost hyoid muscles are relatively small during early ontogeny: the sternohyoideus halves fuse relatively late; the hyohyoidei adductores develop latest of all ventral head muscles. Suspensorial and opercular muscles are among those muscles usually involved in respiration (and feeding). The adductor arcus palatini is relatively large throughout the whole ontogeny, while the levator arcus palatini is minute. It develops in association with the dilatator operculi. The levator and adductor operculi are connected during early ontogeny, and anterior fibres of the latter muscle differentiate into the adductor hyomandibulae, a muscle previously thought to be absent in loricariids. Some functional considerations are made concerning hypothetical free-living embryonic to adult kinematics of the hyoid- suspensorium mechanism. Some considerations are given on the function of the cranial muscles during ontogeny.

Part 6 — Dissections and manipulations of fresh specimens, serial sections, histological examination of some peculiar tissues, and a limited kinematic study using high-speed video was performed on adult Ancistrus cf. triradiatus. Some functional hypotheses have been formulated, anticipating on a more thorough biomechanical investigation in the future. Results of this part of the thesis also serve as an appropriate anatomical basis for such a study. The suspensorium is a rather rigid structure. The hyoid is more movable and associated muscles are more substantial; it appears to be more important in the buccal pump system. The transverse orientation of the hyohyoideus abductor suggests it cannot open the branchiostegal membrane. This movement might be passive. Apart from divisions inserting on the lower and 210 PART 10 — SUMMARY upper jaws, a medial adductor mandibulae division, the retractor veli, inserts on the oral valve. The levator tentaculi and the lateral part of the completely subdivided extensor tentaculi move the maxillary barbel, a structure that allows controlled inspiration via a small lip furrow (preventing substantial leaking and thus failure of the suction system). Rotational movements of the lower and upper jaws result in scraping on the substrate. Antagonistic muscles for the adductor mandibulae divisions inserting on the lower and upper jaws (the adductor mandibulae s.s. and the retractor premaxillae) are the dentary part of the intermandibularis posterior and the medial part of the extensor tentaculi, respectively. The lower jaws are most mobile, not being linked to the hyoid arch medially. A medial cartilage plug acts as a supporting and gliding device for the lower jaws. All aspects studied in Ancistrus cf. triradiatus, except for the kinematic study, have been studied in the loricariids Farlowella acus and Otocinclus vestitus as well. The main morphological differences are listed, and eventual functional consequences are briefly discussed. The comparisons allow a limited appraisal of loricariid anatomical diversity.

Part 7 — Morphology, growth, replacement pattern, ontogeny (in Ancistrus cf. triradiatus) and loricariid diversity of oral teeth and keratinous unicellular epidermal brushes on the lip surface (‘unculi’) are described. Both structures assist in scraping food off substrates. The teeth are exquisitely curved, and are usually asymmetrically bicuspid. Apart from the general tooth form and specific crown shape, the anterior layer of soft tissue on the lower shaft region, present in several species, appears to be a specialization for scraping on irregular surfaces, as it enhances the mobility and even flexibility of individual teeth. Unculi of A. cf. triradiatus are curved in the same direction of the teeth, suggesting their role in scraping. During early ontogeny, a transition from simple conical to mature teeth is observed. The first unculi appear together with the first teeth carrying a bicuspid crown, two days after the first exogenous feeding, but synchronous with the complete resorption of the yolk sac.

Part 8 — In the Loricariidae, the opercle has been decoupled from the lower jaw, and has also lost its function in expiration. While many loricariid species have a small and slightly mobile opercle with reduced opercular musculature, within the hypostomine subfamily a novel opercular mechanism has developed that erects a tuft of enlarged odontodes (‘spines’, tooth-like denticles) anterior to the opercle. This extraordinary defensive mechanism is examined in Ancistrus cf. triradiatus. The opercle has a prominent anterior process and the orientation of the reinforced articulation hinge to the hyomandibular bone has shifted. The well developed opercular musculature includes a hypertrophied dilatator operculi, that extends deep inside the skull roof bones and towards the midline, over the brain, but below the superficial skull roof. Hence several skull roof bones consist of a dorsal, superficial part PART 10 — SUMMARY 211 and a deeper part separating the brain from the muscle: two functional skull roofs are thus formed. The impact on the path of the cranial sensory canals is substantial, moving canals away from the skull surface. Thus the hypertrophy of this muscle drastically modifies the skull, literally hollowing it out. The cranial modifications are greater in males than in females, related to the territorial behaviour of the former, in which the erectile spines are used.

Part 9 — The last part recapitulates the main findings of this study, providing a synthetic overview of the major aspects of the thesis. First, some final functional considerations are made regarding the head morphology and the three main functions performed by the loricariid mouth: suckermouth attachment, respiration, and the scraping feeding mode. It is followed by a section discussing the evolution towards the loricariid head configuration. Existing evolutionary hypotheses are reappraised, critically evaluated, and supplemented with findings and hypotheses arising from the present study. Thirdly, a developmental discussion combines these functional and evolutionary aspects with the ontogenetic data on Ancistrus cf. triradiatus. The early functioning of the loricariid head elements is discussed, and a brief overview is given on the major findings regarding homologies of structures. Finally, a synopsis recapitulates the major results of the whole study. 212 PART 10 — SUMMARY

PART 10 — SAMENVATTING 213

10.2. SAMENVATTING

Deze doctoraatsthesis beschrijft and bespreekt de vroeg-ontogenetische levensstadia, de ontwikkeling en de functionele morfologie van Ancistrus cf. triradiatus, een vertegenwoordiger van de neotropische katvisfamilie Loricariidae, ofwel de pantsermeervallen met een zuigmond. Verscheidene aspecten i.v.m. de adulte morfologie werden eveneens in andere soorten van deze familie onderzocht. De voorliggende samenvatting geeft een opsomming van de belangrijkste resultaten en conclusies van deze studie, en biedt zodoende een antwoord op de doelstellingen geformuleerd in paragraaf 1.1.3. Allereerst worden het inleidende hoofdstuk en het gedeelte over het gebruikte materiaal en methoden hier kort overlopen.

Deel 1 — Hierin wordt een overzicht gegeven van het algemeen kader van deze studie. Een waaier aan onderzoeksaspecten (ontogenie, functionele morfologie, biomechanica, evolutionaire beschouwingen en convergente kenmerken) van de katvistaxa Loricarioidea (Zuid-Amerika) en Mochokidae (Afrika) worden binnen dit kader onderzocht. In deze beide taxa is een onafhankelijke evolutie opgetreden naar een algenschrapend voedingsapparaat met een zuigmond en ventraalwaarts gerichte onderkaken. Een aantal probleemstellingen aangaande de ontogenie en functionele morfologie, die behandeld zullen worden in deze thesis, worden toegelicht. De doelstellingen van dit onderdeel van het project worden opgesomd (zie paragraaf 1.1.3), waarbij ook de structuur van deze thesis wordt uiteengezet. De katvis- (of meerval-)familie Loricariidae wordt geïntroduceerd: haar systematische positie wordt beschreven, en voorafgegaan door een overzicht van de hogere taxa waartoe ze behoort, en van de nauwverwante families die eveneens tot de superfamilie Loricarioidea behoren. De huidige wetenschappelijke kennis over de systematiek binnen de familie zelf wordt bondig overlopen, alsook een aantal aspecten van de familie, zoals de algemene morfologie, de voortplanting en de voedingsecologie.

Deel 2 — Dit gedeelte bevat een lijst van alle specimens die voor deze thesis gebruikt werden. Er wordt kort toegelicht waarom juist deze soorten werden gebruikt. Alle methoden die in de loop van dit onderzoek werden toegepast worden gedetailleerd besproken. Deze methoden omvatten de kweek van Ancistrus cf. triradiatus, observaties en high-speed filmen van levende dieren, de voorbereiding van specimens voor onderzoek, metrische kenmerken, in toto opheldering en kleuring, dissecties, seriële coupereeksen, 3D-reconstructies en scanning electronen microscopie.

214 PART 10 — SAMENVATTING

Deel 3 — Zowel de eitjes als de vroege levensstadia worden beschreven. De eitjes hebben een kleverige schaal, en bevatten een grote hoeveelheid dooier. Embryo’s ontluiken na ongeveer vijf tot zes dagen; de dooierzak van de vrijlevende embryo’s wordt in een bijkomende zes dagen geresorbeerd. Er is geen echt larvaal stadium, noch een metamorfose. De kop- en staartregio’s groeien positief allometrisch, wat het meest opvallend is in de snuitlengte. Intense maar geleidelijke ontogenetische vormveranderingen treden op gedurende de embryonale en vrijlevend embryonale stadia; de zuigmond verschuift geleidelijk van een vrijwel rostrale naar een ventrale positie. De uitwendige vormveranderingen gaan samen met even intense inwendige transformaties. De resulterende transformatie van het habitus kan beschouwd worden als een aanpassing aan zowel de loricariide algenschrapende voedingswijze, als de noodzaak van een zuigmond die functioneel is vanaf het ogenblik van ontluiken, waarbij een ventraalstandige zuigmond nadelig zou zijn, gezien de aanwezigheid van de grote, ventrale dooierzak.

Deel 4 — De skeletale ontwikkeling van zowel het vroege chondrocranium als het latere en definitieve osteocranium worden beschreven. De chondrocraniale (kraakbeen-)schedel is platybasisch vanaf zijn ontstaan, een algemeen kenmerk van katvissen. In vergelijking met andere katvissen waarvan de vroege ontwikkeling gekend is, is de ethmoidregio smal, met een rudimentaire solum nasi. Gedurende de ontogenie verlengt voornamelijk deze ethmoidregio, en de tip ervan groeit ventraalwaarts. Dit is gerelateerd aan de rostroventrale groei van de snuit, die de bovenkaakregio vóór de onderkaak brengt. De ventrale tip van het suspensorium (palatoquadratum) verschuift rostraalwaarts, maar minder uitgesproken dan in andere katvissen. De kraakbeenderen van Meckel zijn mediaal gericht i.p.v. rostraal (wat het geval is in de meeste teleosten). Een laterale commissuur en myodomen voor de oogspieren (gewoonlijk afwezig in katvissen) ontwikkelen. De eerste elementen van het osteocranium die ontstaan zijn het operculare, de meest laterale branchiostegale straal, de boven- en onderkaken, en de mediale beenderen van de schedelbodem. Het supracleithrum en de dermale en perichondrale componenten van het pteroticum vormen één groot, tweelagig been, zonder duidelijk bewijs van de betrokkenheid van een supratemporale. Het ligament van Baudelot verbeent vanaf beide uiteinden, d.w.z. mediaal, vanaf het basioccipitale, en lateraal, vanaf het supracleithrum. Het parurohyale, gevormd door een fusie van een ventraal sesamoidbeen en een dorsaal kraakbenig element geassocieerd met het eerste basibranchiale, bevat een foramen voor een ader, wat niet het geval is bij andere bestudeerde katvissen. Het interhyaal kraakbeen verdwijnt tijdens de ontogenie; mediaal ervan ontstaat een sesamoidbeen in een ligament. De grootste, kanaaldragende kaakplaat is niet homoloog met het interoperculare.

PART 10 — SAMENVATTING 215

Deel 5 — De onwikkeling van alle kopspieren (met uitzondering van de oog- en kieuwboogspieren) wordt beschreven. Ontogenetische evidentie ondersteunt homologie of niet-homologie van een aantal spieren in Loricariidae en niet-Loricariidae. In de bovensnuitregio bevindt zich een uitzonderlijk hoog aantal spieren, afgeleid van het adductor mandibulae complex en de adductor arcus palatini. De terminologie van deze spieren varieert in de literatuur, en er zijn vooralsnog geen gegevens beschikbaar over hun ontogenetische oosprong. Voor deze kaakspieren en spieren van de maxillaire monddraad werd eveneens de ontwikkeling van Corydoras aeneus (Callichthyidae) bestudeerd, wat resulteerde in enkele nieuwe hypotheses i.v.m. spierhomologieën. De adductor mandibulae s.s. (aanhechtend op de onderkaak) is homoloog met de A1-OST en A3’ van basale katvissen. De A3’ heeft bovendien ook aanleiding gegeven tot de retractor veli. De A2 en A3” vormen de retractor tentaculi in Callichthyidae, en de retractor premaxillae in Loricariidae. Deze beide spieren zijn homoloog. In Loricariidae bestaat de extensor tentaculi uit twee gescheiden spieren die insereren op het autopalatinum. Er wordt aangetoond dat de levator tentaculi in Loricariidae (voorheen gekend als retractor tentaculi) ontstaan is uit de extensor tentaculi, en niet uit het adductor mandibulae complex, zoals tot nu toe gedacht werd. De intermandibulaire en hyoidspieren wijken af van de algemene situatie in katvissen. Verscheidene modificaties van deze spieren kunnen gerelateerd worden aan de aanwezigheid van de zuigmond (en in het bijzonder de vergrote onderlip), en de gewijzigde mobiliteit van de onderkaak. De intermandibularis anterior bevat twee delen, aanhechtend op de onderkaak en op de onderlip. Een gelijkaardige onderverdeling in een dentaal en een labiaal deel treedt op in de intermandibularis posterior. Deze laatste spier is bij vele teleosten samen met de interhyoideus vergroeid tot de protractor hyoidei. Verschillende argumenten, met name de afwezigheid van een myocomma en een dubbele innervatie, tonen de afwezigheid van een interhyoideus-gedeelte aan bij Loricariidae. De term protractor hyoidei is dus foutief bij Loricariidae; de naam intermandibularis posterior dient gehanteerd te worden. Een dubbele innervatie werd gevonden in de hyohyoideus inferior. De meest posterieure hyoidspieren zijn relatief klein tijdens de vroege ontwikkeling: de beide helften van de sternohyoideus fusioneren vrij laat; de hyohyoidei adductores ontstaan als laatste van alle ventrale kopspieren. De suspensorium- en operculaire spieren zijn gewoonlijk betrokken bij de respiratie (en voeding). De adductor arcus palatini is relatief groot gedurende de gehele ontogenie, terwijl de levator arcus palatini miniem is. Deze laatste spier ontwikkelt in associatie met de dilatator operculi. De levator en adductor operculi zijn in de vroegste ontogenie verbonden, en anterieure vezels van de adductor operculi differentiëren tot de adductor hyomandibulae, een spier waarvan tot nu toe werd aangenomen dat ze afwezig is in Loricariidae. Een aantal functionele beschouwingen worden geformuleerd i.v.m. de hypothetische kinematica van het 216 PART 10 — SAMENVATTING hyoid-suspensorium-mechanisme van vrijlevende embryo’s tot adulten. Er wordt ook kort ingegaan op de functie van de kopspieren gedurende de ontogenie.

