Phylogenetic analysis reveals that Rhabdopleura is an extant ......ecological character (character...
23
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 deuterosto- mes, 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 sup- ported 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 grapto- lites 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 informa- tion 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 in our efforts to understand the pattern of relation- ships among the three major deuterostome taxa, the nature of the deuterostome ancestor and the evolu- tionary origin of the chordates (Fig. 1). Historically, hemichordates have been divided into four classes: the extant Enteropneusta (acorn worms), Planktosp- haeroidea and Pterobranchia and the extinct Grapto- lithina (Bulman 1970). The relationship among the classes is highly uncertain, however, and this con- founds efforts to reconstruct deuterostome evolution. It is not known whether the pterobranchs are a sister group to the enteropneusts (e.g. Winchell et al. 2002; Cameron 2005) or derived within the enteropneust clade (e.g. Cameron et al. 2000; Cannon et al. 2009). It is not known whether the monotypic Planktosp- haeroidea – rare, large, planktic larvae – represents a distinct clade of animals or are merely hypertrophied tornaria larva of an enteropneust (Spengel 1932). Finally, we do not know the relationship among the tube-building groups: the graptolites and ptero- branchs. This latter question is the topic of our study. Pterobranchs are colonial or pseudocolonial and reproduce via short-lived planula larvae and asexual budding. They produce a collagenous dwelling struc- ture that is commonly called a coenecium in the case of pterobranchs and a rhabdosome in the case of graptolites. For simplicity’s sake, we will use the term tubarium (Lankester 1884) for all of these communal domiciles. The two major orders commonly assigned to the Pterobranchia are the Cephalodiscida and the Rhabdopleurida. Both groups are relatively well known from their living representatives, but have yielded a fairly sparse fossil record (Rickards et al. 1995). Graptolites, in contrast, have a relatively com- plete and rich fossil record of their skeletal details, whereas their soft-part anatomy is almost completely DOI 10.1111/j.1502-3931.2012.00319.x Ó 2012 The Authors, Lethaia Ó 2012 The Lethaia Foundation
Transcript of Phylogenetic analysis reveals that Rhabdopleura is an extant ......ecological character (character...
Phylogenetic analysis reveals that Rhabdopleura is an extantgraptolite
CHARLES E. MITCHELL, MICHAEL J. MELCHIN, CHRIS B. CAMERON AND JORG MALETZ
Mitchell, C.E., Melchin, M.J., Cameron, C.B. & Maletz, J. 2013: Phylogenetic analysisreveals 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 ordersreveals that Rhabdopleura nests among the benthic, encrusting graptolite taxa as it sharesall of the synapomorphies that unite the graptolites. Therefore, rhabdopleurids can beregarded as extant members of the Subclass Graptolithina (Class Pterobranchia). Com-bined with the results of previous molecular phylogenetic studies of extant deuterosto-mes, these results also suggest that the Graptolithina is a sister taxon to the SubclassCephalodiscida. 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 andpalaeobiology 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 sup-ported in this analysis. In addition, we recognize the clade Eugraptolithina (nov.). Thisclade comprises the Graptoloida and all of the other common and well-known grapto-lites of the distinctive Palaeozoic fauna. Most of the graptolites traditionally regarded astuboids and dendroids appear to be paraphyletic groups within the Eugraptolithina;however, Epigraptus is probably not a member of this clade. The Eugraptolithina appearto be derived from an encrusting, Rhabdopleura-like species, but the available informa-tion is insufficient to resolve the phylogeny of basal graptolites. The phylogeneticposition of Mastigograptus and the status of the Dithecoidea and Mastigograptidaalso 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 ofEarth Sciences, St. Francis Xavier University, Antigonish, NS B2G 2W5, Canada; Chris B.Cameron [[email protected]], Departement de sciences biologiques, Universite deMontreal, Montreal QC H2V 2S9, Canada; Jorg Maletz [[email protected]], Institutfur Geologische Wissenschaften, Freie Universitat 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 rolein our efforts to understand the pattern of relation-ships among the three major deuterostome taxa, thenature of the deuterostome ancestor and the evolu-tionary origin of the chordates (Fig. 1). Historically,hemichordates have been divided into four classes:the extant Enteropneusta (acorn worms), Planktosp-haeroidea and Pterobranchia and the extinct Grapto-lithina (Bulman 1970). The relationship among theclasses is highly uncertain, however, and this con-founds efforts to reconstruct deuterostome evolution.It is not known whether the pterobranchs are a sistergroup to the enteropneusts (e.g. Winchell et al. 2002;Cameron 2005) or derived within the enteropneustclade (e.g. Cameron et al. 2000; Cannon et al. 2009).It is not known whether the monotypic Planktosp-haeroidea – rare, large, planktic larvae – represents adistinct clade of animals or are merely hypertrophiedtornaria larva of an enteropneust (Spengel 1932).
Finally, we do not know the relationship among thetube-building groups: the graptolites and ptero-branchs. This latter question is the topic of ourstudy.
Pterobranchs are colonial or pseudocolonial andreproduce via short-lived planula larvae and asexualbudding. They produce a collagenous dwelling struc-ture that is commonly called a coenecium in the caseof pterobranchs and a rhabdosome in the case ofgraptolites. For simplicity’s sake, we will use the termtubarium (Lankester 1884) for all of these communaldomiciles. The two major orders commonly assignedto the Pterobranchia are the Cephalodiscida and theRhabdopleurida. Both groups are relatively wellknown from their living representatives, but haveyielded a fairly sparse fossil record (Rickards et al.1995). Graptolites, in contrast, have a relatively com-plete and rich fossil record of their skeletal details,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
unknown except for a few poorly preserved remnants(Bjerreskov 1978, 1994; Rickards & Stait 1984; Loydellet al. 2004). In addition, the Middle to Late Cambrianfossil record of graptolites, which encompasses theearliest part of their evolutionary history, is much lesscomplete than for the Ordovician or Silurian periodswhen they became diverse and widespread (Rickards& Durman 2006). This disparity in the available infor-mation for the various hemichordate groups has hin-dered a complete understanding of the phylogeneticrelationships between the living and extinct groupsand between different groups of graptolites.
The goals of the present study are: (1) to testwhether Pterobranchia and Graptolithina are mono-phyletic groups and sister taxa as has commonly beensupposed; and, (2) in so doing, create a robust phylo-genetic tree of the major tube-building hemichordatetaxa that can provide a foundation for further palaeo-biological and macroevolutionary studies.
Previous phylogenetic interpretations
The construction of the collagenous tubarium andasexually budded colonial zooids are synapomorphiesof the Pterobranchia (Cameron 2005) and these fea-tures are shared with the graptolites, along with manydetails of the composition and fusellar mode of con-struction of the tubarium (Andres 1977; Dilly 1993;Mierzejewski & Kulicki 2003). Despite these similari-ties, Bulman (1970) regarded the Pterobranchia andGraptolithina as distinct classes. His diagnosis of thePterobranchia, which matches closely with mostwidely used diagnoses for this group, notes that themembers of the class are colonial (Rhabdopleura) orpseudocolonial (Cephalodiscus, where cloned individ-uals detach from one another as adult zooids) andpossess a shared cuticular tubarium, but otherwisefocuses on the soft tissue anatomy of the organisms.In the absence of details of the soft tissue structure of
graptolites, graptolites were excluded from the Ptero-branchia.
On the other hand, Bulman’s (1970, p. V17) defini-tion of the Class Graptolithina is ‘The Graptolithinaare colonial, marine organisms, which secreted a scler-otized exoskeleton with characteristic growth bands(fuselli) and growth lines. The thecae house individualzooids usually arranged in a single or double rowalong the branches (stipes) of the colony (rhabdo-some), rarely in irregular aggregates. In most orders,the thecae are polymorphic and in three they areclearly related to an internal sclerotized stolon system.Rhabdosomes originate by a single bud from the ini-tial zooid, housed in a conical sicula, producing sim-ple, branched or rarely encrusting colonies.’
Cephalodiscid pterobranchs are clearly excludedfrom the Graptolithina, based on this diagnosis, eventaking into account recent evidence that their tubari-um and its spine-like structures show a mode of con-struction similar to that inferred for graptolites (Dilly1993). In contrast, the only respect in which the rhab-dopleurid pterobranchs do not fit this diagnosis is inthe description of the sicula as a conical structure. Thebushy dendroid graptolites, however, possess a tubularprosicula (Fig. 2; Kozłowski 1949) and, moreover,Epigraptus (and possibly the crustoids) possess a vesic-ular prosicula very similar to that of Rhabdopleura(Kozłowski 1971).
Beklemishev (1951, 1970), in contrast, proposedthat all of the pterobranchs and graptolites should beunited in a single Class Graptolithoidea, distinguishedby the presence of a sclerotized tubarium constructedwith fusellar increments that house a colony of zooids.This classification has been more recently employedby Mierzejewski and Urbanek and their co-authors(e.g. Mierzejewski & Kulicki 2002; Mierzejewski &Urbanek 2004; Urbanek 2004). Whereas this system-atic approach acknowledges the common descent ofthe graptolites and pterobranchs within the Hemi-chordata, it significantly expands the concept of
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).
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 35
‘graptolite’ to encompass extant taxa never previouslyincluded within the group and, in the case of thecephalodiscids, organisms that lack both a sicula and aserially budded colony structure. This approach alsorenders the Class Pterobranchia synonymous with theGraptolithoidea.
Most recently, Rickards & Durman (2006) con-ducted a cladistic analysis of a suite of graptolite taxaas well as Cephalodiscus and Rhabdopleura. Theyincluded a number of taxa that are known only fromflattened specimens and for which little information is
available concerning colony development, internaltubarium structure or ultrastructure (e.g. Archaeolafoea,Sotograptus). They also omitted taxa, especially Epi-graptus, about which a wealth of structural detail isknown (Kozłowski 1971). In addition, their study didnot incorporate information from a number of recentultrastructural studies of pterobranchs and benthicgraptolites (e.g. Bates & Urbanek 2002; Mierzejewski& Kulicki 2003; Maletz et al. 2005). Rickards & Dur-man (2006) treated both Cephalodiscus and Rhabdo-pleura as outgroups and their classification recognized
A B
D E F
G
C
Fig. 2. Sicular structure. Prosicula (ps) shaded and boundary with metasicula (ms) marked by x—x. A, B, crustoid? sicula with sac-like prosi-cula 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 firsttube (metasicula), and to full rings in erect tube; note also spiral astogeny with tubes coiled around the sicula (inverted SEM image modifiedfrom Mierzejewski & Kulicki 2003, Recent material).
36 Mitchell et al. LETHAIA 46 (2013)
three classes in the subphylum Pterobranchia: Rhab-dopleurina, Cephalodiscina and Graptolithina. Theirpublished cladogram showed Cephalodiscus as moreclosely related to graptolites than Rhabdopleura; how-ever, it is unclear if they intended this to suggest thatthe serial budding relations and other features sharedby Rhabdopleura and graptolites was primitive for thegroup and subsequently lost in Cephalodiscus, or wereindependently derived. Thus, the relationships amongRhabdopleura, Cephalodiscus and graptolites remainunclear. Below, we present an explicit assessment ofthe degree to which the available data can support aresolution of these relationships.
Phylogenetic analysis
Character analysis and coding
An assessment of the similarities shared among the setof taxa chosen for analysis is fundamental to any phylo-genetic study. We employed the detailed structure ofanatomical features that represent a suite of external,internal, astogenetic and ultrastructural features toconstruct 32 characters – essentially all features of thesecolonies that could be observed in the range of organ-isms included in this analysis. We also included oneecological character (character 15, planktic habit) thatwe believe reflects a phylogenetically meaningfulchange in larval structure and colony ecology (seebelow). Twenty of the characters are binary and 12 aremulti-state. All characters were treated as unorderedfor the main analyses reported here, but for analyses inwhich we attempted to emulate the results of Rickards& Durman (2006), we ordered some characters as theyhad done. We did not apply any a priori weighting ofcharacters. The putative homologies implicit in thecharacter coding inevitably included similarities thatare incongruent with the patterns of relationship sug-gested by other characters. The sole criterion for deter-mining which of these putative homologies are actuallylikely to be apomorphies is their consistency with themost-parsimonious tree, as is standard procedure forcladistic analyses (e.g. Forey et al. 1992, p. 3).
The following is a list of the 32 characters (numberedfrom 0 to 31) employed in this analysis. The datamatrix is shown in Table 1 and the list of data sourcesis given in Table 2. Most of these characters and thestates in which they occur are extensively described andillustrated by Bulman (1970) and Rickards & Durman(2006), but we also include here some discussion of therationale for our coding choices.
0 – Prosicula. 0 – absent; 1 – vesicular (Fig. 2A–D,G); 2 – tubular (Fig. 2E); 3 – caudal (Figs 2F, 3D; seealso Williams & Clarke 1999). Rickards & Durman T
able
1.D
ata
mat
rix
of
32m
orp
ho
logi
calc
har
acte
rsan
d17
taxa
that
wer
ese
lect
edto
inve
stig
ate
grap
toli
te⁄p
tero
bra
nch
rela
tio
nsh
ips.
01
23
45
67
89
1011
1213
1415
1617
1819
2021
2223
2425
2627
2829
3031
Aca
nth
ogra
ptu
s2
⁄3?
