Phylogenetic Analysis Reveals That Rhabdopleura Is an Extant Graptolite

Home , Cephalodiscus, Graptolithina, Pterobranchia, Rhabdopleura, Rhabdopleura compacta, Tetragraptus

Phylogenetic analysis reveals that Rhabdopleura is an extant graptolite

CHARLES E. MITCHELL, MICHAEL J. MELCHIN, CHRIS B. CAMERON AND JO¨ RG MALETZ

Mitchell, C.E., Melchin, M.J., Cameron, C.B. & Maletz, J. 2013: Phylogenetic analysis reveals that Rhabdopleura is an extant graptolite. Lethaia,Vol.46,pp.34–56.

A phylogenetic analysis of morphological data from modern pterobranch hemichordates (Cephalodiscus, Rhabdopleura) and representatives of each of the major graptolite orders reveals that Rhabdopleura nests among the benthic, encrusting graptolite taxa as it shares all of the synapomorphies that unite the graptolites. Therefore, rhabdopleurids can be regarded as extant members of the Subclass Graptolithina (Class Pterobranchia). Com- bined with the results of previous molecular phylogenetic studies of extant deuterostomes, these results also suggest that the Graptolithina is a sister taxon to the Subclass Cephalodiscida. The Graptolithina, as an important component of Early–Middle Palaeo- zoic biotas, provide data critical to our understanding of early deuterostome phylogeny. This result allows one to infer the zooid morphology, mechanics of colony growth and palaeobiology of fossil graptolites in direct relation to the living members of the clade. The Subdivision Graptoloida (nom. transl.), which are all planktic graptolites, is well supported in this analysis. In addition, we recognize the clade Eugraptolithina (nov.). This clade comprises the Graptoloida and all of the other common and well-known graptolites of the distinctive Palaeozoic fauna. Most of the graptolites traditionally regarded as tuboids and dendroids appear to be paraphyletic groups within the Eugraptolithina; however, Epigraptus is probably not a member of this clade. The Eugraptolithina appear to be derived from an encrusting, Rhabdopleura-like species, but the available information is insufﬁcient to resolve the phylogeny of basal graptolites. The phylogenetic position of Mastigograptus and the status of the Dithecoidea and Mastigograptida also remain unresolved. h Biodiversity, Cambrian, Hemichordata, Deuterostomia, Ordovician.

Charles E. Mitchell [[email protected]], Department of Geology, University at Buffalo- SUNY, Buffalo, NY 14260, USA; Michael J. Melchin [[email protected]], Department of Earth Sciences, St. Francis Xavier University, Antigonish, NS B2G 2W5, Canada; Chris B. Cameron [[email protected]], De´partement de sciences biologiques, Universite´ de Montre´al, Montre´al QC H2V 2S9, Canada; Jo¨rg Maletz [[email protected]], Institut fu¨r Geologische Wissenschaften, Freie Universita¨t Berlin, Malteserstr. 74-100, D-12249 Ber- lin, Germany; manuscript received on 8 ⁄ 12 ⁄ 11; manuscript accepted on 7 ⁄ 5 ⁄ 12.

The small phylum Hemichordata plays a pivotal role Finally, we do not know the relationship among the in our efforts to understand the pattern of relation- tube-building groups: the graptolites and ptero- ships among the three major deuterostome taxa, the branchs. This latter question is the topic of our nature of the deuterostome ancestor and the evolu- study. tionary origin of the chordates (Fig. 1). Historically, Pterobranchs are colonial or pseudocolonial and hemichordates have been divided into four classes: reproduce via short-lived planula larvae and asexual the extant Enteropneusta (acorn worms), Planktosp- budding. They produce a collagenous dwelling struc- haeroidea and Pterobranchia and the extinct Grapto- ture that is commonly called a coenecium in the case lithina (Bulman 1970). The relationship among the of pterobranchs and a rhabdosome in the case of classes is highly uncertain, however, and this con- graptolites. For simplicity’s sake, we will use the term founds efforts to reconstruct deuterostome evolution. tubarium (Lankester 1884) for all of these communal It is not known whether the pterobranchs are a sister domiciles. The two major orders commonly assigned group to the enteropneusts (e.g. Winchell et al. 2002; to the Pterobranchia are the Cephalodiscida and the Cameron 2005) or derived within the enteropneust Rhabdopleurida. Both groups are relatively well clade (e.g. Cameron et al. 2000; Cannon et al. 2009). known from their living representatives, but have It is not known whether the monotypic Planktosp- yielded a fairly sparse fossil record (Rickards et al. haeroidea – rare, large, planktic larvae – represents a 1995). Graptolites, in contrast, have a relatively com- distinct clade of animals or are merely hypertrophied plete and rich fossil record of their skeletal details, tornaria larva of an enteropneust (Spengel 1932). whereas their soft-part anatomy is almost completely

DOI 10.1111/j.1502-3931.2012.00319.x 2012 The Authors, Lethaia 2012 The Lethaia Foundation LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 35

Fig. 1. Synthesis of recent molecular analyses of deuterostome phylogeny. Tree topology based on results presented in Winchell et al. (2002), Cameron (2005), Cannon et al. (2009), Delsuc et al. (2006) and Putnam et al. (2008). unknown except for a few poorly preserved remnants graptolites, graptolites were excluded from the Ptero- (Bjerreskov 1978, 1994; Rickards & Stait 1984; Loydell branchia. et al. 2004). In addition, the Middle to Late Cambrian On the other hand, Bulman’s (1970, p. V17) defini- fossil record of graptolites, which encompasses the tion of the Class Graptolithina is ‘The Graptolithina earliest part of their evolutionary history, is much less are colonial, marine organisms, which secreted a scler- complete than for the Ordovician or Silurian periods otized exoskeleton with characteristic growth bands when they became diverse and widespread (Rickards (fuselli) and growth lines. The thecae house individual & Durman 2006). This disparity in the available infor- zooids usually arranged in a single or double row mation for the various hemichordate groups has hin- along the branches (stipes) of the colony (rhabdo- dered a complete understanding of the phylogenetic some), rarely in irregular aggregates. In most orders, relationships between the living and extinct groups thethecaearepolymorphicandinthreetheyare and between different groups of graptolites. clearly related to an internal sclerotized stolon system. The goals of the present study are: (1) to test Rhabdosomes originate by a single bud from the ini- whether Pterobranchia and Graptolithina are mono- tial zooid, housed in a conical sicula, producing sim- phyletic groups and sister taxa as has commonly been ple, branched or rarely encrusting colonies.’ supposed; and, (2) in so doing, create a robust phylo- Cephalodiscid pterobranchs are clearly excluded genetic tree of the major tube-building hemichordate from the Graptolithina, based on this diagnosis, even taxa that can provide a foundation for further palaeo- taking into account recent evidence that their tubari- biological and macroevolutionary studies. um and its spine-like structures show a mode of construction similar to that inferred for graptolites (Dilly 1993). In contrast, the only respect in which the rhab- Previous phylogenetic interpretations dopleurid pterobranchs do not fit this diagnosis is in the description of the sicula as a conical structure. The The construction of the collagenous tubarium and bushy dendroid graptolites, however, possess a tubular asexually budded colonial zooids are synapomorphies prosicula (Fig. 2; Kozłowski 1949) and, moreover, of the Pterobranchia (Cameron 2005) and these fea- Epigraptus (and possibly the crustoids) possess a vesic- tures are shared with the graptolites, along with many ular prosicula very similar to that of Rhabdopleura details of the composition and fusellar mode of con- (Kozłowski 1971). struction of the tubarium (Andres 1977; Dilly 1993; Beklemishev (1951, 1970), in contrast, proposed Mierzejewski & Kulicki 2003). Despite these similari- that all of the pterobranchs and graptolites should be ties, Bulman (1970) regarded the Pterobranchia and united in a single Class Graptolithoidea, distinguished Graptolithina as distinct classes. His diagnosis of the by the presence of a sclerotized tubarium constructed Pterobranchia, which matches closely with most with fusellar increments that house a colony of zooids. widely used diagnoses for this group, notes that the This classification has been more recently employed members of the class are colonial (Rhabdopleura) or by Mierzejewski and Urbanek and their co-authors pseudocolonial (Cephalodiscus, where cloned individ- (e.g. Mierzejewski & Kulicki 2002; Mierzejewski & uals detach from one another as adult zooids) and Urbanek 2004; Urbanek 2004). Whereas this system- possess a shared cuticular tubarium, but otherwise atic approach acknowledges the common descent of focuses on the soft tissue anatomy of the organisms. the graptolites and pterobranchs within the Hemi- In the absence of details of the soft tissue structure of chordata, it significantly expands the concept of 36 Mitchell et al. LETHAIA 46 (2013)

AB C

D EF Fig. 2. Sicular structure. Prosicula (ps) shaded and boundary with metasicula (ms) marked by x—x. A, B, crustoid? sicula with sac-like prosicula and raised primary aperture from which metasicula extended (A–E from Kozłowski 1971; material from Ordovician glacial boulders). C, Epigraptus sp., vesicular prosicula, secondary apertures for metasicula and (b) bitheca. D, Kozlowskitubus erraticus with vase-like prosicula, helical line in distal tubular part and tubular metasicula. E, Dendrograptus communis with entirely tubular prosicula and helical line through- out; also shown is the first daughter theca and its stolotheca, which begins as a tube within the prosicula, passes through a resorbed foramen. F, Tetragraptus reclinatus? with conical nematophorous prosicula and longitudinal rods; zigzag suture between fusellar half rings along mid- line of metasicula (GSC 118638, Mid Ordovician I. victoriae victoriae Zone, Western Newfoundland; from Williams & Clarke 1999; Pl 2, fig. 11). G, Rhabdopleura compacta, recent, showing smooth prosicular dome and distinct change to fusellar half rings in creeping part of first tube (metasicula), and to full rings in erect tube; note also spiral astogeny with tubes coiled around the sicula (inverted SEM image modified from Mierzejewski & Kulicki 2003, Recent material).

‘graptolite’ to encompass extant taxa never previously available concerning colony development, internal included within the group and, in the case of the tubarium structure or ultrastructure (e.g. Archaeolafoea, cephalodiscids, organisms that lack both a sicula and a Sotograptus). They also omitted taxa, especially Epi- serially budded colony structure. This approach also graptus, about which a wealth of structural detail is renders the Class Pterobranchia synonymous with the known (Kozłowski 1971). In addition, their study did Graptolithoidea. not incorporate information from a number of recent Most recently, Rickards & Durman (2006) con- ultrastructural studies of pterobranchs and benthic ducted a cladistic analysis of a suite of graptolite taxa graptolites (e.g. Bates & Urbanek 2002; Mierzejewski as well as Cephalodiscus and Rhabdopleura.They & Kulicki 2003; Maletz et al. 2005). Rickards & Dur- included a number of taxa that are known only from man (2006) treated both Cephalodiscus and Rhabdo- ﬂattened specimens and for which little information is pleura as outgroups and their classiﬁcation recognized LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 37 three classes in the subphylum Pterobranchia: Rhab- dopleurina, Cephalodiscina and Graptolithina. Their 30 published cladogram showed Cephalodiscus as more ⁄ closely related to graptolites than Rhabdopleura;how- 22 ever, it is unclear if they intended this to suggest that ⁄ the serial budding relations and other features shared by Rhabdopleura and graptolites was primitive for the group and subsequently lost in Cephalodiscus,orwere independently derived. Thus, the relationships among Rhabdopleura, Cephalodiscus and graptolites remain unclear. Below, we present an explicit assessment of the degree to which the available data can support a resolution of these relationships.

