Evolution of the Male Genitalia in the Genus Limnebius Leach, 1815 (Coleoptera, Hydraenidae)

Zoological Journal of the Linnean Society, 2016, 178, 97–127. With 16 ﬁgures

Evolution of the male genitalia in the genus Limnebius Leach, 1815 (Coleoptera, Hydraenidae)

ANDREY RUDOY1, ROLF G. BEUTEL2 and IGNACIO RIBERA1,*

1Institute of Evolutionary Biology (CSIC-Universitat Pompeu Fabra), Passeig Maritim de la Barceloneta 37, 08003 Barcelona, Spain 2Institut für Spezielle Zoologie und Evolutionsbiologie mit Phyletischem Museum, Friedrich-Schiller- Universität Jena, Ebertstrasse 1, D-07743 Jena, Germany

Received 10 October 2015; revised 8 January 2016; accepted for publication 25 January 2016

We investigated the evolution of the male genitalia of species of Limnebius (Coleoptera, Hydraenidae). This genus is very homogeneous externally but characterized by highly diverse male genitalia, which are in some cases extraordinarily complex. We reconstructed a molecular phylogeny for 70 of the c. 150 known species of the genus with six fragments of mitochondrial and nuclear genes. We found two main lineages of Miocene origin, largely corresponding to the synonymized subgenera Bilimneus and Limnebius, which are here re-erected. Within the Holarctic Limnebius s.s. we found four well-supported lineages, although with poorly supported relationships between them: the L. piceus, L. gracilipes, L. parvulus and L. nitidus species groups. We describe the aedeagus and its different appendages in detail based on the study of 116 species, including serial histological sections to reconstruct the internal structure of seven of them. Using Bayesian methods we reconstructed the ancestral Limnebius as having small males (c. 1.2 0.5 mm), with a small aedeagus (c. 0.4 0.3 mm) with a free left paramere, probably an externally fused right paramere, and a possible additional appendage. The species of Bilimneus experienced a reduction in size and a simpliﬁcation of their genitalia, without free parameres and with a very simple, homogenous structure. Within Limnebius s.s. several independent increases in male body and genital size took place, with a strong correlation as measured with independent contrasts. There was, however, no overall correlation between genital size and number of appendages, even though smaller genitalia tend to be less complex.

ADDITIONAL KEYWORDS: aedeagus – ancestral state reconstruction – aquatic Coleoptera – molecular phylogeny – morphological complexity.

not necessarily directly relevant in the context of INTRODUCTION speciation. Sexual organs are used for taxonomic purposes in The structure and function of insect genital struc- different groups of organisms. In many cases, insects tures has been extensively studied in an evolutionary and species of other groups of arthropods can only context (e.g. Mayr, 1963; Eberhard, 1985). The stun- be described or recognized by the shape of the male ning complexity in some groups has puzzled zoolo- genitalia, as already pointed out by Sharp & Muir gists and taxonomists, and several hypotheses have (1912) and Sturtevant (1920). Differences in sexual been developed to explain their diversity, among organs are probably key factors in processes of speci- them the lock-and-key mechanism, sexual conﬂict, ation and species maintenance and recognition, cryptic female choice or male competition (for recent while variability in other traits may be due to neu- reviews see Hosken & Stockley, 2004; Masly, 2012). tral geographical variation or local adaptations but There is some evidence supporting different hypotheses in particular systems, such as lock-and-key in groups of Lepidoptera (Mutanen & Kaitala, 2006) or *Corresponding author. E-mail: [email protected] Coleoptera (Sota & Kubota, 1998; Usami et al.,

2006), sperm competition in seed beetles (Hotzy & main types occurring in the genus. Our specific Arnqvist, 2011), ‘one-size-fits-all’ in some insects objectives were (1) to reconstruct the phylogenetic and spiders (Eberhard et al., 1998; House & Sim- relationships between the species of the genus; (2) to mons, 2005), or sexual selection in preying mantis describe the structure of the male genitalia with the (Holwell et al., 2010) or Odonata (McPeek et al., use of histological serial sections; and (3) to trace the 2010). There is also evidence for the role of the evolution of this structure through the phylogeny of shape of the male genitalia in species diversification the genus. (Sota & Tanabe, 2010). Genital characters are regu- larly used in morphological matrixes for phyloge- TAXONOMIC BACKGROUND ON THE GENUS LIMNEBIUS netic reconstruction, sometimes with special attention to their evolution (see, e.g., Lu, Jackman Currently Limnebius includes c. 150 described spe- & Johnson, 1997; Fresneda, Salgado & Ribera, 2007; cies with a body length of 1–3 mm, distributed Sasakawa et al., 2008; Tarasov & Solodovnikov, through all the continents with the exception of 2011; Matsumura et al., 2014). However, there is South America and Antarctica (Hansen, 1998; still a general lack of detailed studies on the evolu- Table S1). The adults of all known species are aqua- tion of the genitalia in speciose lineages. Most con- tic, living in almost all types of continental waters tributions are restricted to single species, or have a with the only exception of saline habitats. Adults very limited taxon sampling. have a very uniform morphology, generally with an The aim of the present study is to reconstruct the oval, drop-like body shape and a flat ventral side; a evolution of the male genitalia in a diverse group of brownish to dark coloration, without well-defined aquatic Coleoptera, Limnebius Leach, a genus marks or design; no marked body sculpture and no belonging to the family Hydraenidae (Staphyli- other conspicuous structural features (Perkins, 1980; noidea). Whereas the external morphology of species Jach,€ 1993; Fig. 1). This body shape is unique among of Limnebius is very uniform (Fig. 1), they vary Hydraenidae, and both morphological (Perkins, 1997; extremely in their male genitalia, ranging from a Beutel, Anton & Jach,€ 2003) and molecular data simple, oar-shaped median lobe without parameres (Abellan & Ribera, 2011; Abellan et al., 2013) sug- to asymmetric complexes with up to seven folded and gest that Limnebius is monophyletic. Its sister genus curved appendices (Perkins, 1980; Jach,€ 1993; is considered to be Laeliaena J. Sahlberg, both form- Fig. 2). The evolutionary origins of this complex pat- ing the tribe Limnebiini. Laeliaena includes three tern and the homology between different structures species in the Himalayan region, and shares with of the adeagus in different species of Limnebius are Limnebius some characters of the head and the tho- presently unknown. rax, and a smooth dorsal habitus (Perkins, 1997; We use a comprehensive molecular phylogeny of Beutel et al., 2003). Limnebius to reconstruct the character evolution of The position of tribe Limnebiini within Hydraeni- the male genitalia. The shape and general structure dae is more controversial. It is currently placed is documented externally and also with histological within the subfamily Hydraeninae (Perkins, 1997; serial sections of seven species representing different Jach,€ 2015), but without phylogenetic evidence to

Figure 1. Habitus of Limnebius. (A) L.(Bilimneus) oblongus Rey. (B) L.(s.s.) hilaris Balfour-Browne. (C) L.(s.s.) hispanicus d’Orchymont. (D) L.(s.s.) truncatellus. Scale bar, 1 mm (from Millan et al., 2014).

Figure 2. Aedeagus of Limnebius. (A) L.(Bilimneus) evanescens. (B), L.(s.s.) gerhardti L. Heyden. (C) L.(s.s.) truncatellus. a1 to a3, additional appendages 1 to 3; a-a3, bifurcating appendage of a3; a-ml, appendage of the median lobe; bf, basal foramen; cml, cavity of the median lobe; fgo, flagellum opening; lp, left paramere; ml, median lobe. a1 in B is the ‘pseudoparamere’ of Jach€ (1993). support this. Its body shape differs in many respects have been rarely addressed using morphological from all other hydraenid genera, and it is also char- characters, and in these rare cases only based on the acterized by some derived features of the head (Beu- male genital morphology. The first systematic tel et al., 2003). In a recent molecular phylogeny of arrangement of the genus was proposed by d’Orchy- Staphylinoidea (McKenna et al., 2015) Limnebius mont (1938) with the creation of two subgenera: was placed as sister to the rest of the sampled Limnebius sensu stricto (s.s.), characterized by the Hydraenidae, including some representatives of presence of at least one paramere (lateral lobe of the Hydraena Kugelann and Ochthebius Leach, thus aedeagus), and Bilimneus Rey lacking parameres. contradicting its placement in Hydraeninae. How- Jach€ (1993) revised the Palaearctic species of the ever, node support was minimal, and the sampling genus taxonomically, describing 16 new to science was again insufficient to evaluate thoroughly inter- and defining species groups based on the male geni- nal relationships within Hydraenidae. talia, but also on some male secondary sexual charac- Due to the homogeneous morphology of the adults, ters of the last abdominal ventrites. In the same work the internal relationships within the genus Limnebius Bilimneus was synonymized, as some species – the

L. mundus group, part of the L. atomus group (as difference between their female genitalia other than defined in Jach,€ 1993) and the North American the degree of sclerotization of the spermatheca. Both species (revised in Perkins, 1980) – may have species have a long, membranous tube between the retained vestigial external parameres and could not spermatheca and the bursa copulatrix (as noted also be clearly fitted in one of the two subgenera based by Perkins, 1980), and multiple ovarioles. on this criterion alone. Jach€ (1993) recognized six informal species groups among the Palaearctic species, three of them based on the presence and type MATERIAL AND METHODS of secondary sexual characters (L. truncatellus, PHYLOGENY OF THE GENUS LIMNEBIUS L. parvulus and L. claviger groups), and three based on different characters of the male genitalia, mainly the number of the appendages and their characteris- Taxon sampling tics (L. atomus, L. nitidus and L. mundus groups). Preliminary results and published data (e.g. Abellan In Jach€ (1993) a long appendage present in about et al., 2013) clearly suggested the separation of the half the species of Limnebius, usually in the position species of Limnebius into two main lineages, one of the right paramere but not articulated with the with only Holarctic species, including those with the median lobe (i.e. with the base fused), was inter- most complex male genitalia, and a second one with preted as a novel structure, formed by the coales- simpler genitalia, comprising all species occurring in cence and fusion of a group of setae and named the southern Hemisphere but also some northern ‘pseudoparamere’ (Fig. 2). However, the homology of ones. The taxonomic knowledge of the Ethiopian, this ‘pseudoparamere’ across different species groups Oriental and Australian species of the genus is very was not well established. A dorsal appendage present incomplete (Jach,€ 1993; Perkins, 2015). As far as pre- in some species was interpreted as a process of the sently known, they are more homogenous both in median lobe and not as a structure formed by coales- body size (around 1 mm) and the structure of the cent setae (a3 in Fig. 2) (Jach,€ 1993). The North male genitalia, always without clearly defined appen- American species were tentatively arranged phyloge- dages. We thus focused on the species of the Holarc- netically in Perkins (1980), based on external simi- tic lineage, obtaining sequence data of 50 of its 91 larities of the aedeagus. However, a formal analysis known species (one of them undescribed, Table S1). was not performed. We included two specimens of a single species, No other works have dealt with the internal sys- L. cordobanus, to test its monophyly and phyloge- tematics of the genus Limnebius. Molecular data are netic position (see Results below). We also obtained a only available for a limited number of species, used comprehensive representation of the mainly southern as outgroups for phylogenies of other genera of lineage, with 20 species (some of them still unidenti- Hydraenidae or in a more general evolutionary con- fied or undescribed) from different geographical text (Hernando, Aguilera & Ribera, 2008; Abellan & regions, including the type species of Bilimneus Ribera, 2011; Abellan et al., 2013; Trizzino et al., (L. atomus (Duftschmid)) (Table S1). 2013; Millan et al., 2014). There is also very limited As noted above, the genus Limnebius has a partic- information on other aspects of the biology of the ular morphology within the family Hydraenidae, and genus. Delgado & Soler (1997) described the larvae its inclusion in Hydraeninae (Perkins, 1997) is pre- of a species of Limnebius for the first time in detail, sently not supported phylogenetically, either by mor- L. cordobanus d’Orchymont. They noted their gen- phological or by molecular data. The taxon sampling eral resemblance to other species of the genus, exam- of available molecular phylogenies (e.g. Abellan ined but not described by the authors. The et al., 2013; McKenna et al., 2015) is insufficient to karyotypes of some species were studied by Angus & clearly resolve its relationships within the family. Dıaz-Pazos (1991), which found an XO sex determi- We therefore used a species of Laeliaena as out- nation system (as in all known Hydraenidae). The group. This genus is considered as sister to species L. papposus Mulsant and L. furcatus Baudi Limnebius based on multiple morphological synapo- di Selve had identical karyotypes, but different from morphies (Hansen, 1991; Jach,€ 1993, 1995; Perkins, that of L. truncatellus Thunberg in the X-chromo- 1997; Beutel et al., 2003). some (fitting with our phylogenetic results, see below). There are only general descriptions of the DNA extraction and sequencing female sexual organs (e.g. Perkins, 1980), but no sys- Specimens were collected in the field and directly pre- tematic comparative studies. In this work we focus served in absolute ethanol. DNA was extracted from on the male genitalia, although some exploratory entire specimens by a standard phenol-chloroform work (mostly with L. furcatus and L. fretalis Peyer- extraction or by commercial extraction kits (most com- imhoff) did not reveal any substantial structural monly DNeasy Tissue Kit columns, Qiagen, Hilden,