Deel 6 — Dissecties en manipulaties van verse specimens, seriële coupereeksen, histologisch onderzoek van enkele bijzondere weefsels, en een beperkte kinematische studie met high-speed video-opnames, werden uitgevoerd op adulte specimens van Ancistrus cf. triradiatus. Enkele functionele hypotheses worden geformuleerd, anticiperend op een meer diepgaande biomechanische studie in de toekomst. Resultaten uit dit deel van de thesis vormen ook een gepaste anatomische basis voor zulke studie. Het suspensorium is een vrij rigide structuur. Het hyoid is beweeglijker, en de geassocieerde spieren zijn groter; het lijkt een grotere rol te hebben in het buccaal pompsysteem. De dwarse oriëntatie van de hyohyoideus abductor suggereert dat het geen opener is van de branchiostegale membraan. Deze beweging zou passief kunnen zijn. De retractor veli insereert op de orale klep of velum, en is een mediale divisie van het adductor mandibulae complex, afgescheiden van de spierdivisies van dit complex die insereren op de boven- en onderkaak. De levator tentaculi en het lateraal deel van de volledig opgedeelde extensor tentaculi bewegen de maxillaire monddraad, een structuur die gecontroleerde inademing toelaat via een nauwe lipplooi (wat het lekken en dus falen van het zuigmondsysteem voorkomt). Draaiende bewegingen van de boven- en onderkaken resulteren in het afschrapen van het substraat. De antagonistische spieren van de adductor mandibulae divisies die op de boven- en onderkaken aanhechten (retractor premaxillae en adductor mandibulae s.s.), zijn respectievelijk het mediale deel van de extensor tentaculi en het dentaal deel van de intermandibularis posterior. De onderkaken zijn het meest beweeglijk, en zijn mediaal niet gekoppeld aan de hyoidboog. Een mediale kraakbeenplug doet dienst als een ondersteunende en glijdende structuur voor de onderkaken. Alle aspecten bestudeerd in Ancistrus cf. triradiatus, met uitzondering van de kinematische studie, werden ook bestudeerd in de Loricariidae Farlowella acus en Otocinclus vestitus. De voornaamste morfologische verschillen worden toegelicht, en eventuele functionele gevolgen van deze verschillen worden kort besproken. Deze vergelijkingen laten een beperkte inschatting van de anatomische diversiteit binnen de familie Loricariidae toe.

Deel 7 — Dit deel behandelt de morfologie, de groei, het vervangingspatroon, de ontogenie (in Ancistrus cf. triradiatus) en de loricariide diversiteit van de orale tanden en gekeratiniseerde ééncellige epidermale borstels of ‘unculi’. Beide structuren assisteren in het afschrapen van voedsel van substraten. De tanden zijn bijzonder gekromd, en zijn doorgaans bicuspied. Naast deze algemene tandvorm en de specifieke vorm van de tandkroon, lijkt de PART 10 — SAMENVATTING 217 anterieure laag van zacht weefsel op het onderste deel van de schacht een specialisatie te zijn voor het schrapen, aangezien het de mobiliteit en zelfs flexibiliteit van individuele tanden bevordert. Unculi van A. cf. triradiatus zijn gekromd in dezelfde richting als de tanden, wat hun rol in het schrapen suggereert. Gedurende de vroege ontogenie wordt een transitie waargenomen van eenvoudige conische tanden naar ‘volwassen’ tanden. De eerste unculi verschijnen tesamen met de eerste tanden met eeen bicuspiede kroon, twee dagen na de eerste exogene voeding, maar synchroon met de volledige resorptie van de dooierzak.

Deel 8 — In Loricariidae is het operculare ontkoppeld van de onderkaak, en heeft het ook zijn functie in het uitademingsproces verloren. Terwijl vele Loricariidae een klein en weinig beweeglijk operculare met gereduceerde operculaire spieren hebben, is er binnen de subfamilie Hypostominae een nieuw operculair mechanisme ontstaan dat een groep sterk vergrote odontoden (tandachtige ‘stekels’) voor het operculare opricht. Dit bijzonder defensie-mechanisme werd onderzocht in Ancistrus cf. triradiatus. Het operculare heeft een opvallende anterieur uitsteeksel, en de oriëntatie van de verstevigde articulatie met het hyomandibulare is verschoven. De sterk ontwikkelde operculaire musculatuur omvat o.m. een gehypertrofieerde dilatator operculi, die mediaalwaarts doordringt binnenin de schedelbeenderen, boven de hersenen maar onder het oppervlakkige schedeldak. Zodus bestaan verschillende schedelbeenderen uit een dorsaal, oppervlakkig deel, en een dieper deel dat de hersenen scheidt van de spier: twee functionele schedeldaken worden zo gevormd. De impact op het craniaal sensorisch kanaalsysteem is zeer groot, aangezien het de kanalen wegvoert van het schedeloppervlak. Op deze manier modificeert de hypertrofie van deze spier de schedel drastisch: de schedel wordt letterlijk uitgehold. De schedelaanpassingen zijn groter in mannetjes dan in wijfjes, wat gerelateerd is aan het territoriaal gedrag van de mannetjes, waarbij zij de oprichtbare stekels gebruiken.

Deel 9 — Het laatste deel recapituleert de voornaamste bevindingen van deze studie, en biedt een overzicht van de aspecten die in deze thesis behandeld worden. Eerst worden een aantal slotopmerkingen geformuleerd over de kopmorfologie en de drie belangrijkste functies uitgevoerd door de mond van Loricariidae: vasthechting met de zuigmond, ademhaling, en de schrapende voedingswijze. Vervolgens wordt een discussie gevoerd over de evolutie naar de craniale configuratie van Loricariidae. Bestaande evolutionaire hypotheses worden geherwaardeerd, kritisch geëvalueerd, en aangevuld met resultaten en hypotheses die in de loop van de huidige studie ontstaan zijn. 218 PART 10 — SAMENVATTING

Het daarop volgende onderdeel van de discussie combineert de functionele en evolutionaire aspecten met de ontogenetische data over Ancistrus cf. triradiatus. Het vroege functioneren van craniale elementen wordt besproken, en een kort overzicht wordt gegeven van de belangrijkste besluiten i.v.m. de homologie van structuren. Tenslotte vat een synopsis de belangrijkste bevindingen van de gehele studie samen. PART 11

REFERENCES

PART 11 — REFERENCES 219

REFERENCES

ADRIAENS D., 1998. On how a larva becomes an adult catfish: a functional-morphological approach to the cranial ontogeny of the African catfish, Clarias gariepinus (Siluriformes, Clariidae). Unpublished PhD Thesis, Ghent University, 240 + 148 p. ADRIAENS D., 2003. Feeding mechanisms. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 2, p. 221-248. ADRIAENS D., AERTS P. & VERRAES W., 2001. Ontogenetic shift in mouth opening mechanisms in a catfish (Clariidae, Siluriformes): a response to increasing functional demands. Journal of Morphology, 247: 197-216. ADRIAENS D. & VANDEWALLE P., 2003. Embryonic and larval development in catfishes. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 1, p. 639-666. ADRIAENS D. & VERRAES W., 1994. On the functional significance of the loss of the interhyal during ontogeny in Clarias gariepinus Burchell, 1822 (Teleostei: Siluroidei). Belgian Journal of Zoology, 124: 139-155. ADRIAENS D. & VERRAES W., 1996. Ontogeny of cranial musculature in Clarias gariepinus (Siluroidei: Clariidae): The adductor mandibulae complex. Journal of Morphology, 229: 255-269. ADRIAENS D. & VERRAES W., 1997A. Ontogeny of the maxillary barbel muscles in Clarias gariepinus (Siluroidei: Clariidae), with some notes on the palatine-maxillary mechanism. Journal of Zoology (London), 241: 117-133. ADRIAENS D. & VERRAES W., 1997B. Some consequences of transformations in siluriform chondrocrania: a case study of Clarias gariepinus (Burchell, 1822) (Siluriformes: Clariidae). Netherlands Journal of Zoology, 47: 349-363. ADRIAENS D. & VERRAES W., 1997C. The ontogeny of the chondrocranium in Clarias gariepinus: trends in siluroids. Journal of Fish Biology, 50: 1221-1257. ADRIAENS D. & VERRAES W., 1997D. Ontogeny of the hyoid musculature in the African catfish, Clarias gariepinus (Burchell, 1822) (Siluroidei: Clariidae). Zoological Journal of the Linnean Society, 121: 105-128. ADRIAENS D. & VERRAES W., 1997E. Ontogeny of the suspensorial and opercular muscles in Clarias gariepinus (Siluroidei: Clariidae), and the consequences for respiratory movements. Netherlands Journal of Zoology, 47: 61-89. ADRIAENS D. & VERRAES W., 1998. Ontogeny of the osteocranium in the African catfish, Clarias gariepinus (1822) (Siluriformes: Clariidae): ossification sequence as a response to functional demands. Journal of Morphology, 235: 183-237. ADRIAENS D. & VERRAES W., 2002. An empirical approach to study the relation between ontogeny, size and age using geometrics. In: Topics in functional and ecological vertebrate morphology (D'Août K., Herrel A. & Van Damme R., eds). Shaker Publishing (Maastricht), p. 293-324. ADRIAENS D., VERRAES W. & TAVERNE L., 1997. The cranial lateral-line system in Clarias gariepinus (Burchell, 1822) (Siluroidei: Clariidae): morphology and development of canal related bones. European Journal of Morphology, 35: 181-208. AERTS P., 1991. Hyoid morphology and movements relative to abducting forces during feeding in Astatotilapia elegans (Teleostei: Cichlidae). Journal of Morphology, 208: 323-345. AERTS P., OSSE J.W.M. & VERRAES W., 1987. Model of jaw depression during feeding in Astatotilapia elegans (Teleostei: Cichlidae): Mechanisms for energy storage and triggering. Journal of Morphology, 194: 85-109. 220 PART 11 — REFERENCES

AERTS P. & VERRAES W., 1984. Theoretical analysis of a planar four bar system in the teleostean skull: the use of mathematics in biomechanics. Annales de la Société Royale Zoologique de Belgique, 114: 273-290. ALEXANDER R.MCN., 1964. The structure of the Weberian apparatus in the Siluri. Proc. zool. Soc. London, 142: 419-440. ALEXANDER R.MCN., 1965. Structure and function in the catfish. Journal of Zoology (London), 148: 88-152. ALEXANDER R.MCN., 1969. Mechanics of the feeding action of a cyprinid fish. Journal of Zoology (London), 159: 1-15. ALEXANDER R.MCN., 1970. Mechanics of the feeding action of various teleost fishes. Journal of Zoology (London), 162: 145-156. ALEXANDER R.MCN., 1975. Evolution of integrated design. American Zoologist, 15: 419-425. ALEXANDER R.MCN., 1996. Biophysical problems of small size in vertebrates. In: Miniature Vertebrates. The implications of small body size (Miller P.J., ed.). Clarendon Press (Oxford), p. 3- 14. ALLIS E.P.JR., 1889. The anatomy and development of the lateral line system in Amia calva. Journal of Morphology, 2: 463-568. ANGELESCU V. & GNERI F.S., 1949. Adaptaciones del aparato digestivo al régimen alimenticio en algunos peces del Río Uruguay y del Río de la Plata. Tipo omnívoro e iliófago en representantes de las familias “Loricariidae” y “Anostomidae”. Revta Inst. nac. Invest. Cienc. nat. Mus. argent. Cienc. nat. Bernardino Rivadavia (zool.), 1: 161-272. ANGERMEIER P.L. & KARR J.R., 1983. Fish Communities along environmental gradients in a system of tropical streams. Environmental Biology of Fishes, 9: 117-135. ANKER G.CH., 1974. Morphology and kinetics of the head of the stickleback, Gasterosteus aculeatus. Transactions of the Zoological Society of London, 32: 311-416. ANKER G.CH., 1989. The morphology of joints and ligaments of a generalised Haplochromis species: H. elegans Trewavas 1933 (Teleostei, Cichlidae). III. The hyoid and the branchiostegal apparatus, the branchial apparatus and the shoulder girdle apparatus. Netherlands Journal of Zoology, 39: 1- 40. AQUINO A.E., 1994. Secondary sexual dimorphism of the dermal skeleton in two species of the hypoptopomatine genus Otocinclus (Siluriformes: Loricariidae). Ichthyological Explorations of Freshwaters, 5: 217-222. AQUINO A.E., 1996. Redescripción de Otocinclus flexilis Cope, 1894 (Siluriformes, Loricariidae, Hypoptopomatinae) con un nuevo sinónimo. Iheringia, Ser. Zool., Porto Alegre, 81: 13-22. AQUINO A.E., 1997. Las especies de Hypoptopomatinae (Pisces, Siluriformes, Loricariidae) en la Argentina. Revista de Ictiología, 5: 5-21. AQUINO A.E., 1998. Topography of the extrinsic eye-muscles in four hypoptopomatine species (Siluriformes: Loricariidae), with comments on myodomes in catfishes. Ichthyological Explorations of Freshwaters, 8: 231-238. AQUINO A.E. & SCHAEFER S.A., 2002. The temporal region of the cranium of loricarioid catfishes (Teleostei: Siluriformes): morphological diversity and phylogenetic significance. Zoologischer Anzeiger, 241: 223-244. ARANHA J.M.R., TAKEUTI D.F. & YOSHIMURA T.M., 1998. Habitat use and food partitioning of the fishes in a coastal stream of Atlantic Forest, Brazil. Revista de Biologia Tropical, 46: 951-959. ARENS W., 1994. Striking convergence in the mouthpart evolution of stream-living algae grazers. Journal of Zoological Systematical and Evolutionary Research, 32: 319-343. PART 11 — REFERENCES 221