Phylogenetic analysis

Character analysis and coding An assessment of the similarities shared among the set of taxa chosen for analysis is fundamental to any phylo- geneticstudy.Weemployedthedetailedstructureof anatomical features that represent a suite of external, internal, astogenetic and ultrastructural features to pterobranch relationships.

construct 32 characters – essentially all features of these ⁄ colonies that could be observed in the range of organisms included in this analysis. We also included one ecological character (character 15, planktic habit) that we believe reﬂects a phylogenetically meaningful change in larval structure and colony ecology (see below). Twenty of the characters are binary and 12 are multi-state. All characters were treated as unordered for the main analyses reported here, but for analyses in which we attempted to emulate the results of Rickards & Durman (2006), we ordered some characters as they had done. We did not apply any aprioriweighting of characters. The putative homologies implicit in the xa that were selected to investigate graptolite 1100211?000102 3211110100201012? 0 character coding inevitably included similarities that ⁄ are incongruent with the patterns of relationship suggested by other characters. The sole criterion for deter- mining which of these putative homologies are actually likely to be apomorphies is their consistency with the most-parsimonious tree, as is standard procedure for cladistic analyses (e.g. Forey et al. 1992, p. 3). The following is a list of the 32 characters (numbered from 0 to 31) employed in this analysis. The data matrix is shown in Table 1 and the list of data sources is given in Table 2. Most of these characters and the ??111110222?000102 3 1 3?2?0111212141101212021? ??111110121?000102 3 1 3?2?01?1212141101102012? ⁄ states in which they occur are extensively described and ⁄ ???21200010000 ? ? ?? ? ? ????1?1211120000001110100010? 0 0? 0 ????1?1311000000001110100001? 1 01211012?1130000001010100111? 1 01211012111100000010100010002 ?1?21310002101 ? ? ? ???11?12?1130100101012110011? 1 022?1021111100000011101020101 20122411001000 3 ? 3 1 2 1100111012100110? 12201112121411010020211000102 2 ? ?2?01?12121411000010110000102 ? ? ????1?12?11? 1 12111111 ?0?22411002 000102 3 3 ? 1 3? 1 ???01?12?2141101102012? 3 12201132021411110120211000102 3 12201132021411112120211000102 2 0 12345678 910111213141516171819202122232425262728293031 2 illustrated by Bulman (1970) and Rickards & Durman

(2006), but we also include here some discussion of the ? ? 1 0 00000000000000000010100000010 CA00000000000000000000000000001? OI0 rationale for our coding choices. 0–Prosicula. 0 – absent; 1 – vesicular (Fig. 2A–D, Data matrix of 32 morphological characters and 17 ta G); 2 – tubular (Fig. 2E); 3 – caudal (Figs 2F, 3D; see Bithecocamara Cysticamara Epigraptus Rhabdopleura Cephalodiscus Cephalodiscus Dendrotubus Bulmanicrusta Dendrograptus Mastigograptus Reticulograptus Kozlowskitubus Koremagraptus Anisograptus Rhabdinopora Dictyonema Table 1. also Williams & Clarke 1999). Rickards & Durman Acanthograptus 38 Mitchell et al. LETHAIA 46 (2013)

Table 2. List of sources from which information on characters states coded in Table 1 and discussed in the text was taken. In addition, information for features of many of these pterobranch taxa was also found in Bulman (1970) and Rickards & Durman (2006). Acanthograptus Bulman 1937; Bulman & Rickards 1966; Kozłowski 1949; Urbanek & Towe 1974; Wiman 1901 Koremagraptus Bulman 1927, 1947 Anisograptus Cooper et al. 1998; Maletz 1992; authors’ unpublished observations Rhabdinopora Bulman 1949; Legrand 1974 Dictyonema Bulman 1933; Kozłowski 1949; Urbanek & Mierzejewski 1984; Urbanek & Towe 1974 Dendrograptus Kozłowski 1949, 1971; Urbanek & Mierzejewski 1986 Mastigograptus Andres 1977, 1980; Bates & Urbanek 2002; Bates et al. 2009; Urbanek & Mierzejewski 1984; Urbanek & Towe 1974 Reticulograptus Whittington & Rickards 1968 Kozlowskitubus Kozłowski 1963, 1971; Mierzejewski 1978 Dendrotubus Kozłowski 1949, 1963, 1971 Bulmanicrusta Kozłowski 1962; Urbanek & Mierzejewski 1984; Mierzejewski et al. 2005 Bithecocamara Kozłowski 1949 Cysticamara Kozłowski 1949 Epigraptus Eisenack 1941, 1974; Kozłowski 1949, 1971; Urbanek & Mierzejewski 1982 Rhabdopleura Dilly 1985a,b, Dilly 1986; Mierzejewski & Kulicki 2003; Stebbing 1970; Urbanek & Dilly 2000 Cephalodiscus Andersson 1907; Dilly 1993; Gilchrist 1915, 1917; Harmer 1905; John 1931, 1932; Schiaparelli et al. 2004; Urbanek 1976; Urbanek & Mierzejewski 1984; authors’ unpublished observations

(2006) summarized the available information con- regard this structure as an independently acquired fea- cerning the variety of forms of prosiculae among ture that is not homologous with the prosicula. With graptolites and Rhabdopleura. Epigraptus and Rhabdo- these observations in mind, we have coded cephalo- pleura possessavesicularprosiculathatlacksahelical discidsaslackingbothaprosiculaandametasicula. line (Fig. 2C, G). Maletz et al. (2005) argued that this ThechoiceofcodingofthesicularstatesforCepha- structure may not be homologous with the more typi- lodiscus described above raises a general issue regard- cal tubular or conical prosiculae seen in dendroids ing the coding of features for the cephalodiscids, (Fig. 2E) and graptoloids (Figs 2F, 3D). The fact that which lack many of the tubarium characters present the prosicula of Kozlowskitubus appears to possess a in Rhabdopleura and graptolites. Among graptolites, proximal vesicular unstructured portion and distal there is considerable variation in the form and struc- tubular portion with a helical line (Fig. 2D) supports ture of the sicula, and also in the structure of the sto- the suggestion that the vesicular structures may repre- lon and related features. We have coded all these sent a slightly earlier stage in the ontogeny of the sicul- features as they are likely to be of phylogenetic signiﬁ- ozooid. A specimen that Kozłowski (1971) described cance, but their absence in cephalodiscids might rea- as a probable crustoid has a vesicular prosicula that sonably be handled in several ways. As all grades into a more tubular distal portion, but lacks a Cephalodiscus species lack a sicula and stolon system, helical line (Fig. 2A–B). the simplest approach (which we follow in our main IntheiranalysisRickards&Durman(2006) analyses) is that all the characters related to these two assumed that Cephalodiscus had a mode of develop- systems are coded as absent for the Cephalodiscus taxa ment similar to that in Rhabdopleura (vesicular prosi- included in our analysis. This risks introducing a cula formed by a settled larva). No structures of this forced correlation, possibly ‘double counting’ the sort have ever been described in this group, however. absence of stolons and a sicula, if these absences are all Gilchrist (1915, 1917) and John (1932) described the forced by the same basic lack. That these joint larval development of Cephalodiscus in detail and did absences are not necessarily forced is indicated by the not note the construction of any structure like a prosi- fact that some of these characters (such as the helical cula at any point of larval development. Rather, both line of the prosicula and stolon diaphragms) are also authors noted that construction of the tubarium absent among some graptolites. We tested for forced began after larval development was complete and the correlationeffectsonourresultsintwoways:(1)For zooid had reached adult form. Thus, newly formed the sets of characters that deal with the sicula (0, 1, 2, cephalodiscid colonies of these species apparently lack 6)andthestolon(5,7,8,9,10),wereducedthe a prosicular stage in their construction. weights of those characters to give each set the same In contrast, Schiaparelli et al. (2004) found larval weight as other individual characters. (2) Characters cocoons within colonies of C. (Orthoecus) densus not present because entire organs are absent may be dredged from near Antarctica. The cocoons are some- thought of as inapplicable (see Kitching et al. 1998, what variable in size and shape, and are all formed of pp. 27–30). The prosicula and stolon were coded as agglutinated sand grains. This structure does not form absent in the Cephalodiscus taxa and the other charac- the initial stage of the tubarium, unlike the prosicula, ters in the sicular and stolonal character sets were and is unique among known cephalodiscids. Thus, we coded as unknown (?), meaning that these characters LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 39

Fig. 3. Colony form and structure in the Pterobranchia. a: autotheca; b: bitheca; bs: bithecal stolon; is: irregular suture; ms: metasicula; ps: prosicula; zs: zig-zag suture; sd: stolon diaphragm (base of autothecal cup); s: stolon. A, C, Cephalodiscus (Idiothecia) levinseni,fragmentofa colony showing isolated tubes inhabited by zooids (darker matter) (Recent specimen, Senckenberg Forschungsinstitut und Naturmuseum, Frankfurt, SMF 75728; photo by JM) and C, inset from 3A showing aperture of thecal tube with irregular fuselli. B, Rhabdopleura compacta (Recent material, image modified from Cavers 2005) with distinct diad nodes and stolon diaphragms at the base of the zooidal tubes. D, Adelograptus tenellus (Early Ordovician, drawing modified from Hutt 1974), a primitive planktic graptoloid; early colony growth showing thecal budding and fusellar growth patterns. are inapplicable in that instance. We describe the tubular transition from prosicula to metasicula, indi- effects of these alternate codings in the results section cating that this pore was a primary structure. below. 3–Metasicularfuselli.0 – absent; 1 – irregular 1–Helicalline.0 – absent (Fig. 2A–C, G); 1 – pres- (Fig. 2D); 2 – regular zigzag suture (Fig. 2E, F). ent(Figs2D,E,3D). 4 – Spiral astogeny. 0–absent,1–present.This 2 – Metasicular opening in prosicula. 0–absent;1– character is manifest as a spiral growth pattern of the resorption; 2 – primary. The vesicular prosicula in first theca around the sicula seen in some encrusting Rhabdopleura is secreted initially as a closed vesicle, taxa such as Epigraptus (Fig. 2C) and Dendrotubus through which a pore is resorbed for growth of the (Kozłowski 1949). Rhabdopleura compacta also shows metasicula (Dilly 1986; Sato et al. 2008a). This results a spiral pattern of growth of the first thecal tube in a sharp angle at the point of transition from the around the sicula (Fig. 2G; Mierzejewski & Kulicki prosicula, a feature also seen in the sicula of Epigraptus 2003). (Kozłowski 1971). All other taxa in this analysis for 5 – Serial budding. 0 – absent (Fig. 3A); 1 – present which the form of the sicula is known show a smooth, (Figs 3B, D, 4A, D, E). Cephalodiscid zooids bud from 40 Mitchell et al. LETHAIA 46 (2013)