Germany), following the instructions of the manufac- combined the results with the use of Logcombiner 1.8 turers. Vouchers and DNA samples are kept in the and Treeanotator 1.8 (Drummond & Rambaut, 2007). collections of the Museo Nacional de Ciencias Natu- Fossils to calibrate the phylogenetic tree are not rales (MNCN, Madrid), and in the Institute of Evolu- available. Therefore, we used the rates estimated in tionary Biology (IBE, Barcelona). DNA extractions Cieslak, Fresneda & Ribera (2014) for a related were non-destructive, to preserve voucher specimens group (familiy Leiodidae, within the same super- for subsequent morphometric and morphological anal- family Staphylinoidea, Beutel & Leschen, 2005; yses. Usually only males were sequenced, and the McKenna et al., 2015) and the same gene combina- male genitalia (used for the identification of the spe- tion based on the tectonic separation of the Sar- cies) were dissected and mounted previous to the dinian plate. We thus set as prior average mean rate extraction to ensure correct identification. a normal distribution with average 0.015 substitu- We amplified and sequenced five fragments, three tions per site Myr1 for the mitochondrial protein mitochondrial (50 end of the cytochrome c oxidase genes, 0.006 for the mitochondrial ribosomal genes subunit 1 [cox1] – the barcode fragment, 30 end of and 0.004 for the nuclear ribosomal genes, all with a cox1, and 30 end of large ribosomal unit plus the Leu- standard deviation of 0.001. cine transfer plus the 50 end of NADH dehydroge- Morphological characters were reconstructed using nase subunit 1 [rrnL+trnL+nad1]), and two nuclear Bayesian probabilities in BEAST 1.8 and parsimony (small ribosomal unit [SSU] and large ribosomal unit in MESQUITE v.3 (Maddison & Maddison, 2015). [LSU]; see Table S2 for the primers used and We used quantitative variables for the size of the Table S3 for accession numbers). For some specimens body and aedeagus and the total number of appen- the 30 end cox1 fragment was amplified using inter- dages, and qualitative for the presence or absence of nal primers to obtain two smaller fragments of each of the individual structures (see Results and ~400 bp each (Table S2). Sequences were assembled Table S1). In BEAST we used a Brownian movement and edited using Geneious v6 (Kearse et al., 2012) or model of evolution for quantitative characters. Sequencher 4.7 (Gene Codes). A total of 218 new To place all known species of Limnebius in the sequences have been deposited in GenBank (EMBL main species groups and to test the sensitivity of our accession numbers LN995192–LN995409, see results to the inclusion of species for which no molec- Table S3). ular data could be obtained, we placed missing species in the molecular phylogenetic tree in their most Phylogenetic analyses likely position according to their morphological fea- For the length-variable regions we used multiple tures. We did not attempt a formal analysis of a pairwise comparisons using the online version of morphological matrix to place these species, as this MAFFT v.6.8 and the G-INS-i algorithm (Katoh & would obviously imply a strong circularity in the Toh, 2008). Phylogenetic analyses were conducted interpretation. However, a certain degree of circular- with Bayesian analyses in BEAST 1.8 (Drummond & ity was unavoidable as genitalia basically provide Rambaut, 2007), as these methods allow the use of a the only reliable set of characters to identify and molecular-clock approach to estimate divergence estimate the relationships between species of Lim- times and to reconstruct the ancestral states of qual- nebius. Some of the characters used to place species itative and quantitative characters. We also analysed were, however, not referring to the male genitalia, the data with maximum likelihood as implemented such as geographical distribution (known to be of in the online version of RAxML (Stamatakis, Hoover phylogenetic relevance in other Hydraenidae beetles, & Rougemont, 2008) with a GTR+G+I model and the Ribera et al., 2010, 2011; Trizzino et al., 2011, 2013), same partitions as in the Bayesian analyses below, or some secondary male sexual characters (see estimating node support with a fast bootstrapping. Results below). In general, only species with a strong Bayesian analyses were conducted on a combined and obvious overall resemblance in their male geni- data matrix using three partitions, the mitochondrial talia were placed as sisters in a well-resolved phylo- protein coding genes (the two cox1 fragments plus genetic position; species without clear siblings were nad1), the mitochondrial ribosomal genes (rrnL generally placed in a polytomy among those with plus trnL) and the nuclear ribosomal genes (SSU similar aedeagal characteristics. Species with no plus LSU), with a Yule speciation process as the tree molecular data were placed in the middle of the cor- prior and an uncorrelated relaxed clock. Analyses responding connecting branch. were run for 100 million generations, ensuring that To estimate the correlation between the size of the the number of generations after convergence was male genitalia, male body size and total number of sufficient as assessed with Tracer v1.6 (Drummond appendages we used a regression of independent con- & Rambaut, 2007) and after removal of the burn-in trasts through the origin using the PDAP package in fraction. We ran two independent analyses and MESQUITE (Midford, Garland & Maddison, 2011),

© 2016 The Linnean Society of London, Zoological Journal of the Linnean Society, 2016, 178, 97–127 102 A. RUDOY ET AL. using both the phylogenetic tree obtained with the Genitalia were dissected, cleaned and dehydrated molecular data only and the estimated phylogeny prior to the inclusion in Araldite CY212 (Agar Scien- with all species included. tiﬁc, Stansted, UK). They were sectioned at 1 lm (cross sections) using an HM 360 microtome (Microm, Thermo Fisher, Waltham, MA, USA) equipped with a MORPHOLOGY OF THE AEDEAGUS diamond blade. Serial sections were stained with tolu- External morphology idine blue and pyronin G (Waldeck GmbH and We studied the morphology of the male genitalia of Co.KG/Division Chroma, Munster, Germany). Sec- 116 of the 147 described species of Limnebius, plus tions were mounted on microscope slides and digitized four undescribed species and the three known spe- using a transmitted light microscope (Zeiss Axioplan, cies of Laeliaena as outgroups (Table S1). Missing Jena, Germany) equipped with a camera (PixeLink species mostly include some for which only very few Capture OEM, Otawa, Canada) using an enlargement specimens are known and could not be accessed. of 2009 or 4009 depending on the size of the geni- Material was obtained mainly from the collections of talia. Images were cleaned and orientated with Gimp the IBE, the Naturhistorisches Museum in Wien v.2 (available at http://www.gimp.org). Only a repre- (NMW) and the Museum of Comparative Zoology in sentative sample of the serial sections was illustrated Harvard (MCZ), the last named for some Nearctic but the described structures were traced through the species. whole series (Table S4). Body length of adults was measured as the sum of the individual maximum lengths of pronotum and Scanning electron microscopy elytra, as the different position of the articulation We observed the aedeagus of some species with scan- between the two could alter the total length when ning electron microscopy (SEM) to complement the measured together. Similarly, the head was not mea- understanding of particular structures (see Results sured, as in many specimens it was partly concealed below). Genitalia were dissected from specimens pre- below the pronotum. Measures were obtained with served in absolute ethanol, glued to hair pins and stereoscope microscopes equipped with an ocular coated with gold (Sputter Coater, Quorum Technolo- micrometer. gies, Lewes, UK). The micrographs were taken with Aedeagi were dissected and mounted on transpar- an ESEM XL30 microscope (Philips, Amsterdam, the ent labels with dimethyl hydantoin formaldehyde Netherlands) and Scandium FIVE software (Olym- (Steedman, 1958), a water-soluble medium with good pus, Tokyo, Japan). optical properties. Male genitalia were always pho- tographed in the same standard positions, orientated according to the foramen in ventral and lateral views RESULTS AND DISCUSSION (Fig. 2). MOLECULAR PHYLOGENY OF LIMNEBIUS Measurements were directly obtained from the dig- ital images using ImageJ v.1.49 (National Institutes Protein coding genes had no indels, and the length of Health; http://imagej.nih.gov/ij/). We estimated variation in ribosomal genes was limited to six nucleo- experimental error by measuring the same specimen tid positions in the rrnL+trnL, 13 in the SSU (in the of three species on three different sessions, using two outgroup Laeliaena) and seven in the LSU fragments. sets of images. For all analyses we used as a single The two BEAST runs converged without difﬁculties value the average of each measure in all studied except for the estimation of the frequency of transver- specimens of the same species (Table S1). sions involving guanine in the rrnL+trnL fragment, which was estimated to be zero in some trees due to Histological serial sections their low frequency. This had, however, no effect on To improve the assessment of the origin of the dif- the topology (as tested with simpler evolutionary mod- ferent appendages we used serial transverse sec- els, see reconstruction of morphological traits below). tions, which are also useful for reconstructing small We applied a conservative burn-in fraction of 10%, internal structural details (e.g. Polilov & Beutel, ensuring that enough trees were sampled, and com- 2009; Jałoszynski, Matsumura & Beutel, 2015). We bined the two runs to obtain a single tree. analysed aedeagi of seven species of the main lin- Results with RAxML and BEAST were very simi- eages within the Holarctic clade, according to pre- lar, and all well-supported clades were shared liminary results of the molecular and morphological between the two methods (Fig. 3). The monophyly of data: L. truncatellus, L. maurus Balfour-Browne, Limnebius was strongly supported, as well as its sep- L. nitiduloides Baudi di Selve, L. pilicauda Guille- aration into two clades, mostly corresponding to the beau, L. fretalis, L. cordobanus and L. furcatus former subgenera Bilimneus and Limnebius s.s. (see (Table S4). Introduction), which are here resurrected as valid