ARMBRUSTER J.W., 1998. Modifications of the digestive tract for holding air in loricariid and scoloplacid catfishes. Copeia, 1998: 663-675. ARMBRUSTER J.W., 2004. Phylogenetic relationships of the suckermouth armoured catfishes (Loricariidae) with emphasis on the Hypostominae and the Ancistrinae. Zoological Journal of the Linnean Society, 141: 1-80. ARMBRUSTER J.W., SABAJ M.H., HARDMAN M., PAGE L.M. & KNOUFT J.H., 2000. Catfish genus Corymbophanes (Loricariidae: Hypostominae) with description of one new species: Corymbophanes kaiei. Copeia, 2000: 997-1006. ARRATIA G., 1987. Description of the primitive family Diplomystidae (Siluriformes, Teleostei, Pisces): morphology, taxonomy and phylogenetic implications. Bonner Zoologische Monographien, 24: 1-120. ARRATIA G., 1990. Development and diversity of the suspensorium of trichomycterids and comparison with loricarioids (Teleostei: Siluriformes). Journal of Morphology, 205: 193-218. ARRATIA G., 1992. Development and variation of the suspensorium of the primitive catfishes (Teleostei: Ostariophysi) and their phylogenetic relationships. Bonner Zoologische Monographien, 32: 1-148. ARRATIA G., 1997. Basal teleosts and teleostean phylogeny. Palaeo Ichthyologica, 7: 5-168. ARRATIA G., 2003. Catfish head skeleton: an overview. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 1, p. 3-46. ARRATIA G. & GAYET M., 1995. Sensory canals and related bones of tertiary siluriform crania from Bolivia and North America and comparison with recent forms. Journal of Vertebrate Paleontology, 15: 482-505. ARRATIA G. & HUAQUÍN L., 1995. Morphology of the lateral line system and of the skin of diplomystid and certain primitive loricarioid catfishes and systematic and ecological considerations. Bonner Zoologische Monographien, 36: 1-110. ARRATIA G. AND MENU-MARQUE S., 1981. Revision of the freshwater catfishes of the genus Hatcheria (Siluriformes, Trichomycteridae) with commentaries on ecology and biogeography. Zoologischer Anzeiger, 207: 88-111. ARRATIA G. & MENU-MARQUE S., 1984. New catfishes of the genus Trichomycterus from the high Andes of South America (Pisces, Siluriformes) with remarks on distribution and ecology. Zoologisches Jahrbuch, Abt. Syst. (Oekol.), Geogr. Biol., 111: 493-520. ARRATIA G. & SCHULTZE H.-P., 1990. The urohyal: development and homology within Osteichthyans. Journal of Morphology, 203: 247-282. ATODA K., 1936. The innervation of the cranial nerves of the catfish, Parasilurus asotus. Science Reports of the Tohoku University (series 4: biology), 11: 91-115. BAILEY R.M. & BASKIN J.N., 1976. Scoloplax dicra, a new armored catfish from the Bolivian Amazon. Occasional Papers of the Museum of Zoology, University of Michigan, no. 674. BALLANTYNE F.M., 1930. Notes on the development of Callichthys littoralis. Transcriptions of the Royal Society of Edinborough, 56: 437-466. BALLINTIJN C.M. & HUGHES G.M., 1965. The muscular basis of the respiratory pumps in the trout. The Journal of Experimental Biology, 42: 349-362. BALON E.K., 1975. Terminology of intervals in fish development. Journal of the Fisheries Research Board of Canada, 32: 1663-1670. BALON E.K., 1979. The theory of saltation and its application in the ontogeny of fishes: steps and thresholds. Environmental Biology of Fishes, 4: 97-101. BALON E.K., 1984. Reflections on some decisive events in the early life of fishes. Transactions of the American Fisheries Society, 113: 178-185. 222 PART 11 — REFERENCES

BALON E.K., 1986. Types of feeding in the ontogeny of fishes and the life-history model. Environmental Biology of Fishes, 16: 11-24. BALON E.K., 1999. Alternative ways to become juvenile or a definitive phenotype (and on some persisting linguistic offenses). Environmental Biology of Fishes, 56: 17-38. BAMFORD T.W., 1948. Cranial development of Galeichthys felis (Ariidae). Proceedings of the Zoological Society (London), 118: 364-391. BARTSCH P., GEMBALLA S. & PIOTROWSKI T., 1997. The embryonic and larval development of Polypterus senegalus Cuvier, 1829: its staging with reference to external and skeletal features, behaviour and locomotory habits. Acta Zoologica (Stockholm), 78: 309-328. BASKIN J.N., 1973. Structure and relationships of the Trichomycteridae. Unpublished PhD Thesis, City University of New York, 389 p. BASKIN J.N., ZARET T.M. & MAGO-LECCIA F., 1980. Feeding of reportedly parasitic catfishes (Trichomycteridae and Cetopsidae) in the Rio Portuguesa Basin, Venezuela. Biotropica, 12: 182- 186. BENJAMIN M., 1986. The oral sucker of Gyrinocheilus aymonieri (Teleostei: Cypriniformes). Journal of Zoology (London), 1: 211-254. BENJAMIN M., 1990. The cranial cartilages of teleosts and their classification. Journal of Anatomy, 169: 153-172. BERKELBACH V.B.S., 1915. The central relations of the cranial nerves in Silurus glanis and Mormyrus cuschive. The Journal of Comparative Neurology, 25: 5-63. BERRA, T.M., 2001. Freshwater fish distribution. Academic Press, San Diego, London, 604 p. BERTMAR G., 1959. On the ontogeny of the chondral skull in Characidae, with a discussion on the chondrocranial base and the visceral chondrocranium in fishes. Acta Zoologica, 40: 37-162. BHARGAVA H.N., 1958. The development of the chondrocranium of Mastacembelus armatus (Cuv. & Val.). Journal of Morphology, 102: 401-426. BHATTI H.K., 1938. The integument and dermal skeleton of Siluroidea. Transactions of the Zoological Society of London, 24: 1-102. BHIMACHAR B.S., 1933. On the morphology of the skull of certain Indian catfishes. Journal of the Mysore University, 7: 233-267. BIRKHEAD W.S., 1972. Toxicity of stings of ariid and ictalurid catfishes. Copeia, 1972: 790-807. BLACKSTONE N.W., 1987. Size and time. Systematic Zoology, 36: 211-215. BLAXTER J.H.S., 1969. Development: eggs and larvae. In: Fish Physiology (Hoar W.S. & Randall D.J., eds). Academic Press (New York), vol. 3, p. 177-252. BLECKMANN H., NIEMANN U. & FRITZSCH B., 1991. Peripheral and central aspects of the acoustic and lateral line system of a bottom dwelling catfish, Ancistrus sp. Journal of Comparative Neurology, 314: 452-466. BOCK W.J. & SHEAR CH.R., 1972. A staining method for gross dissection of vertebrate muscles. Anatomischer Anzeiger, 130: 222-227. BOWEN, S.H., 1984. Detritivory in neotropical fish communities. In: Evolutionary ecology of neotropical freshwater fishes (Zaret T.M., ed.). Dr. W. Junk Publishers (The Hague), p. 59-66. BRANSON B.A., 1962. Observations on the breeding tubercles of some Ozarkian minnows with notes on the barbels of Hybopsis. Copeia, 1962: 532-539. BREAULT J.L., 1991. Candiru: Amazonian parasitic catfish. Journal of Wilderness Medicine, 2: 304- 312. BREWSTER B., 1987. Eye migration and cranial development during flatfish metamorphosis: a reappraisal (Teleostei: Pleuronectiformes). Journal of Fish Biology, 31: 805-833. PART 11 — REFERENCES 223

BRIDGE T.W., 1890. The air-bladder in certain Siluroid fishes. Proceedings of the Philosophical Society of Birmingham, 6: 131-136. BRIDGE T.W. & HADDON A.C., 1893. Contributions to the anatomy of fishes, II: The air-bladder and Weberian ossicles in the siluroid fishes, Philosophical Transactions of the Royal Society of London, 184: 65-333, plates 11-19. BRIDGE T.W. & HADDON A.C., 1894. Note on the production of sounds by the air-bladder of certain siluroid fishes. Proceedings of the Royal Society of London, 55: 439-441. BRITZ R. & HOFFMANN M., 2006. Ontogeny and homology of the claustra in otophysan Ostariophysi (Teleostei). Journal of Morphology, 267: 909-923. BROWMAN M.W. & KRAMER D.L., 1985. Pangasius sutchi (Pangasiidae), an air-breathing catfish uses the swimbladder as an accessory respiratory organ. Copeia, 1985: 994-998. BUCK S. & SAZIMA I., 1995. An assemblage of mailed catfishes (Loricariidae) in southeastern Brazil: distribution, activity, and feeding. Ichthyological Explorations of Freshwaters, 6: 325-332. BURGESS W.E., 1989. An atlas of freshwater and marine catfishes. A preliminary study of the Siluriformes. T.F.H. Publications, Neptune City, New Jersey, 784 p. CARDONA L. & GUERAO G., 1994. Astroblepus riberae, una nueva especie de siluriforme cavernicola del Perú (Osteichthyes: Astroblepidae). Mémoires de Biospéologie, XXI: 21-24. CARTER G.S. & BEADLE L.C., 1931. The fauna of the Paraguayan Chaco. Zoological Journal of the Linnean Society, 37: 327-368. CEAS P.A. & PAGE L.M., 1996. Chaetostoma yurubiense (Teleostei: Siluriformes), a new species of loricariid catfish from the Aroa, Urama and Yaracuy River systems in Venezuela. Copeia, 1996: 671-677. CHAO N.L., 2001. The fishery, diversity, and conservation of ornamental fishes in the Rio Negro Basin, Brazil - A review of project Piaba (1989-99). In: Conservation and management of ornamental fish resources of the Rio Negro Basin, Amazonia, Brazil - Project Piaba (Chao N.L., Petry P., Prang G., Sonneschien L. & Tlusty M., eds). Editoria da Universidade do Amazonas (Manaus), p. 161-204. CHARDON M., 1968. Anatomie comparée de l'appareil de Weber et des structures connexes chez les Siluriformes. Annales du Musée Royal de l'Afrique Centrale, Série IN-8°, 169: 1-277. CHARDON M., PARMENTIER E. & VANDEWALLE P., 2003. Morphology, development and evolution of the Weberian apparatus in catfish. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 1, p. 71-119. CHEN X. & ARRATIA G., 1996. Breeding tubercles of Phoxinus (Teleostei: Cyprinidae): morphology, distribution and phylogenetic implications. Journal of Morphology, 228: 127-144. COBURN M.M. & FUTEY L.M., 1996. The ontogeny of supraneurals and neural arches in the cypriniform Weberian apparatus (Teleostei: Ostariophysi). Zoological Journal of the Linnean Society, 116: 333-346. COBURN M.M. & GRUBACH P.G., 1998. Ontogeny of the Weberian apparatus in the armored catfish Corydoras paleatus (Siluriformes: Callichthyidae). Copeia, 1998: 301-311. COOK A., 1996. Ontogeny of feeding morphology and kinematics in juvenile fishes: a case study of the cottid fish Clinocottus analis. Journal of Experimental Biology, 199: 1961-1971. COPP G.H. & KOVÁČ V., 1996. When do fish with indirect development become juveniles? Canadian Journal of Fisheries and Aquatic Sciences, 53: 746-752. COUGHLIN D.J., 1994. Suction prey capture by clownfish larvae (Amphiprion perideraion). Copeia, 1994: 242-246. 224 PART 11 — REFERENCES

DAGET J., 1964. Le crâne des téléostéens, I. Origine et mise en place des tissus squelettogènes chez l'embryon. Mésomésenchyme et ectomésenchyme. Mémoires du Muséum National d'Histoire Naturelle, Série A. Zoologie, 31: 164-340. DAGET J. & D'AUBENTON F., 1957. Développement et morphologie du crâne d'Heterotis niloticus Ehr. Bulletin de l'Institut Français d'Afrique Noire, Série A: Sciences Naturelles, 193: 881-936. DARWIN C., 1859. On the origin of species by means of natural selection or the preservation of favored races in the struggle for life. J. Murray (London). (Reprint edition: Oxford University Press (Oxford, 1996). DAS D. & NAG T.C., 2006. Fine structure of the organ of attachment of the teleost, Garra gotyla gotyla (Ham). Zoology, 109: 300-309. DATTA MUNSHI J.S. & SINGH B.R., 1967. The cranial muscles and the natural mechanism of opening and closing of mouth in two Indian major carps. Zoologischer Anzeiger, 178: 49-60. DAVENPORT J., LØNNING S. & KJØRSVIK E., 1986. Some mechanical and morphological properties of the chorions of marine teleost eggs. Journal of Fish Biology, 29: 289-301. DE AZEVEDO P., 1938. O cascudo dos acudes nordestinos Plecostomus plecostomus. Arquivos do Instituto Biologico, 9: 211-225. DE BEER G.R., 1927. The early development of the chondrocranium of Salmo fario. Quarterly Journal of Micr. Sciences, 71: 259-312. DE BEER G.R., 1937. Studies on the vertebrate head. Oxford: Clarendon Press. DE PINNA M.C.C., 1989. Redescription of Glanapteryx anguilla, with notes on the phylogeny of Glanapteryginae (Siluriformes, Trichomycteridae). Proceedings of the Academy of Natural Sciences of Philadelphia, 141: 361-374. DE PINNA M.C.C., 1993. Higher-level phylogeny of Siluriformes, with a new classification of the order (Teleostei, Ostariophysi). Unpublished PhD Thesis, City University of New York, 482 p. DE PINNA M.C.C., 1996. A phylogenetic analysis of the Asian catfish families Sisoridae, Akysidae, and Amblycipitidae, with a hypothesis on the relationships of the Neotropical Aspredinidae (Teleostei, Ostariophysi). Fieldiana, 84: 1-83. DE PINNA M.C.C., 1998. Phylogenetic relationships of Neotropical Siluriformes (Teleostei: Ostariophysi): Historical overview and synthesis of hypotheses. In: Phylogeny and classification of neotropical fishes (Malabarba L.R., Reis R.E., Vari R.P., Lucena Z.M. & Lucena C.A.S., eds). Edipucrs (Porto Alegre), p. 279-330. DE SCHEPPER N., ADRIAENS D. & DE KEGEL B., 2005. Moringua edwardsi (Moringuidae: Anguilliformes): cranial specialization for head-first burrowing? Journal of Morphology, 266: 356-368. DECLEYRE D., VERRAES W. & AERTS P., 1991. Structural adaptations for forceful biting in gobies (Teleostei: Gobiidae). Belgian Journal of Zoology, 121 (suppl. 1): 11. DELARIVA R.L. & AGOSTINHO A.A., 2001. Relationship between morphology and diets of six neotropical loricariids. Journal of Fish Biology, 58: 832-847. DELEON M.F., REESE R.O. & CONLEY W.J., 1991. Effects of fixation and dehydration on shrinkage and morphology in common snook yolk-sac larvae. NOAA Technical Report NMFS 95: 121-128. DIETZ P.A., 1914. Beiträge zur Kenntnis der Kiefer- und Kiemenbogenmuskulatur der Teleostier. I. Die Kiefer- und Kiemenbogenmuskeln der Acanthopterygii. Mitteilungen der Zoologischen Station zu Neapel: 22: 99-162. DIOGO, R., 2003. Higher-level phylogeny of Siluriformes – An overview. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 1, p. 353- 384. PART 11 — REFERENCES 225

