Diadectomorphs as amniotes
If Klembara et al.’s (2021) hypothesis that diadectomorphs are crown amniotes is correct, it has two major implications for amniote evolutionary history. The most obvious of these is the taxonomic augmentation of the early Amniota at the expense of stem amniote diversity. What is more impactful, however, is the identification of diadectomorphs as bona fide amniotes would finally cement high-fiber herbivory as a uniquely amniote adaptation: as anamniotes, diadectid diadectomorphs have been an evolutionary puzzle because they have been regarded as the only non-amniote herbivores in all of tetrapod evolutionary history (Sues and Reisz, 1998; but see Das, 1996; Hill et al., 2015). From a research point of view, a broad acceptance of the idea that diadectomorphs are amniotes would inspire wider research on this poorly studied group of early tetrapods (versus the tacit understanding that diadectomorphs are an evolutionary “dead end”).
However, Klembara et al.’s (2021) hypothesis that diadectomorphs are amniotes by way of a sister group relationship with synapsids, while attractive, is weakened by faulty character argumentation and flawed character scoring. One of the most curious faults of many of their characters states is that they are formatted strictly as absent or present states that do not adequately cover the observed range of morphologies for each character, and I was left with the conclusion that the authors either tried to pigeonhole a character state for a particular taxon or were forced to code that taxon as “?”. For example, their characters 157 and 158 concerned the morphology of the ventral surface of the parasphenoid. The former character was constructed such that a parasphenoid with a single median depression on its ventral surface represents the derived state, and any parasphenoid in which this morphology was absent was the primitive state. The latter character was constructed such that a parasphenoid with paired lateral depressions on its ventral surface represents the derived state and any parasphenoid in which this morphology was absent exhibits the primitive state. From my understanding of parasphenoid morphology of early tetrapods, the derived states of these two characters could not both manifest in the same taxon, and they can be combined together to form a multiple-state character. In addition, the primitive morphology for neither character is actually described, and this seems to me to be a curious, if unintentional, means to circumvent the outgroup comparison. This confusing type of character construction may also have led to patent character duplication with regard to Klembara et al.’s (2021) characters 251 (L-shaped proximal tarsal element absent or present) and 274 (tibiale, intermedium, and proximal centrale not fused to form an astragalus or fused to form an astragalus).
Striking problems I found with Klembara et al.’s (2021) study are that their data matrix is, at least with regard to the amniote taxa, riddled with scoring errors, and a seeming lack of awareness of the evolving literature on early amniotes. For example, Klembara et al. (2021) used Paleothyris acadiana as a representative reptile, and I presume (because they do not provide their sources for their scoring choices), they based their scores solely from Carroll (1969) description. However, Modesto (2006) concluded that one of Carroll (1969) referred specimens, one that provided most of the information for Carroll’s reconstruction of the skull of this reptile in the occipital view, is incorrectly assigned to P. acadiana and probably belongs to another taxon. Removal of this questionable specimen from the hypodigm of P. acadiana affects the scoring of 22 characters (i.e., 7% of the total characters) for this reptile. I would recode an additional 17 characters for P. acadiana, resulting in a total of 13% of Klembara et al.’s (2021) 294 phylogenetic characters incorrectly coded for this taxon (see Supplementary Appendix SA1 in online Supplementary Material for my comments on Klembara et al., ’s 2021 scorings for P. acadiana). In another example, Klembara et al. (2021) seem to have relied on Romer and Price (1940) to code the synapsids Eothyris parkeyi and Varanops brevirostris: for the former synapsid, I was able to score 21 out of the 24 characters that I rescored using Reisz et al. (2009), and for the latter, I was able to score 24 out of 26 characters that I re-scored using Williston (1911) and Campione and Reisz (2010, 2011); see Supplementary Appendices SA4, SA5 in the online Supplementary Material. In addition, I would rescore 24 and 26 characters for the reptiles Captorhinus aguti and Labidosaurus hamatus, respectively; these figures represent 8%–9% of Klembara et al.’s (2021) 294 phylogenetic characters.