© 2016 The Linnean Society of London, Zoological Journal of the Linnean Society, 2016, 178, 97–127 EVOLUTION OF THE AEDEAGUS IN LIMNEBIUS 103 subgenera (see a checklist of the known species In general, there was a good agreement between in Table S1). Within Bilimneus there were four our results and the species groups proposed by Jach€ well-supported subclades with a loose geographical (1993) based mostly on the general appearance of structure (Fig. 3) (the incomplete sampling pre- the aedeagus and some secondary sexual characters. vented considering these subclades as species groups, Thus, Bilimneus corresponds to the L. atomus group as it was not possible to confidently assign missing of Jach€ (1993), which is characterized by the absence species to any of them): B1, including mostly species of externally visible parameres or appendages of the from the Oriental region, but also from Madagascar aedeagus (with the possible exception of some short and with one western Palaearctic species (L. atomus, subapical appendages found in some species, which the type species of Bilimneus); B2, including mostly may correspond to the right paramere, see below), Mediterranean species, but also a South African rep- small body size (below 1.3 mm), dark brownish body resentative; B3, including species from the Arabian colour, body surface impunctate and absence of sec- Peninsula with one extending over the entire ondary sexual characters. However, none of these Mediterranean (L. myrmidon Rey); and B4, also with features is restricted to species of Bilimneus.As species mostly occurring in the Oriental region, defined here, a potential unambiguous autapomor- including also the only described Australian species phy of the subgenus is the position in the middle to (L. acupunctus Perkins; Perkins, 2004). apical region of the median lobe of a cavity (the ‘cap- Within Limnebius s.s. two weakly supported nodes sule’ of Jach,€ 1993), and the lack of coils in the flag- effectively collapsed the basal relationships within the ellum (with the only possible exception of clade to a polytomy of four subclades: L1, L. piceus L. feuerborni d’Orchymont, see below) (Fig. 2). The group, including all the studied Nearctic species; L2, shape and some features of the venation of the hind L. parvulus group, with L2.1 L. aluta and wings, and the presence of a strongly sclerotized L2.2 L. parvulus subgroups; L3, L. gracilipes group; sperm pump in the species of Bilimneus, seem also and L4, L. nitidus group, with L4.1 L. mundus, to support the separation of the two subgenera L4.2 L. punctatus, L.4.3 L. nitiduloides, L4.4 L. trun- unambiguously (A. Rudoy & I. Ribera, unpubl. catellus and L4.5 L. nitidus subgroups. Within the data). L. nitidus subgroup there were a series of isolated The L. parvulus group and its two subgroups species with poorly resolved relationships among agree with those defined in Jach€ (1993) to include them, with the only exception of a complex of very clo- species with a fringe of long setae on the male stern- sely related species with a similar aedeagal structure, ite VIII. The L. nitidus group as recognized here the L. nitidus complex (the L. nitidus subgroup sensu includes the L. nitidus, L. mundus and L. truncatel- Jach,€ 1993; see also Fresneda & Ribera, 1998). Note lus groups of Jach€ (1993) (defined for species with an that two of the clades have been named after species apical protuberance in the male sternite VIII), with not included in the molecular phylogeny (L. mundus the exclusion of L. aluta Bedel and L. lusitanus Bal- Baudi di Selve and L. punctatus Wollaston), as these four-Browne (found to be sister to the L. parvulus are the earliest described species that can be confi- subgroup) and L. cordobanus (found here to be sister dently placed in their respective clades according to to L. gracilipes Wollaston and related species, our the morphology of their aedeagus (see below). L. gracilipes group). A particular case is that of Of all the recognized clades, only the L. nitidus L. kocheri Balfour-Browne, included in the L. trun- subgroup (clade L4.5) had weak support, and was catellus group by Jach€ (1993) due to the presence of not recovered in the RAxML analyses. This may a protuberance in the male sternite VIII, although have been due to the inclusion of the isolated the similarity of the aedeagus of L. kocheri with that L. murentius d’Orchymont, of which we could only of the species of the L. nitidus group was recognized. obtain two of the genes (cox1 and LSU, Table S3). According to our molecular data this species belongs We repeated the analyses with the same settings but to the L. nitidus group (Fig. 3), in agreement with excluding L. murentius to test its effect on the topol- the structure of the aedeagus and rendering the sim- ogy and support of the phylogenetic tree. The sup- ilarity of the secondary sexual characters of the male port for the L. nitidus group (clade L4) increased to homoplasic. a bootstrap value (BT) of 93% in RAxML and a pos- According to the a-priori rates used in the calibra- terior probability (pp) of 1 in BEAST. The L. nitidus tion of the divergence times, the stem age of Lim- subgroup (clade L4.5) was recovered as monophyletic nebius was estimated at c. 35 9 Ma (Late Eocene), with both methods although still with low support and the age of the crown group at c. 27 6Ma (BT < 50% in RAxML, pp = 0.9 in BEAST, Fig. S1). (Early Miocene, Fig. 3). All the clades defined above In both cases the clade formed by the L. truncatellus were estimated to be of Miocene origin, with the only and L. nitidus subgroups (clades L4.4 and L4.5) was exception of the L. nitidus complex with a very strongly supported (BT = 86%, pp = 1, Fig. S1). recent, Pleistocene origin.

L. hilaris AI919 0.55/- 0.65/68 L. aguilerai AI910 0.55/<50 L. millani AI920 L. monfortei AF199 0.92/<50 L. maurus AI1317* 0.99/0.62 L. ordunyai AI854 1/69 L. gerhardti AI891 L. nitidus cplx 1/100 L. irmelae AI949 L. nitidus AI377 0.78/65 1/<50 L. montanus AI936 L. ibericus RA88 L. kocheri RA91 0.78/- L. bacchus PA274 L. minoricensis AR23 L4.5 0.70/- L. nitidus sbgr L. corybus AI1280 0.86/- L. corfidius AN33 0.86/- 0.51/- L. hieronymi AR22 L. truncatellus AI451* L4.4 1/100 L. truncatellus sbgr L. mesatlanticus AI913 L. schoenmanni RA1000 0.99/79 L. crassipes RA68 0.98/97 L. simplex AC39 0.93/94 L. nitiduloides AR16* 0.6/- 1/100 L. alibeii RA90 1/100 L. fretalis AI1307* L4.3 1/97 1/100 L. thery AI950 L. nitiduloides sbgr L. hispanicus AI1309 L. spinosus AN54 1/100 L. levantinus RA731 L4-L. nitidus gr 0.98/81 0.68/78 L. ignarus AI993 L. zaerensis AC14 1/100 L4.2 L. pilicauda AR60* L. punctatus sbgr 1/100 L. mucronatus AI704 0.58/- L4.1 L. murentius AN53 L. mundus sbgr 1/100 L. cordobanus AI1024* L. cordobanus PA275 1/99 L3-L. gracilipes gr L. fallaciosus RA1121 0.67/- L gracilipes AI1094

1/100 L. furcatus AI987* L. doderoi AI1174 1/99 1/100 L. stagnalis AI824 L. reuvenortali AN45 L2.2 1/87 L. parvulus sbgr L. papposus RA84 1/99 L. crinifer AR55 Limnebius s.str. 1/96 1/100 1/96 1/100 L. rubropiceus AN32 L2-L. parvulus gr L. parvulus AI706 L2.1 L. aluta AI953 1/100 L. aluta sbgr L. lusitanus AI899 L. arenicolus AI466 1/100 L1-L. piceus gr L. sinuatus AR9 1/100 L. piceus RA1120 Philippines AR12 0.89/82 0.9/78 L. nakanei AC30 B4 L. acupunctus AI582 Limnebius 0.93/71 1/100 India AF190 1/100 Bhutan AI1271 1/100 L. pararabicus RA761 1/100 0.85/76 L. dioscoridus RA728 B3 1/88 L. wewalkai RA108

1/100 L. myrmidon AI703 L. oblongus AI942 1/100 1/100 L. perparvulus AI428 0.83/<50 L. extraneus RA89 B2 Bilimneus 1/100 L. evanescens AI915 1/99 L. endroedyi AN78 China RA627 1/98 0.95/91 L. atomus AI1217 L. feuerborni AR11 B1 0.92/97 Madagascar AI549 0.55/<50 L. pollex RA1019 0.55/ - India AF193 Laeliaena sahlbergi HI19 Ma

35 30 25 20 15 10 5 0

Figure 3. Phylogeny of Limnebius as reconstructed with molecular data using Bayesian probabilities in BEAST. Num- bers in nodes are Bayesian posterior probabilities/bootstrap support values (as obtained with a fast maximum-likelihood algorithm in RAxML). See text for codes of the main clades. In bold and with stars are species for which we obtained histological serial sections of the male genitalia (see Figs 4–10).

Thery, Fig. S2:77,127). However, due to the complete STRUCTURE OF THE AEDEAGUS IN THE SPECIES OF fusion of the right paramere with the cuticle of the LAELIAENA AND LIMNEBIUS median lobe below the origin of this appendage (in species with serial sections) and its different anatom- Parameres ical position, the interpretation of these appendages The general structure of the aedeagus in Coleoptera as a possible continuation of the right paramere is (see, e.g., Beutel & Lawrence, 2005) can still be recog- less convincing. In some species a group of setae is nized in Limnebiini, although it is usually highly mod- present on the right side of the apical part of the ified. The basal piece (phallobase) is fused with the median lobe (Jach,€ 1993), but the homology with a median lobe (penis), and the parameres (lateral lobes) true right paramere remains also uncertain. are also partially or completely fused with them. In all In clear contrast, the distal part of the left para- species with serially sectioned specimens the bases of mere is still recognizable in most species of Lim- the two parameres were clearly recognizable as sepa- nebius s.s. Its shape can vary from relatively long rate cavities, generally symmetrical, on both sides of a and curved, covering most of the dorsal part of the central structure corresponding to the median lobe aedeagus (as in e.g. the species of the L. parvulus (Figs 4–10). In all studied species the lumen of the group; Fig. 11) to short and straight (as in e.g. the base of the parameres is connected to the foramen (see species of the L. nitidus group) (Figs 2, 4–10). Even below; see also e.g. Figs 10:12–13, 6:17, 8:16). when the left paramere is partially fused with the In Laeliaena both parameres are fused with the median lobe (e.g. in the L. piceus group, Perkins, median lobe, although they are still externally recog- 1980; Fig. S2:145,149,158) or even completely (as in nizable, and both have free apices always bearing species of the L. mundus group), there is still a rec- setae (Jach,€ 1995; Fig. S2:1–3). In most species of ognizable group of setae inserted on the apical or Limnebius the right paramere is completely fused subapical region of the entire structure, suggesting with the median lobe and cannot be externally iden- the inclusion of a remnant of a paramere (as shown tified (we follow Jach,€ 1993 in the orientation of the with serial sections, see below). In the species of Bil- aedeagus). In species with serially sectioned speci- imneus, also with a completely fused left paramere, mens the lumen of the right paramere could only be there are also always some setae on the left side of clearly identified at the base of the aedeagus. In the median lobe. They probably correspond to the most studied species the lumen of the right paramere apex of the paramere (Jach,€ 1993). fuses completely with that of the median lobe shortly The middle region of the left paramere can also be after the base (e.g. Figs 4:13, 5:10, 10:9). Externally, partly fused with the median lobe, with the two it can only be recognized as a small apical appendage lumina only separated by a thin layer of cuticle in in some species of Bilimneus (e.g. L. boukali Jach€ some species (e.g. L. fretalis, Fig. 7:8–11, or L. ni- or, to a lesser degree, L. rufipennis Regimbart, tiduloides, Fig. 8:13). In the studied species of the Fig. S2:11,51). In the species of the L. parvulus L. punctatus and L. truncatellus subgroups the cuti- group the fusion of the cuticle of the right paramere cle of the left paramere simply fuses with that of the with that of the median lobe is almost complete, but median lobe (Fig. 9:9), while in the L. nitiduloides there is an appendage on the right side of the med- subgroup the cuticle of the paramere is connected ian lobe immediately above the fused area that could with that of the median lobe in a more complex form be interpreted as the apical part of the right para- (Fig. 8:12). mere. Due to the lack of setae or visible pores on this appendage (as observed with SEM in L. stagnalis Ejaculatory duct (‘flagellum’) Guillebeau, Fig. 11) we have not considered it as a A strongly sclerotized ‘flagellum’ is present in all spe- visible right paramere, in agreement with Jach€ cies of Limnebini. In Limnebius s.s. it is coiled (1993). However, the homology of this structure within a cavity at the base of the median lobe (the remains uncertain (see ‘Additional appendages’ ‘basal capsule’, see below) (Perkins, 1980; Jach,€ 1993, below; Figs 6:9,10, 11). Other species have also an 1995). In species of the L. parvulus subgroup (Fig. 4; appendage on the right side of the aedeagus (e.g. Fig. S2:132,141) this basal capsule is very distinct, L. fretalis and L. nitiduloides, Figs 7, 8; also present, with multiple loops of the flagellum in its interior. but less apparent, in L. ferroi Jach€ or L. mesatlanticus The flagellum can be partly fused with the upper

Figure 4. Histological serial sections of the aedeagus of L. furcatus. a1, additional principal ventral appendage; a2, additional secondary ventral appendage; a3, additional dorsal appendage; a-a1 to a-a3, bifurcating appendage of a1 to a3; a-ml, appendage of the median lobe; aa, apical appendage of the median lobe; b, base; bc, basal capsule; bf, basal foramen; cm, connecting membrane; con, connection; fge, flagellum entrance; fgo, flagellum opening; fu, fused; lp, left paramere; ml, median lobe; op, open; rp, right paramere; s, setae. Green, flagellum and related structures; brown, median lobe; orange, parameres; red, additional appendages.