A B C

E Fig. 4. Stolon development and budding patterns among graptolites. a: autotheca; as:autothecal stolon; b: bitheca; bs: bithecal stolon; g: graptoblast (resting cyst); n: stolon node, sd: stolon diaphragm (base of autothecal cup); ss: stolothecal stolon (axis extending tube); tw: thecal wall. A, SEM image and interpretive diagram of colony segment from dendroid graptolite Desmograptus micronematodes (mid-Silurian, illus- tration after Saunders et al. 2009) showing triad nodes at which colonies formed paired autotheca and bitheca, and branching triad nodes where bitheca was replaced by stolotheca stolon at branch dichotomies. B, sketch of serial section through specimen of Cysticamara accollis (modified after Kozłowski 1949; pl. 29, fig. 15) showing unshared autothecal walls and stolons embedded in spongy extrathecal tissue (i.e. no stolothecal tubes). C, schematic diagram of cross-section of Dictyonema sp. showing shared internal layers of autothecal walls (modified after Urbanek & Towe 1974; fig. 1). D, SEM image of isolated crustoid graptolite stolon fragment (Early Ordovician, image modified from Mierze- jewski et al. 2005) with two closely spaced diad nodes, the first division into a thecal stolon and a stolothecal stolon and the second division of the thecal stolon into an autothecal and a bithecal stolon. E, sketch of fragment of crustoid colony Bulmanicrusta latialata (Mid Ordovi- cian, image after Kozłowski 1962), with preserved closely spaced diad nodes as in C, as well as a graptoblast (resting cyst). F, SEM image of Acanthograptus divergens, showing compound stipes with multiple series of thecae (Mid Ordovician, O¨ land, Sweden, Lund University LO 11414t; photo by JM). a basal disc whereas rhabdopleurids and all graptolites (Fig. 3B; Kozłowski 1949) and homologous lateral bud serially from the stolon system. expansion of the stolon at the base of the autotheca 6 – Internal stolotheca in prosicula. 0–absent;1– canbeobservedinmanygraptolitetaxa(Figs3D,4A, present (Fig. 2E). A number of taxa show the origin D). Although we do have evidence in the form of gra- of the stolotheca as occurring within the prosicula dients in thecal form, branching structures and pat- (e.g. Kozłowski 1971). terns of damage repair, that unsclerotized stolons were 7 – Stolon type. 0 – basal disc (as in Cephalodiscus); retained in planktic graptolites (character 7), there is 1 – tubular (Fig. 3B, D); 2 – beaded (Fig. 4D); 3 – un- no basis from which one could infer that stolon dia- sclerotized. The state ‘unsclerotized’ is inferred where phragms persisted as unsclerotized structures. They thecae are serially interconnected through thecal may have, but lacking any affirmative evidence we tubes, but no stolonal material is found preserved code them as absent in Rhabdinopora and Anisograp- within the tubarium. ‘Beaded’ refers to the state in tus. ‘crustoids’ in which the stolons exhibit regularly 10 – Budding type. 0–basal;1–diad(Fig.3B);2– spaced swellings and constrictions (Fig. 4D; Mierze- triad(Figs3E,4A).Thischaracterfocusesontheanat- jewski et al. 2005). omy of zooid budding – whether from a basal disc (0) 8 – Stolon position. 0 – absent; 1 – embedded in or by division of a persistent stolon into two (diad) basal wall; 2 – central; 3 – embedded in upper wall. segments or simultaneously into three (triad). As 9 – Stolon diaphragms. 0 – absent; 1 – present. noted above, the multiple absences in Cephalodiscus of These structures are well known in Rhabdopleura characters present in the other taxa examined here LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 41 complicates character coding. It happens to be the blind tubes that are not interconnected and hence dif- case that Cephalodiscus zooids both lack serial budding fer structurally from the thecal series seen in grapto- and form buds basally rather than from an elongating lites and Rhabdopleura. stolon. We know of no reason to think that these two 15 – Planktic. 0 – no; 1 – yes. Most graptolites and conditions are forced correlates or are necessarily all the pterobranchs show direct evidence of a benthic redundant. It is conceivable, for instance, that Cepha- mode of life from in situ preservation or their encrust- lodiscus zooids could have evolved to bud serially from ing habit, or both. The nematophorous graptoloids on the basal disc by remaining attached and each budding the other hand, exhibit none of the features associated from the next, in sequence but without an intervening with encrusting habits and available taphonomic evi- stolon, although none are known to do this. Thus, we dence suggests that they could not have lived on or think it appropriate to code budding type separately attached to the seafloor and must have acquired a from serial budding, but we examine the effect of planktic habit and lived within the overlying water alternate coding in the section on sensitivity analyses. column. Although this character is not strictly a mor- Budding in crustoids was not truly triad, but closely phological feature, we believe it captures an evolution- spaced events of diad budding (Fig. 4D, E; Mierzejew- ary innovation in colony development (larval ski et al. 2005). Those species that lack triad budding maturation within the water column) that is relevant have an irregular number of bithecae in relation to to graptolite phylogeny. autothecae; that is, they are not regularly paired, or 16 – Paired dimorphic thecae. 0 – absent; 1 – present they lack bithecae altogether. On the other hand, Mas- (Fig. 4A, F). See the discussion of character 10. tigograptus does possess triad budding, but exhibits no 17 – Stipe connection. 0 – absent; 1 – anastomosis; 2 significant dimorphism between the autothecae and – dissepiments. This character refers to the presence of bithecae, which Bates & Urbanek (2002) considered to intermittent lateral interconnection between adjacent be a primitive trait for this taxon. Triad budding is stipes by either merging of lateral walls (anastomosis) ubiquitous within the dendroids and more derived or presence of extrathecal bars laterally connecting graptolites (e.g. Desmograptus, Fig. 4A; see Saunders stipes (dissepiments). et al. 2009). For these reasons, we have coded paired 18 – Upright planar tubarium. 0–absent;1–pres- dimorphic thecae (character 16) separately from the ent. This character refers to the arrangement of the presence or absence of conothecae or bithecae (char- stipes. It is considered present when stipes are acters 26 and 27 respectively). arranged into a single, two-dimensional sheet (that 11 – Stolothecae. 0 – absent: 1 – present. As noted may be curved into a conical form) and absent if the above, Cephalodiscus lacks interconnected thecal tubes stipes are branched in three dimensions, forming a as well as a serially branching stolon system. Cystica- bushy or irregular tubarium, or if the tubarium is mara possesses a serially branching stolon system, but encrusting and lacking erect stipes. the stolon system is not surrounded by continuous 19 – Thecal construction. 0 – irregular; 1 – tubular thecal tubes (the stolothecae) as seen in all other with unshared walls (Fig. 4B); 2 – tubular with shared graptolites and Rhabdopleura. Rather, the stolon sys- dorsal walls (Fig. 4C). tem is embedded in extrathecal tissue (Fig. 4B) and 20 – Vesicular thecae. 0–absent;1–present open thecal tubes only form at the base of each autot- (Fig. 4E). heca (Kozłowski 1949). 21 – Metathecal ⁄autothecal isolation. 0–non- 12 – Stolothecae location within thecorhiza. 0 – stolo- tubular or irregular; 1 – complete or partial (Figs 3B, thecae absent; 1– encrusting (incorporated in basal 4A, F); 2 – not isolated (Fig. 3D). layer of encrusting mass of thecal tubes – the thecorh- 22–Branchcondition.0 – undefined⁄absent iza); 2 – on top of thecorhiza: 3 – diverse locations (Fig. 3A); 1 – stipes possess single thecal series within thecorhiza (Fig. 4B); 4 – thecorhiza absent (Fig. 3D); 2 – compound (Fig. 4F). Compound stipes (stolothecae within upright-grown stipes). are present among ‘tuboids’ and acanthograptids that 13 – Encrusting. 0–yes;1–no.Anencrustinghabit possess several thecal series fused and growing along a is inferred from evidence of in situ preservation single stipe or branch. and ⁄or presence of a basal membrane. 23 – Fusellar sutures on erect (autothecal) tubes. 0– 14 – Erect series of interconnected thecae. 0 – absent; irregular(Figs2G,3B);1–zigzag(Fig.3D). 1 – present. This character is coded as absent when 24 – Autothecal coiling. 0 – absent; 1 – present. In only individual thecal tubes are erect (Fig. 2G) and some ‘tuboids’ (e.g. Dendrotubus), the autothecae present where there are erect series of multiple inter- show a spiral coiling of the erect thecal tube (Kozłow- connected thecae (stipes; Fig. 4A, F). Thus, although ski 1949). some cephalodiscids [e.g. C.(Idiothecia) negriscens] 25–Closedterminalbuds.0 – absent; 1 – encapsu- have upright bushy colonies, they consist of individual lated; 2 – graptoblast (Fig. 4E). See Urbanek (1984) 42 Mitchell et al. LETHAIA 46 (2013) and Mierzejewski (2000) for descriptions of grapto- (Cephalodiscus CA, hereafter) comprises species that blasts and their possible homology with the closed ter- form encrusting or branching masses that possess a minal buds observed in Rhabdopleura. common internal chamber that houses many individ- 26 – Conothecae. 0 – absent; 1 – present. These are ual zooids (such as Cephalodiscus (Cephalodiscus) large conical thecae irregularly developed in some taxa gracilis Harmer 1905 and Eocephalodiscus polonicus regarded as ‘tuboids’ (Bulman 1970). Kozłowski 1949), or that form reticulate masses with 27 – Bithecae. 0 – absent (Fig. 3A, B); 1 – present no zooecial chambers (such as C.(Acoelothecia) kempi (Figs 3D, 4A, F). Regularly developed thecae that are John 1931). With these two Cephalodiscus composite diminutive as compared with the autothecae. In some taxa, our analysis includes a total of 17 taxa. taxa, autothecae and bithecae are paired (see character TheanatomyoftheuniquepterobranchAtubaria 10, above). heterolopha Sato (1936) fits within the range of form 28 – Spongy extrathecal mass. 0 – present; 1 – exhibited among species of Cephalodiscus; however, in absent. Spongy tissue surrounds the thecal walls in thesingledredgehaulfromwhichthe43knownspec- Cephalodiscus (Fig. 3A) and many encrusting grapto- imens were obtained, none were found associated with lite taxa (e.g. Fig. 4B; Kozłowski 1949). a tubarium, and Sato inferred that they were originally 29 – Endocortex. 0 – pseudocortex; 1 – paracortex; 2 free living and did not form a tubarium. Unfortu- – eucortex. See Urbanek & Mierzejewski (1984) and nately, the absence of tubarial characters for Atubaria references cited therein for descriptions of the ultra- and the lack of detailed soft-part anatomy for the fos- structural terms referred to in characters 29, 30 and sil graptolite taxa means that there is no overlap in the 31. character sets available for these two groups. Although 30 – Ectocortex. 0–absent;1–pseudocortex;2– we could include soft-part characters for the other paracortex; 3 – eucortex. extant pterobranchs (Cephalodiscus and Rhabdople- 31 – Vesicular sheet fabric. 0 – absent; 1 – present. ura), these data would not bear on our main question: whether or not graptolites and pterobranchs form Taxa included in the analysis mutually exclusive groups, and inclusion of Atubaria could not influence the cladistic relationships among Our choice of taxa was guided by the desire to use taxa graptolites and extant pterobranchs. for which most features of the early phases of domicile In addition to the two composite Cephalodiscus construction, structure and ultrastructure are known taxa, we coded Rhabdopleura compacta Hincks (1880). while also including representatives of the full range of The development and structural features (including higher taxa recognized by previous analyses. As we tubarium ultrastructure) are better known in R. com- mentioned previously, Rickards & Durman (2006) pacta colonies (e.g. Stebbing 1968; Dilly 1971, 1973, included taxa for which very little useful information 1975, 1985a, 1986; Stebbing & Dilly 1972; Cavers beyond general colony form is available (Sotograptus 2005; Urbanek & Dilly 2000; Mierzejewski & Kulicki and Archaeolafoea). In addition, they included three 2003; Sato et al. 2008a,b) than in the other commonly genera (Flexicollicamara, Thallograptus and Pala- reported Rhabdopleura species, R. normani Allman eodictyota) that are redundant with better-known taxa. (1869), and for our purposes, the differences between We have omitted all these taxa from our study. In addi- them are of no importance. Some early observations tion to the taxa included by Rickards & Durman of R. normani branching patterns suggested that they (2006), the graptolites Bithecocamara, Bulmanicrusta, produced new zooids behind a persistent terminal Epigraptus and Kozlowskitubus are known from isolated bud that was itself not a fully developed zooid (Sche- collections and we included these taxa in our analysis potieff 1907). This blastozooide inacheve´ was shown as well. Thus, we were able to gather adequate informa- within a pointed, sealed tube that otherwise resembled tion for fourteen unique genera of fossil graptolites. the other tubes of the colony. As Rhabdopleura zooids Among extant pterobranchs, we coded tubarium each build their own tubes through the mortaring characters for representatives of Cephalodiscus and activities of their cephalic disc, it is hard to understand Rhabdopleura.WithinCephalodiscus, we formed two how a terminal bud of the sort described could have composite taxa. The first taxon is species of the sub- formed the tube that contained it. More recent discov- genera C.(Orthoecus)(suchasC. densus Andersson eries of R. normani have not confirmed those observa- 1907; C. solidus Andersson 1907 and C. graptolitoides tions – all tubes are of the more usual form with Dilly 1993) and species of the subgenus C. (Idiothecia) successive budding (Burdon-Jones 1954; Barnes 1977; (e.g. C. nigrescens Lankester 1905). All these species Dilly 1985b; Halanych 1993). Furthermore, observa- (Cephalodiscus OI, hereafter) form pseudocolonial tions of resting cysts in R. compacta (Dilly 1975) and a coenoecia in which the separated individual zooids number of graptolites (Urbanek 1984; Mierzejewski each inhabit their own blind tubes. The second taxon 2000) suggest that the features described in R. normani LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 43 by Schepotieff (1907) are likely to have been faculta- In our second set of analyses, we employed the tive structures, perhaps dormant buds. results of our unrooted analysis and constrained the We did not include the recently discovered Galea- two Cephalodiscus taxa to be a paraphyletic outgroup. plumosus abilus (Hou et al. 2011) in our analysis. This rooted tree differed in several important ways Although this organism has been interpreted as the from the results presented in Rickards & Durman oldest-known fossil pterobranch (lower Cambrian), (2006) and so we conducted a set of experiments in shows soft-bodied preservation and evidence of PAUP 4.10b that were modelled on the methods and growth in a fusellar tube, nothing is known of about data described in Rickards & Durman (2006) in an whether the organism was colonial or what the colony effort to reproduce their results. In these comparisons, organization might have been like (if indeed it was we employed bootstrap resampling analyses with 1000 colonial), whether it had stolonal budding, or a sicula. replicates and random branch addition sequence to Theultrastructureofthetubewallsisunknownas assess group support. well.Therefore,itwasnotpossibletoincludethis We followed up these preliminary analyses with a taxon in a way that adds to our analysis. third and final set of analyses in TNT (Tree Analysis using New Technology, see Goloboff et al. 2008b). Cladistic methods The goals of these analyses were two-fold: first, to test the effects of reweighting characters by their consis- We conducted three sets of analyses, each aimed at tency with the most-parsimonious trees (MPTs) and, achieving slightly different goals. First, we conducted second, to employ TNT’s refined set of measures of an unrooted parsimony analysis to assess the relation- group support and its enhanced methods for under- ships between extant pterobranchs and fossil grapto- standing and comparing trees. Research suggests that lites. Because no other hemichordates are tube weighting by character consistency improves phyloge- builders and because we lack virtually all soft-part netic results (Goloboff et al. 2008a), but this should information (not to mention genetic data) about fossil be done cautiously. Unlike the iterative, ‘reweight by graptolites, there is no usable outgroup by which to character consistency’ option in PAUP 4.10b, the root the analysis and thus to polarize character trans- implied weighting options in TNT implement a search formations for this part of our study. The Enteropn- that seeks to simultaneously maximize character con- eusta are widely regarded as the next most closely sistency and minimize tree length (Goloboff 1993, related member of the Hemichordata, but they possess 1997). This joint optimization reduces the risk posed none of the tube-building features of the pterobranchs in sequential optimization that the initial MPT, which and graptolites (Cameron 2005). More distantly was obtained without regard to character consistency, related taxa also offer no help. Thus, we first present will lead on subsequent reweighting steps to a local the results of an unrooted analysis and discuss its sig- minimum in tree length rather than the globally opti- nificance in light of what is known about the relation- mal solution. TNT provides two alternative methods ships among extant deuterostomes from molecular for weighting by character consistency: implicit phylogenetic studies. It is important to recall in this weighting (IW) and self-weighted optimization (SL). context that an outgroup is not necessary to deter- IW applies weights based on the behaviour of the mine the basic tree structure in Wagner parsimony character over all its transitions, whereas SL estimates analysis (which we employ in all our analyses) or the weights separately for all possible state transitions. other models of character change so long as character Thus, for SL optimization, gains and losses of a partic- transformations are symmetrical: The change from ular character might be weighted differently if losses state a to state b adds to tree length in the same are frequent and gains rare. Both IW and SL weighting amount as a change from b to a (see Felsenstein 2004, require setting a parameter called K, which determines pp. 4–8, 32–34). Rather, the role of outgroup rooting how strongly the optimization discounts the value of is to fix the direction of change: to reveal the relative homoplastic characters (Goloboff et al. 2008a,b; temporal order of taxon divergence, patterns of Mirande 2009). We employed IW weighting and var- monophyly and the sequence of change in characters iedKoverarangefrom3to12inaseriesofruns,as over time. For our initial analysis, we asked a more this was a range over which Mirande (2009) found limited, non-directional question: Do the Cephalodis- noteworthy effects in his analyses. cus taxa and Rhabdopleura form a group to the exclu- Several measures of branch support are imple- sion of graptolites or do one or more of these taxa mented in TNT. Bremer support measures the sup- nestwithinthegraptolitegroup?Weexaminedthis port for each node in terms of the number of question via a branch-and-bound search in PAUP additional steps in tree length that arise from collaps- 4.10b (Swofford 2001). Characters were unweighted ing that node and is equal to F–C, where F is the num- and unordered. ber of characters in favour of that clade and C is the 44 Mitchell et al. LETHAIA 46 (2013) number of conflicting characters. Relative Bremer support measures the relative fit difference, or more specifically, (F–C) ⁄F (Goloboff & Farris 2001). Con- sidered together, these values reveal both the amount of evidence and the degree to which that evidence points toward a particular hypothesis. Resampled group support is similar to bootstrap support, but as implemented in TNT involves a symmetric resampling approach that is said to be less biased than other approaches (Goloboff et al. 2003). Group support also may be expressed as a relative degree of group support, the GC statistic of Goloboff et al. (2003), which for a particular node, is equal to resampled group frequency minus that of the next most fre- quently encountered group for that set of taxa.