2?
01
?1
21
21
41
10
11
02
01
2?
00
01
02
31
Kor
emag
rapt
us
??
??
01
?1
2?
21
41
10
11
02
01
2?
00
01
02
3?
An
isog
rapt
us
31
22
01
13
20
21
41
11
10
12
02
11
00
01
02
31
Rha
bdin
opor
a3
12
20
11
32
02
14
11
11
21
20
21
10
00
10
23
1D
icty
onem
a2
⁄3?
2?
01
11
21
21
41
10
12
12
02
1?
00
01
02
31
Den
drog
rapt
us
21
22
01
11
21
21
41
10
10
02
02
11
00
01
02
3?
Mas
tigo
grap
tus
2?
2?
01
?1
21
21
41
10
00
01
01
10
00
01
02
31
Ret
icu
logr
aptu
s?
??
??
1?
12
?1
1?
11
00
11
10
12
10
01
10
??
?K
ozlo
wsk
itu
bus
11
21
11
11
1⁄2
11
11
01
00
20
10
12
?0
00
10
23
0D
endr
otu
bus
??
??
11
?1
2?
11
30
10
01
01
01
21
10
01
1?
??
Bu
lman
icru
sta
10
22
?1
02
11
11
10
00
00
01
11
01
02
01
01
⁄22
⁄30
Bit
heco
cam
ara
??
??
?1
?1
21
11
20
00
00
01
11
01
00
01
0?
??
Cys
tica
mar
a?
??
??
1?
13
11
00
00
00
00
11
10
10
00
01
??
?E
pigr
aptu
s1
01
21
10
12
?1
13
00
00
00
10
10
10
01
11
??
0R
habd
ople
ura
10
12
11
01
21
11
10
00
00
01
01
00
01
00
02
00
Cep
halo
disc
us
CA
00
00
00
00
00
00
00
00
00
00
00
00
00
00
1?
??
Cep
halo
disc
us
OI
00
00
00
00
00
00
00
00
00
01
01
00
00
00
10
10
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 37
(2006) summarized the available information con-cerning the variety of forms of prosiculae amonggraptolites and Rhabdopleura. Epigraptus and Rhabdo-pleura possess a vesicular prosicula that lacks a helicalline (Fig. 2C, G). Maletz et al. (2005) argued that thisstructure may not be homologous with the more typi-cal tubular or conical prosiculae seen in dendroids(Fig. 2E) and graptoloids (Figs 2F, 3D). The fact thatthe prosicula of Kozlowskitubus appears to possess aproximal vesicular unstructured portion and distaltubular portion with a helical line (Fig. 2D) supportsthe suggestion that the vesicular structures may repre-sent a slightly earlier stage in the ontogeny of the sicul-ozooid. A specimen that Kozłowski (1971) describedas a probable crustoid has a vesicular prosicula thatgrades into a more tubular distal portion, but lacks ahelical line (Fig. 2A–B).
In their analysis Rickards & Durman (2006)assumed that Cephalodiscus had a mode of develop-ment similar to that in Rhabdopleura (vesicular prosi-cula formed by a settled larva). No structures of thissort have ever been described in this group, however.Gilchrist (1915, 1917) and John (1932) described thelarval development of Cephalodiscus in detail and didnot note the construction of any structure like a prosi-cula at any point of larval development. Rather, bothauthors noted that construction of the tubariumbegan after larval development was complete and thezooid had reached adult form. Thus, newly formedcephalodiscid colonies of these species apparently lacka prosicular stage in their construction.
In contrast, Schiaparelli et al. (2004) found larvalcocoons within colonies of C. (Orthoecus) densusdredged from near Antarctica. The cocoons are some-what variable in size and shape, and are all formed ofagglutinated sand grains. This structure does not formthe initial stage of the tubarium, unlike the prosicula,and is unique among known cephalodiscids. Thus, we
regard this structure as an independently acquired fea-ture that is not homologous with the prosicula. Withthese observations in mind, we have coded cephalo-discids as lacking both a prosicula and a metasicula.
The choice of coding of the sicular states for Cepha-lodiscus described above raises a general issue regard-ing the coding of features for the cephalodiscids,which lack many of the tubarium characters presentin Rhabdopleura and graptolites. Among graptolites,there is considerable variation in the form and struc-ture of the sicula, and also in the structure of the sto-lon and related features. We have coded all thesefeatures as they are likely to be of phylogenetic signifi-cance, but their absence in cephalodiscids might rea-sonably be handled in several ways. As allCephalodiscus species lack a sicula and stolon system,the simplest approach (which we follow in our mainanalyses) is that all the characters related to these twosystems are coded as absent for the Cephalodiscus taxaincluded in our analysis. This risks introducing aforced correlation, possibly ‘double counting’ theabsence of stolons and a sicula, if these absences are allforced by the same basic lack. That these jointabsences are not necessarily forced is indicated by thefact that some of these characters (such as the helicalline of the prosicula and stolon diaphragms) are alsoabsent among some graptolites. We tested for forcedcorrelation effects on our results in two ways: (1) Forthe sets of characters that deal with the sicula (0, 1, 2,6) and the stolon (5, 7, 8, 9, 10), we reduced theweights of those characters to give each set the sameweight as other individual characters. (2) Charactersnot present because entire organs are absent may bethought of as inapplicable (see Kitching et al. 1998,pp. 27–30). The prosicula and stolon were coded asabsent in the Cephalodiscus taxa and the other charac-ters in the sicular and stolonal character sets werecoded as unknown (?), meaning that these characters
Table 2. List of sources from which information on characters states coded in Table 1 and discussed in the text was taken. In addition, infor-mation 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 1901Koremagraptus Bulman 1927, 1947Anisograptus Cooper et al. 1998; Maletz 1992; authors’ unpublished observationsRhabdinopora Bulman 1949; Legrand 1974Dictyonema Bulman 1933; Kozłowski 1949; Urbanek & Mierzejewski 1984; Urbanek & Towe 1974Dendrograptus Kozłowski 1949, 1971; Urbanek & Mierzejewski 1986Mastigograptus Andres 1977, 1980; Bates & Urbanek 2002; Bates et al. 2009; Urbanek & Mierzejewski 1984; Urbanek & Towe 1974Reticulograptus Whittington & Rickards 1968Kozlowskitubus Kozłowski 1963, 1971; Mierzejewski 1978Dendrotubus Kozłowski 1949, 1963, 1971Bulmanicrusta Kozłowski 1962; Urbanek & Mierzejewski 1984; Mierzejewski et al. 2005Bithecocamara Kozłowski 1949Cysticamara Kozłowski 1949Epigraptus Eisenack 1941, 1974; Kozłowski 1949, 1971; Urbanek & Mierzejewski 1982Rhabdopleura Dilly 1985a,b, Dilly 1986; Mierzejewski & Kulicki 2003; Stebbing 1970; Urbanek & Dilly 2000Cephalodiscus Andersson 1907; Dilly 1993; Gilchrist 1915, 1917; Harmer 1905; John 1931, 1932; Schiaparelli et al. 2004;
Urbanek 1976; Urbanek & Mierzejewski 1984; authors’ unpublished observations
38 Mitchell et al. LETHAIA 46 (2013)
are inapplicable in that instance. We describe theeffects of these alternate codings in the results sectionbelow.
1 – Helical line. 0 – absent (Fig. 2A–C, G); 1 – pres-ent (Figs 2D, E, 3D).
2 – Metasicular opening in prosicula. 0 – absent; 1 –resorption; 2 – primary. The vesicular prosicula inRhabdopleura is secreted initially as a closed vesicle,through which a pore is resorbed for growth of themetasicula (Dilly 1986; Sato et al. 2008a). This resultsin a sharp angle at the point of transition from theprosicula, a feature also seen in the sicula of Epigraptus(Kozłowski 1971). All other taxa in this analysis forwhich the form of the sicula is known show a smooth,
tubular transition from prosicula to metasicula, indi-cating that this pore was a primary structure.
3 – Metasicular fuselli. 0 – absent; 1 – irregular(Fig. 2D); 2 – regular zigzag suture (Fig. 2E, F).
4 – Spiral astogeny. 0 – absent, 1 – present. Thischaracter is manifest as a spiral growth pattern of thefirst theca around the sicula seen in some encrustingtaxa such as Epigraptus (Fig. 2C) and Dendrotubus(Kozłowski 1949). Rhabdopleura compacta also showsa spiral pattern of growth of the first thecal tubearound the sicula (Fig. 2G; Mierzejewski & Kulicki2003).
5 – Serial budding. 0 – absent (Fig. 3A); 1 – present(Figs 3B, D, 4A, D, E). Cephalodiscid zooids bud from
A
B
D
C
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, fragment of acolony 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 showingthecal budding and fusellar growth patterns.
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 39
a basal disc whereas rhabdopleurids and all graptolitesbud serially from the stolon system.
6 – Internal stolotheca in prosicula. 0 – absent; 1 –present (Fig. 2E). A number of taxa show the originof the stolotheca as occurring within the prosicula(e.g. Kozłowski 1971).
7 – Stolon type. 0 – basal disc (as in Cephalodiscus);1 – tubular (Fig. 3B, D); 2 – beaded (Fig. 4D); 3 – un-sclerotized. The state ‘unsclerotized’ is inferred wherethecae are serially interconnected through thecaltubes, but no stolonal material is found preservedwithin the tubarium. ‘Beaded’ refers to the state in‘crustoids’ in which the stolons exhibit regularlyspaced swellings and constrictions (Fig. 4D; Mierze-jewski et al. 2005).
8 – Stolon position. 0 – absent; 1 – embedded inbasal wall; 2 – central; 3 – embedded in upper wall.
9 – Stolon diaphragms. 0 – absent; 1 – present.These structures are well known in Rhabdopleura
(Fig. 3B; Kozłowski 1949) and homologous lateralexpansion of the stolon at the base of the autothecacan be observed in many graptolite taxa (Figs 3D, 4A,D). Although we do have evidence in the form of gra-dients in thecal form, branching structures and pat-terns of damage repair, that unsclerotized stolons wereretained in planktic graptolites (character 7), there isno basis from which one could infer that stolon dia-phragms persisted as unsclerotized structures. Theymay have, but lacking any affirmative evidence wecode them as absent in Rhabdinopora and Anisograp-tus.
10 – Budding type. 0 – basal; 1 – diad (Fig. 3B); 2 –triad (Figs 3E, 4A). This character focuses on the anat-omy of zooid budding – whether from a basal disc (0)or by division of a persistent stolon into two (diad)segments or simultaneously into three (triad). Asnoted above, the multiple absences in Cephalodiscus ofcharacters present in the other taxa examined here
A
D
F
B
E
C
Fig. 4. Stolon development and budding patterns among graptolites. a: autotheca; as:autothecal stolon; b: bitheca; bs: bithecal stolon; g: gra-ptoblast (resting cyst); n: stolon node, sd: stolon diaphragm (base of autothecal cup); ss: stolothecal stolon (axis extending tube); tw: thecalwall. 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 nodeswhere 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. nostolothecal tubes). C, schematic diagram of cross-section of Dictyonema sp. showing shared internal layers of autothecal walls (modified afterUrbanek & 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 divisionof 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 ofAcanthograptus divergens, showing compound stipes with multiple series of thecae (Mid Ordovician, Oland, Sweden, Lund University LO11414t; photo by JM).
40 Mitchell et al. LETHAIA 46 (2013)
complicates character coding. It happens to be thecase that Cephalodiscus zooids both lack serial buddingand form buds basally rather than from an elongatingstolon. We know of no reason to think that these twoconditions are forced correlates or are necessarilyredundant. It is conceivable, for instance, that Cepha-lodiscus zooids could have evolved to bud serially fromthe basal disc by remaining attached and each buddingfrom the next, in sequence but without an interveningstolon, although none are known to do this. Thus, wethink it appropriate to code budding type separatelyfrom serial budding, but we examine the effect ofalternate coding in the section on sensitivity analyses.
Budding in crustoids was not truly triad, but closelyspaced events of diad budding (Fig. 4D, E; Mierzejew-ski et al. 2005). Those species that lack triad buddinghave an irregular number of bithecae in relation toautothecae; that is, they are not regularly paired, orthey lack bithecae altogether. On the other hand, Mas-tigograptus does possess triad budding, but exhibits nosignificant dimorphism between the autothecae andbithecae, which Bates & Urbanek (2002) considered tobe a primitive trait for this taxon. Triad budding isubiquitous within the dendroids and more derivedgraptolites (e.g. Desmograptus, Fig. 4A; see Saunderset al. 2009). For these reasons, we have coded paireddimorphic thecae (character 16) separately from thepresence or absence of conothecae or bithecae (char-acters 26 and 27 respectively).
11 – Stolothecae. 0 – absent: 1 – present. As notedabove, Cephalodiscus lacks interconnected thecal tubesas well as a serially branching stolon system. Cystica-mara possesses a serially branching stolon system, butthe stolon system is not surrounded by continuousthecal tubes (the stolothecae) as seen in all othergraptolites and Rhabdopleura. Rather, the stolon sys-tem is embedded in extrathecal tissue (Fig. 4B) andopen thecal tubes only form at the base of each autot-heca (Kozłowski 1949).