Figure 4. Continued

loides it is dorsal and with a lamina (Fig. 8:3), while in L. fretalis it is small and placed inside a special fold of the median lobe (Fig. 7:4). In the species of Bilimneus the flagellum opening is always nearly apical, without any well-defined structure of the median lobe beyond it. The flagellum is normally retracted inside the aedeagus and only rarely seen extended. Perkins (1980) noted that among more than 1000 studied specimens only one had it extended (in L. oza- palachicus Perkins, fig. 75F in Perkins, 1980). Even though the holotype of L. endroedyi Perkins is depicted with an everted flagellum in Perkins (2015), it is noted that the usual condition of the species is with this structure retracted. Jach€ (1993) also noted that among all the material he studied only one specimen had an extended flagellum, in the species L. kweichowensis Pu (see fig. 19b in Jach,€ 1993). In Ferro (1989) L. cuspidatus Ferro (currently a syn- onymy of L. atomus,Jach,€ 1993) is illustrated with an everted flagellum, which was interpreted as a fili- form paramere by the author. Among all the material examined we found only one specimen of L. hieronymi Vorst with extended flagellum (Fig. S2:104). Among the species newly described from South Africa by Perkins (2015) one, L. masculi- nus Perkins, seems to have a permanently everted apex of the flagellum. However, as noted by the author, the homology of this exposed structure with the internal flagellum is not well established. In all species with histological sections the flagellum can be distinguished as a tubular structure, hollow and well sclerotized, with sometimes some undifferentiated cells or tissue in its interior (e.g. Figs 6:17, 9:11). The function of the flagellum is thus likely to transfer sperm during the copula, i.e. it can be considered an ejaculatory duct, similar to that found in other beetle groups (e.g. Rodrıguez, Windsor & Eberhard, 2004). It does not seem to be a mechani- cal aid for the copula, as can be the case in other Figure 5. Histological serial sections of the aedeagus of groups of Coleoptera with similar structures (e.g. L. cordobanus. See legend to Figure 4 for definitions. Scydmaeninae, Jałoszynski et al., 2015). Presently there are no data on the mechanics of the copula in part of the basal capsule (see e.g. Figs 5:10,11, Limnebius, and the mechanism by which the flagel- 6:15,16, 10:10,11). lum is everted is unknown. Perkins (1980) noted that The flagellum exits the median lobe through an Limnebius females have a long tube between the opening usually located in its apical third, in some bursa copulatrix and the spermatheca, which agrees cases forming a small channel (Fig. 4:5). In some with our observations (see above). This could suggest species the opening is surrounded by or near to a that at least a part of the flagellum is inserted in the group of setae (e.g. L. parvulus group, Fig. 4:5), or it female genital track during the copula. is located on a small prominence of the median lobe (in the L. truncatellus and L. punctatus subgroups, Median lobe Figs 6:3–5, 9:3,4). In the L. nitiduloides subgroup The median lobe is asymmetrical in all species of the relative position of the flagellum opening is less Laeliaena and Limnebius. In most species of Bilim- apical, without any special prominence, but associ- neus it is either straight or tends to be curved to the ated with different structures. Thus, in L. nitidu- left, while in most species of Limnebius s.s. it tends

Figure 6. Histological serial sections of the aedeagus of L. pilicauda. See legend to Figure 4 for deﬁnitions.

Figure 7. Histological serial sections of the aedeagus of L. fretalis. See legend to Figure 4 for deﬁnitions.

Figure 8. Histological serial sections of the aedeagus of L. nitiduloides. See legend to Figure 4 for deﬁnitions.

Figure 8. Continued

Figure 9. Histological serial sections of the aedeagus of L. truncatellus. See legend to Figure 4 for deﬁnitions.

Figure 10. Histological serial sections of the aedeagus of L. maurus. See legend to Figure 4 for definitions. to be curved to the right (as in e.g. the L. nitidus flattened, divided by thin cuticular walls (as in e.g. subgroup). However, there are exceptions in both lin- some of the species with the most complex aedeagus eages (Fig. 2; Fig. S2:77,78,106). of the L. nitidus group, Figs 7:10, 9:9) or even subdi- In all species of Limnebius a spherical or oval hol- vided completely (e.g. in L. nitiduloides, Fig. 8:12). low capsule is present inside the median lobe. This In the distal part of the median lobe, above the capsule is basal and the flagellum coiled inside in capsule, the lumina of the parameres are usually species of Limnebius s.s. (the ‘basal capsule’). In con- totally (as in e.g. L. furcatus, Fig. 4:12,13) or at least trast, it is placed medially or apically in the species partially fused with the lumen of the median lobe. of Bilimneus, with a different shape and with an The lumen of the right paramere is separated from uncoiled flagellum (Fig. 2). In L. feuerborni (included that of the median lobe by thin cuticular walls in in the clade B1 of Bilimneus, Fig. 3) the main cavity some species, with a poorly defined structure (e.g. of the median lobe is larger and in a more basal posi- L. pilicauda, Fig. 6:10). Usually its distal part is tion within the median lobe than in all other species completely fused with the cuticle of the median lobe of Bilimneus (d’Orchymont, 1932; Fig. S2:27). How- (e.g. L. fretalis, Fig. 7:9,10, or other species of the ever, it has a different structure than in species of L. nitidus group, Figs 10:9, 8:12). It is thus uncertain Limnebius s.s. and does not reach the base of the whether the appendages or other structures originat- aedeagus. ing from the walls of this cavity are homologous with The basal capsule is usually strongly sclerotized in the right paramere (see above). Regardless, we use the species of Limnebius s.s. (e.g. Figs 2, 9). In some the presence of the flagellum to identify the central species the upper part of the basal capsule can be cavity resulting from the complete fusion of the med-

1912). The structure of the basal foramen is very A conserved, with a strongly sclerotized ring with a peg-like structure in the distal part and a basal pointed projection (Jach,€ 1993; Fig. 2). The conservation of this structure also affects the base of the parameres, the lumina of which open into the foramen in all studied species (see above). The only variation refers to its general shape: in Laeliaena and most species of Bilimneus the foramen is oval or somewhat triangular (with some exceptions, e.g. L. pararabicus Jach€ & Delgado, L. rufipennis or L. taiwanensis Jach),€ whereas it is round in most B species of Limnebius s.s. (with the exception of some species of the L. mundus group, see below) (Table S1, Fig. S2). The relative position of the foramen varies with the shape of the base of the aedeagus; in species with a well-developed capsule (e.g. within the L. parvulus group, Figs 4, 11) it is more lateral and in a more distal position, while in species with a straight base (e.g. some species of Bilimneus, Fig. 2A) it is almost at the base of the aedeagus. Figure 11. Aedeagus of L. stagnalis (specimen from Slovenia, Carniola, Cerknisko Polje, river Strzen, I. Rib- Additional appendages era, A. Cieslak & C. Hernando leg. 2007). (A) distal part, Additional appendages of the aedeagus are present ventral view; (B) lateral view. a1, additional appendage 1; in many species of Limnebius s.s., resulting in the bc, basal capsule; lp, left paramere; ml, median lobe. highly complex male genitalia typical for the genus (Jach,€ 1993). These appendages can be distinguished ian lobe and the right paramere. There are some dif- based on their position on the ventral or dorsal side ferences in the anatomical position of this fused right and on their origin. In the species with the most paramere + median lobe among the lineages of Lim- complex aedeagus they may have secondary subdivi- nebius s.s. In the L. parvulus group (represented by sions. One or two main ventral appendages can occur L. furcatus) it lies on the right side of the aedeagus, (a1 and a2, corresponding in most cases to appen- which is strongly flattened dorsoventrally and con- dages A and B in Jach,€ 1993), and one dorsal (a3, stricted in the central region (Fig. 4:13,14). In other corresponding to appendage C in Jach,€ 1993). All of groups it is placed in a more central position (e.g. them can be subdivided. In most cases we have iden- L. nitidus group, Fig. S2:81,96). tified these appendages based on their position but The apical part of the median lobe, beyond the also on structural features. However, the homology opening of the flagellum, extends and forms different between appendages of different groups of Limnebius structures in most species of Limnebius s.s. In some remain uncertain. In some species they may repre- species (as in e.g. L. papposus or L. doderoi Gridelli sent vestiges of the right paramere. within the L. parvulus group, Fig. S2:33,90) they are Appendage a1 is present in all species included in large and cover most of the apex of the aedeagus, the L. gracilipes group and in most species of the but in others (mostly within the L. nitidus group) L. nitidus group (except for the species of the they are thin and more lateral (e.g. L. truncatellus, L. mundus subgroup and some within the L. nitidus Figs 2, 9). In species of Bilimneus the apex of the subgroup, Table S1). In almost all studied species aedeagus is usually simple, although in some the appendage a1 has multiple internal channels (see apical part can have some hook-like small appen- e.g. the base of a1 in L. nitiduloides, Fig. 8:8–14), dages (e.g. L. pollex Jach€ & Delgado; Jach€ & with the only exception of L. fretalis, in which the Delgado, 2013; Fig. S2:47). channels fuse in a single lumen just after the base (Fig. 7:2–14). Appendage a2 usually has a single Basal foramen internal cavity, although in L. nitiduloides the apical In all species of Limnebiini, as in other genera of part shows a complex shape with multiple folds and Hydraenidae, the ejaculatory duct enters the aedea- cavities (Fig. 8:3–7). gus through a ventral opening at the base, the basal In the species of the L. nitidus subgroup appendage foramen (the ‘median foramen’ of Sharp & Muir, a1 forms the ‘pseudoparamere’ described by Jach€

small channels, as for instance in L. fretalis A (Fig. 7:10) or L. nitiduloides (Fig. 8:11). In the species with small channels, apical to the area connected to the central cavity of the median lobe the cuticle of a1 tightly merges with that of the median lobe (Fig. 8:12). In the species of the L. parvulus subgroup an appendage originates from the area where the right paramere fuses with the median lobe (see B ‘Parameres’ above). In Jach€ (1993) this is interpreted as a homologue of the ventral appendage a1 (his appendage A) in other species. Although its position and the internal structure are similar to that of a1 in the species of the L. nitidus group, the histological sections of L. furcatus show that this appendage C probably formed as an extension of the cuticle of the median lobe (Figs 4:8–11, 11). An additional fold of the dorsal part of the median lobe is present in some species of the L. parvulus subgroup. This is appar- ently not an appendage, as its cuticle remains fused to the median lobe over its entire length (Fig. 4:7–12). The secondary ventral appendage (a2) is usually Figure 12. Aedeagus of L. mucronatus (specimen from shorter and less curved than a1, without any specific Corsica, Corte, river Restonica, I. Ribera & A. Cieslak modifications. In most cases it is located more later- leg. 1999). (A) dorsal view; (B) lateral view; (C) distal ally, with the base very close to that of the left para- part, frontal view. a1–a3, additional appendages 1 to 3; mere, or even surrounded by it in some species (e.g. bc, basal capsule; lp, left paramere; ml, median lobe. Fig. 8:12–15). Appendage a2 only occurs if a1 is also present, in the L. nitiduloides, L. punctatus and L. truncatellus subgroups and in some species of the (1993). Its elongate shape and general appearance led L. gracilipes group (Table S1). some authors to consider it as the right paramere, but The lower part of both ventral appendages is usu- histological sections show that this is not the case. ally enclosed in a deep groove of the median lobe, This is in agreement with Jach€ (1993), who based his formed by the base of the parameres and the median interpretation mostly on the lack of setae or microp- lobe itself. This groove is most distinct in the L. ni- ores, which are always present in true parameres. tiduloides subgroup, where ventral appendages are With our data it is not possible to test if this appen- very strongly developed (Figs 7:6–12, 8:8–13). In dage was formed by fusion of setae, as suggested by other species, such as those of the L. punctatus and Jach€ (1993). However, the fact that in some species L. truncatellus subgroups, the base of the parameres the appendage may be hollow for most of its length does not encircle a2, but only forms a widely open suggests that it may have originated as an extension concavity (Figs 6:10, 9:8–11). of the cuticle, as it probably happened with other addi- In some species a1 and a2 intersect, as in L. pili- tional appendages. The similarities in the internal cauda and others within the L. punctatus subgroup structure and the connection with the median lobe of (Fig. 6). In these cases, a1 occupies a more central the ‘pseudoparamere’ of the L. nitidus subgroup with position on the aedeagus. Both a1 and a2 can form the a1 of other studied species of the L. nitidus group different structural types in the apical region, from (in particular L. truncatellus and L. pilicauda, Figs 6, short flap-like expansions of the cuticle (e.g. L. cor- 9) support this interpretation. The ‘pseudoparamere’ dobanus, Fig. 4) to long pectinated extensions (e.g. may be functionally analogous to the right paramere L. truncatellus, Fig. 9). in other species of Coleoptera, as it is placed in the The dorsal appendage a3 is present in all species same position and has a similar elongated structure. of the L. truncatellus, L. punctatus and L. nitidu- However, the function of the parameres (and other loides subgroups, and in some species of the L. gra- appendages) in Limnebius remains unclear. cilipes group (Table S1). Its structure is similar to The lumen of appendage a1 is always connected at that of a1, although its base is different from both a1 some point to the central cavity of the median lobe, and a2: it originates as a flat, thin extension of the either by fusion of the cuticles (as in e.g. L. cor- dorsal side of the median lobe, actually forming its dobanus, Fig. 4:9) or through an area with abundant dorsal wall in the basal and middle regions