Results

Do pterobranchs and graptolites form mutually exclusive groups? Here, we test the hypothesis that pterobranchs and graptolites are each monophyletic groups and sister taxa to one another. As we noted above, no outgroup exists that builds a domicile with even a few characters in common with the tubarium of pterobranchs and graptolites. Consequently, we ﬁrst consider the hypothesis in its more general unrooted form: pterobranchs and graptolites form mutually exclusive groups.Thishypothesisincludestherelationshippos- ited by Bulman (1970; pterobranchs as sister to graptolites, Fig. 5B) as well as that proposed by Rickards & Durman (2006; pterobranchs as paraphyletic ancestral group to graptolites, Fig. 5C). As unrooted trees are less familiar than rooted trees, an explanation of the logic of this test may be helpful. If Cephalodiscus and Rhabdopleura form a clade exclusive of graptolites, and vice versa,eachgroupwillbeclusteredtogether on an unrooted tree (Fig. 5A) with all three extant

Fig. 5. Unrooted trees used to test the hypothesis that Cephalodis- cus and Rhabdopleura form a clade exclusive of graptolites and vice versa. A, an unrooted tree consistent with this hypothesis, shown with three possible rooting locations corresponding to the rooted trees shown in B–D. B, rooted cladogram that corresponds to the phylogenetic relations in Bulman (1970). C, rooted cladogram that corresponds to the phylogenetic interpretation of Rickards & Dur- man (2006). D, a third alternative tree. E, G, two unrooted trees that contradict the hypothesis. F, H, two possible rooted trees corresponding to the unrooted trees in E and G. I, unrooted tree resulting from our ﬁrst round analyses (see text), showing the same geometry among extant pterobranchs and graptolites as that in G and H. Ceph CA = Cephalodiscus (Cephalodiscus)+C.(Aceolothe- cia), Ceph OI = Cephalodiscus (Orthoecus)+C. (Idiothecia), Rhab = Rhabdopleura compacta and Grap 1 and Grap 2 are hypo- thetical graptolites. LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 45 pterobranchs linked directly to one another and none these autothecae are interconnected by a sclerotized will occur between the two graptolites: either group stolon with stolon diaphragms, in common with could be pruned off the tree without removing a Rhabdopleura and the other graptolites. However, in member of the other group. This topological condi- contrast to Rhabdopleura and the other graptolites, tion remains the case in all of the alternate rooted ver- Cysticamara autothecae are separate closed tubes that sions irrespective of which taxon is chosen as are imbedded in spongy, extrathecal tissue through primitive in the rooted trees (e.g. Fig. 5B–D). which they are connected only by their thin stolons Although some outgroup choices cause an extant and not by distinct stolothecal tubes. Thus, Cysticama- pterobranch to share a more recent common ancestor ra is intermediate in its characters between the two with a graptolite than with the other living ptero- pterobranch groups. Accordingly, the unrooted tree branchs (e.g. Fig. 5C), this does not contradict the exhibits the same geometry among extant ptero- hypothesis as the extant pterobranchs always form a branchs and graptolites as that in Figure 5G, H and pterobranch-only paraphyletic group ancestral to consequently, contradicts the hypothesis that grapto- a monophyletic graptolite clade. Even if one picks a lites form a clade exclusive of Cephalodiscus and Rhab- graptolite as the outgroup for trees of this form dopleura. (Fig. 5D), extant pterobranchs become a derived clade within, but still exclusive of, graptolites: in no case is a Rooting the pterobranch–graptolite tree graptolite nested among the extant pterobranchs. Other possible nearest-neighbour relations on unroot- Recent phylogenetic analyses of hemichordates and ed trees do contradict the hypothesis, however (e.g. related taxa based on 18S ribosomal DNA (Cameron Fig. 5E, G): an extant pterobranch is nested between et al. 2000; Cannon et al. 2009) provide important thegraptolitetaxa(andvice versa) and no single prun- constraints on the range of possible relations among ing cut can remove all of one group without also the study taxa. First, these results suggest that the spe- removingamemberoftheothergroup.Inthiscase, cies of Cephalodiscus form a clade that is sister to as the rooted trees illustrate (Fig. 5F, H), any ptero- Rhabdopleura (Fig. 1). In addition, the relationships branch clade that excludes graptolites must be poly- among Cephalodiscus species indicate that the one- phyletic (or vice versa, depending on the rooting). zooid to one-tube habit of colony construction com- Our unrooted simple parsimony analysis produced mon to the C.(Orthoecus)andC.(Idiothecia)species 20 equally most-parsimonious trees (MPTs), an was derived independently from that in Rhabdopleura ensemble consistency index (CI, excluding non-parsi- rather than being an intermediate state between com- mony informative characters) of 0.814 and an ensem- mon tube-dwelling Cephalodiscus and the condition ble rescaled consistency index (RC) of 0.718. An of Rhabdopleura and graptolites. unrooted majority-rule consensus of the 20 equally The results of previous phylogenetic analyses of most-parsimonious trees for this analysis is shown in deuterostomes, including many hemichordates based Figure 5I. Differences among the MPTs involved dif- on both morphological and genetic data sets, as dis- ferences in the branching order of Acanthograptus and cussed above, indicates that pterobranchs are derived Koremagraptus among the ‘dendroids’ and the posi- relative to the enteropneusts (Fig. 1) and that secretion of Kozlowskitubus and Epigraptus among the tion of a collagenous tubarium by use of the special- encrusting graptolites. As these differences did not ized cephalic disc is a synapomorphic feature of the involve radical differences in the basic structure of the Pterobranchia (Cameron 2005). Similarly, the serially tree, the strict and majority rule trees fairly represent budded colony, stolon system and a larvally produced the state of our understanding of the phylogenetic prosicula, which are shared features of Rhabdopleura relations among graptolites. We explore the causes and graptolites, also must be derived features relative and meaning of these uncertainties in the relationships to those of non-pterobranch hemichordates. There- among graptolites more fully below. fore, the results of our unrooted analysis require that The unrooted tree reveals that the extant Rhabdo- either: (1) the larvally produced prosicula, serially pleura nests among encrusting fossil graptolites, sepa- budded colony and stolon system are also all synapo- rately from the extant Cephalodiscus (Fig. 5I). The morphic to the Pterobranchia + graptolites and sub- graptolite Cysticamara occupies a position between sequently were lost in Cephalodiscus;or(2)these Cephalodiscus and Rhabdopleura in every MPT that we shared characters unite Rhabdopleura and the grapto- obtained, regardless of any differences in other taxa lites in a clade that excludes Cephalodiscus.Thelater included, choices of character weighting or outgroup. hypothesis is substantially more parsimonious than Cysticamara occupies this position because it possesses the former. A constraint tree that forces Cephalodiscus (among other things) individual zooidal dwelling and Rhabdopleura to be a monophyletic outgroup (i.e. tubes (autothecae) with erect, isolated apertures and Fig. 5B) is two steps longer (Table 3) than a tree in 46 Mitchell et al. LETHAIA 46 (2013)

Table 3. Fifty percentage majority rule bootstrap percentage values, tree lengths, consistency index (CI) and retention index (RI) obtained for ﬁve sensitivity analyses. Column 1 represents the main analysis. Column 2 represents the traditional view that Rhabdopleura belongs in a monophyletic outgroup with both the Cephalodiscus species groups [i.e. C.(Orthoecus)+C.(Idiothecia) and the C.(Acoelothecia)+C. (Cephalodiscus)]. A bootstrap value is not given for Subclass Graptolithina because Rhabdopleura has been removed from this clade. Column 3 represents the Rickards & Durman (2006) tree with four taxa removed that are not included in their analysis; Bithecocamara, Bulmanicrusta, Kozlowskitubus and Epigraptus. In Column 4, two sets of characters (0–2, 6 and 5, 7–10), which are potentially correlated due to absence of key structures in Cephalodiscus taxa, were down-weighted to 0.25 and 0.20 respectively. In Column 5, the same two sets of characters were treated as inapplicable to the Cephalodiscus taxa and coded as unknown. In Column 6, two interrelated thecal characters (17, 28) were treated as one character by down-weighting each to 0.5. See text for further discussion of these sensitivity analyses.

Sicular & Sicular & budding Thecal Main Rhabdo-pleura as Taxa budding characters characters characters Clade analysis outgroup omitted re-weighted inapplicable re-weighed

1. Subclass Graptolithina 100 N ⁄ A 100 98 67 100 2. Infraclass Eugraptolithina 80 76 93 61 53 81 3. Div. Graptoloida 77 75 76 66 85 77 Tree length 80 82 84 50.95 59 79.5 CI 0.65 0.64 0.63 0.59 0.76 0.64 RI 0.79 0.78 0.66 0.66 0.84 0.78 which only Cephalodiscus is the outgroup and Rhabdo- This can be seen best by excluding Epigraptus and pleura is nested among graptolites (Fig. 5H, I). Taken Dendrotubus from the analysis, each in turn. Exclud- together, these features of hemichordate phylogeny ing Epigraptus (Fig. 6B) results in unambiguous reso- argue for rooting the pterobranch–graptolite tree at lution of the basal portion of the graptolite clade. The the node between Cephalodiscus and all the other taxa clade containing all graptolites + Rhabdopleura (Node (Fig. 6). The remainder of our analyses are based on 1, Fig. 6) in that case has Bremer support of 7 and rel- that rooting. ative Bremer support of 88% (7 of 8 characters support this clade), and resampling group support of Rooted phylogenetic analysis of the Graptolithina 99%. No other hypothesized group appears in the resampling set with a frequency of more than 1%. Branch and bound searches of our dataset using TNT Thus, we may have a high degree of confidence in the produced the same tree topologies as we obtained in reliability of this finding. Cysticamara emerges as sister the earlier analyses, as expected, but provided addi- to Rhabdopleura + all other graptolites, with a Bremer tional opportunities to gauge the effect of choices support of 2, relative Bremer support of 50% (avail- about character weighting, assess group support and able characters support this clade 50% more than they to explore the causes of the two major polytomies that support alternative clades) and strong resampling sup- are present in the strict consensus tree (Fig. 6A). port (79% with no other group found at more than Alternate versions were run with either Cephalodiscus- 11% frequency). Several additional clades emerge with CA (zooids of which inhabit shared common cham- moderate to strong support including the clade con- bers within the tubarium) or Cephalodiscus-OI (zooids sisting of Kozlowskitubus and the more derived grapt- of which inhabit separate individual tubes) composite olites (Node 2, Fig. 6B), which we propose to name taxa employed as outgroups. Both outgroup experi- the Eugraptolithina (see Proposed Phylogenetic Clas- ments resulted in precisely the same results. Conse- sification, below). The effects of excluding Epigraptus quently, we show the two Cephalodiscus taxa on the arise because it exhibits an inflated larval vesicle, like resulting tree as part of an unresolved basal polytomy that in Rhabdopleura (which this analysis suggests is a with the ingroup (Fig. 6). These analyses produced 12 primitive condition for the group). On the other MPTs, were 65 steps in length and had an ensemble hand, Epigraptus also has bithecae, which Rhabdople- CI of 0.633 and RI of 0.707. The strict consensus of ura does not possess, but which are widely shared these trees results in a polytomy that contains Rhabdo- among more derived graptolites (crustoids, Bithecoca- pleura along with six other taxa that have generally mara and the eugraptolithines). These conflicting sim- been regarded as crustoid and tuboid graptolites, as ilarities cause Epigraptus to join the tree with equal well as a clade that consists of Mastigograptus,den- levels of support either above or below Rhabdopleura, droids and graptoloid graptolites. and thus contribute to the large basal polytomy seen Inspection of the alternative trees reveals that the in the all-taxon strict consensus tree (Fig. 6A). phylogenetic location of Epigraptus and Dendrotubus Excluding Dendrotubus from the analysis likewise is very uncertain. Both taxa exhibit several features leads to much better resolution of the phylogenetic that must be regarded as homoplastic under any inter- relationships among the other graptolites (Fig. 6C); pretation of the evolutionary history of this group. however, in this case, the effect is among the LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 47

eugraptolithines (compare Fig. 6B, C). In the absence of Dendrotubus the phylogenetic relationships among the eugraptolithine taxa can be fully resolved although A supports for the several nested groups that emerge are not very high in most cases. Several new features, particularly the appearance of the helical line in the prosicula, presence of an internal stolotheca within the sicula and emergence of eucortex, appear in this part of the tree, but, unfortunately nothing is known about these features (or any other aspect of its sicula or tube ultrastructure) in Dendrotubus. Thus, although its position on the cladogram suggests that Dendrotubus is a member of the Eugraptolithina, its precise phylogenetic position within the clade is unclear. In addition to analyses in which all characters were weighted equally, we ran a set of experiments using both the implied weighting (IW) and self-weighting (SL) options available in TNT over a range of K values (see discussion of methods above). Variation in K from 3 to 12 produced no effect on the results in our case. As expected, reweighting by character consistency B resulted in fewer equally good alternative taxon place- ments and better resolution in the strict consensus as this approach reduced the weight of characters that were inconsistent with the best-ﬁt trees. There were only three MPTs, which had CI > 0.81 and RI > 0.86, under the IW optimization. This effect was even more pronounced for self-weighting. Inspection of group supports for the additional nodes recovered in both the IW and SL solutions revealed that these trees were over-resolved, however. When nodes with low group support were collapsed the resulting trees all exhibited the same topologies as the strict consensus of those obtained in the unweighted analyses (Fig. 6A).

