Previously considered an actinopterygian or an osteichthyan incertae sedis, the Devonian (Givetian–Frasnian) Holopterygius nudus is reinterpreted as a coelacanth. This genus is among the oldest coelacanths known from articulated remains, but its eel-like morphology marks a considerable departure from the conventional coelacanth body plan. A cladistic analysis places Holopterygius as the sister taxon of the Carboniferous (Serpukhovian) genus Allenypterus. Despite the specialized morphology of these genera, they occupy a surprisingly basal position in coelacanth phylogeny; only Diplocercides and Miguashaia are further removed from the crown. A morphometric analysis reveals that coelacanths were anatomically disparate early in their history. Conflicts between this result and those of previous studies challenge the adequacy of systematic character sets for describing historical patterns of morphological variety. Coelacanths have long had an iconic place in the study of vertebrate evolution for their apparent anatomical conservatism over geological time, but Holopterygius provides clear evidence for rapid morphological evolution early in the history of this clade.
The affinities of Holopterygius nudus Jessen, from the latest Givetian–earliest Frasnian Upper Plattenkalk of Bergisch-Gladbach, Germany, have been enigmatic since its description. While a long, diphycercal caudal fin and eel-like body distinguish Holopterygius from its contemporaries, these features have frustrated efforts to assign it to a higher-level taxon. Jessen (1973), followed by Lund & Melton (1982), speculated on a relationship between this genus and Carboniferous tarrasiid actinopterygians (Moy-Thomas 1934), but offered no characters in support of this arrangement beyond similar body shapes. More recent studies have regarded the position of Holopterygius among osteichthyans as uncertain (Taverne 1996; Lund & Poplin 2002). Re-examination of the only known specimen of Holopterygius reveals numerous derived characters linking this genus with coelacanths, including a gracile pectoral girdle, an extracleithrum and a mandible with a greatly expanded posterior coronoid. This new interpretation reveals that early coelacanths were more anatomically diverse than previously realized, and challenges the widely held view of this clade as morphologically stagnant.
2. Material and methods
(a) Systematic methodology
To investigate the systematic placement of Holopterygius within coelacanths, we have included this taxon in an updated version of the matrix assembled by Forey (1998). This matrix was subjected to parsimony analysis using the branch-and-bound algorithm in PAUP* v. 4.0b (Swofford 2002). Initial results were poorly resolved, so characters were re-weighted by their re-scaled consistency indices and the analysis run again. For the complete character matrix and additional details of the cladistic analyses, consult electronic supplementary material B.
(b) Morphometric methodology
We have used relative warps analysis (Bookstein 1991) to summarize variation in coelacanth body form, unlike previous studies, which have examined discrete character data (Schaeffer 1952; Cloutier 1991b; Smith 1994; Forey 1998). Eleven landmarks were digitized from published reconstructions or specimen photographs of 22 genera using the program tpsDig v. 1.40 (Rohlf 2004a). Relative warps analysis was conducted in tpsRelw v. 1.39 (Rohlf 2004b). As this study concerns large- and small-scale morphological variation, the parameter α (which determines the weight of different principal warps) was set to zero. Uniform components of shape change were estimated (Bookstein 1996) and appended to the matrix of principal warps. A principal components analysis of this matrix gives the relative warps and relative warp scores. Coordinate data and the relative warps matrix are given in electronic supplementary material C. Only those axes accounting for over 5% of total variance were examined. Disparity was calculated using four frequently used metrics: sum of variances, product of variances, sum of ranges and geometric mean of ranges (VanValen 1974; Foote 1992).
3. Systematic palaeontology
(a) Revised diagnosis
The following combination of characters distinguishes H. nudus from all other coelacanths:
well-developed premaxillary dentition;
keel scales along ventral margin of body;
strongly asymmetrical caudal lobes; and
well-developed ventral lobe of caudal fin.
Naturhistoriska Riksmuseet P 7789 a-c, incomplete fish in part and counterpart.
(c) Locality and horizon
The type and only specimen of Holopterygius is from the Oberer Plattenkalk in the Heiligenstock quarry, east of Bergisch-Gladbach near Cologne, Germany (Ørvig 1960; Jessen 1973). The depositional environment has been interpreted as lagoonal (Jessen 1973).