(Figs 7:8–12, 9:9–12). In most studied species the with which they are parapatric (Jach,€ 1993). Within lumen of a3 is clearly visible. In L. pilicauda (and the L. gracilipes group, L. canariensis d’Orchymont probably in other species of the L. punctatus sub- was placed as sister to L. gracilipes (d’Orchymont, group, clade L4.2) the lumen is very narrow due to 1940), with a very similar aedeagus and both ende- the dorsoventral flattening (Figs 6:3–7, 12). In mic to the Canary Islands (Jach,€ 1993); and L. pa- L. truncatellus it is reduced to a narrow channel due ganettii Ganglbauer as sister to L. fallaciosus to the strong sclerotization and subdivision of the Ganglbauer, with the same general structure of the appendage (Fig. 9:5,6). In L. pilicauda a3 is cylindri- appendages. Within the L. punctatus subgroup (clade cal at its base. It originates simply as an extension L4.2) L. punctatus and L. similis Wollaston where of the median lobe and flattens only in the apical placed as unresolved, both displaying the same gen- region (Fig. 6), in contrast to the usual shape in all eral structure of the aedeagus as for other species of other species (base flattened and apex cylindrical). In the group (Jach,€ 1993); L. kamali Sainz-Cantero & L. mucronatus Baudi di Selve, also within the Bennas was placed in an unresolved clade with L. punctatus subgroup (Fig. 3), a3 is only visible as a L. pilicauda, L. zaerensis Hernando et al. and L. ig- short flap-like structure on the dorsal side of the narus, the first three occurring in Morocco (Sainz- aedeagus, superficially resembling a short paramere Cantero & Bennas, 2006; Hernando et al., 2008). (Fig. 12). In the L. truncatellus subgroup the distal Within the L. nitiduloides subgroup (clade L4.3), region of a3 is divided to form three separate sub- L. calabricus Jach€ was placed as sister to L. simplex appendages (Fig. 9:5), two with a similar, flattened Baudi di Selve (Jach,€ 1993). Within the L. nitidus shape (Fig. 9:5,6) and a cylindrical one that extends subgroup (clade L4.5), L. nitifarus d’Orchymont was to the apex of the aedeagus (Fig. 9:1). placed in a polytomy with the other species of the The different types of connection of a3 with the L. nitidus complex, with the exclusion of L. nitidus median lobe suggest that it was formed indepen- and L. montanus Balfour-Browne (Jach,€ 1993; Fres- dently in several lineages. It is central in the L. ni- neda & Ribera, 1998) (Figs S2, S3). tiduloides subgroup (Fig. 8:14,15), inserted on the The affinities of other species were more uncertain, left side in the L. truncatellus subgroup without close similarities but with some characters (Fig. 9:11,12), which is also characterized by a longi- linking them to a larger number of species. The two tudinally subdivided apex, and of a less complex Turkish species L. claviger Jach€ and L. setifer structure in the L. punctatus subgroup, where it Iablokoff-Khnzorian seem very close to each other lacks additional structures (Figs 6:10, 11). Its pres- (Jach€ & Skale, 2011; Fig. S2:142,143). Their aedea- ence seems to be strongly correlated with that of a2: gus is similar to that of other species of the L. parvu- of the 23 species of Limnebius with a2, only two lack lus subgroup (clade L2.2), with which they share the a3 (both within the L. gracilipes group, Table S1). presence of a large and well-sclerotized basal capsule with a constriction in the distal part. In Jach€ (1993) L. claviger was placed in its own species group, not- ESTIMATION OF THE RELATIONSHIPS OF SPECIES ing that the male ventrite VI had no protuberance WITHOUT MOLECULAR DATA (in contrast to the species of the L. truncatellus Most of the species for which no molecular data were group) but a depression flanked by ridges. In Jach€ & available had male genitalia strongly resembling Skale (2011) L. setifer is shown to share the same those of other species included in the phylogeny. sexual dimorphism (mentioned also in the original This allowed their unambiguous placement as sister description, Iablokoff-Khnzorian, 1962) and aedeagus group to a single species or, in some cases, in a structure with L. claviger. We conservatively placed polytomy formed by a small number of species L. claviger and L. setifer in a polytomy at the base of (Figs S2, S3). the L. parvulus group (Fig. S3), although the size The three species of Laeliaena (Jach,€ 1995) were and complexity of their male genitalia may indicate placed in an unresolved clade as sister to Limnebius. a closer relationship with the species of the L. parvu- Within Bilimneus, L. arabicus Balfour-Browne was lus subgroup. placed in an unresolved clade with L. pararabicus With the L. nitiduloides subgroup (clade L4.3), and L. dioscoridius Jach€ & Delgado (Jach€ & Del- L. simulans d’Orchymont shares a similar structure gado, 2012), and L. nanus Jach€ as sister to of the apex of the main ventral appendage a1 with L. evanescens Kiesenwetter (Jach,€ 1993). Within the L. crassipes Kuwert and L. schoenmanni Jach,€ but L. parvulus subgroup (clade L2.2), L. gridellii Pret- the structure of the left paramere is similar to the ner was placed in an unresolved clade with L. furca- condition in L. levantinus Jach€ and L. spinosus Jach€ tus and L. doderoi (Jach,€ 1993), and L. shatrovskiyi (Jach,€ 1993; Fig. S2:92,97). It was thus placed in a Jach€ and L. glabriventris Shatrovskiy in an unre- basal polytomy within the clade, leaving five main solved clade with the widespread L. parvulus Herbst, lineages within the subgroup (Fig. S3).

The species L. grandicollis Wollaston, L. nitigeus mundus group includes the only species of Limnebius d’Orchymont and L. graecus Jach€ share many of the s.s. with an oval or roughly triangular basal foramen characters of the L. nitidus subgroup, particularly (e.g. L. murentius, Fig. S2:82). Species with an oval L. grandicollis (the other two are characterized by a or triangular basal foramen also tend to have a more somewhat deviating apical part of the median lobe, straight median lobe of the aedeagus. However, with- Fig. S2:102,108). However, they cannot be reliably out additional data it is not possible to distinguish placed as sister to any of the species (Jach,€ 1993). well-defined clades within the group, with the excep- We thus collapsed the basal nodes within this group tion of the close relationship between L. murentius (which was poorly supported in any case), and placed and L. attalensis Jach€ (Jach,€ 1993; Figs S2, S3). these species in the resulting polytomy (Fig. S3). The highest proportion of species without molecular EVOLUTION OF THE AEDEAGUS data was concentrated in three lineages: Bilimneus, the Nearctic L. piceus group sensu Perkins (1980), We used the molecular phylogeny to reconstruct the and the L. mundus group sensu Jach€ (1993). Within evolution of the aedeagus in Limnebius, including its Bilimneus we tried to include a sample of species size, the presence or absence of externally visible with different types of aedeagus and from different parameres, and the number and type of additional regions. The affinities of the species are in general appendages. To improve statistical convergence in uncertain due to the general simple structure of the BEAST we used an HKY+G model of evolution for aedeagus. Consequently, we either placed them as the ribosomal genes, keeping the same partitions as unresolved or followed a geographical criterion. in the BEAST and RAxML analyses above. The Regardless, the simplicity of the aedeagus makes the resulting topology maintained the same well- precise position irrelevant in the context of the evolu- supported nodes, with variation only in weakly tion of the genitalia, as most species display the same supported clades (e.g. the internal topology of the basic structural features (Table S1, Fig. S2). L. nitidus groups, Figs 11–13). All Nearctic species form a monophyletic lineage according to Perkins (1980), based on the presence of Evolution of the size of the male body and the a group of two setae on the median lobe of the aedea- aedeagus gus considered as vestiges of the fused left paramere. The reconstructed ancestral Limnebius was small We followed his phylogenetic hypothesis with the (c. 1.2 0.5 mm; Fig. 11), with a small (c. 0.4 exception of a closer relationship of L. arenicolus 0.3 mm; Fig. 12), simple aedeagus with an externally Perkins to L. sinuatus (Sharp), instead of to L. visible separate left paramere. Ptiliidae, the sister piceus (Horn), following the results of the molecular group of Hydraenidae (Hunt et al., 2007; McKenna data, and collapsing some of the nodes for which et al., 2015), also includes very small beetles (most of affinities were uncertain. them between 0.5 and 1.5 mm), often with an oval Jach€ (1993) defined the L. mundus group for some body shape, and in many cases a simple aedeagus species with a main eastern Mediterranean distribu- (Hall, 2005). McKenna et al. (2015) placed Limnebius tion and a very simple aedeagus, without visible (with low support) as sister to the rest of the sam- parameres. However, some subapical setae present pled Hydraenidae. This opens the possibility that in all species (usually four) may represent a vestigial small size, simple genitalia and oval body shape may external left paramere (as in the L. piceus group, see be ancestral for the whole family Hydraenidae, above). We follow the definition of this group, for although the largely unclarified internal relation- which molecular data are only available for a single ships in Hydraenidae and Ptiliidae preclude any species (L. murentius, Table S1). We tentatively robust conclusions. include the Himalayan L. nigritus Balfour-Browne in There was a strong positive correlation between this group, which is somewhat similar to L. kaszabi adult male size and aedeagus size, both measured Chiesa from Afghanistan. It shares the complete with independent contrasts (regression through the external fusion of the left paramere with the median origin using the topology of Fig. 3, r = 0.49, n = 66, lobe with the other species of the group, and also the P < 0.0001) or with the reconstructed values at all structure of the basal capsule and the apical part of nodes in the phylogeny (r = 0.93, n = 142, P < 0.0001 the median lobe (Fig. S2:83). Although Balfour- in the reconstructed topology in Figs 13, 14). This Browne (1956) included this species in the subgenus correlation increased when the tree with all species Bilimneus based on the lack of visible parameres, he was used (regression through the origin using the noted in the description the different structure of the topology of Fig. S3, r = 0.66, n = 118, P < 0.0001; aedeagus, as well as the larger size of the specimens, Fig. 15A). Changes of the largest magnitude mostly darker colour and the punctured dorsal surface (all occurred among the terminals in the phylogeny, such features typical of Limnebius s.s., see above). The L. as a strong increase in the aedeagus at the origin of

1.37 1.35 L. hilaris AI919 lm 2.45 1.39 1.38 L. aguilerai AI910 1.44 L. monfortei AF199 1.39 1.41 1.42 L. gerhardti AI891 1.38 1.44 L. maurus AI1317

1.39 1.30 L. ordunyai AI854 1.36 L. millani AI920 L. nitidus cplx 1.43 L. irmelae AI949 1.38 1.36 L. montanus AI936 1.31 1.38 1.40 L. nitidus AI377 1.32 1.21 L. ibericus RA88 1.26 L. bacchus PA274 1.31 L4.5 1.33 L. kocheri RA91 1.33 1.22 L. corybus AI1280 L. nitidus sbgr 1.23 1.27 1.14 L. corfidius AN33 1.28 L. hieronymi AR22 1.28 L4.1 1.09 L. murentius AN53 1.24 1.24 L. minoricensis AR23 2.45 L. nitiduloides AR16 2.17 1.96 L. simplex AC39 0.82 2.14 2.13 L. schoenmanni RA1000 L4-L. nitidus gr 2.20 1.55 2.06 2.33 L. crassipes RA68 1.98 L. spinosus AN54 2.01 L4.3 2.03 2.00 L. levantinus RA731 L. nitiduloides sbgr 2.27 2.39 L. fretalis AI1307 2.23 2.16 L. alibeii RA90 1.87 2.14 2.24 L. thery AI950 2.06 L. hispanicus AI1309 L4.4 2.38 L. truncatellus AI451 2.11 1.59 L. truncatellus sbgr 2.04 L. mesatlanticus AI913 1.31 L. zaerensis AC14 1.29 1.11 L. ignarus AI993 Limnebius s.str. L4.2 1.32 1.41 1.34 1.51 L. pilicauda AR60 L. punctatus sbgr 1.18 L. mucronatus AI704 2.07 L. papposus RA84 1.98 1.93 L. crinifer AR55 1.95 1.87 1.93 L. rubropiceus AN32 1.93 L. parvulus AI706 L2.2 1.69 1.96 L. furcatus AI987 L. parvulus sbgr 2 2.05 L. doderoi AI1174 L2-L. parvulus gr 1.82 1.51 1.94 1.87 L. stagnalis AI824 2.02 L. reuvenortali AN45 L2.1 1.11 L. lusitanus AI899 1.30 L. aluta sbgr 1.08 L. aluta AI953 1.42 0.93 0.92 L. cordobanus AI1024 L3-L. gracilipes gr 0.92 L. cordobanus PA275 1.25 1.32 L. gracilipes AI1094 1.26 1.31 L. fallaciosus RA1121 Limnebius 1.34 1.22 1.26 L. arenicolus AI466 1.22 L1-L. piceus gr 1.22 L. sinuatus AR9 1.21 1.12 L. piceus RA1120 0.86 0.86 L. nakanei AC30 0.89 0.85 L. acupunctus AI582 B4 Philippines AR12 0.92 - - India AF190 0.99 - Bhutan AI1271 0.92 0.83 0.84 L. dioscoridus RA728 0.82 L. pararabicus RA761 B3 0.87 0.89 0.83 L. wewalkai RA108 0.89 L. myrmidon AI703 1.21 0.96 1.01 L. oblongus AI942 0.95 0.92 0.91 L. perparvulus AI428 0.84 L. extraneus RA89 B2 0.93 Bilimneus 0.95 0.88 L. evanescens AI915 1.01 1.03 L. endroedyi AN78 0.99 L. atomus AI1217 0.98 - China RA627 0.98 0.99 India AF193 B1 0.93 0.99 0.83 L. feuerborni AR11 - Madagascar AI549 0.98 0.93 L. pollex RA1019 1.21 Laeliaena sahlbergi HI19 Ma 30 25 20 15 10 5 0

Figure 13. Reconstruction of the length of the male body size (in mm) in BEAST, using a Brownian model of evolution.