Sensitivity analyses We conducted several additional tests of whether the C branching pattern among the various pterobranch clades in our analysis (Figs 5, 6) was robust. First, we

Fig. 6. Cladograms depicting the phylogenetic relationships among living pterobranchs (Rhabdopleura and Cephalodiscus) and fossil graptolite taxa together with suggested higher taxonomy. Trees A– C obtained from analysis of the 32-character morphological data set using TNT and equal character weighting, but with different taxa included. Numbers above branches are characters for which states are apomorphic at that branch ⁄ Bremer support values ⁄ relative Bremer supports. Values below the nodes are symmetric resampling group frequencies ⁄ GC values. All three analyses recovered strong support for Clade 1 (Graptolithina) and Clade 3 (Graptoloida). Support for Clade 2 (Eugraptolithina) depends on the taxa included in the analysis as a result of poor control on the phylogenetic position of Epigraptus and Dendrotubus (see text for discussion) A, strict consensus of 12 equally parsimonious trees obtained from the full 17 taxon set. B, strict consensus of five equally parsimonious trees based on an analysis that excluded Epi- graptus. C, strict consensus of two equally parsimonious trees based on an analysis that excluded Dendrotubus. 48 Mitchell et al. LETHAIA 46 (2013) constrained Cephalodiscus and Rhabdopleura to a monophyletic clades, and sister group to Cephalodiscus monophyletic outgroup status (Table 3, Column 2) holds over this range of experimental manipulations. and found that the 50% majority rule consensus tree length increased by two steps. We then forced the tree into the Rickards & Durman (2006) topology and to Discussion avoid taxon bias we removed four taxa from the analysis (Bithecocamara, Bulmanicrusta, Epigraptus and Interpretation of tree topology Kozlowskitubus) that were not included in their taxon set, and the tree topology was four steps longer (84 vs. The most profound finding is the discovery that Rhab- 80 steps; Table 3, Column 3) than our main analysis dopleura nests within the clade that includes all grapt- tree. olites. Therefore, the Graptolithina must be included To comply with the assumptions of parsimony, within the Class Pterobranchia (Fig. 6). These results characters should be independent. We ran three fur- clearly distinguish the clade that includes all grapto- ther sensitivity analyses to test the effect of grouping lites plus Rhabdopleura from the Cephalodiscus clade. characters that may be perceived as dependent. In Previous hypotheses have suggested that graptolites the first, two sets of characters were down-weighted: are sister to the pterobranchs (Beklemishev 1951, 1, characters 0–2, and 6, which describe various fea- 1970; Bulman 1970; Rickards & Durman 2006). Our tures of the sicula (Cephalodiscus lacks a sicula and result, which is well supported by all the group sup- all characters associated with it) were reduced to a port measures and by sensitivity analyses, suggests that relative weight of 0.25 so that collectively they graptolites are derived within the pterobranch clade, counted as much as a single character; and, 2, charac- and shows a phylogenetic tree topology of hemichor- ters 5 and 7–10, which describe budding of the dates that has not been suggested by other authors. zooids, including the stolon and features associated The clade Rhabdopleura + graptolites shows 99–100% with it (again Cephalodiscus lacks stolons), were resampling support, the strongest support of any clade down-weighted to 0.20, and then the analysis was in the analysis (Fig. 6). The majority rule consensus rerun (Table 3, Column 4). This resulted in a minor tree shows Rhabdopleura arising from within this reduction in bootstrap support for the Graptolithina clade,notatitsbase(Fig.6).Thistreetopology (to 98%) and somewhat greater reductions for the together with the distribution of shared characters Eugraptolithina and Graptoloida, but did not change (see rooting discussion above), suggests that Rhabdo- the topology of the strict consensus tree, which still pleura is derived relative to Cephalodiscus and that the contains all the major clades recognized in our main zooidal structure of graptolites may be inferred from analysis. Thus, our result is not significantly affected the study of Rhabdopleura. This finding is particularly by the potential correlations among characters that important because, as we noted above, none of the are absent in Cephalodiscus. We also tested whether partially preserved pterobranch or graptolite zooids simply retaining equal weights for characters 1, 2, 6– (e.g. Durman & Sennikov 1993 or Rickards et al. 10 and coding their states in Cephalodiscus as 2009), record sufficient detail to provide much insight unknown rather than absent (i.e. inapplicable; see into zooid anatomy other than that they may be Kitching et al. 1998; p. 27–30), would alter our roughly the same size relative to the diameter of their results (Table 3, Column 5). Once again, we obtained tubes, as are Rhabdopleura zooids. The similarities to thesametreestructureaspreviously,althoughinthis Cephalodiscus are numerous and interesting in this case, resampling support for the Graptolithina was connection as well, including their tube and spine- 67% rather than 98–100% as seen in the other tests. building mechanisms, but we disagree with Dilly This reflects the reduced amount of information (1993) and Rigby (1993) that these features, which are retained for seven phylogenetically informative char- clearly plesiomorphic relative to the Graptolithina, acters in this coding scheme. In the third test of char- make Cephalodiscus a graptolite (see also Urbanek acter independence, two characters that describe 1994). thecal characteristics: paired dimorphic thecae (char- There has been a reduction in zooidal size from acter 17), and bithecae (28), were also combined into Cephalodiscus to Rhabdopleura and accompanying this a single character by down-weighing each to a weight reduction in size has been a simplification of body of 0.5 (Table 3, Column 6). plan (Cameron 2005). The evolution of Rhabdopleura In all the sensitivity analysis (except the forced tree, has been characterized by a loss in gill pores, of one Column 2) the Graptolithina, Eugraptolithina and oftwogonadsaswellasalossoftheredbandonthe Graptoloida were well supported. In no case did the cephalic shield and a reduction in the number of feed- tree topology change and thus the hypothesis that ing arms. If comparison of thecal tube diameter can Graptolithina, Eugraptolithina and Graptoloida are be taken as any indication of zooidal body sizes, then LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 49 this suggests that graptolites had body sizes at least as unable to duplicate the Rickards & Durman (2006) small and possibly derived as those of Rhabdople- results (Table 4, Fig. 7). On the basis of a series of ura (Rigby & Sudbury 1995). In addition, the aper- experiments with this matrix, we infer that the differ- tural form of many graptolite thecal tubes suggests ences between their results and ours arise primarily that, like Rhabdopleura, graptolite zooids may have from differences in the character coding schemes and possessed a single pair of feeding arms, although the information included in the analyses. First, it paired feeding arms have been reported from male C. appears from the discussion in Rickards & Durman sibogae (Harmer 1905) zooids. If the Early Cambrian that they coded many of the characters related to the tube-dwelling animal, Galeaplumosus (Hou et al. sicula and budding patterns in Rhabdopleura and 2011) is a pterobranch, then it also presents an exam- Cephalodiscus as being the same. The information ple of a relatively large-bodied pterobranch with only available to us, however, suggests that Cephalodiscus one pair of feeding tentacles. lacks a sclerotized sicula and also lacks serial, stolonal In keeping with these phylogenetic results, we sug- budding. These features are among the key synapo- gest that Rhabdopleura should be regarded as an morphies for the Rhabdopleura +graptolitesclade extant graptolite (Fig. 6). One might argue conversely (Fig. 6). In addition, according to Rickards & Durman that graptolites should be regarded as extinct rhab- (2006), the synapomorphies uniting all graptolites, dopleurids. That alternative is equally consistent with but excluding Rhabdopleura are ‘heteromorphic fuselli the tree and is more consistent with much recent taxo- and cortex,’ ‘connected wavy fusellar microstructure nomic practice (e.g. Donoghue 2005; de Queiroz to the fuselli’ and ‘helical line potentially present.’ 2007). That approach, however, would subsume a However, several graptolite taxa share with Rhabdople- diverse group with a rich and long-studied fossil ura the presence of a vesicular prosicula (Fig. 2G) that record within a relatively little-studied taxon that hap- lacks a helical line (Fig. 2A–C; Chapman et al. 1996). pens to include four extant species (Dawydoff 1948; In addition, there are no important differences in the Stebbing 1970). As the graptolite fossil record under- fusellar and cortical microstructure between Rhabdo- pins a large portion of the Early Palaeozoic timescale pleura and some graptolites (Mierzejewski & Kulicki and is used by many earth scientists and biologists not 2003). Changing the coding for these characters pro- familiar with the details of deuterostome phylogeny duces a tree topology that is more similar to our and taxonomy, we conclude that the interests of taxo- results than in Rickards & Durman’s (2006) strict con- nomic stability and clarity of scientific communica- sensus tree. Most importantly, that change in charac- tion will be best served by retaining the name ter coding causes Rhabdopleura to be relocated within Graptolithina for this group. the Graptolithina, just as in our results, and so also lends support to our rejection of the hypothesis that Comparison to the Rickards & Durman results pterobranchs and graptolites are mutually exclusive clades. The reasons for the differences in results between our analysis and those of Rickards & Durman (2006) are Relationships among the major tube-building difficult to discern because they did not include the hemichordates data matrix in their publication. A matrix is present in the 1992 doctoral dissertation of P. Durman, which The results of our study permit a re-evaluation of also included a set of phylogenetic trees that appear to hemichordate classification (Fig. 6A). In addition to be the same as those in Rickards & Durman (2006). differences in the relationships between pterobranchs We analysed that matrix following the methods and graptolites, our results also differ from those of described in Rickards & Durman (2006), but we were Rickards & Durman (2006) in that their analysis

Table 4. Quantitative summary of emulation of the Rickards & Durman (2006) results based on an updated matrix from Durman (1992) and several experiments to gauge the effect of differences in the choice of taxa, number of taxa and character information given their character set. See text and Figure 7 for further information about experiments.

