Precise occlusion and trophic niche differentiation indicate specialized feeding in Early Devonian jawed vertebrates

Specimens described in this study are housed in the University of Alberta Laboratory for Vertebrate Palaeontology (UALVP) collections. The calcareous matrix of argillaceous limestone in which fossils from MOTH are preserved was removed from the fossils using a solution of dilute acetic acid and soft brushes; fossils were stabilized using a 5% solution of Glyptal cement in acetone. The specimens were whitened using ammonium chloride sublimate and photographed using a Zeiss Discovery V8 stereomicroscope and Nikon NIS-Elements F 2.20 imaging software. Because the density of the matrix for MOTH specimens is quite high and interferes with CT scanning, the fossils were then embedded in Buehler EpoThin Low Viscosity Resin and Hardener and prepared from the buried side to remove the remaining matrix, separating the fossil entirely from the matrix to facilitate CT scanning. This was not completed for UALVP 45848 because of its status as the holotype specimen of E. rugosus.

Specimens UALVP 45648 and UALVP 48487, representing the ischnacanthiform species E. rugosus and T. gannitus, respectively (Fig. 2), were scanned using a SkyScan 1174 μCT scanner at 50 kV and 800 uA, with a slice thickness of 9 μm. Slice data were analysed using OsiriX, Mimics, and Avizo 6.3. A stereolithography (.STL) surface mesh of the three-dimensional geometry was created in Mimics from the stacked DICOM (Digital Imaging and Communications in Medicine format) images acquired by the CT scanner. Each .STL file was then imported into Geomagic Studio 6 software. Using the “Sculpt” tool, Geomagic was used to reconstruct broken or missing teeth, using intact specimens of the same species for reference. In particular, UALVP 45050 (fig. 3D in Blais et al. 2015) was used as a reference to reconstruct the broken teeth of UALVP 45648. To reconstruct the broken teeth and missing jaw cartilage of UALVP 48487, UALVP 45078, and UALVP 41920 were used as references (fig. 5 in Blais et al. 2015). The blank labial side of UALVP 45648 was also filled in as flat, as it was reconstructed on the lingual side only and not prepared out of its matrix because of its status as the holotype of the species.

Three-dimensional hollow surface meshes were created for the flattened specimens and manipulated to determine their occlusion. The surface meshes were then sculpted using Geomagic until the curvature of the cartilages in the ischnacanthiform jaw reconstructions (Fig. 3) more closely resembled the curvature of the jaws in related Devonian fossil specimens that were preserved in three dimensions or dorsoventrally exposed, including Ptomacanthus anglicus (Miles 1973), redescribed by Brazeau (2009); Doliodus problematicus (Woodward 1892), based on CT data presented by Maisey et al. (2009, 2014); and Gogoselachus lynbeazleyae (Long et al. 2015a). Although the more laterally flattened MOTH ischnacanthiform jaws may have been less curved than the jaws of these specimens in life, the purpose of curving the jaw reconstructions was to determine whether and how it affected the occlusion of the jaws, not to attempt to recreate an entirely accurate life position. The curved reconstructions were then re-examined to determine whether the interpretation of their occlusal style was significantly affected by their taphonomic deformation. CT scans are reposited at the University of Alberta and are also available with the original .STL files on figshare.com (Blais 2017a, 2017b).

The anterior and posterior labial flanges are well developed in E. rugosus, and the teeth occlude tightly together in this species (Figs. 4, 5). The large lateral teeth curve slightly medially, even in the curved reconstructions. When the anterior and posterior labial flanges of opposing teeth are in contact with each other, each tooth cusp is slightly “above” (either dorsal or ventral to) and medial to the opposing inter-tooth pit. When the teeth are in contact with each other in the central and anterior parts of the jaw as in the articulated specimens, there is very little open space left between each pair of opposing teeth. The smaller teeth at the anterior and posterior ends of the jaw bones do not contact each other. The anterior and posterior labial flanges form an almost continuous surface when the jaws are tightly closed, coming together like scissor blades. The jaw bones themselves are curved dorsoventrally, so that at the anterior end of the jaws the distance between the upper and lower jaw bone is greatest. Based on articulated ischnacanthiform specimens, this space was probably occupied by large tooth whorls anterior to both the palatoquadrates and Meckel’s cartilages. The tooth whorls are almost never preserved in situ, so it is difficult to determine how they may have affected the closure of the jaws.

The lateral teeth are broad, with longer posterior flanges than anterior flanges, and slightly recurved. When the lower jaw is moved relative to the upper jaw to close the jaws, the teeth occlude in a shearing motion. Each upper tooth cusp contacts and shears down the posterior surface of the opposing lower tooth until the posterior flanges contact each other.