Perhaps the greatest weakness in Klembara et al.’s (2021) phylogenetic conclusions is that these authors neglected to include any “microsaurs,” which is surprising in light of numerous previous studies that recover “microsaurs” (with other lepospondyls) closer to Amniota than to either seymouriamorphs, gephyrostegids, or chroniosuchids (Carroll, 1995; Laurin and Reisz, 1997; Ahlberg and Clack, 1998; Anderson, 2001; Ruta et al., 2003; Huttenlocker et al., 2013). For instance, many, if not most, “microsaurs” exhibit elliptical and elongate nostrils (Klembara et al.’s (2021) character 1), the presence of an oblique orientation of the anterior surface of the premaxilla (Klembara et al.’s (2021) character 5), a lacrimal with a total length that is less than two and a quarter times its maximum pre-orbital length (their character 27), the presence of parietal lateral lappets (Klembara et al.’s (2021) character 40), and the absence of a single median depression on the parasphenoid (Klembara et al.’s (2021) character 157). The inclusion of “microsaurs” might also help resolve the mis-homologizing of certain anatomical features, such as the identity of the anterior and posterior Meckelian foramina in amniotes, because “microsaurs” have only the latter (Carroll and Gaskill, 1978).
To conclude this part of the Discussion, my examination of Klembara et al.’s (2021) scorings for the amniote terminal taxa that appear in their analysis reveals a startingly incomplete phylogenetic survey of early amniotes. The most puzzling weakness of Klembara et al.’s (2021) phylogenetic analysis is the total omission of lepospondyls, particularly “microsaurs,” which have been recovered as close relatives of amniotes in several major phylogenetic evaluations of the interrelationships of early amniotes and their relationships to other Paleozoic tetrapods (e.g., Laurin and Reisz, 1995, 1997, 1999; Anderson, 2001; Vallin and Laurin, 2004). Given these problems, it is premature to accept Klembara et al.’s (2021) hypothesis that diadectomorphs are amniotes; however, I hope that it fosters increased interest in this group of Permo-Carboniferous tetrapods because the more detailed knowledge we have about this group, which is the sister taxon of traditional Amniota, the more confident we will be in elucidating the origins of amniote synapomorphies and the evidence they hold for reconstructing macroevolutionary advances made by the earliest amniotes.
Recumbirostrans as amniotes
Pardo et al.’s (2017) hypothesis that a cohesive group of “microsaurs,” recently identified and named Recumbirostra by Anderson (2007), is nested within Amniota as reptiles is the first cladistic support for mid-20th century taxonomic proposals that “microsaurs” are amniotes and closely allied with captorhinids (as “captorhinomorphs;” Westoll, 1942; Gregory, 1948; Vaughn, 1962).
If Pardo et al. (2017) are correct in identifying recumbirostrans (including lysorophians) as amniotes, their hypothesis not only implies that Permo-Carboniferous amniote taxonomic diversity was greater than generally appreciated but also that Amniota of this time was more ecologically diverse than conventional reconstructions of early amniote adaptations (and that anamniotes were less so). Early amniote ways of life include insectivory (Modesto et al., 2009), carnivory (Brink and Reisz, 2014), high-fiber herbivory (Sues and Reisz, 1998), amphibious (Romer and Price, 1940; but see Laurin and de Buffrénil, 2016), aquatic (Modesto, 2010), and facultatively bipedal (Berman et al., 2000, 2021) adaptations. However, evidence of a fossorial way of life in unequivocal amniotes does not appear until the Late Permian (Smith, 1987). Most recumbirostrans are fossorial forms (Anderson, 2007; Maddin et al., 2011); if they are amniotes, as envisioned by Pardo et al. (2017), this implies that fossoriality evolved in amniotes as early as, and in concert with, the evolution of high-fiber herbivory in amniotes, at the end of the Carboniferous (approximately 300 million years ago).
Recumbirostra includes “microsaurs” such as ostodolepids, gymnarthrids, Pantylus cordatus, Rhynchonkos stovalli, and their close relatives (Anderson, 2007; Szostakiwskyj et al., 2015). Most recumbirostrans exhibit stoutly constructed skulls with upturned snouts that are analogous with those of recent head-digging fossorial amniotes (Maddin et al., 2011), and most recumbirostrans for which we have postcrania exhibit elongate bodies and limb reduction (similar to fossorial squamates). Among unequivocal fossil amniotes, the earliest possible evidence for head-digging appears in the Late Permian with cistecephalid therapsids (synapsids; Kammerer, 2021). Recumbirostrans, as reptiles, would accelerate the acquisition of head-digging fossorial ways of life in amniotes by roughly 40 million years.
