Phylogenetic analysis

The initial analysis (full character sampling) recovered 198 most parsimonious trees (MPTs) with a length of 304 steps. As noted in Materials and methods, leaving multistate characters unordered did not produce changes to the strict consensus topology (as with Schoch, Reference Schoch2019), although it recovered 198 MPTs with a length of 300 steps (four fewer steps than with ordered multistate characters). Nanobamus macrorhinus was recovered as diverging after the Micropholidae (Eoscopus Daly, Reference Daly1994, Micropholis Huxley, Reference Huxley1859, Pasawioops Fröbisch and Reisz, Reference Fröbisch and Reisz2008) and before the Amphibamidae (Amphibamus Cope, Reference Cope1865, Doleserpeton Bolt, Reference Bolt1969) (Fig. 5.1, 5.2; Supp. Fig. 1). The three permutations with removal of character 17 (postparietal length; 72 MPTs, 296 steps), removal of character 30 (postparietal length, distinct from character 17; 144 MPTs, 298 steps), and removal of both characters (54 MPTs, 291 steps), produced similar topologies with comparable bootstrap and Bremer support for major nodes. The consensus trees for the permutation without characters 17 and 30 is presented in the main text (Fig. 5.3, 5.4); the single character removal permutations are included in Supplemental Figure 1. The overall consistency between permutations suggests that including both characters 17 and 30 at least does not exert substantial influence on the tree topology, but the characters both refer to proportions of the postparietal and are thus not fully independent. As such, we recommend that they either be combined into a single character modeled after character 30 (the more quantitative and binary of the two) or more clearly defined in a way that makes them fully independent in future analyses. In none of the permutations is there strong support for the position of N. macrorhinus. Our character change for the naris of Georgenthalia clavinasica did not affect the tree topology compared to analyses where it was coded for the primitive condition as by Schoch (Reference Schoch2019).

These results, including the instabilities and variable position of some taxa, are consistent with the findings of Schoch (Reference Schoch2019). Formally defined clades (e.g., amphibamids, branchiosaurids) are the best supported. Here, Georgenthalia clavinasica was always recovered as the sister taxon to Gerobatrachus hottoni but with weak support in all permutations (Fig. 5.1, 5.2). Platyrhinops lyelli was always recovered as an early diverging amphibamiform rather than at the base of the Amphibamidae, a position found in some MPTs from Schoch's analysis. Tersomius texensis was recovered as the sister taxon to Pasawioops mayi, with this pair being sister to Micropholis stowi (Fig. 5.1, 5.2). This differs slightly from the results of Schoch (Reference Schoch2019) in which Pa. mayi and Micropholis stowi formed the sister clade to T. texensis. Instabilities in some parts of the amphibamiform tree likely reflect widespread convergence and variable combinations of retained plesiomorphic states. Although the position of Nanobamus macrorhinus was not strongly supported, it was recovered as the sister taxon to micropholids and as diverging prior to amphibamids in all permutations of the analysis. Weak support for N. macrorhinus was likely driven by incompleteness of the coding, specifically with respect to many features of greater informative value among amphibamids (e.g., many features of the palate and the dentition). The recovered position is unsurprising based on the combination of retained plesiomorphies (many of which are plesiomorphic for Dissorophoidea at large [e.g., nonposteriorly-truncated postparietals]) and derived features (often shared only with derived amphibamiforms [e.g., slender palatal rami of the pterygoid]).