DIOGO, R., 2005. Morphological evolution, aptations, homoplasies, constraints, and evolutionary trends: catfishes as a case study on general phylogeny and macroevolution. Science Publishers, Enfield (USA), 487 p. DIOGO R. & CHARDON M., 2000A. Homologies among different adductor mandibulae sections of teleostean fishes, with special regards to catfishes (Teleostei: Siluriformes). Journal of Morphology, 243: 193-208. DIOGO R. & CHARDON M., 2000B. The structures associated with catfish (Teleostei: Siluriformes) mandibular barbels: origin, anatomy, function, taxonomic distribution, nomenclature and synonymy. Netherlands Journal of Zoology, 50: 455-478. DIOGO R., CHARDON M. & VANDEWALLE M., 2006. On the osteology and myology of the cephalic region and pectoral girdle of Nematogenys inermis (Ghichenot, 1848), with comments on the autapomorphies and phylogenetic relationships of the Nematogenyidae (Teleostei: Siluriformes). Belgian Journal of Zoology, 136: 15-24. DIOGO R., OLIVEIRA C. & CHARDON M., 2000. On the anatomy and function of the cephalic structures in Phractura (Siluriformes: Amphiliidae), with comments on some striking homoplasies occurring between the Doumeinae and some loricaroid catfishes. Belgian Journal of Zoology, 130: 117-130. DIOGO R., OLIVEIRA C., & CHARDON M., 2001. On the homologies of the skeletal components of catfish (Teleostei: Siluriformes) suspensorium. Belgian Journal of Zoology, 131: 93-109. DIOGO R. & VANDEWALLE P., 2003A. Review of superficial cranial musculature of catfishes, with comments on plesiomorphic states. In: Catfishes (Arratia G., Kapoor A.S., Chardon M. & Diogo R., eds). Science Publishers (Enfield, USA), vol. 1, p. 47-70. DIOGO R. & VANDEWALLE P., 2003B. Catfishes as a case study for discussions on general evolution: the importance of functional uncouplings in morphological macroevolution. European Journal of Morphology, 41: 139-148. DOI M., KOHNO H., TAKI Y. & OHNO A., 1998. Development of swimming and feeding functions in larvae and juveniles of the red snapper, Lutjanus argentimaculatus. Journal of Tokyo University of Fisheries, 85: 81-95. DONASCIMIENTO C. & PROVENZANO F., 2006. The genus Henonemus (Siluriformes: Trichomycteridae) with a description of a new species from Venezuela. Copeia, 2006: 198-205. DROST M.R., OSSE J.W.M. & MULLER M., 1988. Prey capture by fish larvae, water flow patterns and the effect of escape movements of prey. Netherlands Journal of Zoology, 38: 23-45. DROST M. R. & VAN DEN BOOGAART J.G.M., 1986A. A simple method for measuring the changing volume of small biological objects, illustrated by studies of suction feeding by fish larvae and of shrinkage due to histological fixation. Journal of Zoology (London), 209: 239-249. DROST M.R. & VAN DEN BOOGAART J.G.M., 1986B. The energetics of feeding strikes in larval carp, Cyprinus carpio. Journal of Fish Biology, 29: 371-379. DUTTA H.M., 1968. Functional morphology of the head of Anabas testudineus. Doctoral thesis, University of Leiden, 146 p. EATON T.H.JR., 1935. Evolution of the upper jaw mechanisme in teleost fishes. Journal of Morphology, 58:157-172. EATON T.H.JR., 1948. Form and function in the channel catfish, Ictalurus lacustris punctatus. Journal of Morphology, 83:181-194. EDGEWORTH F.H., 1928. The development of some of the cranial muscles of ganoid fishes. Philosophical Transactions of the Royal Society, Series B, 217: 39-89. EDGEWORTH F.H., 1935. The cranial muscles of Vertebrates. University Press (Cambridge). EIGENMANN C.H., 1918. The Pygidiidae. Academy of Science Proceedings, 163: 59-66. 226 PART 11 — REFERENCES

EIGENMANN C.H., 1922. The fishes of western South America. Part I. The fresh-water fishes of north-western South America, including Colombia, Panama, and the Pacific slopes of Ecuador and Peru, together with an appendix upon the fishes of the Río Meta in Colombia. Memoirs of the Carnegie Museum, 9: 1-346. EIGENMANN C.H., 1928. The fresh-water fishes of Chile. Memoir of the National Academy of Science, 22: 1-63. EIGENMANN C.H. & EIGENMANN R.S., 1890. A revision of the South American Nematognathi or cat-fish. Occasional Papers of the Californian Academy of Sciences, 1:1-508. ELSHOUD G.C.A., 1978. Respiration in the three-spined stickleback, Gasterosteus aculeatus L.; an electromyographic approach. Netherlands Journal of Zoology, 28: 524-544. ELSHOUD-OLDENHAVE M.J.W. & OSSE J.W.M., 1976. Functional morphology of the feeding system in the ruff – Gymnocephalus cernua (L. 1758) – (Teleostei, Percidae). Journal of Morphology, 150: 399-422. FAVARETTO A.L.V., PETENUSCI S.O. & LOPES R.A., 1981. Effect of exposure to air on haematological parameters in Hypostomus regani (Pisces: Loricariidae), teleost with aquatic and aerial respiration. Copeia, 1981: 918-920. FERNANDES M.N., PERNA S.A., SANTOS C.T.C. & SEVERI W., 1995. The gill filament muscles in two loricariid fish (genus Hypostomus and Rhinelepis). Journal of Fish Biology, 46: 1082-1085. FERNÁNDEZ L.A. & AQUINO A.E., 1993. Nuevas localidades para peces de la cuenca del Río Sali (Tucuman). Acta Zoologica Lilloana, 42: 305-307. FERRARIS C.J., SCHAEFER S.A., WEBER C. & FISCH-MULLER S., 2003. Family Loricariidae. In: Check list of the freshwater fishes of South and Central America (Reis R.E., Kullander S.O. & Ferraris C.J., eds). Edipucrs (Porto Alegre), p. 318-400. FINE M.L., FRIEL J.P., MCELROY D., KING C.B., LOESSER K.E. & NEWTON S., 1997. Pectoral spine locking and sound production in the channel catfish Ictalurus punctatus. Copeia, 1997: 777- 790. FINK W.L., 1981. Ontogeny and phylogeny of tooth attachment modes in actinopterygian fishes. Journal of Morphology, 167: 167-184. FINK S.V. & FINK W.L., 1981. Interrelationships of the ostariophysan fishes (Teleostei). Zoological Journal of the Linnean Society, 72: 297-353. FINK S.V. & FINK W.L., 1996. Interrelationships of ostariophysan fishes (Teleostei). In: Interrelationships of fishes (Stiassny M.L.J., Parenti L.R. & Johnson G.D., eds). Academic Press (London), p. 209-249. FISCH-MULLER S., 1999. Systématique du genre Ancistrus Kner (Teleostei, Loricariidae): approches morphologique et génétique. Thèse de doctorat, Université de Genève, 300 p., 26 pls. FISCH-MULLER S., 2003. Subfamily Ancistrinae. In: Check list of the freshwater fishes of South and Central America (Reis R.E., Kullander S.O. & Ferraris C.J., eds). Edipucrs (Porto Alegre), p. 373- 400. FISCH-MULLER S., MAZZONI R. & WEBER C., 2001. Genetic and morphological evidence for two new sibling species of Ancistrus (Siluriformes: Loricariidae) in upper rio Tocantins drainage, Brazil. Ichthyological Explorations of Freshwaters, 12: 289-304. FOSTNER H., HINTERLEITNER S., MÄHR K. & WIESER W., 1983. Towards a better definition of ‘metamorphosis’ in Coregonus sp.: biochemical, histological, and physiological data. Can. J. Fish. Aquat. Sci., 40: 1224-1232. FREIHOFER W.C. & NEIL E.H., 1967. Commensalism between midge larvae (Diptera: Chironomidae) and catfishes of the families Astroblepidae and Loricariidae. Copeia, 1967: 39-45. PART 11 — REFERENCES 227

FRIEL J.P., 1995. Acanthobunocephalus nicoi, a new genus and species of miniature banjo-catfish from the upper Orinoco and Casiquiare Rivers, Venezuela (Siluriformes, Aspredinidae). Ichthyological Explorations of Freshwaters, 6: 89-95. FUGI R., HAHN N.S. & AGOSTINHO A.A., 1996. Feeding styles of five species of bottom-feeding fishes of the high Paraná River. Environmental Biology of Fishes, 46: 297-307. FUIMAN L.A., 1983. Growth gradients in fish larvae. Journal of Fish Biology, 23: 117-123. FUIMAN L.A., 1997. What can flatfish ontogenies tell us about pelagic and benthic lifestyles? Journal of Sea Research, 37: 257-267. FUIMAN L.A., POLING K.R. & HIGGS D.M., 1998. Quantifying developmental progress for comparative studies of larval fishes. Copeia, 1998: 602-611. GALIS F., 1993. Interactions between the pharyngeal jaw apparatus, feeding behaviour and ontogeny in the cichlid fish, Haplochromis piceatus: a study of morphological constraints in evolutionary ecology. Journal of Experimental Zoology, 267: 137-254. GALIS F., TERLOUW A. & OSSE J.W.M., 1994. The relation between morphology and behaviour during ontogenetic and evolutionary changes. Journal of Fish Biology, 45 (suppl. A): 13-26. GANS C., 1988. Adaptation and the form-function relation. American Zoologist, 28: 681-697. GAYET M., 1986. About ostariophysan fishes: a reply to S.V. Fink, P.H. Greenwood & W.L. Fink's criticism. Bulletin du Musée National d'Histoire Naturel (Paris), 8: 393-409. GAYET M., 1993. Relations phylogénétiques des Gonorhynchiformes (Ostariophysi). Belgian Journal of Zoology, 123: 165-192. GAYET M. & CHARDON M., 1987. Possible otophysic connections in some fossil and living ostariophysan fishes. Proceedings of the Congress of European Ichthyology (Stockholm, 1985), 31-42. GEE J.H., 1976. Buoyancy and aerial respiration: factors influencing the evolution of reduced swimbladder volume of some Central American catfishes (Trichomycteridae, Callichthyidae, Loricariidae, Astroblepidae). Canadian Journal of Zoology, 54: 1030-1037. GHIOT F., 1978. The barbel movements of three South American pimelodid catfishes. Zoologischer Anzeiger, 56: 395-401. GHIOT F., VANDEWALLE P. & CHARDON M., 1984. Comparaison anatomique et fonctionnelle des muscles et des ligaments en rapport avec les barbillons chez deux familles apparentées de poissons Siluriformes Bagroidei. Annales de la Société Royal Zoologique de Belgique, 114: 261-272. GISBERT E., MERINO G., MUGUET J.B., BUSH D., PIEDRAHITA R.H. & CONKLIN D.E., 2002. Morphological development and allometric growth patterns in hatchery-reared California halibut larvae. Journal of Fish Biology, 61: 1217-1229. GLUCKMANN I., HURIAUX F., FOCANT F. & VANDEWALLE P., 1999. Postembryonic development of the cephalic skeleton in Dicentrarchus labrax (Pisces, Perciformes, Serranidae). Bulletin of Marine Science, 65: 11-36. GOODRICH E.S., 1958. The skull. In: Studies on the structure and development of vertebrates (Goodrich E.S., ed.). Dover Publications (New York), p. 230-395. GOSLINE W.A., 1973. Considerations regarding the phylogeny of cypriniform fishes with special reference to structures associated with feeding. Copeia, 1973: 761-777. GOSLINE W.A., 1975. The palatine-maxillary mechanism in catfishes, with comment on the evolution and zoogeography of modern Siluroids. Occasional Papers of the Californian Academy of Sciences, 120: 1-31. GOSLINE W.A., 1986. Jaw muscle configuration in some higher teleostean fishes. Copeia, 1986: 705- 713. 228 PART 11 — REFERENCES

GOSLINE W.A., 1989. Two patterns of differentiation in the jaw musculature of teleostean fishes. Journal of Zoology (London), 218: 649-661. GOZLAN R.E., COPP G.H. & TOURENQ J.-N., 1999. Comparison of growth plasticity in the laboratory and field, and implications for the onset of juvenile development in sofie, Chondrostoma toxostoma. Environmental Biology of Fishes, 56: 153-165. GRADWELL N., 1971A. A photographic analysis of the air breathing behavior of the catfish, Plecostomus punctatus. Canadian Journal of Zoology, 49: 1089-1094. GRADWELL N., 1971B. A muscular oral valve unique in fishes. Canadian Journal of Zoology, 49: 837-839. GRAHAM J.B., 1997. Air-breathing fishes: Evolution, diversity and adaptation. California: Academic Press, 299 p. GRANDE L., 1987. Redescription of †Hypsidoris farsonensis (Teleostei: Siluriformes), with a reassessment of its phylogenetic relationships. Journal of Vertebrate Paleontology, 7: 24-54. GRANDE L. & EASTMAN J.T., 1986. A review of the antarctic ichthyofaunas in the light of new fossil discoveries. Palaeontology, 29: 113-137. GREENWOOD, P.H., 1971. Hyoid and ventral gill arch musculature in osteoglossomorph fishes. Bulletin of the British Museum of Natural History (Zoology), 22: 1-55. GREENWOOD P.H., ROSEN D.E., WEITZMAN S.H. & MYERS G.S., 1979. Phyletic studies of teleostean fishes, with a provisional classification of living forms. In: Readings in ichthyology (Love M.S. & Caillet G.M., eds). Goodyear Publishing Company (Santa Monica, California), p. 73-94. GREGORY W.K., 1933. Fish skulls: a study of the evolution of natural mechanisms. Transactions of the American Philosophical Society, 23: 75-481. GROSMAN M.F., GONZÁLEZ CASTELAIN J.R. & USUNOFF E.J., 1996. Trophic niches in an Argentine pond as a way to assess functional relationships between fishes and other communities. Water SA, 22: 345-350. HAAS A., 2001. Mandibular arch musculature of anuran tadpoles, with comments on homologies of amphibian jaw muscles. Journal of Morphology, 247: 1-33. HAAS A. & FISCHER M.S., 1997. Three-dimensional reconstruction of histological sections using modern product-design software. The Anatomical Record, 249: 510-516. HANKEN J. & WASSERSUG R., 1981. The visible skeleton. A new double-stain technique reveals the native of the ‘hard’ tissues. Functional Photography, 16: 22-26. HARRINGTON R.W.JR., 1955. The osteocranium of the American cyprinid fish, Notropis bifrenatus, with an annotated synonymy of teleost skull bones. Copeia, 1955: 267-291. HAYLOR G.S., 1992. Terminology for the early developmental stages of the African catfish, Clarias gariepinus (Burchell): working definitions for aquaculture. Aquaculture and Fisheries Management, 23: 511-514. HE S., GAYET M. & MEUNIER F.J., 1999. Phylogeny of the Amphiliidae (Teleostei: Siluriformes). Annales des Sciences Naturelles, 20: 117-146. HELFMAN G.S., COLETTE B.B. & FACEY D.E., 1997. The diversity of fishes. London: Blackwell Science, Inc., 528 p. HERNÁNDEZ L.P., BARRESI M.J.F. & DEVOTO S.H., 2002. Functional morphology and developmental biology of zebrafish: Reciprocal illumination from an unlikely couple. Integrative and Comparative Biology, 42: 222-231. HOEDEMAN J.J., 1960A. Studies on callichthyid fishes: (3) Notes on the development of Callichthys (1) (Pisces - Siluriformes). Bulletin of Aquatic Biology, 1: 53-68. PART 11 — REFERENCES 229