The stage-level dating of the Bergisch-Gladbach fish beds is uncertain. The flags below the Oberer Plattenkalk yield the Givetian goniatite Maenioceras terebratum (Ørvig 1960; Harland et al. 1990). Although the Frasnian goniatite Manticoceras first appears in beds that overlie the Oberer Plattenkalk (Ørvig 1960), the Givetian–Frasnian boundary probably precedes its first appearance (Harland et al. 1990). Therefore, the age of the fish beds is probably latest Givetian–earliest Frasnian (ca 385 Ma; all dates in this article are from Gradstein et al. 2004).
The skull roof of Holopterygius (figure 1b, sr) is preserved primarily as an impression of its visceral surface, but intact areas are perforated by large, irregular pores for the supraorbital sensory canal. The premaxilla (figure 1b, pmx) is squat and bears four large, irregularly shaped teeth. The dorsal surface of the splint-like parasphenoid (figure 1b, psph) is exposed, and is marked by a well-developed hypophysial fossa. Much of the palate is visible (figure 1b, pal), and it has the triangular shape characteristic of coelacanths (Forey 1998). The quadrate is preserved as a thickened region at the posteroventral corner of the palate, while a series of teeth along its ventral margin represent the remains of the ectopterygoid. A large, subspherical structure in the otic region is interpreted as an otolith (figure 1b, ot; cf. Clack 1996). No dermal bones of the cheek can be identified. The opercle (figure 1b, op) is a large rectangular bone with a long vertical axis, and resembles that of Allenypterus (Lund & Lund 1985; Forey 1998).
The lower jaw of Holopterygius bears a long anterior ramus, a well-developed dorsal process at mid length, and another ramus posteriorly. The anterior arm of the lower jaw resembles the slender dentary of many early coelacanths (Lund & Lund 1985; Forey et al. 2000). Although the dentary of the intact jaw is preserved only as an impression, a long tooth-bearing fragment on the opposite side of the skull probably represents a portion of the dentary from the other mandible. The deep dorsal process of the jaw is composed of an expanded posterior coronoid (figure 1b, cop), which is a coelacanth synapomorphy (Forey 1998). Small teeth anterior to the principal coronoid may be the remains of additional members of the series. The posterior ramus of the jaw corresponds to the retroarticular process of other coelacanths, and is overlain by a stocky cylindrical bone that is identified as the symplectic (figure 1b, sym), a major component of the coelacanth tandem jaw joint. The ceratohyals (figure 1b, chy) are flared distally and proximally. Narrow endoskeletal rods represent remains of the gill skeleton (figure 1b, gs).
The cleithrum and clavicles (figure 1b, cle, cla) are well-preserved and exhibit the slender proportions characteristic of coelacanths crownward of Miguashaia (Cloutier 1996; Forey 1998). Holopterygius also possesses an extracleithrum (figure 1b, ex), an additional dermal ossification that is a coelacanth synapomorphy (Forey 1998). Although Jessen (1973) noted the similarities between the pectoral girdles of Holopterygius and coelacanths, he did not consider this resemblance sufficient to justify placement within the group. Fragments of the anocleithrum (figure 1b, ano) are located dorsal to the tip of the cleithrum, and suggest this bone was of the simple unbranched morphology primitive for actinistians.
The distinctive postcranium of Holopterygius is leaf-shaped, with a greatly elongated caudal region. Despite the unusual body shape of this genus, the components of its postcranial skeleton conform to the general coelacanth pattern. There are approximately 80 postcranial segments, just over 20 of which are abdominal. Vertebral centra are unossified, but small ossifications located below the abdominal neural arches may represent parapophyses (figure 1b, pap). There are no ossified ribs. Neural arches and spines (figure 1b, nas) and haemal arches and spines (figure 1b, has) are co-ossified.
The dorsal and ventral lobes of the diphycercal caudal fin are each supported by a single row of endoskeletal radials (figure 1b, rd, rv). There is a one-to-one ratio between radials and neural or haemal spines and a greater than one-to-one relationship between fin rays and radials. Jessen (1973) reconstructed Holopterygius with multiple radials per body segment, but this seems to have been informed by reference to conditions in tarrasiid actinopterygians (Lund & Melton 1982; Taverne 1996). Direct comparison with Tarrasius problematicus (Natural History Museum, London, P.1167, P.18064, P.18065, P.18061) confirms the difference in spine to radial ratios. It is not possible to determine whether the fin rays were segmented in Holopterygius as in most coelacanths, or unsegmented as in Allenypterus (Lund & Lund 1985). The dorsal and ventral lobes of the caudal fin are strongly asymmetrical, with the dorsal lobe extending much further anteriorly, as in Allenypterus (Lund & Lund 1985; Forey 1998). No components of the first dorsal, pectoral, anal or pelvic fins are preserved. An irregularly shaped ossification located dorsal to the neural spines immediately anterior to the first radial of the caudal fin probably represents the basal plate of the second dorsal fin (figure 1b, bp).