0.49 0.48 L. hilaris AI919 lg 1.21 0.48 0.52 L. aguilerai AI910 0.45 L. monfortei AF199 0.48 0.46 L. gerhardti AI891 0.47 0.49 0.49 L. maurus AI1317

0.51 0.45 L. ordunyai AI854 0.48 L. millani AI920 L. nitidus cplx 0.57 L. irmelae AI949 0.53 0.64 L. montanus AI936 0.44 0.55 0.51 L. nitidus AI377 0.41 0.43 L. ibericus RA88 0.33 L. bacchus PA274 0.39 L4.5 0.32 L. kocheri RA91 0.42 0.34 L. corybus AI1280 L. nitidus sbgr 0.36 0.38 0.31 L. corfidius AN33 0.38 L. hieronymi AR22 0.40 L4.1 0.32 L. murentius AN53 0.38 0.42 L. minoricensis AR23 1.00 L. nitiduloides AR16 0.79 0.68 L. simplex AC39 0.21 0.74 0.59 L. schoenmanni RA1000 L4-L. nitidus gr 0.68 0.52 0.72 0.73 L. crassipes RA68 0.65 L. spinosus AN54 0.67 L4.3 0.66 L. levantinus RA731 0.72 L. nitiduloides sbgr 1.12 1.21 L. fretalis AI1307 1.09 L. alibeii RA90 0.97 0.65 0.84 0.84 L. thery AI950 0.71 L. hispanicus AI1309 L4.4 0.70 L. truncatellus AI451 0.72 0.54 L. truncatellus sbgr 0.81 L. mesatlanticus AI913 0.51 L. zaerensis AC14 0.45 0.39 L. ignarus AI993 Limnebius s.str. L4.2 0.45 0.47 0.45 0.46 L. pilicauda AR60 L. punctatus sbgr 0.37 L. mucronatus AI704 0.67 0.67 L. papposus RA84 0.70 L. crinifer AR55 0.66 0.61 0.65 L. rubropiceus AN32 0.62 L. parvulus AI706 L2.2 0.54 0.57 L. furcatus AI987 L. parvulus sbgr 0.58 0.59 L. doderoi AI1174 L2-L. parvulus gr 0.55 0.50 0.53 0.55 L. stagnalis AI824 0.51 L. reuvenortali AN45 L2.1 0.36 L. lusitanus AI899 0.46 L. aluta sbgr 0.48 L. aluta AI953 0.48 0.33 0.33 L. cordobanus AI1024 L3-L. gracilipes gr 0.33 L. cordobanus PA275 0.41 0.40 L. gracilipes AI1094 0.41 0.42 L. fallaciosus RA1121 Limnebius 0.45 0.41 0.47 L. arenicolus AI466 0.44 L1-L. piceus gr 0.36 L. sinuatus AR9 0.45 0.52 L. piceus RA1120 0.22 0.23 L. nakanei AC30

0.24 0.21 L. acupunctus AI582 B4 - Philippines AR12 0.26 - India AF190 0.26 - Bhutan AI1271 0.28 0.25 0.23 L. dioscoridus RA728 0.26 L. pararabicus RA761 B3 0.29 0.28 0.36 L. wewalkai RA108 0.23 L. myrmidon AI703 0.40 0.34 0.43 L. oblongus AI942 0.42 0.39 0.44 L. perparvulus AI428 0.37 L. extraneus RA89 B2 0.37 Bilimneus 0.37 0.32 L. evanescens AI915 0.37 0.41 L. endroedyi AN78 0.41 L. atomus AI1217 0.39 - China RA627 0.37 0.39 India AF193 B1 0.34 0.37 0.25 L. feuerborni AR11 - Madagascar AI549 0.38 0.44 L. pollex RA1019 0.34 Laeliaena sahlbergi HI19 Ma 30 25 20 15 10 5 0

Figure 14. Reconstruction of the length of the aedeagus size (in mm) in BEAST, using a Brownian model of evolution.

A and always to a lesser extent. There was a consistent trend of increase of male size in the L. parvulus subgroup, and in two of the subgroups of the L. nitidus group: the L. truncatellus and L. nitiduloides subgroups. The resolution at the base of the L. nitidus group had no support (Fig. 3). However, if these two subgroups form a single lineage (as reconstructed in Fig. 11) this would imply a single origin of the increase in size within the L. nitidus group. Simi- larly, depending on the phylogenetic placement of the clade formed by the species L. claviger and L. setifer (without molecular data), one or two independent size increases could have occurred within the L. parvulus group (Fig. S3). B The strongest decrease in male size was found in an isolated species, L. cordobanus, but there were also decreases in size in the L. aluta and L. nitidus subgroups, and in the L. mundus subgroup, provided that it does not form a clade with the L. nitidus subgroup (as in the topology in Fig. 3). When an increase in male size occurred in a lineage this was very consistent, with all included known species becoming significantly larger than any of the species in other groups. This trend has resulted in a clearly bimodal size distribution of the males of Limnebius, with large species measuring more than 1.8 mm and small ones less than 1.5 (Table S1). Only two species within the L. gracilipes group, the lineage with the Figure 15. Relationships between: (A) length of male highest size variation, had intermediate male sizes body size (lm) and aedeagus (lg); and (B) lg and total (L. paganettii and L. canariensis; Table S1). number of appendages (TT) as measured with independent contrasts in DPAD, using the phylogenetic tree with Appendages all species placed according to their most likely phyloge- The length of the genitalia was not significantly cor- netic relationships (Fig. S3). Dotted line, regression related with the total number of appendages (includ- through the origin. ing subdivisions of the main additional appendages a1 to a3). Results were the same both with indepen- the species pair L. alibeii Hernando et al. and L. fre- dent contrasts using only the species with molecular talis, followed by a relative increase in the body size data (regression through the origin using the topol- of the latter. ogy of Fig. 3, r = 0.04, n = 66, P > 0.7) or the tree Male body size showed little change through the with all species (regression through the origin using basal diversification of the subgenus. Body size is the topology of Fig. S3, r = 0.06, n = 118, P > 0.5; very homogenous in the species of Bilimneus, rang- Fig. 15B). Although the smallest genitalia always ing from 0.76 mm in L. arabicus to c. 1 mm in some had a low number of appendages, there were some Himalayan species (Table S1). Although the incom- relatively small species with rather complex genital- plete sampling does not allow a definitive conclusion, ia, as for instance in the L. nitidus or L. punctatus the phylogeny does not show any clear trend of size subgroups (Fig. 16). variation among lineages within the subgenus The ancestral aedeagus of Limnebius was recon- (Fig. 13). By contrast, there seems to be a small but structed as a relatively simple structure, with a med- consistent reduction in aedeagus size in the lineage ian lobe and an externally visible left paramere. The including clades B3 and B4 (mainly species from the presence or absence of an additional appendage Middle East, Oriental and Australian regions), with remained ambiguous (Fig. 16). There are no data on the single exception of L. wewalkai Jach€ & Delgado the phylogenetic placement of L. boukali or other (Fig. 14; Table S1). species of Bilimneus with a possible free apex of the Within Limnebius s.s. body size of males increased right paramere. However, if they are derived within substantially in several non-related lineages. A the subgenus, this would imply the complete fusion decrease in size also occurred but in fewer groups, of the right paramere in the ancestor of Bilimneus,

© 2016 The Linnean Society of London, Zoological Journal of the Linnean Society, 2016, 178, 97–127 122 A. RUDOY ET AL. with a secondary re-formation of the free paramere sions of the main appendage, in the lineage leading in some species. Alternatively, these species could be to L. doderoi and L. furcatus and in the one leading sister to the rest of Bilimneus. This would imply that to L. papposus and related species (Fig. 16). the presence of a free right parameral apex may be The L. gracilipes group is the most diverse in body ancestral for the subgenus – and even for the entire size and evolution of genital complexity, with the genus Limnebius, if the appendage on the right side smallest of the species of Limnebius s.s. (L. cor- of the median lobe of species of the L. parvulus sub- dobanus) and some unusual combinations of the pres- group is indeed homologous to the apex of the right ence of ventral and dorsal appendages: they include paramere (see above). the only species with a2 but no a3, and the reverse, The reconstructed groundplan of Limnebius s.s. a3 present but not a2 (see above and Table S1). In includes at least a free left paramere and an addi- the species with ventral appendages but lacking tional appendage. The uncertainty in the relation- those on the dorsal side (L. canariensis and L. gra- ships of the four main lineages within the subgenus cilipes, both endemic to the Canary Islands and with did not allow a precise reconstruction. The position very similar aedeagus morphology, Fig. S2:70,73), the of the only studied species of the L. mundus group main ventral appendage a1 is well developed and (L. murentius), characterized by the complete fusion placed on the central part of the median lobe, origi- of the left paramere, is particularly uncertain due to nating in the base of the left paramere. Both species the small number of sequences obtained (Table S1). share a particular hook-shape of the apex of the left With the available data it was placed within the paramere, and the same general structure of the L. nitidus group with strong support, but in some additional appendages. Within the same group, L. analyses it was placed as sister to the rest of Lim- fallaciosus has a well-developed dorsal and main ven- nebius s.s., which could imply an ancestral condition tral appendages (a3 and a1), but no secondary ventral with a fused left paramere and its secondary re-for- appendage (a2) (Table S1, Fig. S2:72). mation as a free appendage, an unlikely interpreta- Our reconstruction suggests that L. cordobanus tion from a morphological point of view. experienced a size reduction of the entire body and Regardless, our data suggest a simplification of the the genitalia, and also a reduction of the number aedeagus in the ancestor of the L. piceus group, with and complexity of the appendages of the aedeagus the fusion of the base of the left paramere (com- (Fig. 5). Only the main ventral appendage a1 is pre- pletely fused in the derived L. sinuatus and L. uta- sent and other folds or structures of the median lobe hensis Perkins) and the absence of any additional are missing. As a result, the aedeagus is very similar appendages (Fig. 16). The species of the L. piceus to that of other species within the L. nitidus subgroup have no free appendages other than the apex group (e.g. L. corfidius d’Orchymont or L. corybus of the left paramere, and the median lobe has only d’Orchymont, Fig. S2:100,101), although with a more some folds that somewhat resemble those of the spe- complex apex of the median lobe. To ensure that the cies of the L. parvulus group. However, we found no unexpected phylogenetic position of L. cordobanus support for a relationship between these two lineages was not caused by sequencing errors, we included a in our phylogeny. second specimen, which, although with considerable An increase in complexity occurs in the lineage intraspecific variation, confirmed its sister relation- leading to the L. parvulus group, with folds in the ship with L. gracilipes and related species. There are median lobe, and with the formation of an additional also some characters of the aedeagus that indicate appendage of uncertain homology on the right side of that the resemblance with some species of the the aedeagus, which could correspond to the rem- L. nitidus group may be homoplasic, with a similar nants of the right paramere (see above). The basal reduction of the left paramere and a basal fusion with capsule is well developed and sclerotized in all spe- the median lobe. Thus, contrary to the studied species cies, and separated from the continuation of the med- of the L. nitidus group, the main ventral appendage is ian lobe by a constriction. These characters of the straight and strictly ventral (with a more lateral posi- basal capsule are probably autapomorphies of the tion in the species of the L. nitidus group); has a sim- L. parvulus group. The species of the L. parvulus ple internal structure, with fewer channels and no subgroup developed a highly complex adeagus, with perforated structure (see above); and has a very short a strongly curved median lobe and appendages with free base placed in a more apical position (Fig. 5:8,9). fan-like setae (Fig. 4). Within the subgroup there are The area of the median lobe facing the basal part of a1 two independent instances of development of subdivi- has a thinner cuticle (Fig. 5:8,9), similar to what hap-

Figure 16. Reconstruction of the total number of appendages in BEAST, using a Brownian model of evolution. Num- bers at nodes are the integer interval of the quantitative values reconstructed, with values with a probability below c. 0.3 in parentheses.