12 – Stolothecae location within thecorhiza. 0 – stolo-thecae absent; 1– encrusting (incorporated in basallayer of encrusting mass of thecal tubes – the thecorh-iza); 2 – on top of thecorhiza: 3 – diverse locationswithin thecorhiza (Fig. 4B); 4 – thecorhiza absent(stolothecae within upright-grown stipes).
13 – Encrusting. 0 – yes; 1 – no. An encrusting habitis inferred from evidence of in situ preservationand ⁄ or presence of a basal membrane.
14 – Erect series of interconnected thecae. 0 – absent;1 – present. This character is coded as absent whenonly individual thecal tubes are erect (Fig. 2G) andpresent where there are erect series of multiple inter-connected thecae (stipes; Fig. 4A, F). Thus, althoughsome cephalodiscids [e.g. C. (Idiothecia) negriscens]have upright bushy colonies, they consist of individual
blind tubes that are not interconnected and hence dif-fer structurally from the thecal series seen in grapto-lites and Rhabdopleura.
15 – Planktic. 0 – no; 1 – yes. Most graptolites andall the pterobranchs show direct evidence of a benthicmode of life from in situ preservation or their encrust-ing habit, or both. The nematophorous graptoloids onthe other hand, exhibit none of the features associatedwith encrusting habits and available taphonomic evi-dence suggests that they could not have lived on orattached to the seafloor and must have acquired aplanktic habit and lived within the overlying watercolumn. Although this character is not strictly a mor-phological feature, we believe it captures an evolution-ary innovation in colony development (larvalmaturation within the water column) that is relevantto graptolite phylogeny.
16 – Paired dimorphic thecae. 0 – absent; 1 – present(Fig. 4A, F). See the discussion of character 10.
17 – Stipe connection. 0 – absent; 1 – anastomosis; 2– dissepiments. This character refers to the presence ofintermittent lateral interconnection between adjacentstipes by either merging of lateral walls (anastomosis)or presence of extrathecal bars laterally connectingstipes (dissepiments).
18 – Upright planar tubarium. 0 – absent; 1 – pres-ent. This character refers to the arrangement of thestipes. It is considered present when stipes arearranged into a single, two-dimensional sheet (thatmay be curved into a conical form) and absent if thestipes are branched in three dimensions, forming abushy or irregular tubarium, or if the tubarium isencrusting and lacking erect stipes.
19 – Thecal construction. 0 – irregular; 1 – tubularwith unshared walls (Fig. 4B); 2 – tubular with shareddorsal walls (Fig. 4C).
20 – Vesicular thecae. 0 – absent; 1 – present(Fig. 4E).
21 – Metathecal ⁄ autothecal isolation. 0 – non-tubular or irregular; 1 – complete or partial (Figs 3B,4A, F); 2 – not isolated (Fig. 3D).
22 – Branch condition. 0 – undefined ⁄ absent(Fig. 3A); 1 – stipes possess single thecal series(Fig. 3D); 2 – compound (Fig. 4F). Compound stipesare present among ‘tuboids’ and acanthograptids thatpossess several thecal series fused and growing along asingle stipe or branch.
23 – Fusellar sutures on erect (autothecal) tubes. 0 –irregular (Figs 2G, 3B); 1 – zigzag (Fig. 3D).
24 – Autothecal coiling. 0 – absent; 1 – present. Insome ‘tuboids’ (e.g. Dendrotubus), the autothecaeshow a spiral coiling of the erect thecal tube (Kozłow-ski 1949).
25 – Closed terminal buds. 0 – absent; 1 – encapsu-lated; 2 – graptoblast (Fig. 4E). See Urbanek (1984)
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 41
and Mierzejewski (2000) for descriptions of grapto-blasts and their possible homology with the closed ter-minal buds observed in Rhabdopleura.
26 – Conothecae. 0 – absent; 1 – present. These arelarge conical thecae irregularly developed in some taxaregarded as ‘tuboids’ (Bulman 1970).
27 – Bithecae. 0 – absent (Fig. 3A, B); 1 – present(Figs 3D, 4A, F). Regularly developed thecae that arediminutive as compared with the autothecae. In sometaxa, autothecae and bithecae are paired (see character10, above).
28 – Spongy extrathecal mass. 0 – present; 1 –absent. Spongy tissue surrounds the thecal walls inCephalodiscus (Fig. 3A) and many encrusting grapto-lite taxa (e.g. Fig. 4B; Kozłowski 1949).
29 – Endocortex. 0 – pseudocortex; 1 – paracortex; 2– eucortex. See Urbanek & Mierzejewski (1984) andreferences cited therein for descriptions of the ultra-structural terms referred to in characters 29, 30 and31.
30 – Ectocortex. 0 – absent; 1 – pseudocortex; 2 –paracortex; 3 – eucortex.
31 – Vesicular sheet fabric. 0 – absent; 1 – present.
Taxa included in the analysis
Our choice of taxa was guided by the desire to use taxafor which most features of the early phases of domicileconstruction, structure and ultrastructure are knownwhile also including representatives of the full range ofhigher taxa recognized by previous analyses. As wementioned previously, Rickards & Durman (2006)included taxa for which very little useful informationbeyond general colony form is available (Sotograptusand Archaeolafoea). In addition, they included threegenera (Flexicollicamara, Thallograptus and Pala-eodictyota) that are redundant with better-known taxa.We have omitted all these taxa from our study. In addi-tion to the taxa included by Rickards & Durman(2006), the graptolites Bithecocamara, Bulmanicrusta,Epigraptus and Kozlowskitubus are known from isolatedcollections and we included these taxa in our analysisas well. Thus, we were able to gather adequate informa-tion for fourteen unique genera of fossil graptolites.
Among extant pterobranchs, we coded tubariumcharacters for representatives of Cephalodiscus andRhabdopleura. Within Cephalodiscus, we formed twocomposite taxa. The first taxon is species of the sub-genera C. (Orthoecus) (such as C. densus Andersson1907; C. solidus Andersson 1907 and C. graptolitoidesDilly 1993) and species of the subgenus C. (Idiothecia)(e.g. C. nigrescens Lankester 1905). All these species(Cephalodiscus OI, hereafter) form pseudocolonialcoenoecia in which the separated individual zooidseach inhabit their own blind tubes. The second taxon
(Cephalodiscus CA, hereafter) comprises species thatform encrusting or branching masses that possess acommon internal chamber that houses many individ-ual zooids (such as Cephalodiscus (Cephalodiscus)gracilis Harmer 1905 and Eocephalodiscus polonicusKozłowski 1949), or that form reticulate masses withno zooecial chambers (such as C. (Acoelothecia) kempiJohn 1931). With these two Cephalodiscus compositetaxa, our analysis includes a total of 17 taxa.
The anatomy of the unique pterobranch Atubariaheterolopha Sato (1936) fits within the range of formexhibited among species of Cephalodiscus; however, inthe single dredge haul from which the 43 known spec-imens were obtained, none were found associated witha tubarium, and Sato inferred that they were originallyfree living and did not form a tubarium. Unfortu-nately, the absence of tubarial characters for Atubariaand the lack of detailed soft-part anatomy for the fos-sil graptolite taxa means that there is no overlap in thecharacter sets available for these two groups. Althoughwe could include soft-part characters for the otherextant pterobranchs (Cephalodiscus and Rhabdople-ura), these data would not bear on our main question:whether or not graptolites and pterobranchs formmutually exclusive groups, and inclusion of Atubariacould not influence the cladistic relationships amonggraptolites and extant pterobranchs.
In addition to the two composite Cephalodiscustaxa, we coded Rhabdopleura compacta Hincks (1880).The development and structural features (includingtubarium ultrastructure) are better known in R. com-pacta colonies (e.g. Stebbing 1968; Dilly 1971, 1973,1975, 1985a, 1986; Stebbing & Dilly 1972; Cavers2005; Urbanek & Dilly 2000; Mierzejewski & Kulicki2003; Sato et al. 2008a,b) than in the other commonlyreported Rhabdopleura species, R. normani Allman(1869), and for our purposes, the differences betweenthem are of no importance. Some early observationsof R. normani branching patterns suggested that theyproduced new zooids behind a persistent terminalbud that was itself not a fully developed zooid (Sche-potieff 1907). This blastozooide inacheve was shownwithin a pointed, sealed tube that otherwise resembledthe other tubes of the colony. As Rhabdopleura zooidseach build their own tubes through the mortaringactivities of their cephalic disc, it is hard to understandhow a terminal bud of the sort described could haveformed the tube that contained it. More recent discov-eries of R. normani have not confirmed those observa-tions – all tubes are of the more usual form withsuccessive budding (Burdon-Jones 1954; Barnes 1977;Dilly 1985b; Halanych 1993). Furthermore, observa-tions of resting cysts in R. compacta (Dilly 1975) and anumber of graptolites (Urbanek 1984; Mierzejewski2000) suggest that the features described in R. normani
42 Mitchell et al. LETHAIA 46 (2013)
by Schepotieff (1907) are likely to have been faculta-tive structures, perhaps dormant buds.
We did not include the recently discovered Galea-plumosus abilus (Hou et al. 2011) in our analysis.Although this organism has been interpreted as theoldest-known fossil pterobranch (lower Cambrian),shows soft-bodied preservation and evidence ofgrowth in a fusellar tube, nothing is known of aboutwhether the organism was colonial or what the colonyorganization might have been like (if indeed it wascolonial), whether it had stolonal budding, or a sicula.The ultrastructure of the tube walls is unknown aswell. Therefore, it was not possible to include thistaxon in a way that adds to our analysis.
Cladistic methods
We conducted three sets of analyses, each aimed atachieving slightly different goals. First, we conductedan unrooted parsimony analysis to assess the relation-ships between extant pterobranchs and fossil grapto-lites. Because no other hemichordates are tubebuilders and because we lack virtually all soft-partinformation (not to mention genetic data) about fossilgraptolites, there is no usable outgroup by which toroot the analysis and thus to polarize character trans-formations for this part of our study. The Enteropn-eusta are widely regarded as the next most closelyrelated member of the Hemichordata, but they possessnone of the tube-building features of the pterobranchsand graptolites (Cameron 2005). More distantlyrelated taxa also offer no help. Thus, we first presentthe results of an unrooted analysis and discuss its sig-nificance in light of what is known about the relation-ships among extant deuterostomes from molecularphylogenetic studies. It is important to recall in thiscontext that an outgroup is not necessary to deter-mine the basic tree structure in Wagner parsimonyanalysis (which we employ in all our analyses) orother models of character change so long as charactertransformations are symmetrical: The change fromstate a to state b adds to tree length in the sameamount as a change from b to a (see Felsenstein 2004,pp. 4–8, 32–34). Rather, the role of outgroup rootingis to fix the direction of change: to reveal the relativetemporal order of taxon divergence, patterns ofmonophyly and the sequence of change in charactersover time. For our initial analysis, we asked a morelimited, non-directional question: Do the Cephalodis-cus taxa and Rhabdopleura form a group to the exclu-sion of graptolites or do one or more of these taxanest within the graptolite group? We examined thisquestion via a branch-and-bound search in PAUP4.10b (Swofford 2001). Characters were unweightedand unordered.
In our second set of analyses, we employed theresults of our unrooted analysis and constrained thetwo Cephalodiscus taxa to be a paraphyletic outgroup.This rooted tree differed in several important waysfrom the results presented in Rickards & Durman(2006) and so we conducted a set of experiments inPAUP 4.10b that were modelled on the methods anddata described in Rickards & Durman (2006) in aneffort to reproduce their results. In these comparisons,we employed bootstrap resampling analyses with 1000replicates and random branch addition sequence toassess group support.
We followed up these preliminary analyses with athird and final set of analyses in TNT (Tree Analysisusing New Technology, see Goloboff et al. 2008b).The goals of these analyses were two-fold: first, to testthe effects of reweighting characters by their consis-tency with the most-parsimonious trees (MPTs) and,second, to employ TNT’s refined set of measures ofgroup support and its enhanced methods for under-standing and comparing trees. Research suggests thatweighting by character consistency improves phyloge-netic results (Goloboff et al. 2008a), but this shouldbe done cautiously. Unlike the iterative, ‘reweight bycharacter consistency’ option in PAUP 4.10b, theimplied weighting options in TNT implement a searchthat seeks to simultaneously maximize character con-sistency and minimize tree length (Goloboff 1993,1997). This joint optimization reduces the risk posedin sequential optimization that the initial MPT, whichwas obtained without regard to character consistency,will lead on subsequent reweighting steps to a localminimum in tree length rather than the globally opti-mal solution. TNT provides two alternative methodsfor weighting by character consistency: implicitweighting (IW) and self-weighted optimization (SL).IW applies weights based on the behaviour of thecharacter over all its transitions, whereas SL estimatesthe weights separately for all possible state transitions.Thus, for SL optimization, gains and losses of a partic-ular character might be weighted differently if lossesare frequent and gains rare. Both IW and SL weightingrequire setting a parameter called K, which determineshow strongly the optimization discounts the value ofhomoplastic characters (Goloboff et al. 2008a,b;Mirande 2009). We employed IW weighting and var-ied K over a range from 3 to 12 in a series of runs, asthis was a range over which Mirande (2009) foundnoteworthy effects in his analyses.