No squamation is preserved on the flanks of Holopterygius, but a series of keel scales run along the ventral margin of the body from just behind the skull to approximately the midpoint of the abdominal region (figure 1b, vs). It is possible that the absence of scales covering the body of Holopterygius is size-related, as scales are often not preserved in small fossil coelacanths (Lund & Lund 1985). There is no indication of the ossified swim bladder characteristic of derived coelacanths.
Widely cited studies place coelacanths as the most basal extant sarcopterygian radiation (Cloutier & Ahlberg 1996; Forey 1998), but little is known about their early history. While the first crown-group sarcopterygians are known from Lower Devonian (Lochkovian; 411–416 Ma) deposits (Cloutier & Ahlberg 1996), the oldest unequivocal coelacanths are of Givetian age (385–392 Ma; Forey 1998; Long 1999; Forey et al. 2000). Previously described Devonian coelacanths fall into two morphological categories: primitive forms (Gavinia, Miguashaia) with postcrania resembling those of other plesiomorphic sarcopterygians, and more crownward taxa (Chagrinia, Diplocercides) whose postcranial anatomy is similar to that of stratigraphically younger and phylogenetically more derived coelacanths, including the Recent Latimeria.
Holopterygius marks a considerable anatomical departure from other Devonian forms, and a cladistic analysis places it as the sister taxon of Allenypterus, another early coelacanth with unusual postcranial morphology (analysis and results are described in figure 2a). Two unequivocal synapomorphies link these genera: a series of ventral keel scales and an asymmetrical caudal fin in which the dorsal lobe extends far beyond the anterior extremity of the ventral lobe. Both of these fishes have unstable taxonomic histories resulting from their unconventional body profiles; like Holopterygius, Allenypterus was identified as an actinopterygian (Melton 1969) before being recognized as a coelacanth (Lund & Lund 1984).
Although Holopterygius and Allenypterus are unusual among coelacanths in having greatly elongated caudal regions, they are easily distinguished. While the deep body of Allenypterus is strongly arched dorsally and comparatively straight ventrally, the postcranium of Holopterygius is shallower with more symmetrical dorsal and ventral margins. Small specimens of Allenypterus have a less exaggerated dorsal hump than larger individuals (Lund & Lund 1985; Field Museum, Chicago, PF10942, PF 10943a,b), but they retain greatly elongated neural spines above the abdominal region, unlike the short structures in Holopterygius. Both genera share asymmetrical lobes of the caudal fin, but the ventral lobe is more extensive in Holopterygius than in Allenypterus, where it is reduced to a minor series of shortened fin rays near the tip of the tail. Additional dissimilarities are present in the skulls. The premaxillae and mandibles of Allenypterus are edentulous (Lund & Lund 1985; Forey 1998) but bear well-developed teeth in Holopterygius. Additionally, lower jaw proportions of Holopterygius are unlike those of Allenypterus, but are similar to those of other early coelacanths (Lund & Lund 1985; Forey et al. 2000), suggesting that it had a more conventional skull shape than Allenypterus, in which the cranial region is drastically foreshortened.
Previous studies have placed Allenypterus crownward of one or more of its contemporaries from Bear Gulch (Cloutier 1991a,b; Forey 1991; Forey 1998), but our preferred solution (figure 2a) reconstructs the clade containing it and Holopterygius as basal to all other coelacanths in the analysis with the exception of Miguashaia and Diplocercides. This result implies that the lineages leading to Allenypterus and to Hadronector plus all more crownward taxa diverged by the Middle Devonian. Previous minimum age estimates for these lineage-splitting events were centred on the Lower Carboniferous (Cloutier 1991a; Forey 1998). Consequently, our findings extend taxon ranges by between 27 and 44 million years, and indicate that much of early coelacanth diversity remains unsampled.