The lateral teeth of T. gannitus are less broad than the teeth of E. rugosus, with smaller anterior and posterior labial flanges. The teeth are also longer relative to the overall size of the jaws and more widely spaced than those of E. rugosus. Because of this, the anteriormost teeth do not occlude as closely and there is space left between the opposing anteriormost teeth when the jaws are closed (Figs. 4, 6). The teeth near the center of the jaws occlude tightly, but the posteriormost teeth do not contact the posteriormost teeth on the opposing jaw. The anteriormost and the largest tooth cusps are prevented from contacting the inter-tooth pits on the opposing jaw bone by their associated anterior and posterior labial flanges, which come into contact with the base of the opposing tooth. If the jaw joint is disarticulated (Fig. 6), the space between the anteriormost teeth is reduced but the teeth still do not contact the opposing inter-tooth pits. The main points of contact of the anteriormost teeth appear to be the lower half of each of the tooth cusps with the anterior and posterior labial flanges of the opposing teeth. The teeth of T. gannitus also appear to occlude via a more vertical motion than the teeth of E. rugosus, but there is still an element of shear in the anteriormost teeth. There is also less distance between the anterior ends of the upper and lower jaws than in E. rugosus, which is likely related to the smaller anterior tooth whorls in this species.

Interestingly, the robust medial teeth of T. gannitus never come into contact with the medial teeth of the opposing jaw, regardless of how the jaws are manipulated. If the lateral teeth occlude in the interlocking manner indicated by the articulated specimens, the medial teeth cannot occlude. Instead, they project either dorsally or ventrally, and slightly medially, into the mouth cavity (Fig. 6).

The three-dimensional reconstructions created for the two species in this study reveal two different and distinct styles of jaw and tooth occlusion. As these species were coeval, the hypothesis that ischnacanthiforms can be grouped chronologically based on their jaw occlusion types (Burrow 2004) must, therefore, be rejected. Although the occlusion in T. gannitus is closer to the cog-like vertical tooth closure suggested by Burrow (2004) than is the occlusion in E. rugosus, neither species exactly fits the model proposed for Silurian-Middle Devonian ischnacanthiforms. The teeth of both species occlude in an interlocking fashion similar to that proposed by Ørvig (1973), with a precision unusual for fishes. Because these animals retained their teeth throughout their lives without replacing them, very precisely aligning teeth may have been advantageous. Conversely, if a single tooth was out of alignment with the rest of the teeth in the jaw, the animal’s ability to bite effectively would likely have been greatly impacted. In this sense, tooth occlusion in ischnacanthiform acanthodians may be more more comparable to that in most mammals and some other tooth-retaining tetrapods than that in polyphyodont taxa like chondrichthyans and most bony fishes.

Similar to the condition in mammals, it is possible that the differences in the tooth morphology and occlusion between the two species of ischnacanthiforms reflect differences in biting and feeding strategies, which were evidently more varied than expected in at least some Early Devonian ischnacanthiforms. The employment of different biting or chewing strategies by coeval ischnacanthiforms at the MOTH locality provides support for the hypothesis that these two species could have coexisted by partitioning trophic niches and specializing in prey sources. This also supports the hypothesis proposed by Choo et al. (2014) that early gnathostomes had likely undergone significant trophic niche differentiation by the Devonian.

In addition to the indication that the two species of ischnacanthiforms could have specialized in different food sources, the three-dimensional reconstructions also provide some insight into the possible ecological role of each species. The shearing, interlocking occlusion, fang-like curved tooth cusps, and sharp, smooth anterior and posterior labial tooth flanges present in E. rugosus (Fig. 5) suggest that it could have been an effective predator of large prey (relative to its body size), including soft-bodied animals. Tightly interlocking teeth have been suggested as an indicator of macrophagy in crocodylians (Young et al. 2012), and the anterior and posterior flanges would have formed an almost continuous slicing or cutting surface. The teeth of E. rugosus also have longitudinal ridges (Blais et al. 2015), which, in other taxa, have been suggested to indicate the presence of plicidentine (Meunier et al. 2014). Plicidentine is currently unknown in ischnacanthiforms, and a histological analysis of the teeth would be required to determine whether these ridges are the result of infolding of the dentine in Euryacanthus. Plicidentine has been hypothesized to strengthen the attachment of the teeth to the tooth-bearing element of the jaw, particularly in active predators (Scanlon and Lee 2002; Modesto and Reisz 2008; Maxwell et al. 2011a, 2011b; Meunier et al. 2014). Euryacanthus rugosus was also one of the largest ischnacanthiform acanthodians from the Early Devonian MOTH locality, although compared with ischnacanthiforms from other localities and ages, it was a small fish. Taken together, the relative size, jaw occlusion, and tooth morphology suggest an actively predatory ecological role for E. rugosus, which would be further supported if plicidentine was determined to be present.