Several measures of branch support are imple-mented in TNT. Bremer support measures the sup-port for each node in terms of the number ofadditional steps in tree length that arise from collaps-ing that node and is equal to F–C, where F is the num-ber of characters in favour of that clade and C is the
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 43
number of conflicting characters. Relative Bremersupport measures the relative fit difference, or morespecifically, (F–C) ⁄ F (Goloboff & Farris 2001). Con-sidered together, these values reveal both the amountof evidence and the degree to which that evidencepoints toward a particular hypothesis. Resampledgroup support is similar to bootstrap support, but asimplemented in TNT involves a symmetricresampling approach that is said to be less biased thanother approaches (Goloboff et al. 2003). Group sup-port also may be expressed as a relative degree ofgroup support, the GC statistic of Goloboff et al.(2003), which for a particular node, is equal to resam-pled group frequency minus that of the next most fre-quently encountered group for that set of taxa.
Results
Do pterobranchs and graptolites form mutuallyexclusive groups?
Here, we test the hypothesis that pterobranchs andgraptolites are each monophyletic groups and sistertaxa to one another. As we noted above, no outgroupexists that builds a domicile with even a few charactersin common with the tubarium of pterobranchs andgraptolites. Consequently, we first consider thehypothesis in its more general unrooted form: ptero-branchs and graptolites form mutually exclusivegroups. This hypothesis includes the relationship pos-ited by Bulman (1970; pterobranchs as sister to grapt-olites, Fig. 5B) as well as that proposed by Rickards &Durman (2006; pterobranchs as paraphyletic ancestralgroup to graptolites, Fig. 5C). As unrooted trees areless familiar than rooted trees, an explanation of thelogic of this test may be helpful. If Cephalodiscus andRhabdopleura form a clade exclusive of graptolites,and vice versa, each group will be clustered togetheron 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 viceversa. A, an unrooted tree consistent with this hypothesis, shownwith three possible rooting locations corresponding to the rootedtrees shown in B–D. B, rooted cladogram that corresponds to thephylogenetic relations in Bulman (1970). C, rooted cladogram thatcorresponds to the phylogenetic interpretation of Rickards & Dur-man (2006). D, a third alternative tree. E, G, two unrooted treesthat contradict the hypothesis. F, H, two possible rooted trees cor-responding to the unrooted trees in E and G. I, unrooted treeresulting from our first round analyses (see text), showing the samegeometry among extant pterobranchs and graptolites as that in Gand 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.
44 Mitchell et al. LETHAIA 46 (2013)
pterobranchs linked directly to one another and nonewill occur between the two graptolites: either groupcould be pruned off the tree without removing amember of the other group. This topological condi-tion remains the case in all of the alternate rooted ver-sions irrespective of which taxon is chosen asprimitive in the rooted trees (e.g. Fig. 5B–D).Although some outgroup choices cause an extantpterobranch to share a more recent common ancestorwith a graptolite than with the other living ptero-branchs (e.g. Fig. 5C), this does not contradict thehypothesis as the extant pterobranchs always form apterobranch-only paraphyletic group ancestral toa monophyletic graptolite clade. Even if one picks agraptolite as the outgroup for trees of this form(Fig. 5D), extant pterobranchs become a derived cladewithin, but still exclusive of, graptolites: in no case is agraptolite nested among the extant pterobranchs.Other possible nearest-neighbour relations on unroot-ed trees do contradict the hypothesis, however (e.g.Fig. 5E, G): an extant pterobranch is nested betweenthe graptolite taxa (and vice versa) and no single prun-ing cut can remove all of one group without alsoremoving a member of the other group. In this case,as the rooted trees illustrate (Fig. 5F, H), any ptero-branch clade that excludes graptolites must be poly-phyletic (or vice versa, depending on the rooting).
Our unrooted simple parsimony analysis produced20 equally most-parsimonious trees (MPTs), anensemble consistency index (CI, excluding non-parsi-mony informative characters) of 0.814 and an ensem-ble rescaled consistency index (RC) of 0.718. Anunrooted majority-rule consensus of the 20 equallymost-parsimonious trees for this analysis is shown inFigure 5I. Differences among the MPTs involved dif-ferences in the branching order of Acanthograptus andKoremagraptus among the ‘dendroids’ and the posi-tion of Kozlowskitubus and Epigraptus among theencrusting graptolites. As these differences did notinvolve radical differences in the basic structure of thetree, the strict and majority rule trees fairly representthe state of our understanding of the phylogeneticrelations among graptolites. We explore the causesand meaning of these uncertainties in the relationshipsamong graptolites more fully below.
The unrooted tree reveals that the extant Rhabdo-pleura nests among encrusting fossil graptolites, sepa-rately from the extant Cephalodiscus (Fig. 5I). Thegraptolite Cysticamara occupies a position betweenCephalodiscus and Rhabdopleura in every MPT that weobtained, regardless of any differences in other taxaincluded, choices of character weighting or outgroup.Cysticamara occupies this position because it possesses(among other things) individual zooidal dwellingtubes (autothecae) with erect, isolated apertures and
these autothecae are interconnected by a sclerotizedstolon with stolon diaphragms, in common withRhabdopleura and the other graptolites. However, incontrast to Rhabdopleura and the other graptolites,Cysticamara autothecae are separate closed tubes thatare imbedded in spongy, extrathecal tissue throughwhich they are connected only by their thin stolonsand not by distinct stolothecal tubes. Thus, Cysticama-ra is intermediate in its characters between the twopterobranch groups. Accordingly, the unrooted treeexhibits the same geometry among extant ptero-branchs and graptolites as that in Figure 5G, H andconsequently, contradicts the hypothesis that grapto-lites form a clade exclusive of Cephalodiscus and Rhab-dopleura.
Rooting the pterobranch–graptolite tree
Recent phylogenetic analyses of hemichordates andrelated taxa based on 18S ribosomal DNA (Cameronet al. 2000; Cannon et al. 2009) provide importantconstraints on the range of possible relations amongthe study taxa. First, these results suggest that the spe-cies of Cephalodiscus form a clade that is sister toRhabdopleura (Fig. 1). In addition, the relationshipsamong Cephalodiscus species indicate that the one-zooid to one-tube habit of colony construction com-mon to the C. (Orthoecus) and C. (Idiothecia) specieswas derived independently from that in Rhabdopleurarather than being an intermediate state between com-mon tube-dwelling Cephalodiscus and the conditionof Rhabdopleura and graptolites.
The results of previous phylogenetic analyses ofdeuterostomes, including many hemichordates basedon both morphological and genetic data sets, as dis-cussed above, indicates that pterobranchs are derivedrelative to the enteropneusts (Fig. 1) and that secre-tion of a collagenous tubarium by use of the special-ized cephalic disc is a synapomorphic feature of thePterobranchia (Cameron 2005). Similarly, the seriallybudded colony, stolon system and a larvally producedprosicula, which are shared features of Rhabdopleuraand graptolites, also must be derived features relativeto those of non-pterobranch hemichordates. There-fore, the results of our unrooted analysis require thateither: (1) the larvally produced prosicula, seriallybudded colony and stolon system are also all synapo-morphic to the Pterobranchia + graptolites and sub-sequently were lost in Cephalodiscus; or (2) theseshared characters unite Rhabdopleura and the grapto-lites in a clade that excludes Cephalodiscus. The laterhypothesis is substantially more parsimonious thanthe former. A constraint tree that forces Cephalodiscusand Rhabdopleura to be a monophyletic outgroup (i.e.Fig. 5B) is two steps longer (Table 3) than a tree in
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 45
which only Cephalodiscus is the outgroup and Rhabdo-pleura is nested among graptolites (Fig. 5H, I). Takentogether, these features of hemichordate phylogenyargue for rooting the pterobranch–graptolite tree atthe node between Cephalodiscus and all the other taxa(Fig. 6). The remainder of our analyses are based onthat rooting.
Rooted phylogenetic analysis of the Graptolithina
Branch and bound searches of our dataset using TNTproduced the same tree topologies as we obtained inthe earlier analyses, as expected, but provided addi-tional opportunities to gauge the effect of choicesabout character weighting, assess group support andto explore the causes of the two major polytomies thatare present in the strict consensus tree (Fig. 6A).Alternate versions were run with either Cephalodiscus-CA (zooids of which inhabit shared common cham-bers within the tubarium) or Cephalodiscus-OI (zooidsof which inhabit separate individual tubes) compositetaxa employed as outgroups. Both outgroup experi-ments resulted in precisely the same results. Conse-quently, we show the two Cephalodiscus taxa on theresulting tree as part of an unresolved basal polytomywith the ingroup (Fig. 6). These analyses produced 12MPTs, were 65 steps in length and had an ensembleCI of 0.633 and RI of 0.707. The strict consensus ofthese trees results in a polytomy that contains Rhabdo-pleura along with six other taxa that have generallybeen regarded as crustoid and tuboid graptolites, aswell as a clade that consists of Mastigograptus, den-droids and graptoloid graptolites.
Inspection of the alternative trees reveals that thephylogenetic location of Epigraptus and Dendrotubusis very uncertain. Both taxa exhibit several featuresthat must be regarded as homoplastic under any inter-pretation of the evolutionary history of this group.
This can be seen best by excluding Epigraptus andDendrotubus from the analysis, each in turn. Exclud-ing Epigraptus (Fig. 6B) results in unambiguous reso-lution of the basal portion of the graptolite clade. Theclade containing all graptolites + Rhabdopleura (Node1, Fig. 6) in that case has Bremer support of 7 and rel-ative Bremer support of 88% (7 of 8 characters sup-port this clade), and resampling group support of99%. No other hypothesized group appears in theresampling set with a frequency of more than 1%.Thus, we may have a high degree of confidence in thereliability of this finding. Cysticamara emerges as sisterto Rhabdopleura + all other graptolites, with a Bremersupport of 2, relative Bremer support of 50% (avail-able characters support this clade 50% more than theysupport alternative clades) and strong resampling sup-port (79% with no other group found at more than11% frequency). Several additional clades emerge withmoderate to strong support including the clade con-sisting of Kozlowskitubus and the more derived grapt-olites (Node 2, Fig. 6B), which we propose to namethe Eugraptolithina (see Proposed Phylogenetic Clas-sification, below). The effects of excluding Epigraptusarise because it exhibits an inflated larval vesicle, likethat in Rhabdopleura (which this analysis suggests is aprimitive condition for the group). On the otherhand, Epigraptus also has bithecae, which Rhabdople-ura does not possess, but which are widely sharedamong more derived graptolites (crustoids, Bithecoca-mara and the eugraptolithines). These conflicting sim-ilarities cause Epigraptus to join the tree with equallevels of support either above or below Rhabdopleura,and thus contribute to the large basal polytomy seenin the all-taxon strict consensus tree (Fig. 6A).
Excluding Dendrotubus from the analysis likewiseleads to much better resolution of the phylogeneticrelationships among the other graptolites (Fig. 6C);however, in this case, the effect is among the
Table 3. Fifty percentage majority rule bootstrap percentage values, tree lengths, consistency index (CI) and retention index (RI) obtainedfor five sensitivity analyses. Column 1 represents the main analysis. Column 2 represents the traditional view that Rhabdopleura belongs in amonophyletic 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. Column3 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 ofkey structures in Cephalodiscus taxa, were down-weighted to 0.25 and 0.20 respectively. In Column 5, the same two sets of characters weretreated as inapplicable to the Cephalodiscus taxa and coded as unknown. In Column 6, two interrelated thecal characters (17, 28) were treatedas one character by down-weighting each to 0.5. See text for further discussion of these sensitivity analyses.
CladeMainanalysis
Rhabdo-pleura asoutgroup
Taxaomitted
Sicular &budding charactersre-weighted
Sicular & buddingcharactersinapplicable
Thecalcharactersre-weighed
1. Subclass Graptolithina 100 N ⁄ A 100 98 67 1002. Infraclass Eugraptolithina 80 76 93 61 53 813. Div. Graptoloida 77 75 76 66 85 77Tree length 80 82 84 50.95 59 79.5CI 0.65 0.64 0.63 0.59 0.76 0.64RI 0.79 0.78 0.66 0.66 0.84 0.78
46 Mitchell et al. LETHAIA 46 (2013)
eugraptolithines (compare Fig. 6B, C). In the absenceof Dendrotubus the phylogenetic relationships amongthe eugraptolithine taxa can be fully resolved althoughsupports for the several nested groups that emerge arenot very high in most cases. Several new features, par-ticularly the appearance of the helical line in the prosi-cula, presence of an internal stolotheca within thesicula and emergence of eucortex, appear in this partof the tree, but, unfortunately nothing is known aboutthese features (or any other aspect of its sicula or tubeultrastructure) in Dendrotubus. Thus, although itsposition on the cladogram suggests that Dendrotubusis a member of the Eugraptolithina, its precise phylo-genetic position within the clade is unclear.
In addition to analyses in which all characters wereweighted equally, we ran a set of experiments usingboth 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 Kfrom 3 to 12 produced no effect on the results in ourcase. As expected, reweighting by character consistencyresulted in fewer equally good alternative taxon place-ments and better resolution in the strict consensus asthis approach reduced the weight of characters thatwere inconsistent with the best-fit trees. There wereonly three MPTs, which had CI > 0.81 and RI > 0.86,under the IW optimization. This effect was even morepronounced for self-weighting. Inspection of groupsupports for the additional nodes recovered in boththe IW and SL solutions revealed that these trees wereover-resolved, however. When nodes with low groupsupport were collapsed the resulting trees all exhibitedthe same topologies as the strict consensus of thoseobtained in the unweighted analyses (Fig. 6A).