Perhaps the most surprising result of this study is the discovery of an unusual coelacanth so early in the history of the clade, particularly since this group is often presented an exemplar of morphological conservatism (Simpson 1944, 1953; Stanley 1998). In contrast to this standard view, several authors have inferred high evolutionary rates early in coelacanth history followed by a near monotonic decrease. However, the magnitude of this shift is disputed, with some studies finding a rapid exponential decay (Schaeffer 1952; Cloutier 1991b) while others report a more gradual decline (Forey 1998). Nevertheless, the repeated recovery of an early peak in evolutionary rate reflects a clear biological signal: the early accumulation of characters uniting coelacanths to the exclusion of other sarcopterygians. Interestingly, existing estimates of coelacanth morphological disparity are not positively correlated with evolutionary rates, but instead show an increase throughout the Palaeozoic and much of the Mesozoic (Smith 1994; Forey 1998), peaking during an interval of low rates of anatomical change.
The postcranial morphologies of Allenypterus and Holopterygius—and presumably their ecologies—differ radically from those of ‘typical’ coelacanths, but this is not reflected in previous measures of disparity. This appears to result from methods used previously to calculate morphological diversity. The pairwise comparisons made by Smith (1994) and Forey (1998) were based upon data matrices that avoided phylogenetically uninformative characters. Prior to our reinterpretation of Holopterygius, most of the unusual features of Allenypterus were unique to that genus and, therefore, not brought to bear on disparity estimates. This presents a major problem: while autapomorphies are uninformative for parsimony-based methods of phylogenetic inference, they are essential for calculations of dissimilarity. Consequently, previous measurements of disparity in coelacanths are incomplete because they omit these critical data.
Here, rather than manipulating matrices designed to resolve phylogeny or revising such data to include autapomorphies, we have adopted a morphometric approach in order to describe anatomical variety in coelacanths. The results of this analysis are summarized in figure 2c. The apparent conservatism of coelacanth postcrania has contributed profoundly to notions of this clade as a bradytelic lineage (Schaeffer 1952), and, once again, our results show that most taxa cluster together in morphospace, suggesting that this impression of coelacanths is not entirely unwarranted. Similar regions are occupied by both Devonian/Carboniferous and post-Carboniferous samples, but there are notable outliers early in coelacanth history (figure 2c). Miguashaia, which is the plesiomorphic sister taxon of all other coelacanths (Cloutier 1991a,b, 1996; Forey 1998; Forey et al. 2000), lies far from the major cluster. This reflects the unusual postcranium of this genus, which is probably closer to that of a generalized, and perhaps primitive, sarcopterygian than a typical coelacanth. Likewise, Holopterygius and Allenypterus are distantly removed from the cluster of ‘conventional’ coelacanths, but these two genera are also placed far from one another. As this separation reflects differences in body depth and caudal fin morphology, it is not an artefact of uncertainties surrounding the positions of the anal and dorsal fins of Holopterygius. These results clearly contradict those of previous studies (Smith 1994; Forey 1998) that found low levels of morphological disparity early in coelacanth history.
More remarkable than the degree of morphological separation between Holopterygius, Miguashaia and other early coelacanths is the short interval over which this variety was generated. With a probable origin of actinistians during the Early Devonian or Late Silurian, each of these distinctive morphologies was established by the close of the Middle Devonian. In contrast, post-Carboniferous forms show no evidence of the radical morphological shifts that characterize the initial phases of the coelacanth radiation, despite higher levels of taxonomic diversity (Forey 1998) and a longer interval over which to develop anatomical novelties. The most extreme departure during this later interval is the Jurassic Libys, which, although deep-bodied, conforms to the stereotypical coelacanth body plan. Such rapid accumulation of morphological variety early in clade histories is a recurrent theme in studies of disparity in the fossil record (Foote 1997), but the causes underlying this pattern remain uncertain (Foote 1996). What is clear, however, is that Holopterygius and its disparate contemporaries provide evidence for early anatomical experimentation in coelacanths that is strongly at odds with the textbook portrayal of the clade as morphologically invariant over geological time.
We thank L. Verdelin for the loan of Holopterygius, P. Forey for providing an updated version of his data matrix, and M. LaBarbera for use of his photographic equipment. MF is supported by the National Science Foundation, award number DGE-0228235; MIC is supported by the University of Chicago faculty research fund.