Redo from Nearest Dmatrix; Updated D This Rickards & Durman duplicate common matrix; R&D Updated D Analysis paper Durman matrix to R&D taxa unique taxa matrix; all taxa

Number of taxa 17 22 22 22 13 19 21 Tree length 61 73 66 72 65 70 74 Number of MPT’s 4 8 524 8 12 3 335 CI* 0.7455 (0.653) 0.6984 0.6471 0.6724 0.6885 0.6714 RC 0.6569 0.6080 0.5509 0.5375 0.6132 0.5743

CI* : ensemble consistency index (excluding parsimony uninformative characters); RC: rescaled ensemble consistency index. 50 Mitchell et al. LETHAIA 46 (2013)

Fig. 7. Comparison of our results with those of Rickards & Durman (2006). Emulation of the results of Rickards & Durman are based on matrix from (Durman 1992) and several experiments to gauge the effect of differences in the choice of taxa, number of taxa and character information given their character set (all matrices available upon request). A–C are strict consensus trees and D is a 50% majority rule consensus tree; numbers below branches are bootstrap support values based on runs of 100 iterations, and those above branches in D are biparti- tion frequencies. See text for discussion and Table 4 for quantitative results. A, strict consensus tree with closest approximation to the Rickards & Durman result. Settings required to produce this result are described in text. B, strict consensus tree from analysis of 13 taxa shared between the Rickards & Durman taxon set and those employed here. C, strict consensus tree from analysis based on 19 unique taxa in the Rickards & Durman taxon set (omitting three redundant taxa: Thallograptus, Palaeodictyota,andFlexicollicamara) and with some character states updated, primarily to reﬂect new information about Rhabdopleura. D, tree presenting the 50% majority rule consensus of MPT’s obtained with updated Durman matrix (as in C) for the 19 unique Rickards & Durman taxa plus two additional taxa (Epigraptus and Bitheco- camara) unique to our 17-taxon set. united the encrusting camaroids and crustoids as sis- tubarial features such as Camarotubus (Mierzejewski ter taxa in a distinct clade that was distinguished by 2001). We did not include these taxa due to a lack of possession of: 1, vesicular thecae; and, 2, encrusting available information concerning either the early bithecae and autothecae. The crustoid and camaroid stages of tubarium development or ultrastructure, or taxa did not form a clade exclusive of other graptolites both. in our analyses. Rather, Cysticamara occupies a posi- As in the case of crustoids and camaroids, Rickards tion below the common ancestor of Rhabdopleura and & Durman (2006) found ‘tuboids’ to be a distinct other encrusting graptolites, removed from Bithecoca- clade, whereas our results suggest that the ‘tuboids’ mara and Bulmanicrusta, which do however, form a form a paraphyletic group from which all more weakly supported clade. This suggests that vesicular derived graptolites emerge. In addition, our results thecae may have evolved repeatedly among encrusting support suggestions by Maletz et al. (2005) that Epi- graptolites. Uniquely among graptolites, Cysticamara graptus should not be regarded as a tuboid graptolite lacks stolothecae and our analysis suggests this may be (Kozłowski 1949; Bulman 1970), but appears instead a primitive trait. Stolothecae (or their anatomical to be closely allied with Rhabdopleura (Fig. 6). equivalent) are present in Rhabdopleura and most We have only included one of the taxa that Ric- other graptolites. The specimens of Cysticamara also kards & Durman (2006) included in the Dithecoidea lack bithecae, as do colonies of Rhabdopleura. (Mastigograptus), so it was not possible to test their This study supports the suggestion of Mierzejewski hypothesis that the dithecoids are a clade that includes (2001) that the cysticamarids may be primitive with Mastigograptus or whether Mastigograptus should be respect to the tuboids and that some characters may assigned to its own order, the Mastigograptida (Bates have appeared in a mosaic fashion among these & Urbanek 2002). Our results provide only weak sup- encrusting graptolite groups. Further resolution of the port for previous suggestions that the acanthograptids relationships among these primitive graptolites will maybeacladesistertothedendrograptids+grapto- require more information about the proximal devel- loids (Chapman et al. 1996; Rickards & Durman opment and ultrastructure of a wider range of taxa, 2006) and also indicate that the dendrograptids, as particularly forms that appear to show a mosaic of traditionally recognized, are paraphyletic. The LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 51

graptoloids, however, appear to be monophyletic, in keeping with several previous suggestions (e.g. Erdt- mann 1982; Fortey & Cooper 1986). Perhaps the most exciting outcome of this study is that it provides a hypothesis of living graptolites, 1001 rhabdopleurids, from which we can better understand 1200?

⁄ zooidal structure of extinct groups and observe tube construction in hemichordates. To date, there exists

rman (2006) with four coding almost no information on the mechanics of tube building in living hemichordates (Dilly 1986). Perti- nent questions include, from what part of the cephalic

???0 shield is the tube material secreted? How is the dark rind of the sclerotized stolon (pectocaulus) secreted and what is the nature of its connection to the con- tractile stalk (gymnocaulus) of the zooids? Are the outer fuselli secreted before or after the inner fibre layers? What role does the proximity to other zooids play in new tube formation? How does water flow, particularly in the benthic groups, affect tube ontogeny and astogeny? How does Rhabdopleura exit the larval sicula? These are just a few of the questions, long thought intractable by the graptolite community that may finally be attained by a closer examination of living shown in the phylogenetic analysis of Rickards & Du rhabdopleurids. Finally, if our phylogenetic hypothesis is correct, it strongly supports the contention that the tubarium of rhabdopleurids provides a homologue for the graptolite tubarium in terms of its mode of construction and relationship to the zooids. If this is so, then the graptolite tubarium was an engineered domicile built 0000000000000001 0000 ? 0? 001? 100000001 0000 1000003000000000 1120 0110022010011? ? 0 2001 1000003000000000 1130 by the zooids from secretions of the cephalic shield 1 0 1 1 1 entirely externally to the body of the animals – more akin to the honeycombs of bees than an echinoid test, for example. These results directly contradict previous 12 12 ⁄ ⁄ suggestions (Bulman 1970; Kirk 1972; Urbanek 1978; Bates & Kirk 1986) that extrathecal tissue played a 10 10 ⁄ ⁄ role in secretion of the tubarium walls and their derivative structures (such as spines, retiolitid lists, Durman (1992), modified to include only those taxa etc.). This idea has lately lost much of its appeal, especially as the fundamental unity of pterobranch tubarium structure and ultrastructure (including that of fossil graptolites) has become clearer (e.g. Urbanek 1994; Mierzejewski & Kulicki 2003). Therefore, we take this opportunity to emphasize that the sicular and thecal wall material and its derivatives cannot be 0221 121132? 2202? 100000 2001 3? 1132122001000000 2001 ?01212011 ?01212011 regarded in any biological sense as ‘periderm’ and 011? 1 3? 1132122001000000 2001 1001212?1 ⁄ recommend that usage of this term in reference to 1 1 graptolite tubarial material, as has traditionally been done,beabandoned.Thismaterialcansimplybe

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19referred 20 21 22 to 23 as 24 ‘tubarium’ 25 26 27 28 29 or 30 ‘wall 31 32 material’ 33 depending on the specific context. In contrast, it is probable that the black stolon Matrix of taxa, characters and states employed by (pectocaulus) of rhabdopleurids and the benthic graptolite taxa was secreted by the soft tissue of the stolon hbolua0 ?00021?0 0 1 Table 5. Rhabdopleura000 edorpu1?00101121121022000000000 2001 000112011 1 121121022020100000 1 2001 1? 0000003000000001 00 1110 ? 00020? 0 ?00112?1 ?00011?1 0 0 1 1? 011002200001? 1? 0 1001 1 1?errors 0000003000000001 corrected (bold 1110 face). ?00011?1 Cephalodiscus? Dendrograptus11? Dictyonema1?? Rhabdinopora ?00021?0 Anisograptus 2Calyxdendrum ? 2 2 ? 1 0 Dendrotubus01001001??1 Cysticamara0?? Flexicollicamara0?? Wimanicrusta0?? leirsa0?? ?00021?0 Ellesicrusta ylgats? ?0?11 1? 001002210001 ??001??1 Cyclograptus0?? cnhgats? 01?11 121021012100000000 2001 ?0011??1 1 1 1210210? 2110000000 121021012110000000 2001 2001 ?0011??1 ?0011??1 1 1210210? 2100000000 2001 ?0011??1 1 1? 011002210000? 110 2001 ??001??1 1 ??001??1 3 Acanthograptus1?? Koremagraptus??? Palaeodictyota??? Thallograptus??? Tubidendrum0?? Reticulograptus0?? atggats? 01011 ? ? 1021001000000000 ? ? ? 0021011000000001 ?001?0?1 ? 00? ?001???1 ? ? ? 0021001000000001 ? 00? ?001?0?1 Archaeolafoea??? Mastigograptus??? Sotograptus??? (Urbanek & Dilly 2000) and if true, this material 52 Mitchell et al. LETHAIA 46 (2013) could be correctly referred to as periderm. Whether or 2009). These relationships support the hypothesis that not the graptoloid nema was secreted in a way analo- pterobranchs form a sister clade to the Enteropneusta. gous to the rhabdopleurid black stolon or from the The oldest fossil Pterobranchia occur in mid-Cam- cephalic shield in the same manner as cephalodiscid brian strata, dated at approximately 510–505 Ma spines (Dilly 1993) and the other tubarium wall mate- (Maletz et al. 2005; Rickards & Durman 2006). Recent rial remains unsettled (Hutt 1974; Mitchell & Carle discoveries of a single specimen of a possible ptero- 1986; Bates 1987; Rickards 1996). branch (Hou et al. 2011) from the Chengjiang Kons- ervat-Lagersta¨tte of southern China may extend the range of the group into the Early Cambrian Proposed phylogenetic classification (525 Ma). In addition, poorly preserved taxa identi- fied as the relatively derived graptolites Mastigograptus Phylum Hemichordata Bateson 1885; simplex and Dendrograptus sp. (Division Eugraptolith- Class Pterobranchia Lankester 1877 ina, nov.) appear in strata of the Eccaparadoxides insu- laris Zone (lower part of Cambrian Series 3) in Definition. – The least inclusive clade containing Tasmania (Quilty 1971; reviewed in Rickards & Dur- Rhabdopleura normani Allman 1869 (in Norman man 2006), but we are in doubt about whether the 1869), and Cephalodiscus dodecalophus M’Intosh 1887. material is sufficient to support those identifications. These rocks are nearly the same age as the Burgess Discussion. – This is a node-based definition in which Shale, which among other soft-bodied fossils, bears both specifiers are extant; it identifies the pterobranch remains of several enteropneusts (Caron et al. 2010). crown clade. Animals in this clade include both colo- Thus, both the origin of pterobranchs and the split nial and pseudocolonial forms. They share a three- between cephalodiscids and graptolites must predate part body organization with a U-shaped gut, a bilater- 505 Ma and considerable diversification must have ally symmetric, multi-armed suspension feeding organ taken place within the group during the Cambrian that is derived from the mesosome and a prominent explosion as appears to be thecaseallacrosstheBila- cephalic shield (probably homologous with the pro- teria. boscis of the enteropneusts), which they employ to secrete a collagenous, communal domicile (variously Subclass Graptolithina Bronn 1849, emend. referred to as a tubarium, coenecium or rhabdosome). The majority of taxa examined here form small assem- Emended Definition. – Graptolithina is defined as a blages of individuals that inhabit a tubarium attached lineage-based taxon that includes all taxa sharing a to firm substrates, but their most-derived clade, the more recent common ancestry with Rhabdopleura Graptoloida, is planktic, moderately species-rich and than with Cephalodiscus. diverse in colony form. Most of the species included in the pterobranch crown clade are extinct. Discussion. – This definition identifies the total clade The Pterobranchia includes two subgroups: the that includes graptolites and Rhabdopleura and allows Cephalodiscida (including the Cephalodiscidae Har- the taxon to be mapped unambiguously onto molecu- mer 1905 and Eocephalodiscidae Kozłowski 1949) lar phylogenetic trees. On present evidence, the Grap- and the Graptolithina (defined below). The precise tolithina includes all pterobranchs with zooids serially form of the relations between the Pterobranchia and budded from an interconnected stolon system and the Enteropneusta are somewhat uncertain. Genetic, none that lack these features, thus also permitting morphological and total-evidence phylogenies lead to unambiguous identification of fossil graptolites. Syna- conflicting inferences about whether these are sister pomorphic features also appear to include the fact taxa or whether pterobranchs were derived from that the increments that form the creeping tubes of within the harrimanid enteropneust clade (e.g. Hala- the tubarium (fuselli) have a regular zigzag pattern nych 1995; Cameron et al. 2000; Winchell et al. 2002; (Figs 2G, 3B) and those on the upright tubes are simi- Cameron 2005; Sato et al. 2008b; Swalla & Smith lar half rings or regularly arranged full rings. We sus- 2008; Cannon et al. 2009). Recent discovery of a pect that they also are united by possession of a degenerate form of the major body patterning gene sclerotized larval vesicle (prosicula; Figs 2A–E, 3D), hedgehog in Saccoglossus kowalevskii and comparison butitisnotknownwhetherCysticamara,themost with its less-modified ortholog in Rhabdopleura com- basal of our studied taxa, possesses a sclerotized sicula pacta suggests that this gene has a uniquely derived or not. This graptolite differs from all more derived form in S. kowalevskii, the evolution of which has par- graptolites, including Rhabdopleura,inthatitszooidal alleled the evolution of the unique morphological fea- tubes appear to be blind – that is, they are imbedded tures of the harrimanid enteropneusts (Sato et al. in a poorly organized extrathecal tissue and are LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 53 interconnected only by tubular stolons, not continu- rank, however, this name has traditionally retained ous thecal tubes (Fig. 4B). In this respect Cysticamara the ‘…oidea,’ suffix that is now fixed in Linnaean zoo- appears to be intermediate in its organization between logical nomenclature as specific to the rank of super- pseudocolonial Cephalodiscus and the fully intercon- family [International Commission on Zoological nected colonial tubarium of Rhabdopleura and all Nomenclature (ICZN) 1999, Article 29.2]. Thus, we more derived graptolites. follow Maletz et al. (2009) and employ a minor spell- ing alteration to avoid unintentionally suggesting Infraclass Eugraptolithina, nov. superfamily rank for this taxon. The Graptoloida is an apomorphy-based taxon that Definition. – The Eugraptolithina is the holophyletic, includes all planktic graptolites. In addition to the apomorphy-based taxon that includes the first grapto- nematophorous sicula, basal graptoloids share the lite that acquired a prosicula with a helical line and all presence of regularly dichotomously branched stipes its descendants. that form horizontally spreading to conical tubarium and retain paired autothecae and bithecae (see the Discussion. – The taxon includes the paraphyletic recent revision in Maletz et al. 2009; who provide a set of graptolites historically referred to the Tuboi- more complete discussion of the features and content dea (but excluding Epigraptus, as discussed above), of the group). The concept of the Graptoloida pro- the Dendroidea (including Mastigograptus) and the posed herein is identical to the Graptoloidea of Fortey Graptoloidea (see Bulman 1970). Members of the & Cooper (1986) and the Rhabdophora of Allman Eugraptolithina share a classically ‘graptolite’-type (1872) as used by Lapworth (1873a,b), but differs sicula with a basally attached, vase-shaped to tubu- from the Graptoloidea of Bulman (1955, 1970), which lar prosicula that exhibits a helical line. Present evi- excluded the planktic Anisograptidae. Graptoloids dence suggests that this group also shares the first appear immediately above the base of the Ordovi- presence of upright stipes, by which we mean free- cian System, approximately 490 Ma (Sadler et al. standing branches formed by a series of intercon- 2009). Indeed, this event provides a supplemental nected thecae that grew upward from their encrust- guide to the identification of the beginning of this per- ing holdfast into the water column. Many iod of Earth’s history. eugraptolithines formed complexly branched, bushy Acknowledgements. – We wish to acknowledge the financial sup- or conical colonies that contained many hundreds port of Natural Sciences and Engineering Research Council of Can- of zooids. This group includes virtually all of the ada Discovery Grants to MJM and CBC, a St. Francis Xavier macroscopic graptolites that are commonly encoun- University James Chair Visiting Professorship to JM and US National Science Foundation research grant EAR 0418790 to CEM teredinthefossilrecordandthatdiversifiedto and MJM, which provided partial support for this work. become a significant component of the preserved diversity of the early to mid-Palaeozoic faunas. As mentioned above, the Eugraptolithina appear in the early part of Cambrian Series 3 (510 to 505 Ma), References are the oldest-known graptolites (Rickards & Dur- Allman, G.J. 1869: On Rhabdopleura. Quarterly Journal of Micro- scopical Science 9, 57–63. man 2006) and went extinct in the Carboniferous Allman, G.J. 1872: On the morphology and affinities of graptolites. Period (Bulman 1970). As an entirely extinct fossil Annals and Magazine of Natural History, Fourth Series 9, 364–380. group, definition with reference to a key structural Andersson, K.A. 1907: Die Pterobranchier der schwedischen Su¨d- polar-Expedition 1901–1903 nebst Bemerkungen u¨ber Rhabdo- synapomorphy provides a stable and effective pleura normanni Allman. Wissenschaftliche Ergebnisse der means to identify the clade. Schwedischen Su¨dpolar-Expedition 1901-1903 5, 1–122. Andres, D. 1977: Graptolithen aus ordovizischen Geschieben und die fru¨he Stammesgeschichte der Graptolithen. Palaöntologische Division Graptoloida Lapworth, 1875 (in Hop- Zeitschrift 51,52–93. Andres, D. 1980: Feinstrukturen und Verwandtschaftsbeziehungen kinson & Lapworth 1875) emend., nom. transl. der Graptolithen. Palaontologische Zeitschrift 54, 129–170. Barnes, R.D. 1977: A new record of a pterobranch hemichordate Emended definition. – The Graptoloida is the total from the western Hemisphere. Bulletin of Marine Sciences 27, clade descended from the first graptolite to possess a 340–343. sicula with a caudal apex (nematophorous sicula; Bates, D.E.B. 1987: The density of graptoloid skeletal tissue, and its implication for the volume and density of the soft tissue. Lethaia Figs 2F, 3D; see also Williams & Clarke 1999). 20, 149–156. Bates, D.E.B. & Kirk, N.H. 1986: Mode of secretion of graptolite Discussion.– . – ThenameGraptoloidawasestab- periderm, in normal and retiolite graptolites. Geological Society Special Publication 20, 221–236. lished as Section Graptoloidea by Lapworth (in Hop- Bates, D.E.B. & Urbanek, A. 2002: The ultrastructure, develop- kinson & Lapworth 1875), which is a family-group ment, and systematic position of the graptolite genus Mastigo- rank. Ruedemann (1904) raised the taxon to ordinal graptus. Acta Palaeontologica Polonica 47, 445–458. 54 Mitchell et al. LETHAIA 46 (2013)