HOEDEMAN J.J., 1960B. Studies on callichthyid fishes: (4) Development of the skull of Callichthys and Hoplosternum (1) (Pisces - Siluriformes). Bulletin of Aquatic Biology, 1: 73-84. HOEDEMAN J.J., 1960C. Studies on callichthyid fishes: (5) Development of the skull of Callichthys and Hoplosternum (2) (Pisces - Siluriformes). Bulletin of Aquatic Biology, 2: 21-36. HOFER H., 1938. Bau und Mechanik des Schädels von Zeus pungio. Zoologisches Jahrbuch (Anatomie), 64: 482-510. HOFFMANN M. & BRITZ R., 2006. Ontogeny and homology of the neural complex of otophysan Ostariophysi. Zoological Journal of the Linnean Society, 147: 301-330. HOLDEN K.K. & BRUTON M.N., 1994. The early ontogeny of the mouthbrooder, Pseudocrenilabrus philander (Pisces, Cichlidae). Environmental Biology of Fishes, 41: 311-329. HOLMQVIST O., 1910. Der Musculus protractor hyoidei und der Senkungsmechanismus des Unterkiefers bei den Knochenfischen. Acta Universitatis Lundensis (n. ser. - Zool.), 6: 1-27. HOLMQVIST O., 1911. Studien in der von den NN. Trigeminus und Facialis innervierten Muskulatur der Knochenfische. Acta Universitatis Lundensis (n. ser. - Zool.), 7: 1-83. HORA S.L., 1930. Ecology, bionomics and evolution of the torrential fauna, with special reference to the organs of attachment. Philosophical Transactions of the Royal Society of London, Biological Sciences, 218: 171-282. HOWES G.J., 1983A. The cranial muscles of loricarioid catfishes, their homologies and value as taxonomic characters. Bulletin of the British Museum of Natural History (Zoology), 45: 309-345. HOWES G.J., 1983B. Problems in catfish anatomy and phylogeny exemplified by the Neotropical Hypophthalmidae (Teleostei: Siluroidei). Bulletin of the British Museum of Natural History (Zoology), 45: 1-39. HOWES G.J. & SANFORD C.P.J., 1987. Oral ontogeny of the Ayu, Pleccoglossus altivelis and comparisons with the jaws of other salmoniform fishes. Zoological Journal of the Linnean Society, 89: 133-169. HOWES G.J. & TEUGELS G.G., 1989. Observations and homology of the pterygoid bones in Corydoras paleatus and some other catfishes. Journal of Zoology (London), 219: 441-456. HUGHES G.M., 1960. A comparative study of gill ventilation in marine fishes. Journal of Experimental Biology, 37: 28-45. HUGHES G.M., 1970. A comparative approach to fish respiration. Experientia, 26: 113-224. HUNT VON HERBING I., 2001. Development of feeding structures in larval fish with different life histories: winter flounder and Atlantic cod. Journal of Fish Biology, 59: 767-782. HUNT VON HERBING I., MIYAKE T., HALL B.K. & BOUTILIER R.G., 1996A. Ontogeny of feeding and respiration in larval Atlantic cod Gadus morhua (Teleostei, Gadiformes): I. Morphology. Journal of Morphology, 227: 15-35. HUNT VON HERBING I., MIYAKE T., HALL B.K. & BOUTILIER R.G., 1996B. Ontogeny of feeding and respiration in larval Atlantic cod Gadus morhua (Teleostei, Gadiformes): II. Function. Journal of Morphology, 227: 37-50. HUYSENTRUYT F. & ADRIAENS D., 2005A. Adhesive structures of the eggs of Corydoras aeneus (Gill, 1858; Siluriformes; Callichthyidae). Journal of Fish Biology, 66: 871-876. HUYSENTRUYT F. & ADRIAENS D., 2005B. Descriptive osteology of Corydoras aeneus (Gill, 1858) (Siluriformes: Callichthyidae). Cybium, 29: 261-273. HUYSENTRUYT F., GEERINCKX T. & ADRIAENS D, ——. A descriptive myology of Corydoras aeneus (Gill, 1858) (Siluriformes: Callichthyidae), with a brief discussion on adductor mandibulae homologies. Submitted. HUYSSEUNE A., 1990. Development of the anterior part of the mandible and the mandibular dentition in two species of Cichlidae (Teleostei). Cybium, 14: 327-344. 230 PART 11 — REFERENCES

HUYSSEUNE A. & SIRE J.-Y., 1996. Structure and development of the odontodes in an armoured catfish, Corydoras aeneus (Siluriformes, Callichthyidae). Acta Zoologica (Stockholm), 77: 51-72. HUYSSEUNE A. & SIRE J.-Y., 1997A. Structure and development of teeth in three armoured catfish, Corydoras aeneus, C. arcuatus and Hoplosternum littorale (Siluriformes, Callichthyidae). Acta Zoologica (Stockholm), 78: 69-84. HUYSSEUNE A. & SIRE J.-Y., 1997B. Structure and development of first-generation teeth in the cichlid Hemichromis bimaculatus (Teleostei, Cichlidae). Tissue & Cell, 29: 679-697. HUYSSEUNE A. & SIRE J.-Y., 1998. Evolution of patterns and processes in teeth and tooth-related tissues in non-mammalian vertebrates. European Journal of Oral Sciences, 106 (suppl. 1): 437- 481. ISBRÜCKER I.J.H., 1980. Classification and catalogue of the mailed Loricariidae. Amsterdam: Universiteit Amsterdam, (Verslagen en Technische Gegevens (22), 181 p. JARVIK E., 1980. Basic structure and evolution of vertebrates. Vol. 1. Academic Press (London). 575 p. JAYARAM K.C., 1950. Respiratory and other adaptive modifications in the genus Gyrinocheilus Vaillant and their significance in constituting the family Gyrinocheilidae Hora. Zoological Survey of India, Indian Museum, Calcutta, 48: 107-112. JÉGU M., KEITH P., FISCH-MULLER S., FOSSATI O. & MEUNIER F.J., 2004. Les rapides et leur peuplement. In: Piranhas enivrés, des poissons et des hommes en Guyane (Meunier F.J., coord.). SFI/RMN, p. 27-33. JOHNSON R.D.O., 1912. Notes on the habits of a climbing catfish (Arges marmoratus) from the Republic of Colombia. Annals of the New York Academy of Sciences, 22: 327-333. JUGE M., 1899. Recherches sur les nerfs cérébraux et la musculature céphalique du Silurus glanis L. Doctoral thesis, Université de Genève, 171 p. KAATZ I. & LOBEL P.S., 1999. Acoustic behavior and reproduction in five species of Corydoras catfishes (Callichthyidae). Biological Bulletin, 197: 241-242. KAATZ I. & STEWART, D.S., 1997. The evolutionary origin and functional divergence of sound production in fish: catfish stridulation mechanisms. Journal of Morphology, 232: 272. KAMPF W., 1961. Vergleichende funktionsmorphologische Untersuchungen an den Viscerocranien einiger rauberisch lebender Knochenfische. Zoologische Beiträge (N.F.), 6: 391-496. KEENLEYSIDE, M.A., 1979. Diversity and adaptation in fish behaviour. Springer-Verlag (Berlin Heidelberg): 208 p. KELLEY W.E. & ATZ J.W., 1964. A pygiid catfish that can suck blood from goldfish. Copeia, 1964: 703-704. KINDRED J.E., 1919. The skull of Amiurus. Illinois Biological Monographs, 5: 1-120. KIRCHHOFF H., 1958. Funktionell-anatomische Untersuchung des Visceralapparates von Clupea harengus L. Zoologisches Jahrbuch, 76: 461-540. KLUGE A.G. & STRAUSS R.E., 1985. Ontogeny and systematics. Annual Review of Ecology and Systematics, 16: 247-268. KOBAYAKAWA M., 1992. Comparative morphology and development of bony elements in the head region in three species of Japanese catfishes (Silurus: Siluridae; Siluriformes). Japanese Journal of Ichthyology, 39: 25-36. KOHNO H., ORDONIO-AGUILAR R., OHNO A. & TAKI Y., 1996. Osteological development of the feeding apparatus in early stage larvae of the seabass, Lates calcarifer. Ichthyological Research, 43:1-9. KRAMER D.L. & BRYANT M.J., 1995. Intestine length in the fishes of a tropical stream: 1. Ontogenetic allometry. Environmental Biology of Fishes, 42: 115-127. PART 11 — REFERENCES 231

KRAMER D.L. & GRAHAM J.B., 1976. Synchronous air breathing, a social component of respiration in fishes. Copeia, 1976: 689-697. KRAMER D.L., LINDSEY C.C., MOODIE G.E.E. & STEVENS E.D., 1978. The fishes and the aquatic environment of the central Amazon basin, with particular reference to respiratory patterns. Canadian Journal of Zoology, 56: 717-729. KUNZ Y.W., 2004. Developmental biology of fishes. Springer, Dordrecht, 636 p. LAALE H.W., 1980. The perivitelline space and egg envelopes of bony fishes: a review. Copeia, 1980: 210-226. LACHNER E.A., ROBINS C.R. & COURTENAY W.R.JR., 1970. Exotic fishes and other aquatic organisms introduced into North America. Smithsonian Contributions to Zoology, 59: 1-29. LADICH F. & BASS A.H., 1996. Sonic/vocal-acousticolateralis pathways in teleost fishes: a transneuronal biocytin study in mochokid catfish. The Journal of Comparative Neurology, 374: 493-505. LARSON P.M. & REILLY S.M., 2003. Functional morphology of feeding and gill irrigation in the anuran tadpole: electromyography and muscle function in larval Rana catesbeiana. Journal of Morphology, 255: 202-214. LAUDER G.V., 1979. Feeding mechanics in primitive teleosts and in the halecomorph fish Amia calva. Journal of Zoology (London), 187: 543-578. LAUDER G.V., 1980. Evolution of the feeding mechanism in primitive actinopterygian fishes: a functional anatomical analysis of Polypterus, Lepisosteus and Amia. Journal of Morphology, 163: 283-317. LAUDER G.V., 1983. Functional and morphological bases of trophic specialization in sunfishes (Teleostei, Centrarchidae). Journal of Morphology, 178: 1-21. LAUDER G.V., 1985. Aquatic feeding in lower vertebrates. In: Functional vertebrate morphology (Hildebrand M., Bramble D.M., Liem K.F. & Wake D.B., eds). The Belknap Press (Harvard Univ. Press, Cambridge, Massachusetts), p. 210-229. LAUDER G.V., CROMPTON A.W., GANS C., HANKEN J., LIEM K.F., MAIER W.P., MEYER A., PRESLEY R., RIEPPEL O.C., ROTH G., SCHLUTER D. & ZWEERS G.A., 1989. Group report: How are feeding systems integrated and how have evolutionary innovations been introduced? In: Complex organismal functions: integration and evolution in vertebrates (Wake D.B. & Roth G., eds). Wiley & Sons (New York), p. 97-115. LAUDER G.V. & LIEM K.F., 1980. The feeding mechanism and cephalic myology of Salvelinus fontenalis: form, function, and evolutionary significance. In: Charrs, salmonid fishes of the genus Salvelinus (Balon E.K., ed.). Dr. W. Junk Publishers (The Hague), p. 365-390. LAVOUÉ S., MIYA M., INOUE J.G., SAITOH K., ISHIGURO N.B. & NISHIDA M., 2005. Molecular systematics of the gonorhynchiform fishes (Teleostei) based on whole mitogenome sequences: Implications for higher-level relationships within the Otocephala. Molecular Phylogenetics and Evolution, 37: 165-177. LEGENDRE M., TEUGELS G.G., CAUTY C. & JALABERT B., 1992. A comparative study on morphology, growth rate and reproduction of Clarias gariepinus (Burchell 1822), Heterobranchus longifilis Valenciennes, 1840, and their reciprocal hybrids. Journal of Fish Biology, 40: 59-79. LIEM K.F., 1979. Modulatory multiplicity in the feeding mechanism in cichlid fishes, as exemplified by the invertebrate pickers of Lake Tanganyika. Journal of Zoology (London), 189: 93-125. LIEM K.F., 1980A. Adaptive significance of intra- and interspecific differences in the feeding repertoires of cichlid fishes. American Zoologist, 20: 295-314. LIEM K.F., 1980B. Acquisition of energy by teleosts: adaptive mechanisms and evolutionary patterns. In: Environmental physiology of fishes (Ali M.A., ed.). Plenum Publishing Corporation (New York), p. 299-324. 232 PART 11 — REFERENCES

LIEM K.F., 1991. A functional approach to the development of the head of teleosts: implications on constructional morphology and constraints. In: Constructional morphology and evolution (Schmidt-Kittler N. & Vogel K., eds). Springer-Verlag (Berlin), p. 231-249. LOIR M., CAUTY C., PLANQUETTE P. & LE BAIL P.Y., 1989. Comparative study of the male reproductive tract in seven families of South-American catfishes. Aquatic Living Resources, 2: 45- 56. LOPEZ-ROJAS H. & MACHADO-ALLISON A., 1975. Algunas aspectos del desarollo y crescimiento de Loricaria laticeps (Osteichthyes, Siluriformes, Loricariidae). Acta Biologica Venezuelica, 9: 51-76. LOWE R.H., 1964. The fishes of the Rupununi savanna district of British Guyana, South America. Zoological Journal of the Linnean Society, 45: 103-144. LOWE-MCCONNELL R.H., 1969. Speciation in tropical freshwater fishes. Biological Journal of the Linnean Society, 1: 51-75. LOWE-MCCONNELL R.H., 1984. The status of studies of South American freshwater food fishes. In: Evolutionary ecology of neotropical freshwater fishes (Zaret T.M., ed.). Dr. W. Junk Publishers (The Hague), p. 139-156. LUBOSCH W., 1929. Vergleichende Anatomie der Kaumuskulatur der Wirbeltiere, in fünf Teilen. Zweiter Teil (Fortsetzung). Die Kaumuskeln der Teleosteer. Morphologische Jahrbuch, 61: 49- 220. LUBOSCH W., 1938. Muskeln des Kopfes: Viscerale Muskulatur (Fortsetzung). B. Teleostier. In: Handbuch der vergleichenden Anatomie der Wirbeltiere (Bolk L., Göppert E., Kallius E. & Lubosch W., eds). Urban & Schwarzenberg (Berlin & Wien), vol. 5, p. 1011-1024. LUNDBERG J.G., 1975. Homologies of the upper shoulder girdle and temporal region bones in catfishes (Order Siluriformes), with comments on the skull of the Helogeneidae. Copeia, 1975: 66- 74. LUNDBERG J.G., 1982. The comparative anatomy of the toothless blindcat, Trogloglanis pattersoni Eigenmann, with a phylogenetic analysis of the ictalurid catfishes. Miscellaneous Publications Museum of Zoology, University of Michigan, 163: 1-85. LUNDBERG J.G., 1998. The temporal context for the diversification of neotropical fishes. In: Phylogeny and classification of neotropical fishes (Malabarba L.R., Reis R.E., Vari R.P., Lucena Z.M. & Lucena C.A.S., eds). Edipucrs (Porto Alegre), p. 49-68. LUNDBERG J.G., 2001. Freshwater riches of the Amazon. Natural History, 9: 35-43. LUNDBERG J.G. & BASKIN J.N., 1969. The caudal skeleton of the catfishes, order Siluriformes. American Museum Novitates, 2398: 1-19. LUNDBERG J.G., KOTTELAT M. SMITH G.R., STIASSNY M.L.J. & GILL A.C., 2000. So many fishes, so little time: an overview of recent ichthyological discovery in continental waters. Annals of the Missouri Botanical Garden, 87: 26-62. LUNDBERG J.G., MARSHALL L.G., GUERRERO J., HORTON B., MALABARBA M.C.S.L. & WESSELINGH F., 1998. The stage for neotropical fish diversification: a history of tropical South American Rivers. In: Phylogeny and classification of neotropical fishes (Malabarba L.R., Reis R.E., Vari R.P., Lucena Z.M. & Lucena C.A.S., eds). Edipucrs (Porto Alegre), p. 13-48. MABEE P.M. & TRENDLER T.A., 1996. Development of the cranium and paired fins in Betta splendens (Teleostei: Percomorpha): Intraspecific variation and interspecific comparisons. Journal of Morphology, 227: 249-287. MACHADO-ALLISON A. & LOPEZ-ROJAS H., 1975. Etapas de desarollo de Loricariichthys typus (Bleeker) 1964 (Osteichthyes, Siluriformes, Loricariidae). Acta Biologica Venezuelica, 9: 93-119. MAZZONI R. & CARAMASCHI E.P., 1997. Observations on the reproductive biology of female Hypostomus luetkeni Lacèpéde 1803. Ecology of Freshwater Fish, 6: 53-56. PART 11 — REFERENCES 233