Sensitivity analyses
We conducted several additional tests of whether thebranching pattern among the various pterobranchclades in our analysis (Figs 5, 6) was robust. First, we
A
B
C
Fig. 6. Cladograms depicting the phylogenetic relationships amongliving pterobranchs (Rhabdopleura and Cephalodiscus) and fossilgraptolite taxa together with suggested higher taxonomy. Trees A–C obtained from analysis of the 32-character morphological dataset using TNT and equal character weighting, but with differenttaxa included. Numbers above branches are characters for whichstates are apomorphic at that branch ⁄ Bremer support values ⁄ rela-tive Bremer supports. Values below the nodes are symmetric re-sampling group frequencies ⁄ GC values. All three analysesrecovered strong support for Clade 1 (Graptolithina) and Clade 3(Graptoloida). Support for Clade 2 (Eugraptolithina) depends onthe taxa included in the analysis as a result of poor control on thephylogenetic position of Epigraptus and Dendrotubus (see text fordiscussion) A, strict consensus of 12 equally parsimonious treesobtained from the full 17 taxon set. B, strict consensus of fiveequally parsimonious trees based on an analysis that excluded Epi-graptus. C, strict consensus of two equally parsimonious trees basedon an analysis that excluded Dendrotubus.
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 47
constrained Cephalodiscus and Rhabdopleura to amonophyletic outgroup status (Table 3, Column 2)and found that the 50% majority rule consensus treelength increased by two steps. We then forced the treeinto the Rickards & Durman (2006) topology and toavoid taxon bias we removed four taxa from the anal-ysis (Bithecocamara, Bulmanicrusta, Epigraptus andKozlowskitubus) that were not included in their taxonset, and the tree topology was four steps longer (84 vs.80 steps; Table 3, Column 3) than our main analysistree.
To comply with the assumptions of parsimony,characters should be independent. We ran three fur-ther sensitivity analyses to test the effect of groupingcharacters that may be perceived as dependent. Inthe first, two sets of characters were down-weighted:1, characters 0–2, and 6, which describe various fea-tures of the sicula (Cephalodiscus lacks a sicula andall characters associated with it) were reduced to arelative weight of 0.25 so that collectively theycounted as much as a single character; and, 2, charac-ters 5 and 7–10, which describe budding of thezooids, including the stolon and features associatedwith it (again Cephalodiscus lacks stolons), weredown-weighted to 0.20, and then the analysis wasrerun (Table 3, Column 4). This resulted in a minorreduction in bootstrap support for the Graptolithina(to 98%) and somewhat greater reductions for theEugraptolithina and Graptoloida, but did not changethe topology of the strict consensus tree, which stillcontains all the major clades recognized in our mainanalysis. Thus, our result is not significantly affectedby the potential correlations among characters thatare absent in Cephalodiscus. We also tested whethersimply retaining equal weights for characters 1, 2, 6–10 and coding their states in Cephalodiscus asunknown rather than absent (i.e. inapplicable; seeKitching et al. 1998; p. 27–30), would alter ourresults (Table 3, Column 5). Once again, we obtainedthe same tree structure as previously, although in thiscase, resampling support for the Graptolithina was67% rather than 98–100% as seen in the other tests.This reflects the reduced amount of informationretained for seven phylogenetically informative char-acters in this coding scheme. In the third test of char-acter independence, two characters that describethecal characteristics: paired dimorphic thecae (char-acter 17), and bithecae (28), were also combined intoa single character by down-weighing each to a weightof 0.5 (Table 3, Column 6).
In all the sensitivity analysis (except the forced tree,Column 2) the Graptolithina, Eugraptolithina andGraptoloida were well supported. In no case did thetree topology change and thus the hypothesis thatGraptolithina, Eugraptolithina and Graptoloida are
monophyletic clades, and sister group to Cephalodiscusholds over this range of experimental manipulations.
Discussion
Interpretation of tree topology
The most profound finding is the discovery that Rhab-dopleura nests within the clade that includes all grapt-olites. Therefore, the Graptolithina must be includedwithin the Class Pterobranchia (Fig. 6). These resultsclearly distinguish the clade that includes all grapto-lites plus Rhabdopleura from the Cephalodiscus clade.Previous hypotheses have suggested that graptolitesare sister to the pterobranchs (Beklemishev 1951,1970; Bulman 1970; Rickards & Durman 2006). Ourresult, which is well supported by all the group sup-port measures and by sensitivity analyses, suggests thatgraptolites are derived within the pterobranch clade,and shows a phylogenetic tree topology of hemichor-dates that has not been suggested by other authors.The clade Rhabdopleura + graptolites shows 99–100%resampling support, the strongest support of any cladein the analysis (Fig. 6). The majority rule consensustree shows Rhabdopleura arising from within thisclade, not at its base (Fig. 6). This tree topologytogether with the distribution of shared characters(see rooting discussion above), suggests that Rhabdo-pleura is derived relative to Cephalodiscus and that thezooidal structure of graptolites may be inferred fromthe study of Rhabdopleura. This finding is particularlyimportant because, as we noted above, none of thepartially preserved pterobranch or graptolite zooids(e.g. Durman & Sennikov 1993 or Rickards et al.2009), record sufficient detail to provide much insightinto zooid anatomy other than that they may beroughly the same size relative to the diameter of theirtubes, as are Rhabdopleura zooids. The similarities toCephalodiscus are numerous and interesting in thisconnection as well, including their tube and spine-building mechanisms, but we disagree with Dilly(1993) and Rigby (1993) that these features, which areclearly plesiomorphic relative to the Graptolithina,make Cephalodiscus a graptolite (see also Urbanek1994).
There has been a reduction in zooidal size fromCephalodiscus to Rhabdopleura and accompanying thisreduction in size has been a simplification of bodyplan (Cameron 2005). The evolution of Rhabdopleurahas been characterized by a loss in gill pores, of oneof two gonads as well as a loss of the red band on thecephalic shield and a reduction in the number of feed-ing arms. If comparison of thecal tube diameter canbe taken as any indication of zooidal body sizes, then
48 Mitchell et al. LETHAIA 46 (2013)
this suggests that graptolites had body sizes at least assmall and possibly derived as those of Rhabdople-ura (Rigby & Sudbury 1995). In addition, the aper-tural form of many graptolite thecal tubes suggeststhat, like Rhabdopleura, graptolite zooids may havepossessed a single pair of feeding arms, althoughpaired feeding arms have been reported from male C.sibogae (Harmer 1905) zooids. If the Early Cambriantube-dwelling animal, Galeaplumosus (Hou et al.2011) is a pterobranch, then it also presents an exam-ple of a relatively large-bodied pterobranch with onlyone pair of feeding tentacles.
In keeping with these phylogenetic results, we sug-gest that Rhabdopleura should be regarded as anextant graptolite (Fig. 6). One might argue converselythat graptolites should be regarded as extinct rhab-dopleurids. That alternative is equally consistent withthe tree and is more consistent with much recent taxo-nomic practice (e.g. Donoghue 2005; de Queiroz2007). That approach, however, would subsume adiverse group with a rich and long-studied fossilrecord within a relatively little-studied taxon that hap-pens to include four extant species (Dawydoff 1948;Stebbing 1970). As the graptolite fossil record under-pins a large portion of the Early Palaeozoic timescaleand is used by many earth scientists and biologists notfamiliar with the details of deuterostome phylogenyand taxonomy, we conclude that the interests of taxo-nomic stability and clarity of scientific communica-tion will be best served by retaining the nameGraptolithina for this group.
Comparison to the Rickards & Durman results
The reasons for the differences in results between ouranalysis and those of Rickards & Durman (2006) aredifficult to discern because they did not include thedata matrix in their publication. A matrix is present inthe 1992 doctoral dissertation of P. Durman, whichalso included a set of phylogenetic trees that appear tobe the same as those in Rickards & Durman (2006).We analysed that matrix following the methodsdescribed in Rickards & Durman (2006), but we were
unable to duplicate the Rickards & Durman (2006)results (Table 4, Fig. 7). On the basis of a series ofexperiments with this matrix, we infer that the differ-ences between their results and ours arise primarilyfrom differences in the character coding schemes andthe information included in the analyses. First, itappears from the discussion in Rickards & Durmanthat they coded many of the characters related to thesicula and budding patterns in Rhabdopleura andCephalodiscus as being the same. The informationavailable to us, however, suggests that Cephalodiscuslacks a sclerotized sicula and also lacks serial, stolonalbudding. These features are among the key synapo-morphies for the Rhabdopleura + graptolites clade(Fig. 6). In addition, according to Rickards & Durman(2006), the synapomorphies uniting all graptolites,but excluding Rhabdopleura are ‘heteromorphic fuselliand cortex,’ ‘connected wavy fusellar microstructureto the fuselli’ and ‘helical line potentially present.’However, several graptolite taxa share with Rhabdople-ura the presence of a vesicular prosicula (Fig. 2G) thatlacks a helical line (Fig. 2A–C; Chapman et al. 1996).In addition, there are no important differences in thefusellar and cortical microstructure between Rhabdo-pleura and some graptolites (Mierzejewski & Kulicki2003). Changing the coding for these characters pro-duces a tree topology that is more similar to ourresults than in Rickards & Durman’s (2006) strict con-sensus tree. Most importantly, that change in charac-ter coding causes Rhabdopleura to be relocated withinthe Graptolithina, just as in our results, and so alsolends support to our rejection of the hypothesis thatpterobranchs and graptolites are mutually exclusiveclades.
Relationships among the major tube-buildinghemichordates
The results of our study permit a re-evaluation ofhemichordate classification (Fig. 6A). In addition todifferences in the relationships between pterobranchsand graptolites, our results also differ from those ofRickards & 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 charac-ter set. See text and Figure 7 for further information about experiments.
AnalysisThispaper
Rickards &Durman
Redo fromDurmanmatrix
Nearestduplicateto R&D
D matrix;commontaxa
Updated Dmatrix; R&Dunique taxa
Updated Dmatrix; all taxa
Number of taxa 17 22 22 22 13 19 21Tree length 61 73 66 72 65 70 74Number of MPT’s 4 8 524 8 12 3 335CI* 0.7455 (0.653) 0.6984 0.6471 0.6724 0.6885 0.6714RC 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.
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 49
united the encrusting camaroids and crustoids as sis-ter taxa in a distinct clade that was distinguished bypossession of: 1, vesicular thecae; and, 2, encrustingbithecae and autothecae. The crustoid and camaroidtaxa did not form a clade exclusive of other graptolitesin our analyses. Rather, Cysticamara occupies a posi-tion below the common ancestor of Rhabdopleura andother encrusting graptolites, removed from Bithecoca-mara and Bulmanicrusta, which do however, form aweakly supported clade. This suggests that vesicularthecae may have evolved repeatedly among encrustinggraptolites. Uniquely among graptolites, Cysticamaralacks stolothecae and our analysis suggests this may bea primitive trait. Stolothecae (or their anatomicalequivalent) are present in Rhabdopleura and mostother graptolites. The specimens of Cysticamara alsolack bithecae, as do colonies of Rhabdopleura.
This study supports the suggestion of Mierzejewski(2001) that the cysticamarids may be primitive withrespect to the tuboids and that some characters mayhave appeared in a mosaic fashion among theseencrusting graptolite groups. Further resolution of therelationships among these primitive graptolites willrequire more information about the proximal devel-opment and ultrastructure of a wider range of taxa,particularly forms that appear to show a mosaic of
tubarial features such as Camarotubus (Mierzejewski2001). We did not include these taxa due to a lack ofavailable information concerning either the earlystages of tubarium development or ultrastructure, orboth.
As in the case of crustoids and camaroids, Rickards& Durman (2006) found ‘tuboids’ to be a distinctclade, whereas our results suggest that the ‘tuboids’form a paraphyletic group from which all morederived graptolites emerge. In addition, our resultssupport suggestions by Maletz et al. (2005) that Epi-graptus should not be regarded as a tuboid graptolite(Kozłowski 1949; Bulman 1970), but appears insteadto be closely allied with Rhabdopleura (Fig. 6).
We have only included one of the taxa that Ric-kards & Durman (2006) included in the Dithecoidea(Mastigograptus), so it was not possible to test theirhypothesis that the dithecoids are a clade that includesMastigograptus or whether Mastigograptus should beassigned to its own order, the Mastigograptida (Bates& Urbanek 2002). Our results provide only weak sup-port for previous suggestions that the acanthograptidsmay be a clade sister to the dendrograptids + grapto-loids (Chapman et al. 1996; Rickards & Durman2006) and also indicate that the dendrograptids, astraditionally recognized, are paraphyletic. The
Fig. 7. Comparison of our results with those of Rickards & Durman (2006). Emulation of the results of Rickards & Durman are based onmatrix from (Durman 1992) and several experiments to gauge the effect of differences in the choice of taxa, number of taxa and characterinformation given their character set (all matrices available upon request). A–C are strict consensus trees and D is a 50% majority rule con-sensus 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 theRickards & Durman result. Settings required to produce this result are described in text. B, strict consensus tree from analysis of 13 taxashared between the Rickards & Durman taxon set and those employed here. C, strict consensus tree from analysis based on 19 unique taxa inthe Rickards & Durman taxon set (omitting three redundant taxa: Thallograptus, Palaeodictyota, and Flexicollicamara) and with some charac-ter states updated, primarily to reflect new information about Rhabdopleura. D, tree presenting the 50% majority rule consensus of MPT’sobtained 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.