4 4 L. hilaris AI919 TT 4 7 4 L. aguilerai AI910 4 L. monfortei AF199 4 4 4 L. gerhardti AI891 4 4 L. maurus AI1317 3-4 3-4 4 L. ordunyai AI854 3 L. millani AI920 L. nitidus cplx 3-4 3 L. irmelae AI949 3 L. montanus AI936 3 3-4 4 L. nitidus AI377 3 3 L. ibericus RA88 3 L. bacchus PA274 3 L4.5 3 L. kocheri RA91 3 3 L. corybus AI1280 L. nitidus sbgr 3 3 L. corfidius AN33 (2)-3 2 L. hieronymi AR22 3 L4.1 2 L. murentius AN53 2-3 2 L. minoricensis AR23 7 L. nitiduloides AR16 (5)-6 5 L. simplex AC39 1 5-6 5 L. schoenmanni RA1000 L4-L. nitidus gr 5 4-(5) 5-(6) 5 L. crassipes RA68 5 L. spinosus AN54 5-6 L4.3 6 L. levantinus RA731 5-(6) L. nitiduloides sbgr 5 5 L. fretalis AI1307 5 L. alibeii RA90 5-6 5-(6) 5-(6) 6 L. thery AI950 5 L. hispanicus AI1309 L4.4 7 L. truncatellus AI451 6-7 4-5 L. truncatellus sbgr 7 L. mesatlanticus AI913 6 L. zaerensis AC14 5-(6) 5 L. ignarus AI993 Limnebius s.str. L4.2 5-(6) 3-4 (4)-5 5 L. pilicauda AR60 L. punctatus sbgr 5 L. mucronatus AI704 5 L. papposus RA84 4-5 4 L. crinifer AR55 4-(5) (3)-4 4 L. rubropiceus AN32 L2.2 3 L. parvulus AI706 (3)-4 L. parvulus sbgr 5-6 6 L. furcatus AI987 5 L. doderoi AI1174 L2-L. parvulus gr 4 3-4 3 3 L. stagnalis AI824 3 L. reuvenortali AN45 L2.1 3 L. lusitanus AI899 3-(4) L. aluta sbgr 3 L. aluta AI953 3-4 3 3 L. cordobanus AI1024 L3-L. gracilipes gr 3 L. cordobanus PA275 3-4 4 L. gracilipes AI1094 4 5 L. fallaciosus RA1121 Limnebius 3-4 2-3 2 L. arenicolus AI466 (1)-2 L1-L. piceus gr 1 L. sinuatus AR9 (1)-2 2 L. piceus RA1120 1 1 L. nakanei AC30

1 1 L. acupunctus AI582 B4 - Philippines AR12 1 - India AF190 1 - Bhutan AI1271 1 1 1 L. dioscoridus RA728 1 L. pararabicus RA761 B3 1 1 1 L. wewalkai RA108 1 L. myrmidon AI703 2-3 1 1 L. oblongus AI942 1 1 1 L. perparvulus AI428 1 L. extraneus RA89 B2 1 Bilimneus 1 1 L. evanescens AI915 1-2 1 L. endroedyi AN78 1 L. atomus AI1217 1 - China RA627 1 - India AF193 B1 1 1 1 L. feuerborni AR11 - Madagascar AI549 1 1 L. pollex RA1019 3 Laeliaena sahlbergi HI19

30 25 20 15 10 5 0

© 2016 The Linnean Society of London, Zoological Journal of the Linnean Society, 2016, 178, 97–127 124 A. RUDOY ET AL. pens in the L. parvulus group. The opening of the flag- middle region of the median lobe but a complex ellum, very apical and without any setae, is also differ- apical part (Fresneda & Ribera, 1998). The opening ent from that typical of the species of the L. nitidus of the flagellum, which is small and lacks setae or subgroup (Fig. 5:4). The well-developed basal capsule, other specific structures, is located basal to the com- with a strongly coiled flagellum, is also more similar plex structures of the apical region of the median to those of the species of the L. parvulus than those of lobe (Fig. 10:6). the L. nitidus group. The species of the L. nitidus complex are of very The lack of resolution within the L. nitidus group recent origin, probably in the middle to late Pleis- prevented the detailed reconstruction of the struc- tocene, and in our phylogeny their relationships were tural complexity of their aedeagus. The L. punctatus, poorly resolved. A tentative phylogenetic arrange- L. nitiduloides and L. truncatellus subgroups share ment was proposed in Fresneda & Ribera (1998), the presence of several additional appendages and based mostly on the structure of the apex of the med- secondary sexual characters in the male sternite VIII ian lobe. This was only partly supported by our phy- (Jach,€ 1993). However, our reconstructed topology logeny. The position of L. montanus as sister to the (Fig. 3) suggests that they are paraphyletic with rest of the species of the group (except for L. niti- respect to the L. nitidus subgroup. This would imply farus, for which no molecular data were available) is a single origin of the increase in complexity, but a compatible with our results. However, we found it secondary loss in the L. nitidus subgroup, also corre- sister to L. nitidus (Fig. 3), which in Fresneda & lated with a size reduction (see above). However, the Ribera (1998) was assumed to be more closely related three subgroups with more complex aedeagus were to the other Iberian species of the complex. placed as monophyletic in the alternative topology in Figure 16, although also with negligible node sup- CONCLUDING REMARKS port. If this were the case, a single origin of a complex aedeagus in the three groups would still be Characterization of the male insect genitalia is of possible. In this case, the less complex condition in great relevance for the study of sexual selection and the L. nitidus subgroup (with the inclusion of the speciation, but there are few works including old, only studied species of the L. mundus group, diverse lineages with a diversity of genital structures L. murentius) would be plesiomorphic. comparable to that found in Limnebius. We have The L. nitiduloides subgroup includes the species shown that the genus includes two main clades, con- with the largest and most complex male genitalia, all sidered here to be subgenera (Bilimneus and Lim- of them with three additional appendages, which in nebius s.s.), with contrasting evolution of the body some species may be longer than the median lobe size and aedeagus. In Bilimneus there was probably (e.g. a2 in L. fretalis, Fig. 7; or a3 in L. nitiduloides, a simplification of the genital structure, with a very Fig. 8). Two of the well-supported clades within the limited variation among the extant species in body subgroup (those including respectively L. spinosus and genital size and in the shape of the aedeagus. and L. nitiduloides, Figs 3, 16) have ramifications of By contrast, in Limnebius s.s. some lineages have the ventral appendages or the median lobe, while in independently developed extremely complex genital- the third clade (including L. fretalis) species have a ia, sometimes associated with a considerable increase simpler apical part of the median lobe, without rami- in size. But other lineages within Limnebius s.s. fications of the ventral appendages. The relation- have maintained the ancestral, simpler genital struc- ships between these three clades were, however, not ture, or have developed it secondarily. The contrast- well supported. Therefore, an assessment of the ing evolution of the aedeagus between the two homology of these modifications is not possible. subgenera of Limnebius points strongly to the exis- The L. nitidus subgroup includes a series of iso- tence of different selection forces acting on them, lated species with a simple aedeagus and poorly study of which could provide valuable insights into resolved relationships among them, plus the L. niti- the origin of their extraordinary genital variation. dus complex, with a peculiar a1, long and with a more lateral position (Figs 2b, 10) (the ‘pseu- ACKNOWLEDGEMENTS doparamere’ of Jach,€ 1993). The secondary simplification of the aedeagus reaches its maximum in L. We especially thank the collectors listed in Table S2 hieronymi, lacking any apical setae on the median for sending us their material for study. We are also lobe or any other structures on the apical part, pre- grateful to M.A. Jach€ (NMW, Wien) and P. Perkins sent in most species of Limnebius s.s. (Vorst, 2006; (MCZ, Harvard) for allowing us to study the collec- Fig. S2:104). tions of their institutions and for supporting this The L. nitidus complex includes 11 species with a work in many ways. AR thanks M.A. Jach€ for his very similar structure of the aedeagus, with a flat help during his stay in the NMW and for technical

© 2016 The Linnean Society of London, Zoological Journal of the Linnean Society, 2016, 178, 97–127 EVOLUTION OF THE AEDEAGUS IN LIMNEBIUS 125 support in doing the histological sections in the and other body parts in twenty species of insects and spi- Phyletisches Museum (Jena). We also thank Ana ders. Evolution 52: 415–431. Izquierdo, Rocıo Alonso and Anabela Cardoso for Ferro G. 1989. Note su qualche Limnebiinae (Coleoptera laboratory work, and two anonymous referees for Hydraenidae). XXII contributo alla conoscenza degli useful comments to the manuscript. This work was Hydraenidae. Bulletin et Annales de la Societ e Royale Belge partly funded by a JAE PhD studentship (CSIC) to D’Entomologie 125: 277–281. AR, and projects from the Spanish Government Fresneda J, Ribera I. 1998. Revision of the Limnebius niti- (Ministerio de Economıa y Competitividad) CGL dus (Marsham) subgroup (Coleoptera: Hydraenidae), with 2010-15755 and CGL2013-48950-C2-1-P and a Salva- description of two new species and comments on their phylogeny and biogeography. Entomologica Scandinavica 29: dor de Madariaga stage (PRX14/00583) to IR. 395–409. Fresneda J, Salgado JM, Ribera I. 2007. Phylogeny of REFERENCES western Mediterranean Leptodirini, with an emphasis on genital characters (Coleoptera: Leiodidae: Cholevinae). Sys- Abellan P, Ribera I. 2011. Geographic location and phy- tematic Entomology 32: 332–358. logeny are the main determinants of the size of the geo- Hall WE. 2005. 11.2 Ptiliidae. In: Beutel RG, Leschen graphical range in aquatic beetles. BMC Evolutionary RAB, eds. Handbook of zoology, part 38: volume 1: mor- Biology 11: 344. phology and systematics (Archostemata, Adephaga, Myx- Abellan P, Sanchez-Fern andez D, Picazo F, Millan A, ophaga, Polyphaga partim). Berlin: Walter de Gruyter, Lobo JM, Ribera I. 2013. Preserving the evolutionary his- 251–261. tory of freshwater biota in Iberian National Parks. Biologi- Hansen M. 1991. A review of the genera of the beetle family cal Conservation 162: 116–126. Hydraenidae (Coleoptera). Steenstrupia 17: 1–52. Angus RB, Dıaz-Pazos JA. 1991. A chromosomal investiga- Hansen M 1998. World catalogue of insects. 1. Hydraenidae tion of some European Hydraenidae (Coleoptera: Hydraeni- (Coleoptera). Stenstrup, DK: Apollo Books. dae). Koleopterologische Rundschau 61: 95–103. Hernando C, Aguilera P, Ribera I. 2008. Limnebius Balfour-Browne J. 1956. On the Indian species of Lim- zaerensis, a new species from the Pays Zaer-Za€ €ıane, central nebius Leach (Coleoptera: Hydrophilidae). Proceedings of Morocco (Coleoptera: Hydraenidae). Koleopterologische the Royal Entomological Society of London Series B, Taxon- Rundschau 78: 195–198. omy 25: 103–107. Holwell GI, Winnick C, Tregenza T, Herberstein ME. Beutel RG, Lawrence JF. 2005. 4. Coleoptera, morphology. 2010. Genital shape correlates with sperm transfer success In: Beutel RG, Leschen RAB, eds. Handbook of zoology, in the praying mantis Ciulﬁna klassi (Insecta: Mantodea). part 38: volume 1: morphology and systematics (Archostem- Behavioral Ecology and Sociobiology 64: 617–625. ata, Adephaga, Myxophaga, Polyphaga partim). Berlin: Hosken DJ, Stockley P. 2004. Sexual selection and genital Walter de Gruyter, 23–27. evolution. Trends in Ecology and Evolution 19: 87–93. Beutel RG, Leschen RAB (eds). 2005. Handbook of zool- Hotzy C, Arnqvist G. 2011. Sperm competition favors ogy, part 38: volume 1: morphology and systematics harmful males in seed beetles. Current Biology 19: 404– (Archostemata, Adephaga, Myxophaga, Polyphaga partim). 407. Berlin: Walter de Gruyter. House CM, Simmons LW. 2005. The evolution of male geni- Beutel RG, Anton E, Jach€ MA. 2003. On the evolution of talia: patterns of genetic variation and covariation in the adult head structures and the phylogeny of Hydraenidae genital sclerites of the dung beetle Onthophagus taurus. (Coleoptera, Staphyliniformia). Journal of Zoological Sys- Journal of Evolutionary Biology 18: 1281–1292. tematics and Evolutionary Research 41: 256–275. Hunt T, Bergsten J, Levkanicova Z, Papadopoulou A, Cieslak A, Fresneda J, Ribera I. 2014. Life-history spe- St John O, Wild R, Hammond PM, Ahrens D, Balke M, cialization was not an evolutionary dead-end in Pyrenean Caterino MS, Gomez-Zurita J, Ribera I, Barraclough cave beetles. Proceedings of the Royal Society of London, TG, Bocakova M, Bocak L, Vogler AP. 2007. A compre- Series B: Biological Sciences 281: 20132978. hensive phylogeny of beetles reveals the evolutionary ori- Delgado JA, Soler AG. 1997. Morphology and chaetotaxy of gins of a superradiation. Science 318: 1913–1916. larval Hydraenidae (Coleoptera). I: genus Limnebius Leach, Iablokoff-Khnzorian SM. 1962. [New species of Coleoptera 1815 based on a description of Limnebius cordobanus from Transcausus (Insecta-Coleoptera)]. Zoologicheskiı˘ d’Orchymont. Aquatic Insects 19: 37–49. Sbornik. Biologicheskiı˘ (Zoologicheskiı˘) Institut, Akademiya Drummond AJ, Rambaut A. 2007. BEAST: Bayesian evo- Nauk Armyanskoı˘ SSSR 12: 99–124. [in Russian]. lutionary analysis by sampling trees. BMC Evolutionary Jach€ MA. 1993. Taxonomic revision of the Palearctic species Biology 7: 214. of the genus Limnebius Leach, 1815 (Coleoptera: Hydraeni- Eberhard WG. 1985. Sexual selection and animal genitalia. dae). Koleopterologische Rundschau 63: 99–187. Cambridge, MA: Harvard University Press. Jach€ MA. 1995. Taxonomic synopsis of the genus Laeliaena Eberhard WG, Huber BA, Rodrıguez RL, Briceno~ RD, Sahlberg, 1900 (Coleoptera: Hydraenidae). Elytron 8: 35–41. Salas I, Rodrıguez V. 1998. One size ﬁts all? Relation- Jach€ MA. 2015. Hydraenidae. In: L I, L D, eds. Catalogue of ships between the size and degree of variation in genitalia palaearctic coleoptera, vol. 2/1. Revised and updated edi-