MCALLISTER D.E., 1968. The evolution of branchiostegals and associated opercular, gular, and hyoid bones and the classification of teleostome fishes, living and fossil. Bulletin of the National Museum of Canada (Biological Series), 221: 1-239. MCELMAN J.F. & BALON E.K., 1980. Early ontogeny of white sucker, Catostomus commersoni, with steps of saltatory development. Environmental Biology of Fishes, 5: 191-224. MCKAYE K.R. & MARSH A., 1983. Food switching by two-specialized algae-scraping cichlid fishes in Lake Malawi, Africa. Oecologia, 56: 245-248. MCMURRICH J.P., 1884. On the osteology of Amiurus catus (L.) Gill. Zoologischer Anzeiger, 168: 296-299. MÉRIGOUX S. & PONTON D., 1998. Body shape, diet and ontogenetic diet shifts in young fish of the Sinnamary River, French Guiana, South America. Journal of Fish Biology, 52: 556-569. MILLER R.M., 1966. Geographical distribution of Central American freshwater fishes. Copeia, 1966: 773-802. MILLER P.J., 1979. Adaptiveness and implications of small size in teleosts. Symposium of the Zoological Society of London, 44: 263-306. MILLER P.J., 1996. The functional ecology of small fish: some opportunities and consequences. In: Miniature Vertebrates. The implications of small body size (Miller P.J., ed.). Clarendon Press (Oxford), p. 175-200. MIQUILARENA A.M., LOPEZ H.L. & AQUINO A.E., 1994. Los Ancistrinae (Pisces: Loricariidae) de Argentina. In: Fauna de agua dulce de la República Argentina (De Castellanos Z.A., ed.). Museo de la Plata, 40, p. 1-49. MITHEL M., 1964A. The cranial nerves of Mystus seenghala (Sykes). Agra University Journal of Research (Sci.), 13: 67-78. MITHEL M., 1964B. The cranial nerves of the sisorid catfish Bagarius bagarius. Copeia, 1964: 673- 678. MIYAKE T., MCEACHRAN J.D. & HALL B.K., 1992. Edgeworth's legacy of cranial muscle development with an analysis of muscles in the ventral gill arch region of batoid fishes (Chondrichthyes: Batoidea). Journal of Morphology, 212: 213-256. MO T., 1991. Anatomy and systematics of Bagridae (Teleostei), and siluroid phylogeny. Koeltz Scientific Books, Koenigstein. 216 p. MOL J.H., 1995. Ontogenetic diet shifts and diet overlap among three closely related neotropical armoured catfishes. Journal of Fish Biology, 47: 788-807. MONTOYA-BURGOS J.-I., MULLER S., WEBER C. & PAWLOWSKI J., 1997. Phylogenetic relationships between Hypostomatinae and Ancistrinae (Siluroidei: Loricariidae): first results from mitochondrial 12S and 16S rRNA gene sequences. Revue Suisse de Zoologie, 104: 185-198. MONTOYA-BURGOS J.-I.; MULLER S.; WEBER C. & PAWLOWSKI J., 1998. Phylogenetic relationships of the Loricariidae (Siluriformes) based on mitochondrial rRNA gene sequence. In: Phylogeny and classification of neotropical fishes (Malabarba L.R., Reis R.E., Vari R.P., Lucena Z.M. & Lucena C.A.S., eds). Edipucrs (Porto Alegre), p. 363-374. MOODIE G.E.E. & POWER M., 1982. The reproductive biology of an armoured catfish, Loricaria uracantha, from Central America. Environmental Biology of Fishes, 7: 143-148. MOTTA P.J., 1984. Mechanics and functions of jaw protrusion in teleost fishes: a review. Copeia, 1984: 1-18. MOTTA P.J., NORTON S.F. & LUCZKOVICH J.J., 1995. Perspectives on the ecomorphology of bony fishes. Environmental Biology of Fishes, 44: 11-20. 234 PART 11 — REFERENCES

MÜLLER G.B. & WAGNER G.P., 2003. Innovation. in: Keywords & concepts in evolutionary developmental biology (Hall B.K. & Olson W.M., eds). Harvard University Press (Cambridge, MS, USA and London, UK), p. 218-227. MULLER M., 1987. Optimization principles applied to the mechanism of neurocranium levation and mouth bottom depression in bony fishes (Halecostomi). Journal of Theoretical Biology, 126: 343- 368. MULLER M., 1989A. A quantitative theory of expected volume changes of the mouth during feeding in teleost fishes. Journal of Zoology (London), 217: 639-662. MULLER M. & OSSE J.W.M., 1984. Hydrodynamics of suction feeding in fish. Transactions of the Zoological Society of London, 37: 51-135. MULLER S., 1989B. Déscription de deux nouvelles espèces paraguayennes du genre Ancistrus Kner, 1854 (Pisces, Siluriformes, Loricariidae). Revue Suisse de Zoologie, 96: 885-904. MULLER S., 1990. Étude méristique et morphométrique d' Ancistrus piriformis Muller et Ancistrus pirareta Muller (Pisces, Siluriformes, Loricariidae). Revue Suisse de Zoologie, 97: 153-168. MULLER S., PY-DANIEL L.H.R. & ZUANON J., 1994. Ancistrus ranunculus, a new species of loricariid fish (Siluriformes: Loricariidae) from the Xingú and Tocantins Rivers, Brazil. Ichthyological Explorations of Freshwaters, 5: 289-296. MULLER S. & WEBER C., 1992. Les dents de sous-familles Hypostominae et Ancistrinae (Pisces, Siluriformes, Loricariidae) et leur valeur taxonomique. Revue Suisse de Zoologie, 99: 747-754. MYERS G.S., 1966. Derivation of the freshwater fish fauna of Central America. Copeia, 1966: 765- 773. NAKAE M. & SASAKI K., 2004. Homologies of the adductor mandibulae muscles in Tetraodontiformes as indicated by nerve branching patterns. Ichthyological Research, 51: 327-336. NAKATANI K., AGOSTINHO A.A., BAUMGARTNER G., BIALETZKI A., SANCHES P.V., MAKRAKIS M.C. & PAVANELLI C.S., 2001. Ovos e larvas de peixes de água doce. Eduem (Maringá, Brasil), 378 p. NAUWELAERTS S., WILGA C.D., LAUDER G.V. & SANFORD C.P., 2006. Fluid dynamics of suction feeding in bamboo sharks. Unpublished abstract, Joint Meeting of Ichthyologists and Herpetologists, 12-17 June 2006. NAWAR G., 1954. On the anatomy of Clarias lazera: I. Osteology. Journal of Morphology, 94: 551- 585. NAWAR G., 1955A. On the anatomy of Clarias lazera: II. The muscles of the head and the pectoral girdle. Journal of Morphology, 97: 23-38. NAWAR G., 1955B. On the anatomy of Clarias lazera: III. The vascular system. Journal of Morphology, 97: 179-214. NELSON J.A., 2002. Metabolism of three species of herbivorous loricariid catfishes: influence of size and diet. Journal of Fish Biology, 61: 1586-1599. NELSON J.A., WUBAH D.A., WHITMER M.E., JOHNSON E.A. & STEWART D.J., 1999. Wood-eating catfishes of the genus Panaque: gut microflora and cellulolytic enzyme activities. Journal of Fish Biology, 54: 1069-1082. NELSON J.S., 1994. Fishes of the world. New York: John Wiley & Sons, Inc., 600 p. NIKOLSKY G.V., 1963. The ecology of fishes. London: Academic Press. OLIVEIRA RIBEIRO C.A., GUIMARÃES J.R.D. & PFEIFFER W.C., 1996. Accumulation and distribution of inorganic mercury in a tropical fish (Trichomycterus zonatus). Ecotoxicology and Environmental Safety, 34: 190-195. OLSON M.H., 1996. Ontogenetic niche shifts in largemouth bass: variability and consequences for first-year growth. Ecology, 77: 179-190. PART 11 — REFERENCES 235

OLSON K.R., MUNSHI J.S.D., GHOSH T.K. & OJHA J., 1990. Vascular organization of the head and respiratory organs of the air-breathing catfish, Heteropneustes fossilis. Journal of Morphology, 203: 165-179. ONO R.D., 1980. Fine structure and distribution of epidermal projections associated with taste buds on the oral papillae in some loricariid catfishes (Siluroidei: Loricariidae). Journal of Morphology, 164: 139-159. OOZEKI Y., HWANG P.-P. & HIRANO R., 1992. Larval development of the Japanese whiting, Sillago japonica. Japanese Journal of Ichthyology, 39: 59-66. ORTON G.L., 1953. The systematics of vertebrate larvae. Systematic Zoology, 2: 63-75. ORTON G.L., 1955. Some aspects of ecology and ontogeny in the fishes and amphibians. The American Naturalist, 89: 193-203. ØRVIG T., 1967. Phylogeny of tooth tissues: evolution of some calcified tissues in early vertebrates. In: Structural and chemical organization of teeth (Miles A.E.W., ed.), Academic Press (London), p. 45-110. ØRVIG T., 1977. A survey of odontodes (“dermal teeth”) from developmental, structural, functional, and phyletic points of view. In: Problems in vertebrate evolution (Andrews S.M., Miles R.S. & Walker A.D., eds). Academic Press (London), p. 53-75. OSSE J.W.M., 1969. Functional morphology of the head of the perch (Perca fluviatilis L.): an electromyographical study. Netherlands Journal of Zoology, 19: 289-392. OSSE J.W.M. & VAN DEN BOOGAART, J.G.M., 1995. Fish larvae, development, allometric growth, and the aquatic environment. ICES Marine Science Symposium, 201: 21-34. OSSE J.W.M., VAN DEN BOOGAART J.G.M., VAN SNIK G.M.J. & VAN DER SLUYS L., 1997. Priorities during early growth of fish larvae. Aquaculture, 155: 249-258. OTTEN E., 1982. The development of a mouth-opening mechanism in a generalized Haplochromis species: H. elegans Trewavas 1933 (Pisces, Cichlidae). Netherlands Journal of Zoology, 32: 31- 48. OTTEN E., 1983. The jaw mechanism during growth of a generalized Haplochromis species: H. elegans Trewavas 1933 (Pisces, Cichlidae). Netherlands Journal of Zoology, 33: 44-98. PAGE L.M., MOTTESI G.B., RETZER M.E., CEAS P.A. & TAPHORN D.C., 1993. Spawning habitat and larval development of Chaetostoma stannii (Loricariidae) from Río Crucito, Venezuela. Ichthyological Explorations of Freshwaters, 4: 93-95. PATTERSON C., 1975. The braincase of the pholidophorid and leptolepid fishes, with a review of the actinopterygian braincase. Philosophical Transactions of the Royal Society of London, Biological Sciences, 269 (899): 275-579. PATTERSON C., 1977. Cartilage bones, dermal bones and membrane bones, or the exoskeleton versus the endoskeleton. In: Problems in vertebrate evolution (Andrews S.M., Miles R.S. & Walker A.D., eds). Academic Press (London), p. 77-121. PAVLOV D.A., 1999. Features of transition from larva to juvenile in fishes with different types of early ontogeny. Environmental Biology of Fishes, 56: 41-52. PEARSE A.G.E., 1985. Histochemistry. Theoretical and applied. Fourth Edition. Churchill Livingstone, New York, 1055 p. PERNA S.A. & FERNANDES M.N., 1996. Gill morphometry of the facultative air-breathing loricariid fish, Hypostomus plecostomus (Walbaum) with special emphasis on aquatic respiration. Fish Physiology and Biochemistry, 15: 213-220. PFEIFFER W., 1977. The distribution of fright reaction and alarm substance cells in fishes. Copeia, 1977: 653-665. 236 PART 11 — REFERENCES