50 Mitchell et al. LETHAIA 46 (2013)
graptoloids, however, appear to be monophyletic, inkeeping with several previous suggestions (e.g. Erdt-mann 1982; Fortey & Cooper 1986).
Perhaps the most exciting outcome of this study isthat it provides a hypothesis of living graptolites,rhabdopleurids, from which we can better understandzooidal structure of extinct groups and observe tubeconstruction in hemichordates. To date, there existsalmost no information on the mechanics of tubebuilding in living hemichordates (Dilly 1986). Perti-nent questions include, from what part of the cephalicshield is the tube material secreted? How is the darkrind of the sclerotized stolon (pectocaulus) secretedand what is the nature of its connection to the con-tractile stalk (gymnocaulus) of the zooids? Are theouter fuselli secreted before or after the inner fibre lay-ers? What role does the proximity to other zooids playin new tube formation? How does water flow, particu-larly in the benthic groups, affect tube ontogeny andastogeny? How does Rhabdopleura exit the larval sicu-la? These are just a few of the questions, long thoughtintractable by the graptolite community that mayfinally be attained by a closer examination of livingrhabdopleurids.
Finally, if our phylogenetic hypothesis is correct, itstrongly supports the contention that the tubarium ofrhabdopleurids provides a homologue for the grapto-lite tubarium in terms of its mode of constructionand relationship to the zooids. If this is so, then thegraptolite tubarium was an engineered domicile builtby the zooids from secretions of the cephalic shieldentirely externally to the body of the animals – moreakin to the honeycombs of bees than an echinoid test,for example. These results directly contradict previoussuggestions (Bulman 1970; Kirk 1972; Urbanek 1978;Bates & Kirk 1986) that extrathecal tissue played arole in secretion of the tubarium walls and theirderivative structures (such as spines, retiolitid lists,etc.). This idea has lately lost much of its appeal,especially as the fundamental unity of pterobranchtubarium structure and ultrastructure (including thatof fossil graptolites) has become clearer (e.g. Urbanek1994; Mierzejewski & Kulicki 2003). Therefore, wetake this opportunity to emphasize that the sicularand thecal wall material and its derivatives cannot beregarded in any biological sense as ‘periderm’ andrecommend that usage of this term in reference tograptolite tubarial material, as has traditionally beendone, be abandoned. This material can simply bereferred to as ‘tubarium’ or ‘wall material’ dependingon the specific context.
In contrast, it is probable that the black stolon(pectocaulus) of rhabdopleurids and the benthic grap-tolite taxa was secreted by the soft tissue of the stolon(Urbanek & Dilly 2000) and if true, this materialT
able
5.M
atri
xo
fta
xa,
char
acte
rsan
dst
ates
emp
loye
db
yD
urm
an(1
992)
,m
od
ified
toin
clu
de
on
lyth
ose
taxa
sho
wn
inth
ep
hyl
oge
net
ican
alys
iso
fR
icka
rds
&D
urm
an(2
006)
wit
hfo
ur
cod
ing
erro
rsco
rrec
ted
(bo
ldfa
ce).
01
23
45
67
89
1011
1213
1415
1617
1819
2021
2223
2425
2627
2829
3031
3233
Rh
abd
op
leu
ra0
00
?0
00
21
?0
01
10
00
00
00
00
00
00
00
10
00
0C
eph
alo
dis
cus
?0
0?
00
02
0?
00
00
?0
?0
01
?1
00
00
00
01
00
00
Den
dro
grap
tus
11
?0
00
11
20
11
12
11
21
02
20
00
00
00
00
20
01
Dic
tyo
nem
a1
??
?0
01
12
?1
11
21
12
10
22
02
01
00
00
02
00
1R
hab
din
op
ora
2?
1?
01
21
20
11
12
11
32
?2
20
2?
10
00
00
20
01
An
iso
grap
tus
2?
1?
01
21
20
11
3?
11
32
12
20
01
00
00
00
20
01
Cal
yxd
end
rum
21
0⁄1
00
12
12
?1
13
?1
13
21
22
00
10
00
00
02
00
1A
can
tho
grap
tus
1?
??
00
11
??
11
12
10
21
01
21
00
00
00
00
20
01
Ko
rem
agra
ptu
s?
??
?0
01
1?
?1
11
21
02
10
12
11
00
00
00
02
00
1P
alae
od
icty
ota
??
??
00
11
??
11
12
10
21
0?
21
10
00
00
00
20
01
Th
allo
grap
tus
??
??
00
11
??
11
12
10
21
0?
21
00
00
00
00
20
01
Tu
bid
end
rum
0?
??
?0
01
??
11
1?
01
10
02
21
00
00
?1
10
20
01
Ret
icu
logr
aptu
s0
??
??
00
1?
?1
13
10
11
00
22
01
00
11
??
02
00
1C
yclo
grap
tus
0?
??
?0
01
??
11
1?
00
10
02
21
00
01
??
?0
10
01
Den
dro
tubu
s0
10
01
00
1?
?1
11
?0
11
00
22
00
00
1?
1?
01
00
1C
ysti
cam
ara
0?
??
00
01
1?
11
1?
00
00
00
30
00
00
00
01
11
10
Flex
ico
llica
mar
a0
??
?0
00
11
?1
11
?0
00
00
03
00
00
00
00
11
11
0W
iman
icru
sta
0?
??
00
02
1?
0⁄1
0⁄1
21
10
00
00
30
00
00
00
00
11
20
Elle
sicr
ust
a0
??
?0
00
21
?0
⁄10
⁄12
11
00
00
03
00
00
00
00
01
13
0A
rch
aeo
lafo
ea?
??
?0
01
??
?1
??
?0
02
10
11
00
00
00
00
1?
00
?M
asti
gogr
aptu
s?
??
?0
01
?0
?1
1?
?1
02
10
01
00
00
00
00
0⁄1
20
0?
Soto
grap
tus
??
??
00
1?
0?
1?
??
00
21
00
10
00
00
00
01
?0
0?
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 51
could be correctly referred to as periderm. Whether ornot the graptoloid nema was secreted in a way analo-gous to the rhabdopleurid black stolon or from thecephalic shield in the same manner as cephalodiscidspines (Dilly 1993) and the other tubarium wall mate-rial remains unsettled (Hutt 1974; Mitchell & Carle1986; Bates 1987; Rickards 1996).
Proposed phylogenetic classification
Phylum Hemichordata Bateson 1885;Class Pterobranchia Lankester 1877
Definition. – The least inclusive clade containingRhabdopleura normani Allman 1869 (in Norman1869), and Cephalodiscus dodecalophus M’Intosh 1887.
Discussion. – This is a node-based definition in whichboth specifiers are extant; it identifies the pterobranchcrown clade. Animals in this clade include both colo-nial and pseudocolonial forms. They share a three-part body organization with a U-shaped gut, a bilater-ally symmetric, multi-armed suspension feeding organthat is derived from the mesosome and a prominentcephalic shield (probably homologous with the pro-boscis of the enteropneusts), which they employ tosecrete a collagenous, communal domicile (variouslyreferred to as a tubarium, coenecium or rhabdosome).The majority of taxa examined here form small assem-blages of individuals that inhabit a tubarium attachedto firm substrates, but their most-derived clade, theGraptoloida, is planktic, moderately species-rich anddiverse in colony form. Most of the species includedin the pterobranch crown clade are extinct.
The Pterobranchia includes two subgroups: theCephalodiscida (including the Cephalodiscidae Har-mer 1905 and Eocephalodiscidae Kozłowski 1949)and the Graptolithina (defined below). The preciseform of the relations between the Pterobranchia andthe Enteropneusta are somewhat uncertain. Genetic,morphological and total-evidence phylogenies lead toconflicting inferences about whether these are sistertaxa or whether pterobranchs were derived fromwithin the harrimanid enteropneust clade (e.g. Hala-nych 1995; Cameron et al. 2000; Winchell et al. 2002;Cameron 2005; Sato et al. 2008b; Swalla & Smith2008; Cannon et al. 2009). Recent discovery of adegenerate form of the major body patterning genehedgehog in Saccoglossus kowalevskii and comparisonwith its less-modified ortholog in Rhabdopleura com-pacta suggests that this gene has a uniquely derivedform in S. kowalevskii, the evolution of which has par-alleled the evolution of the unique morphological fea-tures of the harrimanid enteropneusts (Sato et al.
2009). These relationships support the hypothesis thatpterobranchs form a sister clade to the Enteropneusta.
The oldest fossil Pterobranchia occur in mid-Cam-brian strata, dated at approximately 510–505 Ma(Maletz et al. 2005; Rickards & Durman 2006). Recentdiscoveries of a single specimen of a possible ptero-branch (Hou et al. 2011) from the Chengjiang Kons-ervat-Lagerstatte of southern China may extend therange of the group into the Early Cambrian(�525 Ma). In addition, poorly preserved taxa identi-fied as the relatively derived graptolites Mastigograptussimplex and Dendrograptus sp. (Division Eugraptolith-ina, nov.) appear in strata of the Eccaparadoxides insu-laris Zone (lower part of Cambrian Series 3) inTasmania (Quilty 1971; reviewed in Rickards & Dur-man 2006), but we are in doubt about whether thematerial is sufficient to support those identifications.These rocks are nearly the same age as the BurgessShale, which among other soft-bodied fossils, bearsremains of several enteropneusts (Caron et al. 2010).Thus, both the origin of pterobranchs and the splitbetween cephalodiscids and graptolites must predate�505 Ma and considerable diversification must havetaken place within the group during the Cambrianexplosion as appears to be the case all across the Bila-teria.
Subclass Graptolithina Bronn 1849, emend.
Emended Definition. – Graptolithina is defined as alineage-based taxon that includes all taxa sharing amore recent common ancestry with Rhabdopleurathan with Cephalodiscus.
Discussion. – This definition identifies the total cladethat includes graptolites and Rhabdopleura and allowsthe taxon to be mapped unambiguously onto molecu-lar phylogenetic trees. On present evidence, the Grap-tolithina includes all pterobranchs with zooids seriallybudded from an interconnected stolon system andnone that lack these features, thus also permittingunambiguous identification of fossil graptolites. Syna-pomorphic features also appear to include the factthat the increments that form the creeping tubes ofthe tubarium (fuselli) have a regular zigzag pattern(Figs 2G, 3B) and those on the upright tubes are simi-lar half rings or regularly arranged full rings. We sus-pect that they also are united by possession of asclerotized larval vesicle (prosicula; Figs 2A–E, 3D),but it is not known whether Cysticamara, the mostbasal of our studied taxa, possesses a sclerotized siculaor not. This graptolite differs from all more derivedgraptolites, including Rhabdopleura, in that its zooidaltubes appear to be blind – that is, they are imbeddedin a poorly organized extrathecal tissue and are
52 Mitchell et al. LETHAIA 46 (2013)
interconnected only by tubular stolons, not continu-ous thecal tubes (Fig. 4B). In this respect Cysticamaraappears to be intermediate in its organization betweenpseudocolonial Cephalodiscus and the fully intercon-nected colonial tubarium of Rhabdopleura and allmore derived graptolites.
Infraclass Eugraptolithina, nov.
Definition. – The Eugraptolithina is the holophyletic,apomorphy-based taxon that includes the first grapto-lite that acquired a prosicula with a helical line and allits descendants.
Discussion. – The taxon includes the paraphyleticset of graptolites historically referred to the Tuboi-dea (but excluding Epigraptus, as discussed above),the Dendroidea (including Mastigograptus) and theGraptoloidea (see Bulman 1970). Members of theEugraptolithina share a classically ‘graptolite’-typesicula with a basally attached, vase-shaped to tubu-lar prosicula that exhibits a helical line. Present evi-dence suggests that this group also shares thepresence of upright stipes, by which we mean free-standing branches formed by a series of intercon-nected thecae that grew upward from their encrust-ing holdfast into the water column. Manyeugraptolithines formed complexly branched, bushyor conical colonies that contained many hundredsof zooids. This group includes virtually all of themacroscopic graptolites that are commonly encoun-tered in the fossil record and that diversified tobecome a significant component of the preserveddiversity of the early to mid-Palaeozoic faunas. Asmentioned above, the Eugraptolithina appear in theearly part of Cambrian Series 3 (�510 to 505 Ma),are the oldest-known graptolites (Rickards & Dur-man 2006) and went extinct in the CarboniferousPeriod (Bulman 1970). As an entirely extinct fossilgroup, definition with reference to a key structuralsynapomorphy provides a stable and effectivemeans to identify the clade.
Division Graptoloida Lapworth, 1875 (in Hop-kinson & Lapworth 1875) emend., nom. transl.
Emended definition. – The Graptoloida is the totalclade descended from the first graptolite to possess asicula with a caudal apex (nematophorous sicula;Figs 2F, 3D; see also Williams & Clarke 1999).