Because E. rugosus has a single row of teeth on each jaw bone, the tooth occlusion in this species is relatively simple compared with that of T. gannitus. Tricuspicanthus gannitus has multiple rows of teeth on each jaw bone and is monognathically heterodont, with striking differences in tooth morphology and occlusion between its lateral and medial dentition (Figs. 4, 6). The lateral teeth of T. gannitus are long, widely spaced, and caniniform, typical of the “piercing graspers” of Bellwood et al. (2014). These teeth could have been well suited to puncturing soft prey. The medial teeth, however, are blunt, robust, and low crowned, typical of animals interpreted to be durophagous (Pregill 1984; Motani 2005; Choo et al. 2014; Crofts and Summers 2014). Two very different forms of teeth on the same tooth-bearing element in the jaw could be indicative of several possible feeding strategies. Perhaps T. gannitus exploited a wide range of food sources, using its lateral dentition to puncture softer prey and using its robust medial teeth to crush hard-shelled animals such as bivalves or arthropods, which were certainly present at the MOTH locality during the Early Devonian. Alternatively, it may have specialized in hard-shelled prey, grasping with its lateral teeth and crushing with its medial teeth. This type of strategy has been proposed for Megamastax amblyodus (Choo et al. 2014) and could have also been employed by T. gannitus.

It is interesting to note, however, that the medial teeth of the upper and lower jaws in this species cannot come into contact with each other, indicating that T. gannitus would have only been able to effectively crush shelled prey that was larger than the gap between the upper and lower medial tooth rows. This suggests that not only might T. gannitus have specialized in hard-shelled prey, but specifically large hard-shelled prey (“large” being relative to the size of T. gannitus, one of the smallest ischnacanthiforms from MOTH). Of the possible candidates for such prey known from the MOTH locality, eurypterids and molluscs seem the most likely. It is also possible that T. gannitus could have ingested small, subadult heterostracans or osteostracans, fossils of which have been recovered from the MOTH locality (Greeniaus and Wilson 2003; Hawthorn et al. 2008). Finally, it is possible that T. gannitus specialized in hard-shelled prey until it reached a certain size; perhaps while its lateral dentition remained relatively small, its medial dentition would have been more effective on smaller prey such as ostracods, which were also relatively abundant at the MOTH locality (Zorn et al. 2005). Judging from the relative abundance of specimens of T. gannitus compared with the other ischnacanthiform species from MOTH (Blais et al. 2015), whatever feeding strategy it employed seems to have been a successful one.

Tricuspicanthus gannitus has very tightly packed, needle-like lip scales that cover the posterior half of the jaws in articulated specimens (Blais et al. 2011). These lip scales appear to form a continuous cover, intermeshing with no obvious dividing line between the posterior part of the upper and lower halves of the jaws (Fig. 7). This indicates that the posterior region of the mouth in these animals was enveloped by a layer of skin, which would have supported the scales. Effectively, T. gannitus may have had cheeks. They also possessed tiny, sharp scales lining the inside of the oral cavity (Blais et al. 2015). If these ischnacanthiforms were eating soft-bodied prey or separating the soft parts from hard-shelled prey, perhaps the combination of a membranous cheek and sandpaper-like roughened internal oral surface combined to prevent the prey tissue from sliding out of the mouth during processing and position it more precisely for chewing. Taken together, these features suggest that postcapture processing of prey may have been present in T. gannitus.

According to the hypothesis posited by Ørvig (1973), the posterior tooth cusps in ischnacanthiform acanthodians were worn down to “stumps” by regularly coming into contact with the inter-tooth pit on the occlusal surface of the opposing jaw bone. If the tips of the posterior tooth cusps were being ground away by such contact, one would expect the surficial orthodentine layer of the apices of the posteriormost tooth cusps to be truncated and worn away, exposing the core of the tooth. Thin sections of ischnacanthiform dentigerous jaw bones from the MOTH locality have revealed that the surficial orthodentine layer covering the posteriormost teeth is continuous with that covering the anteriormost teeth (MacKenzie 2008). Although they are lower and more rounded than the anteriormost teeth, the posteriormost tooth cusps do not appear to be significantly worn relative to the anteriormost tooth cusps in MOTH ischnacanthiforms, and the cores of the posteriormost teeth are not exposed.

This lack of significant wear on the posteriormost tooth cusps relative to the anteriormost tooth cusps can be explained using the model of jaw and tooth occlusion proposed in this study. The main points of contact between teeth according to the three-dimensional reconstructions of T. gannitus and E. rugosus jaws were not the apices of the tooth cusps, but the anterior and posterior medial surfaces of the tooth cusps. These surfaces are smooth in most ischnacanthiforms from the MOTH locality; in species with medial denticles or cuspules associated with each tooth cusp, these are medial to the center of the tooth cusp, with smooth unornamented regions on the anterior and posterior surfaces.

A newly discovered Silurian ischnacanthiform specimen from the MOTH locality (Blais, in review) exhibits evidence of tooth-on-tooth wear in these animals. Although not a member of one of the species for which the three-dimensional reconstructions were created, it was likely closely related, based on similar tooth and jaw morphology. It is likely that its jaws could have occluded in an interlocking, shearing manner similar to that of E. rugosus. It has prominent wear facets on the anterior and posterior lingual surfaces of its lateral tooth cusps that do not extend to the apex of the tooth cusp. This wear pattern supports the manner of occlusion proposed here, in which the teeth interlock and the tooth cusps do not come into contact with the inter-tooth pits of the opposing jaw bone. This evidence of interlocking tooth–tooth contact also supports the hypothesis that these fishes were likely active predators consuming relatively large prey, and that acanthodians were specialized, predatory feeders by the Late Silurian.