PODKOWA D. & GONIALOWSKI-WITALINSKA L., 2003. Morphology of the air-breathing stomach of the catfish Hypostomus plecostomus. Journal of Morphology, 257: 147-163. POWER M.E., 1984. Grazing response of tropical freshwater fishes to different scales of variation in their food. In: Evolutionary ecology of neotropical freshwater fishes (Zaret T.M., ed.). Dr. W. Junk Publishers (The Hague), p. 25-37. RADERMAKER F., SURLEMONT C., SANNA P., CHARDON M. & VANDEWALLE P., 1989. Ontogeny of the Weberian apparatus of Clarias gariepinus (Pisces, Siluriformes). Canadian Journal of Zoology, 67: 2090-2097. RAPP PY-DANIEL L.H., 2000. Tracking evolutionary trends in Siluriformes (Teleostei: Ostariophysi). In: International Congress on the Biology of Fish, Aberdeen. Evolution Symposium Proceedings. University of Aberdeen (Aberdeen), p. 57-72. REESE R.O., DELEON M.F. & CONLEY W.J., 1991. Histological effects from long-term storage of common snook yolk-sac larvae in fixatives and alcohol. NOAA Technical Report NMFS 95: 129- 137. REGAN C.T., 1904. A monograph of the fishes of the family Loricariidae. Transactions of the Zoological Society of London, 17: 191-351. REGAN C.T., 1911A. The classification of the teleostean fishes of the order Ostariophysi: 1. Cyprinoidea. Annals and Magazine of Natural History, 8: 13-32. REGAN C.T., 1911B. The classification of the teleostean fishes of the order Ostariophysi: 2. Siluroidea. Annals and Magazine of Natural History, 8: 553-577. REIF W.E., 1982. Evolution of dermal skeleton and dentition in vertebrates. The odontode regulation theory. Evolutionary Biology, 15: 287-368. REIS R.E., 1998. Anatomy and phylogenetic analysis of the neotropical callichthyid catfishes (Ostariophysi, Siluriformes). Zoological Journal of the Linnean Society, 124: 105-168. REIS R.E. & PEREIRA E.H.L., 2000. Three new species of the loricariid catfish genus Loricariichthys (Teleostei: Siluriformes) from southern South America. Copeia, 2000: 1029-1047. REIS R.E., PEREIRA, E.H.L. & ARMBRUSTER J.W., 2006A. Delturinae, a new loricariid catfish subfamily (Teleostei, Siluriformes), with revisions of Delturus and Hemipsilichthys. Zoological Journal of the Linnean Society, 147: 277-299. REIS, R.E., TRAJANO, E., HINGST-ZAHER, E., 2006B. Shape variation in surface and cave populations of the armoured catfishes Ancistrus (Siluriformes: Loricariidae) from the São Domingo karst area, upper Tocantins River, Brazil. Journal of Fish Biology, 68: 414-429. REISSNER E., 1859. Über die Schwimmblase und den Gehörapparat einiger Siluroiden. Archiv. für Anatomie, Phys., und wiss. Medicin, Leipzig, 421-438. RETZER M.E., NICO L.G. & PROVENZANO F.R., 1999. Two new species of Acestridium (Siluriformes: Loricariidae) from southern Venezuela, with observations on camouflage and color change. Ichthyological Explorations of Freshwaters, 10: 313-326. RETZER M.E. & PAGE L.M., 1996. Systematics of the stick catfishes, Farlowella Eigenmann & Eigenmann (Pisces: Loricariidae). Proceedings of the Academy of Natural Sciences of Philadelphia, 147: 33-88. RIEHL R., 1999. Minireview. The micropyle of teleost fish eggs: morphological and functional aspects. Proceedings of the 5th Indo-Pacific Fish Conference, Nouméa, 589-599. RIEHL R. & PATZNER R.A., 1991. Breeding, egg structure and larval morphology of the catfish Sturisoma aureum (Steindachner) (Teleostei, Loricariidae). Journal of Aquariculture and Aquatic Sciences, 6: 1-6. RIEHL R. & PATZNER R.A., 1998. The modes of egg attachment in teleost fishes. Italian Journal of Zoology, 65 (Suppl.): 415-420. PART 11 — REFERENCES 237

RIZZO E., SATO Y., BARRETO B.P. & GODINHO H.P., 2002. Adhesiveness and surface patterns of eggs in neotropical freshwater teleosts. Journal of Fish Biology, 61: 615-632. ROBERTS T.R., 1973. Interrelationships of ostariophysans. In: Interrelationships of fishes (Greenwood P.H., Miles R.S. & Patterson C., eds). Linnean Society of London, Academic Press (London), p. 373-395. ROBERTS T.R., 1975. Geographical distribution of African freshwater fishes. Zoological Journal of the Linnean Society, 57: 249-319. ROBERTS T.R., 1982. Unculi (horny projections arising from single cells), an adaptive feature of the epidermis of ostariophysan fishes. Zoologica Scripta, 11: 55-76. ROBERTS T.R. & KOTTELAT M., 1993. Revision of the Southeast Asian freshwater family Gyrinocheilidae. Ichthyological Explorations of Freshwaters, 4: 375-383. RODILES-HERNÁNDEZ R., HENDRICKSON D.A., LUNDBERG J.G. & HUMPHRIES J.M., 2005. Lacantunia enigmatica (Teleostei: Siluriformes) a new and phylogenetically puzzling freshwater fish from Mesoamerica. Zootaxa, 1000: 1-24. ROJO A.L., 1991. Dictionary of evolutionary fish osteology. CRC Press (Florida, USA), 273 p. ROMÁN-VALENCIA C., 2001. Ecología trófica y reproductiva de Trichomycterus caliense y Astroblepus cyclopus (Pisces: Siluriformes) en el río Quindio, Alto Cauca, Colombia. Revista de Biología Tropical, 49: 657-666. SABAJ M.H., ARMBRUSTER J.W. & PAGE L.M., 1999. Spawning in Ancistrus (Siluriformes: Loricariidae) with comments on the evolution of snout tentacles as a novel reproductive strategy: larval mimicry. Ichthyological Explorations of Freshwaters, 10: 217-229. SALAZAR F.J.M., ISBRÜCKER I.J.H. & NIJSSEN H., 1982. Dentectus barbarmatus, a new genus and species of mailed catfish from the Orinoco basin of Venezuela (Pisces, Siluriformes, Loricariidae). Beaufortia, 32: 125-137. SAUL W.G., 1975. An ecological study of fishes at a site in Upper Amazonian Ecuador. Proceedings of the Academy of Natural Sciences of Philadelphia, 127: 93-134. SAXENA S.C. & CHANDY M., 1966. Adhesive apparatus in Indian hill stream fishes. Journal of Zoology (London), 148: 315-340. SCHAEFER S.A., 1984. Mechanical strength of the pectoral spine/girdle complex in Pterygoplichthys (Loricariidae: Siluroidei). Copeia, 1984: 1005-1006 SCHAEFER S.A., 1987. Osteology of Hypostomus plecostomus (Linnaeus) with a phylogenetic analysis of the loricariid subfamilies (Pisces: Siluroidei). Contributions in Science, 394: 1-31. SCHAEFER S.A., 1988. Homology and evolution of the opercular series in the loricarioid catfishes (Pisces: Siluroidei). Journal of Zoology (London), 214: 81-93. SCHAEFER S.A., 1990. Anatomy and relationships of the scoloplacid catfishes. Proceedings of the Academy of Natural Sciences of Philadelphia, 142: 167-210. SCHAEFER S.A., 1991. Phylogenetic analysis of the loricariid subfamily Hypoptopomatinae (Pisces: Siluroidei: Loricariidae), with comments on generic diagnoses and geographic distribution. Zoological Journal of the Linnean Society, 102: 1-41. SCHAEFER S.A., 1997. The neotropical cascudinhos: systematics and biogeography of the Otocinclus catfishes (Siluriformes: Loricariidae). Proceedings of the Academy of Natural Sciences of Philadelphia, 148: 1-120. SCHAEFER S.A., 1998. Conflict and resolution: impact of new taxa on phylogenetic studies of the Neotropical Cascudinhos (Siluroidei: Loricariidae). In: Phylogeny and classification of neotropical fishes (Malabarba L.R., Reis R.E., Vari R.P., Lucena Z.M. & Lucena C.A.S., eds), Edipucrs (Porto Alegre), p. 375-400. 238 PART 11 — REFERENCES

SCHAEFER S.A. & AQUINO A.E., 2000. Postotic laterosensory canal and pterotic branch homology in catfishes. Journal of Morphology, 246: 212-227. SCHAEFER S.A. & BUITRAGO-SUÁREZ U.A., 2002. Odontode morphology and skin surface features of Andean astroblepid catfishes (Siluriformes, Astroblepidae). Journal of Morphology, 254: 139- 148. SCHAEFER S.A. & LAUDER G.V., 1986. Historical transformation of functional design: evolutionary morphology of feeding mechanisms in loricarioid catfishes. Systematic Zoology, 35: 489-508. SCHAEFER S.A. & LAUDER G.V., 1996. Testing historical hypotheses of morphological change: biomechanical decoupling in loricarioid catfishes. Evolution, 50: 1661-1675. SCHAEFER S.A. & STEWART D.J., 1993. Systematics of the Panaque dentex species group (Siluriformes: Loricariidae), wood-eating armored catfishes from tropical South America. Ichthyological Explorations of Freshwaters, 4: 309-342. SCHAEFER S.A., WEITZMAN S.H. & BRITSKI H.A., 1989. Review of the Neotropical catfish genus Scoloplax (Pisces: Loricarioidea: Scoloplacidae) with comments on reductive characters in phylogenetic analysis. Proceedings of the Academy of Natural Sciences of Philadelphia, 141: 181- 211. SCHAEFFER B. & ROSEN D.E., 1961. Major adaptive levels in the evolution of the actinopterygian feeding mechanism. American Zoologist, 1: 187-204. SCHILLING T.F. & KIMMEL C.B., 1997. Musculoskeletal patterning in the pharyngeal segments of the zebrafish embryo. Development, 124: 2945-2960. SCHMIDT R.E., 2001. Loricaria cataphracta: parental care and description of early larvae. Ichthyological Explorations of Freshwaters, 12: 235-240. SHELDEN F.F., 1937. Osteology, myology, and probable evolution of the nematognath pelvic girdle. Annals of the New York Academy of Sciences, 37: 1-96. SILVA J.M., HENRANDEZ-BLAQUEZ F.J. & JULIO H.F.JR., 1997. A new accessory respiratory organ in fishes: morphology of the respiratory purses of Loricariichthys platymetopon (Pisces, Loricariidae). Annales des Sciences Naturelles, Zoologie, Paris, 13e Série, 18: 93-103. SINGH B.R., 1967. Movements of barbels in some siluroid fishes. Zoologischer Anzeiger, 178: 402- 412. SIRE J.-Y., DAVIT-BEAL T., DELGADO S., VAN DER HEYDEN C. & HUYSSEUNE A., 2002. First- generation teteh in nonmammalian lineages: evidence for a conserved ancestral character? Microscopy Research and Technique, 59: 408-434. SIRE J.-Y. & ALLIZARD F., 2001. A fourth teleost lineage possessing extra-oral teeth: the genus Atherion (Teleostei; Atheriniformes). European Journal of Morphology, 39: 295-305. SIRE J.-Y. & HUYSSEUNE A., 1996. Structure and development of the odontodes in an armoured catfish, Corydoras aeneus (Siluriformes, Callichthyidae). Acta Zoologica, 77: 51-72. SIRE J.-Y. & MEUNIER F.J., 1993. Ornamentation superficielle et structure des plaques osseuses dermiques de quelques Siluriformes cuirassés (Loricariidae, Callichthyidae, Doradidae). Annales des Sciences Naturelles, Zoologie, Paris, 13e Serie, 14: 101-123. SMITH H.M., 1945. The fresh-water fishes of Siam, or Thailand. Washington: Smithsonian Insitution (Bulletin 188), 620 p. SOARES-PORTO L.M., WALSH S.J., NICO L.G. & NETTO J.M., 1999. A new species of Geloglanis from the Orinocco and Amazon river basins, with comments on miniaturization within the genus (Siluriformes: Auchenipteridae: Centromochlinae). Ichthyological Explorations of Freshwaters, 10: 63-72. SPOTTE S., PETRY P. & ZUANON J.A.S., 2001. Experiments on the feeding behavior of the hematophagous candiru, Vandellia cf plazaii. Environmental Biology of Fishes, 60: 459-464. PART 11 — REFERENCES 239

SRINIVASACHAR H.R., 1957A. Development of the skull in catfishes: Part I - Development of chondrocranium in Silonia, Pangasius and Ailia (Schilbeidae). Proceedings of the National Institute of Science, India, 6: 335-356. SRINIVASACHAR H.R., 1957B. Development of the skull in catfishes: Part II - Development of chondrocranium in Mystus and Rita (Bagridae). Gegenbauers Morphologisches Jahrbuch, 98: 244- 261. SRINIVASACHAR H.R., 1958A. Development of the skull in catfishes: Part IV - The development of the chondrocranium in Arius jella Day (Ariidae) and Plotosus canius Ham. (Plotosidae) with an account for their interrelationships. Gegenbauers Morphologisches Jahrbuch, 99: 986-1016. SRINIVASACHAR H.R., 1958B. Development of the skull in catfishes: Part V - Development of the skull in Heteropneustes fossilis (Bloch). Proceedings of the National Institute of Science, India, 24: 165-190. SRINIVASACHAR H.R., 1959. Development of the skull in catfishes: Part III - The development of the chondrocranium in Heteropneustes fossilis (Bloch) (Heteropneustidae) and Clarias batrachus (Linn) (Clariidae). Gegenbauers Morphologisches Jahrbuch, 101: 373-405. STEINDACHNER F., 1879. Fisch-fauna des Magdalenen-Stromes. Denkschriften von der Akademie des Wissenschaften, Wien, 41: 1-52. STRAUSS R.E., 1984. Allometry and functional feeding morphology in haplochromine cichlids. In: Evolution of fish species flocks (Echelle A.A. & Kornfield I., eds). Orono Press, (Maine, USA), p. 217-229. STRAUSS R.E., 1987. On allometry and relative growth in evolutionary studies. Systematic Zoology, 36: 72-75. STRAUSS R.E., 1995. Metamorphic growth-gradient changes in South American loricariid catfishes Lorcariichthys maculatus and Pseudohemiodon laticeps. Studies on Neotropical Fauna and Environment, 30: 177-191. SVÄRDSON G., 1948. Natural selection and egg number in fish. Reports of the Institute of Freshwater Research, Drottningholm, 29: 115-122. SULLIVAN J.P., LUNDBERG J.G., HARDMAN M., 2006. A phylogenetic analysis of the major groups of catfishes (Teleostei: Siluriformes) using rag1 and rag2 nuclear gene sequences. Molecular Phylogenetics and Evolution, 41: 636-662. SURLEMONT C., CHARDON M. & VANDEWALLE P., 1989. Skeleton, muscles and movements of the head of a 5,2 mm fry of Clarias gariepinus (Burchell) (Pisces Siluriformes). In: Trends in vertebrate zoology (Splechtna H. & Hilgers H., eds). Gustav Fischer Verlag (Stuttgart), p. 459- 462. SURLEMONT C. & VANDEWALLE P., 1991. Développement postembryonnaire du squelette et de la musculature de la tête de Clarias gariepinus (Pisces, Siluriformes) depuis l'eclosion jusqu'à 6,8 mm. Canadian Journal of Zoology, 69: 1094-1103. SUZUKI H.I., AGOSTINHO A.A. & WINEMILLER K.O., 2000. Relationship between oocyte morphology and reproductive strategy in loricariid catfishes of the Paraná River, Brazil. Journal of Fish Biology, 57: 791-807. SWINNERTON H.H., 1902. A contribution to the morphology of the teleostean head skeleton, based upon a study of developing skull of the three-spined stickleback (Gasterosteus aculeatus). Quarterly Journal of Microscopic Sciences, 45: 503-597. TAKAHASI N., 1925. On the homology of the cranial muscles of cypriniform fishes. Journal of Morphology and Physiology, 40: 1-109. TAKASUSUKI J., FERNANDES M.N. & SEVERI W., 1998. The occurrence of aerial respiration in Rhinelepis strigosa during progressive hypoxia. Journal of Fish Biology, 52: 369-379. 240 PART 11 — REFERENCES