tion. Hydrophiloidea – Staphylinoidea, Vol. 1. Leiden: Brill, coleopteros acuaticos de Espana~ peninsular. Madrid: Minis- 130–162. terio de Agricultura, Alimentacion y Medio Ambiente. Jach€ MA, Delgado JA. 2012. Limnebius dioscoridus sp. Mutanen M, Kaitala A. 2006. Genital variation in a dimor- nov. from Socotra Island (Coleoptera: Hydraenidae). Acta phic moth Selenia tetralunaria (Lepidoptera, Geometridae). Entomologica Musei Nationalis Prague 52: 131–134. Biological Journal of the Linnean Society 87: 297–307. Jach€ MA, Delgado JA. 2013. Taxonomic revision of the spe- Orchymont Ad. 1932. Zur Kenntnis der Kolbenwasserkafer€ cies of Limnebius Leach from Mauritius and Reunion (Mas- (Palpicornia) von Sumatra, Java und Bali. Archiv fur€ carene Islands, Indian Ocean) (Coleoptera: Hydraenidae). Hydrobiologie, Supplement 9: 623–714. Koleopterologische Rundschau 83: 53–71. Orchymont Ad. 1938. Notes sur quelques Limnebius Jach€ MA, Skale A. 2011. Annotated checklist of the Hydrae- (Coleoptera Palpicornia). Bulletin et Annales de la Societ e nidae of Armenia (Coleoptera: Hydraenidae). Koleopterolo- Royale Belge D’Entomologie 76: 425–438. gische Rundschau 81: 93–111. Orchymont Ad. 1940. Les Palpicornia des Iles Atlantiques. Jałoszynski P, Matsumura Y, Beutel RG. 2015. Evolution Memoires du Musee Royale D’Histoire Naturelle de Belgique of a giant intromittent organ in Scydmaeninae (Coleoptera: 2: 1–87. Staphylinidae): functional morphology of the male postab- Perkins PD. 1980. Aquatic beetles of the family Hydraeni- domen in Mastigini. Arthropod Structure & Development dae in the Western Hemisphere: classification, biogeogra- 44: 77–98. phy and inferred phylogeny (Insecta: Coleoptera). Katoh K, Toh H. 2008. Recent developments in the MAFFT Quaestiones Entomologicae 16: 3–554. multiple sequence alignment program. Briefings in Bioin- Perkins PD. 1997. Life on the effective bubble: exocrine formatics 9: 286–298. secretion delivery systems (ESDS) and the evolution and Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung classification of beetles in the family Hydraenidae (Insecta: M, Sturrock S, Buxton S, Cooper A, Markowitz S, Coleoptera). Annals of the Carnegie Museum 66: 89–207. Duran C, Thierer T, Ashton B, Mentjies P, Drummond Perkins PD. 2004. Limnebius acupunctus, a new species of A. 2012. Geneious Basic: an integrated and extendable water beetle from Australia and Papua New Guinea desktop software platform for the organization and analysis (Coleoptera: Hydraenidae). Zootaxa 749: 1–12. of sequence data. Bioinformatics 28: 1647–1649. Perkins PD. 2015. Taxonomy of the water beetle genus Lim- Lu W, Jackman JA, Johnson PW. 1997. Male genitalia nebius Leach in southern Africa (Coleoptera: Hydraenidae). and phylogenetic relationships in North American Mordelli- Zootaxa 3948: 41–59. dae (Coleoptera). Annals of the Entomological Society of Polilov AA, Beutel RG. 2009. Miniaturisation effects in lar- America 90: 742–767. vae and adults of Mikado sp. (Coleoptera: Ptiliidae), one of Maddison WP, Maddison DR. 2015. Mesquite: a modular the smallest free-living insects. Arthropod Structure & system for evolutionary analysis. Version 3.04 http:// Development 38: 247–270. mesquiteproject.org. Ribera I, Fresneda J, Bucur R, Izquierdo A, Vogler AP, Masly JP. 2012. 170 years of ‘Lock-and-Key’: genital mor- Salgado JM, Cieslak A. 2010. Ancient origin of a Western phology and reproductive isolation. International Journal of Mediterranean radiation of subterranean beetles. BMC Evolutionary Biology 2012: 247352. Evolutionary Biology 10: 29. Matsumura S, Yao I, Beutel RG, Yoshizawa K. 2014. Ribera I, Castro A, Dıaz JA, Garrido J, Izquierdo A, Molecular phylogeny of the leaf beetle subfamily Crioceri- Jach€ MA, Valladares LF. 2011. The geography of specia- nae (Insecta: Coleoptera: Chrysomelidae) and the correlated tion in narrow-range endemics of the ‘Haenydra’ lineage evolution of reproductive organs. Arthropod Systematics & (Coleoptera, Hydraenidae, Hydraena). Journal of Biogeog- Phylogeny 72: 95–110. raphy 38: 502–516. Mayr E. 1963. Animal species and evolution. Cambridge, Rodrıguez V, Windsor DM, Eberhard WG. 2004. Tortoise MA: Harvard University Press. beetle genitala and demonstrations of a sexually selected McKenna DD, Farrel BD, Caterino MS, Farnum CW, advantage for flagellum length in Chelymorpha alternans David CH, Maddison DR, Seago AI, Short AEZ, New- (Chrysomelidae, Cassidini, Stolaini). In: Jolivet P, Santiago- ton AF, Thayer MK. 2015. Phylogeny and evolution of Blay JA, Schmitt M, eds. New developments in the biology of Staphyliniformia and Scarabaeiformia: forest litter as a Chrysomelidae. The Hague: SBP Academic Publisher, 739–748. stepping stone for diversification of nonphytophagous bee- Sainz-Cantero CE, Bennas N. 2006. Limnebius kamali tles. Systematic Entomology 40: 35–60. sp.n. from Northern Morocco (Coleoptera, Hydraenidae). McPeek MA, Symes LB, Zong DM, McPeek CL. 2010. Revue Suisse de Zoologie 113: 559–563. Species recognition and patterns of population variation in Sasakawa K, Kim JL, Kim JK, Kubota K. 2008. Taxo- the reproductive structures of a Damselfly genus. Evolution nomic studies of Pterostichus ishikawai Nemoto (Coleop- 62: 419–428. tera: Carabidae). Entomological Science 11: 173–178. Midford PE, Garland T Jr, Maddison WP. 2011. PDAP Sharp DS, Muir F. 1912. The comparative anatomy of the Package of Mesquite. Version 1.16. http://mesquitepro- male genital tube in Coleoptera. Transactions of the Ento- ject.org/pdap_mesquite/. mological Society of London 60: 477–642. Millan A, Sanchez-Fernandez D, Abellan P, Picazo F, Sota T, Kubota K. 1998. Genital lock-and-key as a selective Carbonell JA, Lobo JM, Ribera I. 2014. Atlas de los agent against hybridization. Evolution 52: 1507–1513.

Sota T, Tanabe T. 2010. Multiple speciation events in an Trizzino M, Audisio PA, Antonini G, Manchini E, Rib- arthropod with divergent evolution in sexual morphology. era I. 2011. Molecular phylogeny and diversiﬁcation of the Proceedings of the Royal Society of London, Series B: Bio- ‘Haenydra’ lineage (Hydraenidae, genus Hydraena), a logical Sciences 277: 689–696. north-Mediterranean endemic-rich group of rheophilic Stamatakis A, Hoover P, Rougemont J. 2008. A rapid Coleoptera. Molecular Phylogenetics and Evolution 61: 772– bootstrap algorithm for the RAxML web servers. Systematic 783. Biology 57: 758–771. Trizzino M, Jach€ MA, Audisio P, Alonso R, Ribera I. Steedman HF. 1958. Dimethyl Hydantoin Formaldehyde: a 2013. A molecular phylogeny of the cosmopolitan hyperdi- new water-soluble resin for use as a mounting medium. verse genus Hydraena Kugelann (Coleoptera, Hydraenidae). Quarterly Journal of Microscopical Science 99: 451–452. Systematic Entomology 38: 192–208. Sturtevant AH. 1920. Genetic studies on Drosophila simu- Usami T, Yokoyama J, Kubota K, Kawata M. 2006. Genital lans. I. Introduction. Hybrids with Drosophila melanoga- lock-and-key system and premating isolation by mate prefer- ster. Genetics 5: 488–500. ence in carabid beetles (Carabus subgenus Ohomopterus). Tarasov SI, Solodovnikov AY. 2011. Phylogenetic analyses Biological Journal of the Linnean Society 87: 145–154. reveal reliable morphological markers to classify mega- Vorst O. 2006. Limnebius hieronymi, a new species from diversity in Onthophagini dung beetles (Coleoptera: Central Italy (Coleoptera: Hydraenidae). Koleopterologische Scarabaeidae: Scarabaeinae). Cladistics 27: 490–528. Rundschau 76: 75–78.

SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s web- site: Table S1. Checklist of the species of Laeliaena and Limnebius, with main distribution, subgenus and species group according to our phylogenetic results. Table S2. Primers used for ampliﬁcation and sequencing. Table S3. List of specimens used for the molecular data, with subgenus, species group and subgroup, voucher, locality, collector and GenBank accession numbers. Table S4. Specimens used for the histological sections, with locality, collector and number of sections. Figure S1. Phylogenetic trees obtained with the exclusion of L. murentius. Figure S2. Aedeagus of the species of Limnebius, grouped according to their phylogenetic position in Figure 3. Figure S3. Estimated relationships of the species of Limnebius for which no molecular data were available (species without voucher numbers), based mostly on the general shape and structure of the male genitalia (see main text for details).