Bates, D. E. B., Kozłowska, A., Loydell, D., Urbanek, A. & Wade, S. Chapman, A.J., Durman, P.N. & Rickards, R.B. 1996: A provisional 2009: Ultrastructural observations on some dendroid and gra- classification of the graptolite Order Dendroidea. Palaöntologi- ptoloid graptolites and on Mastigograptus. Bulletin of Geosciences sche Zeitschrift 70, 189–202. 84,21–26. Cooper, R.A., Maletz, J., Wang, H. & Erdtmann, B.-D. 1998: Bateson, W. 1885: The later stages in the development of Balanog- Taxonomy and evolution of earliest Ordovician graptoloids. lossus kowalevskii, with a suggestion as to the affinities of the Norsk Geologisk Tidsskrift 78, 3–32. Enteropneusta. Quarterly Journal of Microscopical Science 25,81– Dawydoff, C. 1948: Classe des Pterobranches. In Grasse´,P.P.(ed.): 122. Traite´ de Zoologie, 454–489. Masson et Cie, Paris. Beklemishev, V.N. 1951: Towards establishing a zoological system. Delsuc, F., Brinkmann, H., Chourrout, D. & Philippe, H. 2006: Deuterostomians (Deuterostomia), their origin and composi- Tunicates and not cephalochordates are the closest living relation. Uspekhi Sovremennoi Biologii 32,256–270.[inRussian] tives of vertebrates. Nature 439, 965–968. Beklemishev, V.N. 1970: Principles of Comparative Anatomy of Dilly, P.N. 1971: keratin-like fibres in the hemichordate Rhabdople- Invertebrates. Oliver and Boyd, Edinburgh, Vol. 1, 490 pp, Vol. ura compacta. Zeitschrift fu¨r Zellforschung 117, 502–515. 2, 529 pp. Dilly, N.P. 1973: The larva of Rhabdopleura compacta (Hemichor- Bjerreskov, M. 1978: Discoveries on graptolites by X-ray studies. data). Marine Biology 18, 69–86. Acta Palaeontologica Polonica 23, 463–471. Dilly, P.N. 1975: The dormant buds of Rhabdopleura compacta Bjerreskov, M. 1994: Pyrite diagenesis of graptolites from Born- (Hemichordata). Cell and Tissue Research 159, 387–397. holm, Denmark. In Chen, X., Erdtmann, B.-D. & Ni, Y-N. (eds): Dilly, N.P. 1985a: The prosicular stage of Rhabdopleura (Ptero- Graptolite Research Today, 217–222. Nanjing University Press, branchia: Hemichordata). Journal of Zoology, London (A) 206, Nanjing. 163–174. Bronn, H. G. 1849: Index Palaeontologicus oder U¨ bersicht der bis Dilly, P.N. 1985b: A note on the status of Rhabdopleura from Ber- jetzt bekannten fossilen Organismen, Prof. H. R. Go¨ppert und muda. Journal of the Marine Biological Association of the United Herm. V. Meyer. In Bronn, H.G. (ed.): Zweite Abtheilung. B. Kingdom 65, 987–991. Enumerator palaeontologicus: Systematische Zusammenstellung Dilly, P.N. 1986: Modern pterobranchs: observations on their und geologische Entwicklungs-Gesetze der organischen Reiche,1– behaviour and tube building. In Hughes, C.P. & Rickards, R.B. 976. E. Schweizerbart’sche Verlagsbuchhandlung, Stuttgart. (eds): Palaeoecology and Biostratigraphy of Graptolites. Geological Bulman,O.M.B.1927:UberKoremagraptus fibratus sp. nov. Society Special Publication 20: 261–269. (Grapt., Dendr.). Senkenbergiana 9,38–40. Dilly, P.N. 1993: Cephalodiscus graptolitoides sp. nov. a probable Bulman, O. M. B. 1933: On the graptolites prepared by Holm 4. extant graptolite. Journal of Zoology (London) 229, 69–78. Structural characters of some Dictyonema and Desmograptus Donoghue, P.C.J. 2005: Saving the stem group — a contradiction species from the Ordovician and Silurian rocks of Sweden and in terms? Paleobiology 31, 553–558. the East Baltic Region. Arkiv fo¨r Zoologi 26A,1–52. Durman, P.N. 1992: The Origin and Early Evolution of Graptolites Bulman, O. M. B. 1937: The structure of Acanthograptus suecicus, and Related Hemichordates. Unpublished PhD thesis, Cambridge and the affinities of Acanthograptus. Geologiska Fo¨reningen i University, Cambridge. Stockholm Fo¨rhandlingar 59, 182–188. Durman, P.N. & Sennikov, N.V. 1993: A new rhabdopleurid hemi- Bulman, O. M. B. 1947: A monograph of the Caradoc (Balclatchie) chordate from the Middle Cambrian of Siberia. Palaeontology graptolites from limestones in Laggan Burn, Ayrshire, Part 3. 36, 283–296. Palaeontographical Society Monographs 98, 59–78. Eisenack, A. 1941: Epigraptus bidens n.g. n.sp., eine neue graptoli- Bulman, O. M. B. 1949: A re-interpretation of the structure of thenart des baltischen Ordoviziums. Zeitschrift fu¨r Geschiebefors- Dictyonema flabelliforme Eichwald. Geologiska Fo¨reningen i Stock- chung und Flachlandsgeologie 17,24–28. holm Fo¨rhandlingar 71, 33–40. Eisenack, A. 1974: Einige graptolithen aus dem Ordovizium des Bal- Bulman, O.M.B. 1955: Graptolithina with sections on Enteropneu- tikums,ferneru¨ber EpigraptusundandereIdiotubidae.Neues Jahr- sta and Pterobranchia. In Moore, R.M. (ed.): Treatise on Inverte- buch fu¨rGeologieundPalaöntologie, Monatshefte 66,4–674. brate Paleontology, Part V, 101 pp. Geological Society of America Erdtmann, B.-D. 1982: A reorganization and proposed phyloge- and University of Kansas Press, Lawrence. netic classification of planktic Tremadoc (early Ordovician) den- Bulman, O.M.B. 1970: Graptolithina with sections on Enteropneu- droid graptolites. Norsk Geologisk Tidsskrift 62, 121–145. sta and Pterobranchia In Teichert, K. & Moore, R.M. (eds): Felsenstein, J. 2004: Inferring Phylogenies. Sinauer Associates, Sun- Treatise on Invertebrate Paleontology, Part V (2nd edition), 1– derland. 163. Geological Society of America and University of Kansas Forey, P.L., Humphries, C.J., Kitching, I.L., Scotland, R.W., Siebert, Press, Boulder and Lawrence. D.J. & Williams, D.M. 1992: Cladistics: A Practical Course in Sys- Bulman, O.M.B. & Rickards, R.B. 1966: A revision of Wiman’s tematics. Systematics Association Publication No. 10, Oxford dendroid and tuboid graptolites. Bulletin of the Geological Insti- Science Publications, Oxford. tution of Uppsala 43,1–72. Fortey, R.A. & Cooper, R.A. 1986: A phylogenetic classification of Burdon-Jones, C. 1954: The habitat and description of Rhabdople- the graptoloids. Palaeontology 29, 631–654. ura normani Allman. Universitetet Bergen Arbeits 11,1–17. Gilchrist, J.D.F. 1915: Observations on the Cape Cephalodiscus (C. Cameron, C.B. 2005: A phylogeny of the hemichordates based on gilchristi, Ridewood) and some of its early stages. The Annals morphological characters. Canadian Journal of Zoology 83, 196– and Magazine of Natural History, 8th Series 16: 23, 3–243. 215. Gilchrist, J.D.F. 1917: On the development of the Cape Cephalodis- Cameron, C.B., Garey, J.R. & Swalla, B.J. 2000: Evolution of the cus (C. gilchristi, Ridewood). Quarterly Journal of Microscopical chordate body plan: new insights from phylogenetic analysis of Science 62, 189–211. deuterostome phyla. Proceedings of the National Academy of Sci- Goloboff, P.A. 1993: Estimating character weights during tree ences USA 97, 4469–4474. search. Cladistics 9,83–91. Cannon, J.T., Rychel, A.L., Eccleston, H., Halanych, K.M. & Swalla, Goloboff, P.A. 1997: Self-weighted optimization: tree searches and B.J. 2009: Molecular phylogeny of Hemichordata, with updated character state reconstructions under implied transformation status of deep-sea enteropneusts. Molecular Phylogenetics and costs. Cladistics 13, 225–245. Evolution 52,17–24. Goloboff, P.A. & Farris, J.S. 2001: Methods for quick consensus Caron, J.-B., Conway Morris, S. & Shu, D. 2010: Tentaculate fossils estimation. Cladistics 17, S26–S34. from the Cambrian of Canada (British Columbia) and China Goloboff, P.A., Farris, J.S., Ka¨llersjo¨, M., Oxelman, B., Ramı´rez, M. (Yunnan) interpreted as primitive deuterostomes. PLoS One 5, J. & Szumik, C.A. 2003: Improvements to resampling measures e9586. of group support. Cladistics 19, 324–332. Cavers, P. 2005: Quantifying Variation in a Population of Rhabdo- Goloboff, P.A., Carpenter, J.M., Arias, J.S. & Esquivel, D.R.M. pleura Compacta. Unpublished Zoology Honours thesis, Univer- 2008a: Weighting against homoplasy improves phylogenetic sity of Edinburgh, Edinburgh. analysis of morphological data sets. Cladistics 24, 758–773. LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 55