TAVERNE L., 1999. Les poissons Crétacés de Nardò. 8°. Sorbininardus apuliensis, gen. nov., sp. nov. (Teleostei, Ostariophysi, Anotophysi, Sorbininardiformes, nov. ord.). In: Studi e ricerche sui giacimenti terziairi di Bolca (Sorbini L., ed.). Museo Civico di Storia Naturale (Verona), p. 77- 103. TAVOLGA W.N., 1962. Mechanisms of sound production in the ariid catfishes Galeichthys and Bagre. Bulletin of the American Museum of Natural History, 24: 5-30. TAYLOR J.N., 1983. Field observations on the reproductive ecology of three species of armored catfishes (Loricariidae: Loricariinae) in Paraguay. Copeia, 1983: 254-259. TAYLOR W.R. & VAN DYKE G.C., 1985. Revised procedures for staining and clearing small fishes and other vertebrates for bone and cartilage study. Cybium, 9: 107-119. TEUGELS G.G., 1996. Taxonomy, phylogeny and biogeography of catfishes (Ostariophysi, Siluroidei): an overview. Aquatic Living Resources, 9 (hors série): 9-34. THIELE H., 1963. Vergleichend-morphologische Untersuchungen uber die Funktion der Nahrungserwerbsapparate von Anarrhichas lupus L. und einigen Blenniidae (Teleostei, Blennioidei). Zoologische Beiträge (N.F.), 9: 275-440. TILAK R., 1976. The adhesive thoracic apparatus in the evolution of glyptothoracoid fishes (Sisoridae, Siluriformes). Zoologischer Anzeiger, 196: 255-261. TILNEY R.L. & HECHT T., 1993. Early ontogeny of Galeichthys feliceps from the south east coast of South Africa. Journal of Fish Biology, 43: 183-212. TRAJANO E., 1997. Food and reproduction of Trichomycterus itacarambiensis, cave catfish from south-eastern Brazil. Journal of Fish Biology, 51: 53-63. TRAPANI J., 2001. Position of developing replacement teeth in teleosts. Copeia, 2001: 35-51. UEHARA K. & MIYOSHI S., 1993. Structure of comblike teeth of the Ayu sweetfish, Plecoglossus altivelis (Teleostei: Isospondyli): I. Denticles and tooth attachment. Journal of Morphology, 217: 229-238. UFERMANN A., 1998. Ein weiterer blinder Ancistrus. Deutz Aquarien und Terrarien Zeitschrift, 51: 278. VAL A.L., SILVA M.N.P. & ALMEIDA-VAL V.M.F., 1998. Hypoxia adaptation in fish of the Amazon: a never-ending task. South African Journal of Zoology, 33: 107-114. VAN SNIK G.M.J., VAN DEN BOOGAART J.G.M. & OSSE J.W.M., 1997. Larval growth patterns in Cyprinus carpio and Clarias gariepinus with attention to the finfold. Journal of Fish Biology, 50: 1339-1352. VANDEN BERGHE W., AERTS P., CLAEYS H. & VERRAES W., 1986. A microcomputer-based graphical reconstruction technique for serial sectioned objects, with hidden line removal. The Anatomical Record, 215: 84-91. VANDERVENNET E., WAUTIER K., VERHEYEN E. & HUYSSEUNE A., 2006. From conical to spatulate: intra- and interspecific changes in tooth shape in closely related cichlids (Teleostei; Cichlidae: Eretmodini). Journal of Morphology, 267: 516-525. VANDEWALLE P., BRUNIN P. & CHARDON M., 1986. Functional approach to the morphology of the buccal region of Cteniloricaria platystoma (Günther) (Pisces, Ostariophysi, Loricariidae) with respect to a peculiar respiration. Zoologischer Anzeiger, 217: 363-373. VANDEWALLE P., CHIKOU A. & LALÈYÉ P., 1999. Early development of the chondrocranium in Chrysichthys auratus. Journal of Fish Biology, 55: 795-808. VANDEWALLE P., FOCANT B., HURIAUX F. & CHARDON M., 1992. Early development of the cephalic skeleton of Barbus barbus (Teleostei, Cyprinidae), Journal of Fish Biology, 41: 43-62. PART 11 — REFERENCES 241

VANDEWALLE P., GLUCKMAN I., BARAS E., HURIAUX F. & FOCANT B., 1997. Postembryonic development of the cephalic region in Heterobranchus longifilis. Journal of Fish Biology, 50: 227- 253. VANDEWALLE, P., GLUCKMAN, I. & WAGEMANS, F., 1998. A critical assessment of the Alcian blue/Alizarine double staining in fish larvae and fry. Belgian Journal of Zoology, 128: 93-95. VANDEWALLE P., HUYSSEUNE A., AERTS P. & VERRAES W., 1994. The pharyngeal apparatus in teleost feeding. In: Biomechanics of feeding in vertebrates, Advances in comparative and environmental physiology (Bels V., Chardon M. & Vandewalle P., eds), vol.18, p.59-92. VANDEWALLE P., LALÈYÉ P. & FOCANT B., 1995. Early development of cephalic bony elements in Chrysichthys auratus (Geoffroy Sint-Hilaire, 1808) (Pisces, Siluriformes, Claroteidae). Belgian Journal of Zoology, 125: 329-347. VANDEWALLE P., SURLEMONT C., SANNA P. & CHARDON M., 1985. Interprétation fonctionnelle de modifications du splanchnocrâne pendant le développement post-embryonnaire de Clarias gariepinus (Téléostéens, Siluriformes). Zoologische Jahrbuch der Anatomie, 113: 91-100. VAZ-FERREIRA R. & SEÑORANS J.S., 1971. Oviposición e incubación de Plecostomus alatus Castelnau, en cuevas. Boletín de la Sociedade Zoológica de Uruguay, 1:12-17. VERRAES W., 1974A. Discussion on the shape of the eye and the influence of its size on shape and position of surrounding structures in normal and abnormal conditions during postembryonic development in Salmo gairdneri Richardson, 1836 (Teleostei: Salmonidae). Forma et Functio, 7: 125-138. VERRAES W., 1974B. Discussion on some functional-morphological relations between some parts of the chondrocranium and the osteocranium in the skull base and the skull roof, and of some head parts during postembryonic development of Salmo gairdneri Richardson, 1836 (Teleostei: Salmonidae). Forma et Functio, 7: 281-292. VETTER B., 1878. Untersuchungen zur vergleichenden Anatomie der Kiemen- und Kiefermusculatur der Fische. Jena Zeitschrift für Naturwissenschaften, 12: 431-550. VON IHERING R., BARROS J.C. & PLANET N., 1928. Os óvulos e a desova dos peixes d’água do Brasil. Bolletim Biólogo Brasileiro, São Paulo, 4: 97-109. WAGEMANS F., FOCANT B. & VANDEWALLE P., 1998. Early development of the cephalic skeleton in the turbot. Journal of Fish Biology, 52: 166-204. WAKE D.B. & ROTH G., 1989. The linkage between ontogeny and phylogeny in the evolution of complex systems. In: Complex organismal functions: integration and evolution in vertebrates (Wake D.B. & Roth G., eds). Wiley & Sons (New York), p. 361-377. WAINWRIGHT P.C., 1996. Ecological explanation through functional morphology: the feeding biology of sunfishes. Ecology, 77: 1336-1343. WAINWRIGHT P.C. & LAUDER G.V., 1992. The evolution of feeding biology in sunfishes (Centrarchidae). In: Systematics, historical ecology, and North American fishes (Mayden R.L., ed.). Stanford University Press (Stanford, USA), p. 472-491. WAINWRIGHT P.C., WESTNEAT M.W. & BELLWOOD D.R., 2000. Linking feeding behaviour and jaw mechanisms in fishes. In: Biomechanics in animal behaviour (Domenici P. & Blake R.W., eds). BIOS Scientific Publishers (Oxford), p. 207-221. WASSERSUG R.J. & YAMAHITA M., 2001. Plasticity and constraints on feeding kinematics in anuran larvae. Comparative Biochemistry and Physiology Part A, 131: 183-195. WATANABE K., 1994. Mating behavior and larval development of Pseudobagrus ichikawai (Siluriformes: Bagridae). Japanese Journal of Ichthyology, 41: 243-251. WEBER C., 1987. Hypostomus microstomus sp. nov. et autres poissons-chats cuirassés du Rio Paraná (Pisces, Siluriformes, Loricariidae). Archives des Sciences, Genève, 40: 273-284. 242 PART 11 — REFERENCES

WEBER C., 1992. Révision du genre Pterygoplichthys sensu lato (Pisces, Siluriformes, Loricariidae). Revue Française d'Aquariologie, 19: 1-36. WEISEL G.F., 1967. Early ossification in the skeleton of the sucker (Catostomus macrocheilus) and the guppy (Poecilia reticulata). Journal of Morphology, 121: 1-18. WEITZMAN S.H. & VARI R.P., 1988. Miniaturization in South American freshwater fishes; an overview and discussion. Proceedings of the Biological Society of Washington, 101: 444-465. WERNER E.E., 1977. Species packing and niche complementarity in three sunfishes. American Naturalist, 111: 553-578. WERNER E.E. & GILLIAM J.F., 1984. The ontogenetic niche and species interactions in size- structured populations. Annual Review of Ecology and Systematics, 15: 393-425. WESTNEAT M.W., 1990. Feeding mechanics of Teleost fishes (Labridae; Perciformes): a test of four- bar linkage models. Journal of Morphology, 205: 269-295. WILEY J.L. & COLLETTE B.B., 1970. Breeding tubercles and contact organs in fishes: their occurrence, structure and significance. Bulletin of the American Museum of Natural History, 149: 143-216. WINEMILLER K.O., 1989. Obligate mucus-feeding in a South American trichomycterid catfish (Pisces: Ostariophysi). Copeia, 1989: 511-514. WINEMILLER K.O. & JEPSEN D.B., 1998. Effects of seasonality and fish movement on tropical river food webs. Journal of Fish Biology, 53 (Supl. A): 267-296. WINTERBOTTOM R., 1974. A descriptive synonymy of the striated muscles of the Teleostei. Proceedings of the Academy of Natural Sciences of Philadelphia, 125 (12): 225-317. WITTE F. & VAN OIJEN M.J.P., 1990. Taxonomy, ecology and fishery of Lake Victoria haplochromine trophic groups. Zoologische Verhandelingen (Leiden), 262: 1-47. WORKMAN I.S., 1900. The ophthalmic and eye muscle nerves of the cat fish (Ameiurus). Journal of Comparative Neurology, 10: 403-410. WU K.-Y. & SHEN S.-C., 2004. Review of the teleostean adductor mandibulae and its significance to the systematic positions of the Polymyxiiformes, Lampridiformes, and Triacanthoidei. Zoological Studies, 43: 712-736. YADA O. & FURUKAWA A., 1999. Relationship between external and internal morphological changes and feeding habits in the fry stage of japanese catfish Silurus asotus. UJNR Technical Report, 28: 157-162. YAMAOKA K., 1983. Feeding behaviour and dental morphology of algae scraping cichlids (Pisces: Teleostei) in Lake Tanganyika. African Study Monographs, 4: 77-89. ZARET T.M. & SMITH E.P., 1984. On measuring niches and not measuring them. In: Evolutionary ecology of neotropical freshwater fishes (Zaret T.M., ed.). Dr. W. Junk Publishers (The Hague), p. 127-137. ZELDITCH M.L., SHEETS H.D. & FINK W.L., 2000. Spatiotemporal reorganization of growth rates in the evolution of ontogeny. Evolution, 54: 1363-1371.

PUBLICATION LIST

PUBLICATION LIST 243

PUBLICATION LIST

Updated December 1, 2006.

- Geerinckx T., Adriaens D., Teugels G.G. & Verraes W., 2003. Taxonomic evaluation and redescription of Anaspidoglanis akiri (Risch, 1987) (Siluriformes: Claroteidae). Cybium, 27 (1): 17-25. - Geerinckx T., Adriaens D., Teugels G.G. & Verraes W., 2004. A systematic revision of the African catfish genus Parauchenoglanis Boulenger, 1911 (Siluriformes: Claroteidae). Journal of Natural History, 38 (6): 775-803. - Geerinckx T., Brunain M. & Adriaens D., 2005. Development of the chondrocranium in the suckermouth armored catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Journal of Morphology, 266: 331-355. - Adriaens D., Geerinckx T., Huysentruyt F., Schaefer, S.A. & Herrel A., 2005. Evolution of trophic specialisations in neotropical catfishes: more than a mouthful. Integrative and Comparative Biology 45 (6): 955. - Geerinckx T. & Adriaens D., 2006. The erectile cheek-spine apparatus in the bristlenose catfish Ancistrus (Loricariidae, Siluriformes), and its relation to the formation of a secondary skull roof. Zoology, 109: 287-299. - Geerinckx T., Risch L., Vreven E., Adriaens D. & Teugels G.G., 2007. Claroteidae. In Faune des poissons d’eaux douces et saumâtres de l’Ouest de l’Afrique Centrale / Fauna of the fresh and brackish water fishes of Lower Guinea (Teugels G.G., Stiassny M.L.J. & Hopkins C.D., eds). IRD (Paris) & RMCA (Tervuren), in press. - Geerinckx T., Adriaens D. & Teugels G.G., 2007. Auchenoglanidinae. In Faune des poissons d’eaux douces et saumâtres de l’Ouest de l’Afrique Centrale / Fauna of the fresh and brackish water fishes of Lower Guinea (Teugels G.G., Stiassny M.L.J. & Hopkins C.D., eds). IRD (Paris) & RMCA (Tervuren), in press. - Geerinckx T., Brunain M. & Adriaens D., 2007. Development of the osteocranium in the suckermouth armored catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Journal of Morphology, in press. - Geerinckx T., Brunain M., Herrel A., Aerts P. & Adriaens D., 2007. A head with a suckermouth: a functional-morphological study of the head of the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Belgian Journal of Zoology, in press. - Geerinckx T. & Adriaens D., 2007. Ontogeny of the intermandibular and hyoid musculature in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Animal Biology, in press. - Geerinckx T., Huysentruyt F. & Adriaens D., ——. Ontogeny of the jaw and maxillary barbel musculature in the armoured catfish families Loricariidae and Callichthyidae (Loricarioidea, Siluriformes), with a discussion on muscle homologies. Submitted to the Zoological Journal of the Linnean Society. 244 PUBLICATION LIST

- Geerinckx T. & Adriaens D., ——. Ontogeny of the suspensorial and opercular musculature in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes). Submitted to Zoomorphology. - Geerinckx T., De Poorter J. & Adriaens D., ——. Morphology and development of teeth and epidermal brushes in loricariid catfishes. Submitted to Zoology. - Geerinckx T., Verhaegen Y. & Adriaens D., ——. Ontogenetic allometries and shape changes in the suckermouth armoured catfish Ancistrus cf. triradiatus (Loricariidae, Siluriformes), related to suckermouth attachment and yolk sac size. Submitted to the Journal of Fish Biology. - Huysentruyt F., Geerinckx T. & Adriaens D., ——. A descriptive myology of Corydoras aeneus (Gill, 1858) (Siluriformes: Callichthyidae), with a brief discussion on adductor mandibulae homologies. Submitted to Animal Biology.