Discussion.– . – The name Graptoloida was estab-lished as Section Graptoloidea by Lapworth (in Hop-kinson & Lapworth 1875), which is a family-grouprank. Ruedemann (1904) raised the taxon to ordinal
rank, however, this name has traditionally retainedthe ‘…oidea,’ suffix that is now fixed in Linnaean zoo-logical nomenclature as specific to the rank of super-family [International Commission on ZoologicalNomenclature (ICZN) 1999, Article 29.2]. Thus, wefollow Maletz et al. (2009) and employ a minor spell-ing alteration to avoid unintentionally suggestingsuperfamily rank for this taxon.
The Graptoloida is an apomorphy-based taxon thatincludes all planktic graptolites. In addition to thenematophorous sicula, basal graptoloids share thepresence of regularly dichotomously branched stipesthat form horizontally spreading to conical tubariumand retain paired autothecae and bithecae (see therecent revision in Maletz et al. 2009; who provide amore complete discussion of the features and contentof the group). The concept of the Graptoloida pro-posed herein is identical to the Graptoloidea of Fortey& Cooper (1986) and the Rhabdophora of Allman(1872) as used by Lapworth (1873a,b), but differsfrom the Graptoloidea of Bulman (1955, 1970), whichexcluded the planktic Anisograptidae. Graptoloidsfirst appear immediately above the base of the Ordovi-cian System, approximately 490 Ma (Sadler et al.2009). Indeed, this event provides a supplementalguide to the identification of the beginning of this per-iod of Earth’s history.
Acknowledgements. – We wish to acknowledge the financial sup-port of Natural Sciences and Engineering Research Council of Can-ada Discovery Grants to MJM and CBC, a St. Francis XavierUniversity James Chair Visiting Professorship to JM and USNational Science Foundation research grant EAR 0418790 to CEMand MJM, which provided partial support for this work.
References
Allman, G.J. 1869: On Rhabdopleura. Quarterly Journal of Micro-scopical Science 9, 57–63.
Allman, G.J. 1872: On the morphology and affinities of graptolites.Annals and Magazine of Natural History, Fourth Series 9, 364–380.
Andersson, K.A. 1907: Die Pterobranchier der schwedischen Sud-polar-Expedition 1901–1903 nebst Bemerkungen uber Rhabdo-pleura normanni Allman. Wissenschaftliche Ergebnisse derSchwedischen Sudpolar-Expedition 1901-1903 5, 1–122.
Andres, D. 1977: Graptolithen aus ordovizischen Geschieben unddie fruhe Stammesgeschichte der Graptolithen. PalaontologischeZeitschrift 51, 52–93.
Andres, D. 1980: Feinstrukturen und Verwandtschaftsbeziehungender Graptolithen. Palaontologische Zeitschrift 54, 129–170.
Barnes, R.D. 1977: A new record of a pterobranch hemichordatefrom the western Hemisphere. Bulletin of Marine Sciences 27,340–343.
Bates, D.E.B. 1987: The density of graptoloid skeletal tissue, and itsimplication for the volume and density of the soft tissue. Lethaia20, 149–156.
Bates, D.E.B. & Kirk, N.H. 1986: Mode of secretion of graptoliteperiderm, in normal and retiolite graptolites. Geological SocietySpecial Publication 20, 221–236.
Bates, D.E.B. & Urbanek, A. 2002: The ultrastructure, develop-ment, and systematic position of the graptolite genus Mastigo-graptus. Acta Palaeontologica Polonica 47, 445–458.
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 53
Bates, D. E. B., Kozłowska, A., Loydell, D., Urbanek, A. & Wade, S.2009: Ultrastructural observations on some dendroid and gra-ptoloid graptolites and on Mastigograptus. Bulletin of Geosciences84, 21–26.
Bateson, W. 1885: The later stages in the development of Balanog-lossus kowalevskii, with a suggestion as to the affinities of theEnteropneusta. Quarterly Journal of Microscopical Science 25, 81–122.
Beklemishev, V.N. 1951: Towards establishing a zoological system.Deuterostomians (Deuterostomia), their origin and composi-tion. Uspekhi Sovremennoi Biologii 32, 256–270. [in Russian]
Beklemishev, V.N. 1970: Principles of Comparative Anatomy ofInvertebrates. Oliver and Boyd, Edinburgh, Vol. 1, 490 pp, Vol.2, 529 pp.
Bjerreskov, M. 1978: Discoveries on graptolites by X-ray studies.Acta Palaeontologica Polonica 23, 463–471.
Bjerreskov, M. 1994: Pyrite diagenesis of graptolites from Born-holm, Denmark. In Chen, X., Erdtmann, B.-D. & Ni, Y-N. (eds):Graptolite Research Today, 217–222. Nanjing University Press,Nanjing.
Bronn, H. G. 1849: Index Palaeontologicus oder Ubersicht der bisjetzt bekannten fossilen Organismen, Prof. H. R. Goppert undHerm. V. Meyer. In Bronn, H.G. (ed.): Zweite Abtheilung. B.Enumerator palaeontologicus: Systematische Zusammenstellungund geologische Entwicklungs-Gesetze der organischen Reiche, 1–976. E. Schweizerbart’sche Verlagsbuchhandlung, Stuttgart.
Bulman, O. M. B. 1927: Uber Koremagraptus fibratus sp. nov.(Grapt., Dendr.). Senkenbergiana 9, 38–40.
Bulman, O. M. B. 1933: On the graptolites prepared by Holm 4.Structural characters of some Dictyonema and Desmograptusspecies from the Ordovician and Silurian rocks of Sweden andthe East Baltic Region. Arkiv for Zoologi 26A, 1–52.
Bulman, O. M. B. 1937: The structure of Acanthograptus suecicus,and the affinities of Acanthograptus. Geologiska Foreningen iStockholm Forhandlingar 59, 182–188.
Bulman, O. M. B. 1947: A monograph of the Caradoc (Balclatchie)graptolites from limestones in Laggan Burn, Ayrshire, Part 3.Palaeontographical Society Monographs 98, 59–78.
Bulman, O. M. B. 1949: A re-interpretation of the structure ofDictyonema flabelliforme Eichwald. Geologiska Foreningen i Stock-holm Forhandlingar 71, 33–40.
Bulman, O.M.B. 1955: Graptolithina with sections on Enteropneu-sta and Pterobranchia. In Moore, R.M. (ed.): Treatise on Inverte-brate Paleontology, Part V, 101 pp. Geological Society of Americaand University of Kansas Press, Lawrence.
Bulman, O.M.B. 1970: Graptolithina with sections on Enteropneu-sta and Pterobranchia In Teichert, K. & Moore, R.M. (eds):Treatise on Invertebrate Paleontology, Part V (2nd edition), 1–163. Geological Society of America and University of KansasPress, Boulder and Lawrence.
Bulman, O.M.B. & Rickards, R.B. 1966: A revision of Wiman’sdendroid and tuboid graptolites. Bulletin of the Geological Insti-tution of Uppsala 43, 1–72.
Burdon-Jones, C. 1954: The habitat and description of Rhabdople-ura normani Allman. Universitetet Bergen Arbeits 11, 1–17.
Cameron, C.B. 2005: A phylogeny of the hemichordates based onmorphological characters. Canadian Journal of Zoology 83, 196–215.
Cameron, C.B., Garey, J.R. & Swalla, B.J. 2000: Evolution of thechordate body plan: new insights from phylogenetic analysis ofdeuterostome phyla. Proceedings of the National Academy of Sci-ences USA 97, 4469–4474.
Cannon, J.T., Rychel, A.L., Eccleston, H., Halanych, K.M. & Swalla,B.J. 2009: Molecular phylogeny of Hemichordata, with updatedstatus of deep-sea enteropneusts. Molecular Phylogenetics andEvolution 52, 17–24.
Caron, J.-B., Conway Morris, S. & Shu, D. 2010: Tentaculate fossilsfrom the Cambrian of Canada (British Columbia) and China(Yunnan) interpreted as primitive deuterostomes. PLoS One 5,e9586.
Cavers, P. 2005: Quantifying Variation in a Population of Rhabdo-pleura Compacta. Unpublished Zoology Honours thesis, Univer-sity of Edinburgh, Edinburgh.
Chapman, A.J., Durman, P.N. & Rickards, R.B. 1996: A provisionalclassification of the graptolite Order Dendroidea. Palaontologi-sche Zeitschrift 70, 189–202.
Cooper, R.A., Maletz, J., Wang, H. & Erdtmann, B.-D. 1998:Taxonomy and evolution of earliest Ordovician graptoloids.Norsk Geologisk Tidsskrift 78, 3–32.
Dawydoff, C. 1948: Classe des Pterobranches. In Grasse, P.P. (ed.):Traite de Zoologie, 454–489. Masson et Cie, Paris.
Delsuc, F., Brinkmann, H., Chourrout, D. & Philippe, H. 2006:Tunicates and not cephalochordates are the closest living rela-tives of vertebrates. Nature 439, 965–968.
Dilly, P.N. 1971: keratin-like fibres in the hemichordate Rhabdople-ura compacta. Zeitschrift fur Zellforschung 117, 502–515.
Dilly, N.P. 1973: The larva of Rhabdopleura compacta (Hemichor-data). Marine Biology 18, 69–86.
Dilly, P.N. 1975: The dormant buds of Rhabdopleura compacta(Hemichordata). Cell and Tissue Research 159, 387–397.
Dilly, N.P. 1985a: The prosicular stage of Rhabdopleura (Ptero-branchia: Hemichordata). Journal of Zoology, London (A) 206,163–174.
Dilly, P.N. 1985b: A note on the status of Rhabdopleura from Ber-muda. Journal of the Marine Biological Association of the UnitedKingdom 65, 987–991.
Dilly, P.N. 1986: Modern pterobranchs: observations on theirbehaviour and tube building. In Hughes, C.P. & Rickards, R.B.(eds): Palaeoecology and Biostratigraphy of Graptolites. GeologicalSociety Special Publication 20: 261–269.
Dilly, P.N. 1993: Cephalodiscus graptolitoides sp. nov. a probableextant graptolite. Journal of Zoology (London) 229, 69–78.
Donoghue, P.C.J. 2005: Saving the stem group — a contradictionin terms? Paleobiology 31, 553–558.
Durman, P.N. 1992: The Origin and Early Evolution of Graptolitesand Related Hemichordates. Unpublished PhD thesis, CambridgeUniversity, Cambridge.
Durman, P.N. & Sennikov, N.V. 1993: A new rhabdopleurid hemi-chordate from the Middle Cambrian of Siberia. Palaeontology36, 283–296.
Eisenack, A. 1941: Epigraptus bidens n.g. n.sp., eine neue graptoli-thenart des baltischen Ordoviziums. Zeitschrift fur Geschiebefors-chung und Flachlandsgeologie 17, 24–28.
Eisenack, A. 1974: Einige graptolithen aus dem Ordovizium des Bal-tikums, ferner uber Epigraptus und andere Idiotubidae. Neues Jahr-buch fur Geologie und Palaontologie, Monatshefte 66, 4–674.
Erdtmann, B.-D. 1982: A reorganization and proposed phyloge-netic classification of planktic Tremadoc (early Ordovician) den-droid graptolites. Norsk Geologisk Tidsskrift 62, 121–145.
Felsenstein, J. 2004: Inferring Phylogenies. Sinauer Associates, Sun-derland.
Forey, P.L., Humphries, C.J., Kitching, I.L., Scotland, R.W., Siebert,D.J. & Williams, D.M. 1992: Cladistics: A Practical Course in Sys-tematics. Systematics Association Publication No. 10, OxfordScience Publications, Oxford.
Fortey, R.A. & Cooper, R.A. 1986: A phylogenetic classification ofthe graptoloids. Palaeontology 29, 631–654.
Gilchrist, J.D.F. 1915: Observations on the Cape Cephalodiscus (C.gilchristi, Ridewood) and some of its early stages. The Annalsand Magazine of Natural History, 8th Series 16: 23, 3–243.
Gilchrist, J.D.F. 1917: On the development of the Cape Cephalodis-cus (C. gilchristi, Ridewood). Quarterly Journal of MicroscopicalScience 62, 189–211.
Goloboff, P.A. 1993: Estimating character weights during treesearch. Cladistics 9, 83–91.
Goloboff, P.A. 1997: Self-weighted optimization: tree searches andcharacter state reconstructions under implied transformationcosts. Cladistics 13, 225–245.
Goloboff, P.A. & Farris, J.S. 2001: Methods for quick consensusestimation. Cladistics 17, S26–S34.
Goloboff, P.A., Farris, J.S., Kallersjo, M., Oxelman, B., Ramırez, M.J. & Szumik, C.A. 2003: Improvements to resampling measuresof group support. Cladistics 19, 324–332.
Goloboff, P.A., Carpenter, J.M., Arias, J.S. & Esquivel, D.R.M.2008a: Weighting against homoplasy improves phylogeneticanalysis of morphological data sets. Cladistics 24, 758–773.
54 Mitchell et al. LETHAIA 46 (2013)
Goloboff, P.A., Farris, J.S. & Nixon, K.C. 2008b: TNT, a free pro-gram for phylogenetic analysis. Cladistics 24, 774–786.
Halanych, K.M. 1993: Suspension feeding by the lophophore-likeapparatus of the pterobranch hemichordate Rhabdopleura nor-manni. Biological Bulletin 185, 417–427.