Goloboff, P.A., Farris, J.S. & Nixon, K.C. 2008b: TNT, a free pro- Maletz, J., Steiner, M. & Fatka, O. 2005: Middle Cambrian ptero- gram for phylogenetic analysis. Cladistics 24, 774–786. branchs and the Question: what is a graptolite? Lethaia 38,73– Halanych, K.M. 1993: Suspension feeding by the lophophore-like 85. apparatus of the pterobranch hemichordate Rhabdopleura nor- Maletz, J., Carlucci, J. & Mitchell, C.E. 2009: Graptoloid cladistics, manni. Biological Bulletin 185, 417–427. taxonomy and phylogeny. Bulletin of Geosciences 84, 7–19. Halanych, K.M. 1995: The phylogenetic position of the ptero- Mierzejewski, P. 1978: Tuboid graptolites from erratic boulders of branch hemichordates based on 18S rDNA sequence data. Poland. Acta Palaeontologica Polonica 23, 585–575. Molecular Phylogenetics and Evolution 4, 72–76. Mierzejewski, P. 2000: On the nature and development of grapto- Harmer, S.F. 1905: The Pterobranchia of the Siboga-Expedition blasts. Acta Palaeontologica Polonica 45, 227–238. with an account of other species. Siboga Expedition Monograph Mierzejewski, P. 2001: A new graptolite intermediate between the 26,1–133. Tuboidea and the Camaroidea. Acta Palaeontologica Polonica 46, Hincks, T. 1880: History of British Marine Polyzoa, 601 pp. Van 367–376. Voorst, London. Mierzejewski, P. & Kulicki, C. 2002: Discovery of Pterobranchia Hopkinson, J. & Lapworth, C. 1875: Descriptions of the graptolites (Graptolithoidea) in the Permian. Acta Palaeontologica Polonica of the Arenig and Llandeilo rocks of St. David‘s. Quarterly Jour- 47, 169–175. nal of the Geological Society of London 31, 631–672. Mierzejewski, P. & Kulicki, C. 2003: Cortical fibrils and secondary Hou, X., Aldridge, R.J., Siveter, D.J., Siveter, D.J., Williams, M., deposits in periderm of the hemichordate Rhabdopleura (Gra- Zalasiewicz, J. & Ma, X. 2011: An Early Cambrian hemichordate ptolithoidea). Acta Palaeontologica Polonica 48, 99–111. zooid. Current Biology 21, 612–616. Mierzejewski, P. & Urbanek, A. 2004: The morphology and fine Hutt, J.E. 1974: The development of Clonograptus tenellus and structure of the Ordovician Cephalodiscus-like genus Melanostro- Adelograptus hunnebergensis. Lethaia 7, 79–92. phus. Acta Palaeontologica Polonica 49, 519–528. International Commission on Zoological Nomenclature (ICZN) Mierzejewski, P., Kulicki, C. & Urbanek, A. 2005: The world’s old- 1999: International Code of Zoological Nomenclature (4th edi- est crustoid graptolites from the upper Tremadocian of Poland. tion). http://www.nhm.ac.uk/hosted-sites/iczn/code/ [accessed Acta Palaeontologica Polonica 50, 721–724. 28 November 2011]. M’Intosh, W.C. 1887: Report on Cephalodiscus dodecalophus, John, C.C. 1931: Cephalodiscus. Discovery Reports 3, 223–260. M’Intosh, a new type of the Polyzoa, procured on the voyage of John, C.C. 1932: On the development of Cephalodiscus. Discovery H.M.S. Challenger during the years 1873–76. Challenger Reports, Reports 6, 191–204. Zoology 20,1–37. Kirk, N.H. 1972: Some Thoughts on the Construction of the Rhabdo- Mirande, J. M. 2009: Weighted parsimony phylogeny of the family some in the Graptolithina, with Special Reference to Extrathecal Characidae (Teleostei: Characiformes). Cladistics 25, 574–613. Tissue and its Bearing on the Theory of Automobility. University Mitchell, C.E. & Carle, K.J. 1986: The nematularium of Pseudocli- of Wales, Aberystwyth, Institute of Earth Studies, Aberystwyth, macograptus scharenbergi (Lapworth) and its secretion. Palaeon- 1–21. tology 29, 373–390. Kitching, I. J., Forey, P. L., Humphries, C. J. & Williams, D. M. Norman, A.M. 1869: Shetland final dredging report. Part 2. On the 1998: Cladistics: Theory and Practice of Parsimony Analysis (2nd Crustacea, Tunicata, Polyzoa, Echinodermata, Actinozoa, Hy- edition). Oxford University Press, Oxford. drozoa and Porifera. Reports of the British Association of the Kozłowski, R. 1949: Graptolites and some new groups of Advancement of Science 38, 247–336. animals from the Tremadoc of Poland. Palaeontologia Polonica Putnam, N.H., Butts, T., Ferrier, D.E. et al. 2008: The amphioxus 3, 1–235. [in French] genome and the evolution of the chordate karyotype. Nature Kozłowski, R. 1962: Crustoidea – a new group of graptolites. Acta 453, 1064–1072. Palaeontologica Polonica 7,1–52.[inFrench] de Queiroz, K. 2007: Toward an integrated system of clade names. Kozłowski, R. 1963: Le de´veloppement d’un graptolite tuboı¨de. Systematic Biology 56, 956–974. Acta Palaeontologica Polonica 8, 103–134. Quilty, P.G. 1971: Cambrian and Ordovician dendroids and hy- Kozłowski, R. 1971: Early development stages and the mode of life droids of Tasmania. Journal of the Geological Society of Australia of graptolites. Acta Palaeontologica Polonica 16, 313–343. 17, 171–189. Lankester, E.R. 1877: Notes on the embryology and classifica- Rickards, R.B. 1996: The graptolite nema: problem to all our solution of the animal kingdom: comprising a revision of specu- tions. Geological Magazine 133, 343–346. lations relative to the origin and significance of the germ- Rickards, R.B. & Durman, P.N. 2006: Evolution of the earliest layers. Quarterly Journal of Microscopical Science, New Series graptolites and other hemichordates. National Museum of Wales 17, 399–454. Geological Series 25,5–92. Lankester, E. R. 1884: A contribution to the knowledge of Rickards, R.B. & Stait, B.A. 1984: Psigraptus, its classification, evo- Rhabdopleura. Quarterly Journal of Microscopical Science 24, 622– lution and zooid. Alcheringa 8, 101–111. 647. Rickards, R.B., Durman, P.N. & Chapman, A.J. 1995: Rhabdopleu- Lankester, E.R. 1905: On a new species of Cephalodiscus (C. nigres- ran hemichordates: new fossil forms and review. Proceedings of cens) from the Antarctic Ocean. Proceedings of the Royal Society the Geologists’ Association 106, 293–303. of London Series B-Biological Sciences 76, 400–402. Rickards, R.B., Wright, A.J. & Thomas, G. 2009: Late Llandovery Lapworth, C. 1873a: Notes on the British graptolites and their (Early Silurian) dendroid graptolites from the Cotton Formation allies. 1. On an improved classification of the Rhabdophora. Part near Forbes, New South Wales. Proceedings of the Linnaean Soci- 1. Geological Magazine 10, 500–504. ety of New South Wales 130,63–76. Lapworth, C. 1873b: Notes on the British graptolites and their Rigby, S. 1993: Palaeontology – Graptolites come to life. Nature allies. 1. On an improved classification of the Rhabdophora. Part 362, 209–210. 2. Geological Magazine 10, 555–560. Rigby, S. & Sudbury, M. 1995: Graptolite ontogeny and the size of Legrand, P. 1974: Development of rhabdosomes with four primary the graptolite zooid. Geological Magazine 132, 427–433. branches in the group Dictyonema flabelliforme (Eichwald). Spe- Ruedemann, R. 1904: Graptolites of New York, part 1. New York cial Papers in Palaeontology 13, 19–34. State Museum Memoir 7, 455–803. Loydell, D.K., Orr, P.J. & Kearns, S. 2004: Preservation of soft tis- Sadler, P.M., Cooper, R.A. & Melchin, M.J. 2009: High-resolution, sues in Silurian graptolites from Latvia. Palaeontology 47,503– early Paleozoic (Ordovician-Silurian) time scales. Geological 513. Society of America Bulletin 121, 887–906. Maletz, J. 1992: The proximal development in anisograptids (Gra- Sato, T. 1936: Vorlaufige mitteilung fiber Atubaria heterolopha gen. ptoloidea, Anisograptidae). Palaöntologische Zeitschrift 66, 297– nov., sp. nov., einen in freiem zustand aufgefundenen Ptero- 309. branchier aus dem stillen ozean. Zoologisehe Anzeiger 15,97–106. 56 Mitchell et al. LETHAIA 46 (2013)

Sato, A., Bishop, J. D. D. & Holland, P. W. H. 2008a: Developmen- Urbanek, A. 1984: The significance of graptoblasts in the life cycle tal biology of pterobranch hemichordates: history and perspec- of crustoid graptolites. Acta Palaeontologica Polonica 28, 313– tives. Genesis 46, 587–591. 326. Sato, A., Rickards, R.B. & Holland, P.W.H. 2008b: The origins of Urbanek, A. 1994: When is a pterobranch a graptolite? Lethaia 27, graptolites and other pterobranchs: a journey from ‘Polyzoa’. Le- 324. thaia 41, 303–316. Urbanek, A. 2004: Morphogenetic gradients in graptolites and bry- Sato, A., White-Cooper, H., Dogget, K. & Holland, P.W.H. 2009: ozoans. Acta Palaeontologica Polonica 49, 485–504. Degenerate evolution of the hedgehog gene in a hemichordate Urbanek, A. & Dilly, N.P. 2000: The stolon system in Rhabdopleura lineage. Proceedings of the National Academy of Sciences USA compacta (Hemichordata) and its phylogenetic implications. 106, 7491–7494. Acta Palaeontologica Polonica 45, 201–226. Saunders,K.M.,Bates,D.E.B.,Kluessendorf,J.,Loydell,D.K.&Mi- Urbanek, A. & Mierzejewski, P. 1982: Ultrastructure of the tuboid kulic, D.G. 2009: Desmograptus micronematodes,aSilurianden- graptolite tubotheca. Palaöntologische Zeitschrift 56,87–93. droid graptolite, and its ultrastructure. Palaeontology 52, 541–559. Urbanek, A. & Mierzejewski, P. 1984: The ultrastructure of the Schepotieff, A. 1907: Die Pterobranchier. I. Rhabdopleura normani Crustoidea and the evolution of graptolite skeletal tissue. Lethaia Allman. Zoologische Jahrbu¨cher (Anatomie) 24, 193–238. 17,73–91. Schiaparelli, S., Cattaneo-Vietti, R. & Mierzejewski, P. 2004: A Urbanek, A. & Mierzejewski, P. 1986: A possible new pattern of ‘protective shell’ around the larval cocoon of Cephalodiscus den- cortical deposits in Tremadoc dendroid graptolites from chert sus Andersson, 1907 (Graptolithoidea, Hemichordata). Polar nodules. Geological Society Special Publication 20, 13–19. Biology 27, 813–817. Urbanek, A. & Towe, K.M. 1974: Ultrastructural studies on grapto- Spengel, J.W. 1932: Planctosphaera. Report on the Scientific Results lites 1: the periderm and its derivatives in the Dendroidea and in of the ‘Michael Sars’ North Atlantic Deep-Sea Expedition 1910, 5, Mastigograptus. Smithsonian Contributions to Paleobiology 20,1– 590–600. 48. Stebbing, A.R.D. 1968: Discovery of Rhabdopleura (Hemichordata) Whittington, H.B. & Rickards, R.B. 1968: New tuboid graptolite at Plymouth. Nature 217, 1287. from the Ordovician of Ontario. Journal of Paleontology 42,61– Stebbing, A.R.D. 1970: The status and ecology of Rhabdopleura 69. compacta (Hemichordata) from Plymouth. Journal of the Marine Williams, S. H. & Clarke, L. C. 1999: Structure and secretion of the Biology Association of the United Kingdom 50, 209–221. graptolite prosicula, and its application for biostratigraphical Swalla, B.J. & Smith, A.B. 2008: Deciphering deuterostome phylog- and evolutionary studies. Palaeontology 42, 1003–1015. eny: molecular, morphological and palaeontological perspectives. Wiman, C. 1901: U¨ ber die Borkholmer Schicht im Mittelbaltischen Philosophical Transactions of the Royal Society B 363, 1557–1568. Silurgebiet. Bulletin of the Geological Institute of Uppsala 5, 151– Swofford, D.L. 2001: PAUP: Phylogenetic Analysis using Parsimony, 222. version 4.0b10. Sinauer Associates, Sunderland. Winchell, C.J., Sullivan, J., Cameron, C.B., Swalla, B.J. & Mallatt, J. Urbanek, A. 1976: The problem of graptolite affinities in the light 2002: Evaluating hypothesis of deuterostome phylogeny and of ultrastructural studies on peridermal derivatives in ptero- chordate evolution with new LSU and SSU ribosomal DNA branchs. Acta Palaeontologica Polonica 21,3–36. data. Molecular Biology and Evolution 19, 762–776. Urbanek, A. 1978: Significance of ultrastructural studies for graptolite research. Acta Palaeontologica Polonica 23, 595–625.