There is, however, some doubt regarding the phylogenetic evidence for placing Recumbirostra within Amniota. As noted above, Marjanović and Laurin (2019) remarked that whereas Pardo et al. (2017) introduced dozens of new braincase characters of phylogenetic utility, their analysis is relatively poor in postcranial characters. Marjanović and Laurin (2019) also mused whether the placement of Recumbirostra within Amniota might be an artifact of taxon and character sampling, noting that the sample of “microsaurs” was not representative of their traditional diversity (i.e., microbrachids were not included) and that nectridean lepospondyls were absent altogether.
In the course of this review, I examined Pardo et al.’s (2017) scorings for the animals with which I am most familiar: captorhinid reptiles. Interestingly, Pardo et al. (2017) used only two captorhinid species (Captorhinus laticeps and Opisthodontosaurus carrolli) as terminal taxa in their work, ostensibly those of “well-preserved and completely described neurocranial anatomy” (Pardo et al., 2017, Methods). I am somewhat puzzled by this statement with regard to the inclusion of Opisthodontosaurus carrolli because actually, very little of the braincase of this captorhinid is available in the literature (Reisz et al., 2015). I am also concerned with Pardo et al.’s (2017) conclusion that the braincase of Captorhinus laticeps is well-preserved and completely described because the anatomy of this captorhinid has a checkered history of which those authors appear to be unaware.
With regard to O. carrolli, the parabasisphenoid is the only braincase element that was described and illustrated by Reisz et al. (2015), and it is exposed only in ventral view in those specimens in which it is preserved. Although technically not a braincase element, the stapes is known from a single specimen and only in the ventral view. The preservation of these two elements in a single view each belies Pardo et al.’s (2017) statement of a “well-preserved and completely described neurocranial anatomy” for O. carrolli, and as a result, Pardo et al. (2017) were able to code this captorhinid for only 3 of their 58 new braincase characters. All characters considered, Pardo et al. (2017) were able to code O. carrolli for 124 (33.5%) of their 370 phylogenetic characters. I would recode 64 of these characters (see Supplementary Appendix SA6 in the online Supplementary Material).
Pardo et al.’s (2017) use of Captorhinus laticeps as a terminal taxon, particularly one that purportedly has a well-preserved and well-documented braincase material, is problematic in as much these authors appear to be unaware of the background concerning Heaton’s (1979) work on the cranial anatomy of this captorhinid. Heaton’s (1979) work was carried out in a pre-cladistic context, in which he regarded C. laticeps (then “Eocaptorhinus” laticeps) as morphologically indistinguishable from C. aguti, apart from the presence of multiple rows of marginal teeth in the latter. As a result, he freely used the material of small captorhinids from the Richards Spur locality to supplement his description of C. laticeps. However, C. laticeps is not listed in faunal accounts for the locality (e.g., Evans et al., 2009), and the Supplementary Material probably (as remarked by Heaton, 1979) belongs to C. aguti. Accordingly, some of the cranial anatomy attributed to C. laticeps is the suspect: the prootic, the dorsal surface of the parabasisphenoid, the cavum cranii, the anterodorsal portion of the supraoccipital, the membranous labyrinth of the ear, and the ventral surface of the parietal are among the cranial regions of C. laticeps that were described and reconstructed by Heaton (1979) but should not be used to code this captorhinid for phylogenetic characters concerning such areas. I have noted scorings concerning these cranial regions in the online Supplementary Material. Looking through Pardo et al.’s (2017) line of scoring for C. laticeps, I would correct 64 (17%) of their 370 characters (see Supplementary Appendix SA7 in the online Supplementary Material).