Halanych, K.M. 1995: The phylogenetic position of the ptero-branch hemichordates based on 18S rDNA sequence data.Molecular Phylogenetics and Evolution 4, 72–76.
Harmer, S.F. 1905: The Pterobranchia of the Siboga-Expeditionwith an account of other species. Siboga Expedition Monograph26, 1–133.
Hincks, T. 1880: History of British Marine Polyzoa, 601 pp. VanVoorst, London.
Hopkinson, J. & Lapworth, C. 1875: Descriptions of the graptolitesof the Arenig and Llandeilo rocks of St. David‘s. Quarterly Jour-nal of the Geological Society of London 31, 631–672.
Hou, X., Aldridge, R.J., Siveter, D.J., Siveter, D.J., Williams, M.,Zalasiewicz, J. & Ma, X. 2011: An Early Cambrian hemichordatezooid. Current Biology 21, 612–616.
Hutt, J.E. 1974: The development of Clonograptus tenellus andAdelograptus hunnebergensis. Lethaia 7, 79–92.
International Commission on Zoological Nomenclature (ICZN)1999: International Code of Zoological Nomenclature (4th edi-tion). http://www.nhm.ac.uk/hosted-sites/iczn/code/ [accessed28 November 2011].
John, C.C. 1931: Cephalodiscus. Discovery Reports 3, 223–260.John, C.C. 1932: On the development of Cephalodiscus. Discovery
Reports 6, 191–204.Kirk, N.H. 1972: Some Thoughts on the Construction of the Rhabdo-
some in the Graptolithina, with Special Reference to ExtrathecalTissue and its Bearing on the Theory of Automobility. Universityof Wales, Aberystwyth, Institute of Earth Studies, Aberystwyth,1–21.
Kitching, I. J., Forey, P. L., Humphries, C. J. & Williams, D. M.1998: Cladistics: Theory and Practice of Parsimony Analysis (2ndedition). Oxford University Press, Oxford.
Kozłowski, R. 1949: Graptolites and some new groups ofanimals from the Tremadoc of Poland. Palaeontologia Polonica3, 1–235. [in French]
Kozłowski, R. 1962: Crustoidea – a new group of graptolites. ActaPalaeontologica Polonica 7, 1–52. [in French]
Kozłowski, R. 1963: Le developpement d’un graptolite tuboıde.Acta Palaeontologica Polonica 8, 103–134.
Kozłowski, R. 1971: Early development stages and the mode of lifeof graptolites. Acta Palaeontologica Polonica 16, 313–343.
Lankester, E.R. 1877: Notes on the embryology and classifica-tion of the animal kingdom: comprising a revision of specu-lations relative to the origin and significance of the germ-layers. Quarterly Journal of Microscopical Science, New Series17, 399–454.
Lankester, E. R. 1884: A contribution to the knowledge ofRhabdopleura. Quarterly Journal of Microscopical Science 24, 622–647.
Lankester, E.R. 1905: On a new species of Cephalodiscus (C. nigres-cens) from the Antarctic Ocean. Proceedings of the Royal Societyof London Series B-Biological Sciences 76, 400–402.
Lapworth, C. 1873a: Notes on the British graptolites and theirallies. 1. On an improved classification of the Rhabdophora. Part1. Geological Magazine 10, 500–504.
Lapworth, C. 1873b: Notes on the British graptolites and theirallies. 1. On an improved classification of the Rhabdophora. Part2. Geological Magazine 10, 555–560.
Legrand, P. 1974: Development of rhabdosomes with four primarybranches in the group Dictyonema flabelliforme (Eichwald). Spe-cial Papers in Palaeontology 13, 19–34.
Loydell, D.K., Orr, P.J. & Kearns, S. 2004: Preservation of soft tis-sues in Silurian graptolites from Latvia. Palaeontology 47, 503–513.
Maletz, J. 1992: The proximal development in anisograptids (Gra-ptoloidea, Anisograptidae). Palaontologische Zeitschrift 66, 297–309.
Maletz, J., Steiner, M. & Fatka, O. 2005: Middle Cambrian ptero-branchs and the Question: what is a graptolite? Lethaia 38, 73–85.
Maletz, J., Carlucci, J. & Mitchell, C.E. 2009: Graptoloid cladistics,taxonomy and phylogeny. Bulletin of Geosciences 84, 7–19.
Mierzejewski, P. 1978: Tuboid graptolites from erratic boulders ofPoland. Acta Palaeontologica Polonica 23, 585–575.
Mierzejewski, P. 2000: On the nature and development of grapto-blasts. Acta Palaeontologica Polonica 45, 227–238.
Mierzejewski, P. 2001: A new graptolite intermediate between theTuboidea and the Camaroidea. Acta Palaeontologica Polonica 46,367–376.
Mierzejewski, P. & Kulicki, C. 2002: Discovery of Pterobranchia(Graptolithoidea) in the Permian. Acta Palaeontologica Polonica47, 169–175.
Mierzejewski, P. & Kulicki, C. 2003: Cortical fibrils and secondarydeposits in periderm of the hemichordate Rhabdopleura (Gra-ptolithoidea). Acta Palaeontologica Polonica 48, 99–111.
Mierzejewski, P. & Urbanek, A. 2004: The morphology and finestructure of the Ordovician Cephalodiscus-like genus Melanostro-phus. Acta Palaeontologica Polonica 49, 519–528.
Mierzejewski, P., Kulicki, C. & Urbanek, A. 2005: The world’s old-est crustoid graptolites from the upper Tremadocian of Poland.Acta Palaeontologica Polonica 50, 721–724.
M’Intosh, W.C. 1887: Report on Cephalodiscus dodecalophus,M’Intosh, a new type of the Polyzoa, procured on the voyage ofH.M.S. Challenger during the years 1873–76. Challenger Reports,Zoology 20, 1–37.
Mirande, J. M. 2009: Weighted parsimony phylogeny of the familyCharacidae (Teleostei: Characiformes). Cladistics 25, 574–613.
Mitchell, C.E. & Carle, K.J. 1986: The nematularium of Pseudocli-macograptus scharenbergi (Lapworth) and its secretion. Palaeon-tology 29, 373–390.
Norman, A.M. 1869: Shetland final dredging report. Part 2. On theCrustacea, Tunicata, Polyzoa, Echinodermata, Actinozoa, Hy-drozoa and Porifera. Reports of the British Association of theAdvancement of Science 38, 247–336.
Putnam, N.H., Butts, T., Ferrier, D.E. et al. 2008: The amphioxusgenome and the evolution of the chordate karyotype. Nature453, 1064–1072.
de Queiroz, K. 2007: Toward an integrated system of clade names.Systematic Biology 56, 956–974.
Quilty, P.G. 1971: Cambrian and Ordovician dendroids and hy-droids of Tasmania. Journal of the Geological Society of Australia17, 171–189.
Rickards, R.B. 1996: The graptolite nema: problem to all our solu-tions. Geological Magazine 133, 343–346.
Rickards, R.B. & Durman, P.N. 2006: Evolution of the earliestgraptolites and other hemichordates. National Museum of WalesGeological Series 25, 5–92.
Rickards, R.B. & Stait, B.A. 1984: Psigraptus, its classification, evo-lution and zooid. Alcheringa 8, 101–111.
Rickards, R.B., Durman, P.N. & Chapman, A.J. 1995: Rhabdopleu-ran hemichordates: new fossil forms and review. Proceedings ofthe Geologists’ Association 106, 293–303.
Rickards, R.B., Wright, A.J. & Thomas, G. 2009: Late Llandovery(Early Silurian) dendroid graptolites from the Cotton Formationnear Forbes, New South Wales. Proceedings of the Linnaean Soci-ety of New South Wales 130, 63–76.
Rigby, S. 1993: Palaeontology – Graptolites come to life. Nature362, 209–210.
Rigby, S. & Sudbury, M. 1995: Graptolite ontogeny and the size ofthe graptolite zooid. Geological Magazine 132, 427–433.
Ruedemann, R. 1904: Graptolites of New York, part 1. New YorkState Museum Memoir 7, 455–803.
Sadler, P.M., Cooper, R.A. & Melchin, M.J. 2009: High-resolution,early Paleozoic (Ordovician-Silurian) time scales. GeologicalSociety of America Bulletin 121, 887–906.
Sato, T. 1936: Vorlaufige mitteilung fiber Atubaria heterolopha gen.nov., sp. nov., einen in freiem zustand aufgefundenen Ptero-branchier aus dem stillen ozean. Zoologisehe Anzeiger 15, 97–106.
LETHAIA 46 (2013) Phylogeny of living and fossil graptolites 55
Sato, A., Bishop, J. D. D. & Holland, P. W. H. 2008a: Developmen-tal biology of pterobranch hemichordates: history and perspec-tives. Genesis 46, 587–591.
Sato, A., Rickards, R.B. & Holland, P.W.H. 2008b: The origins ofgraptolites and other pterobranchs: a journey from ‘Polyzoa’. Le-thaia 41, 303–316.
Sato, A., White-Cooper, H., Dogget, K. & Holland, P.W.H. 2009:Degenerate evolution of the hedgehog gene in a hemichordatelineage. Proceedings of the National Academy of Sciences USA106, 7491–7494.
Saunders, K.M., Bates, D.E.B., Kluessendorf, J., Loydell, D.K. & Mi-kulic, D.G. 2009: Desmograptus micronematodes, a Silurian den-droid graptolite, and its ultrastructure. Palaeontology 52, 541–559.
Schepotieff, A. 1907: Die Pterobranchier. I. Rhabdopleura normaniAllman. Zoologische Jahrbucher (Anatomie) 24, 193–238.
Schiaparelli, S., Cattaneo-Vietti, R. & Mierzejewski, P. 2004: A‘protective shell’ around the larval cocoon of Cephalodiscus den-sus Andersson, 1907 (Graptolithoidea, Hemichordata). PolarBiology 27, 813–817.
Spengel, J.W. 1932: Planctosphaera. Report on the Scientific Resultsof the ‘Michael Sars’ North Atlantic Deep-Sea Expedition 1910, 5,590–600.
Stebbing, A.R.D. 1968: Discovery of Rhabdopleura (Hemichordata)at Plymouth. Nature 217, 1287.
Stebbing, A.R.D. 1970: The status and ecology of Rhabdopleuracompacta (Hemichordata) from Plymouth. Journal of the MarineBiology Association of the United Kingdom 50, 209–221.
Swalla, B.J. & Smith, A.B. 2008: Deciphering deuterostome phylog-eny: molecular, morphological and palaeontological perspectives.Philosophical Transactions of the Royal Society B 363, 1557–1568.
Swofford, D.L. 2001: PAUP: Phylogenetic Analysis using Parsimony,version 4.0b10. Sinauer Associates, Sunderland.
Urbanek, A. 1976: The problem of graptolite affinities in the lightof ultrastructural studies on peridermal derivatives in ptero-branchs. Acta Palaeontologica Polonica 21, 3–36.
Urbanek, A. 1978: Significance of ultrastructural studies for grapto-lite research. Acta Palaeontologica Polonica 23, 595–625.
Urbanek, A. 1984: The significance of graptoblasts in the life cycleof crustoid graptolites. Acta Palaeontologica Polonica 28, 313–326.
Urbanek, A. 1994: When is a pterobranch a graptolite? Lethaia 27,324.
Urbanek, A. 2004: Morphogenetic gradients in graptolites and bry-ozoans. Acta Palaeontologica Polonica 49, 485–504.
Urbanek, A. & Dilly, N.P. 2000: The stolon system in Rhabdopleuracompacta (Hemichordata) and its phylogenetic implications.Acta Palaeontologica Polonica 45, 201–226.
Urbanek, A. & Mierzejewski, P. 1982: Ultrastructure of the tuboidgraptolite tubotheca. Palaontologische Zeitschrift 56, 87–93.
Urbanek, A. & Mierzejewski, P. 1984: The ultrastructure of theCrustoidea and the evolution of graptolite skeletal tissue. Lethaia17, 73–91.
Urbanek, A. & Mierzejewski, P. 1986: A possible new pattern ofcortical deposits in Tremadoc dendroid graptolites from chertnodules. Geological Society Special Publication 20, 13–19.
Urbanek, A. & Towe, K.M. 1974: Ultrastructural studies on grapto-lites 1: the periderm and its derivatives in the Dendroidea and inMastigograptus. Smithsonian Contributions to Paleobiology 20, 1–48.
Whittington, H.B. & Rickards, R.B. 1968: New tuboid graptolitefrom the Ordovician of Ontario. Journal of Paleontology 42, 61–69.
Williams, S. H. & Clarke, L. C. 1999: Structure and secretion of thegraptolite prosicula, and its application for biostratigraphicaland evolutionary studies. Palaeontology 42, 1003–1015.
Wiman, C. 1901: Uber die Borkholmer Schicht im MittelbaltischenSilurgebiet. Bulletin of the Geological Institute of Uppsala 5, 151–222.
Winchell, C.J., Sullivan, J., Cameron, C.B., Swalla, B.J. & Mallatt, J.2002: Evaluating hypothesis of deuterostome phylogeny andchordate evolution with new LSU and SSU ribosomal DNAdata. Molecular Biology and Evolution 19, 762–776.
56 Mitchell et al. LETHAIA 46 (2013)