Further investigating Pardo et al.’s (2017) scoring of C. laticeps, they appear to have relied entirely on Heaton (1979) for information on C. laticeps but were not critical of his usage of Richards Spur specimens to supplement his description. They also appear to have misinterpreted certain aspects of the cranial anatomy of C. laticeps, leading, in particular, to mis-homologizing the dorsal anatomy of the supraoccipital of this captorhinid. This area of the braincase is not fully accessible in material of C. laticeps that was available to Heaton (1979) because it comprised partial skeletons and articulated skulls. One skull exhibited a transverse break that afforded Heaton (1979) fortuitous access to the inner cranial morphology, such as the ventral surface of the frontals and some of the surrounding skull roof bones, and access to the dorsal surface to the palate and its interaction with the antorbital buttress. However, the dorsal region of the braincase and the cranial surface were not accessible, and as a result, the braincase reconstruction in figure 30 of Heaton (1979) was supplemented with information from Richards Spur specimens. The braincase of C. laticeps in the occipital view is largely based on that published by Price (1935); Heaton (1979) even reconstructed the posterior profile of the supraoccipital of C. laticeps as relatively low, although the opisthotic–supraoccipital specimen that Price (1935) used was formerly broken and then poorly repositioned when repaired. As a result, Price’s (1935) reconstruction of the supraoccipital presents a dorsoventrally shorter element than he would have had to work with if he had repositioned the opisthotic and the supraoccipital together prior to joining them (figure 7 in Modesto, 1998). A problem with Pardo et al.’s (2017) examination of Heaton’s (1979) C. laticeps braincase is that they (their characters 218 and 219; see online Supplementary Material) have misinterpreted the median and lateral ascending processes of the supraoccipital, extensive processes that rise directly from the dorsal surface of the supraoccipital: the median process extends upward as a somewhat V-shaped extension (see figure 8 in Modesto, 1998) that makes contact with the posteroventral surface of the skull table; the lateral ascending processes extend dorsolaterally from the dorsolateral corners of the supraoccipital plate to contact the posterior margin of the skull roof (figure 8 in Modesto, 1998). Richards Spur specimens show this morphology much more completely than the material available to Heaton (1979). However, Szostakiwskyj et al. (2015) equated the median ascending process of the supraoccipital in C. laticeps with a feature that they identified as median ascending processes in Rhynchonkos stovalli, Aletrimyti gaskillae, and Dvellecanus carrolli. However, the structure in these recumbirostrans and that reconstructed for C. laticeps look nothing alike; the median ascending process of these microsaurs appears to be a low peak of the bone, judging from figures 6a, 9a, and 15f, g in Szostakiwskyj et al. (2015), rather than the conspicuous process seen in captorhinids (Heaton, 1979: fig. 27; Modesto, 1998: fig. 8). Furthermore, Szostakiwskyj et al. (2015) mistakenly identified a median ascending process on the supraoccipital of the diapsid reptile Petrolacosaurus kansensis, but this is merely the medial portion of the dorsal margin of the bone (figure 7b in Reisz, 1981) that looks nothing like the dorsomedial process in captorhinids (figure 8 in Modesto, 1998). Notably, Szostakiwskyj et al. (2015) identified a region on the supraoccipital that extends anteriorly and horizontally as the lateral ascending process and homologized it with the lateral ascending process in C. laticeps, which extends dorsolaterally from the main body of the supraoccipital in this captorhinid to underlie the ventral portion of the postparietal and form the dorsomedial border of the post-temporal fenestra. The “lateral ascending process” appears to form the posterolateral wall of the braincase in these recumbirostrans (figure 12 in Szostakiwskyj et al., 2015) and thus differs greatly in shape and in its contacts with neighboring elements. Given the structural differences in these two features in recumbirostrans and C. laticeps, I believe that the “lateral ascending process” in the latter taxon should not have been homologized with that attributed to recumbirostrans by Szostakiwskyj et al. (2015).
To conclude this part of the discussion, I infer that Pardo et al.’s (2017) phylogenetic recovery of recumbirostrans within Amniota is partly attributable to narrow taxon selection among both “microsaurs” and amniotes, particularly captorhinids, misinterpretation of captorhinid anatomy, and haphazard scoring of the captorhinid terminal taxa. It is striking that Pardo et al. (2017, page 5) stated that their taxon selection prioritized tetrapods with “well-preserved and completely described neurocranial anatomy,” but one of the captorhinid species they included in their analysis (Opisthodontosaurus carrolli) is one of the most poorly known early reptiles in this respect. As discussed above, I believe that Pardo et al. (2017) also appear to have mis-homologized the median and lateral ascending processes of captorhinids with quite different structures on the supraoccipitals of the recumbirostrans with which they were familiar. A more rigorous examination of the recumbirostrans-as-amniotes hypothesis will require the inclusion of additional captorhinid terminal taxa, particularly species such as Captorhinus aguti and Labidosaurus hamatus, for which we have a better preserved and documented braincase material than the poorly known O. carrolli and the problematic C. laticeps, as well as a selection of non-recumbirostran “microsaurs,” notably Microbrachis and Tuditanus, for a fair assessment of the amniote affinities of the Recumbirostra.