It is important to note the continued instability in certain areas of the amphibamiform phylogeny, as evidenced by disparity between studies of the past decade (e.g., Fröbisch and Reisz, Reference Fröbisch and Reisz2008; Anderson and Bolt, Reference Anderson and Bolt2013; Maddin et al., Reference Maddin, Fröbisch, Evans and Milner2013). For example, Georgenthalia clavinasica has been variably recovered as: (1) stemward of Eoscopus locklardi and most other amphibamids (Fröbisch and Reisz, Reference Fröbisch and Reisz2008); (2) the sister taxon of branchiosaurids (Maddin et al., Reference Maddin, Fröbisch, Evans and Milner2013); (3) the sister taxon to a grouping of Plemmyradytes shintoni and Tersomius dolesensis (Anderson and Bolt, Reference Anderson and Bolt2013); (4) the sister taxon to the amphibamids sensu stricto (Sigurdsen and Bolt, Reference Sigurdsen and Bolt2010); and (5) the sister taxon of the stem batrachian Gerobatrachus hottoni (see Schoch, Reference Schoch2019). This might in part relate to incomplete taxon sampling within and between analyses and a correspondingly low number of parsimony-informative characters for differentiating the terrestrial nonlissamphibian amphibamiforms (27 in this analysis). The matrix of Schoch (Reference Schoch2019) that we used in this study does not include the poorly known amphibamiform genera, e.g., Rubeostratilia texensis (see Bourget and Anderson, Reference Bourget and Anderson2011) and P. shintoni (see Huttenlocker et al., Reference Huttenlocker, Pardo and Small2007), which Schoch (Reference Schoch2019) suggested as possible micropholids, or the poorly known species Tersomius mosesi Olson, Reference Olson1970 and Platyrhinops fritschi. Instability could also relate to the various combinations of retained plesiomorphies and apomorphies within the terrestrial amphibamiforms. For example, Geo. clavinasica and Nanobamus macrorhinus were not recovered as closely related, but this result poses an intriguing evolutionary scenario in which narial elongation evolved in an identical but apparently independent morphological fashion. Further work to assess the mosaicism and potential parallelism seen in amphibamiforms will be important for further resolution of their phylogenetic relationships.

Comparative amphibamiform morphology

Nanobamus macrorhinus is characterized by a unique combination of retained plesiomorphies and derived features. The particular combination observed in N. macrorhinus fills a morphological gap between primitive and derived amphibamiforms. For example, the postparietals (as interpreted here) are relatively long anteroposteriorly, the choana forming a long oval slit without prominent medial expansion along its length, and there is at least one pair of vomerine ‘fangs’ (Figs. 1, 3.1, 3.2). This is in contrast to more derived amphibamiforms and lissamphibians, which have foreshortened postparietals, more circular and medially expanded choanae, and vomers lacking ‘fangs’ (e.g., Sigurdsen and Bolt, Reference Sigurdsen and Bolt2010; Schoch, Reference Schoch2019). However, N. macrorhinus also shares a number of features with more derived amphibamiforms. Examples include a posterolaterally expanded parasphenoid with denticles extending onto the cultriform process and a slender palatal ramus of the pterygoid without a developed transverse flange.

Other than the elongate naris (discussed below), most differences between the skull roof of Nanobamus macrorhinus and those of other amphibamiforms pertain to minor variation in the proportions of elements and the composite effect that this has on overall skull shape and profile. Individual features that are rare among amphibamiforms but shared with at least one other taxon include: (1) narrow, posteriorly tapering frontals (shared with Rubeostratilia texensis; Bourget and Anderson, Reference Bourget and Anderson2011); (2) excavation of the lacrimal by a posteriorly expanded naris (shared with Georgenthalia clavinasica; Anderson et al., Reference Anderson, Henrici, Sumida, Martens and Berman2008a); and (3) prominent posteromedial expansion of the postfrontal to markedly constrict the parietals anteriorly (most similar to R. texensis and Platyrhinops lyelli in this regard; Clack and Milner, Reference Clack and Milner2010; Bourget and Anderson, Reference Bourget and Anderson2011). Collectively, this produces a relatively long and slender skull with a narrow interorbital region, in contrast to most other amphibamiforms (but similar in this regard to, e.g., Plemmyradytes shintoni and R. texensis). The orbits are proportionately large relative to the skull and noticeably longer than they are wide. They are similar in this regard to derived amphibamiforms, being slightly larger than those of Doleserpeton annectens and slightly smaller than those of G. clavinasica.

The palate of Nanobamus macrorhinus is most comparable to that of the amphibamids, with a large patch of denticles extending from the base of the cultriform process onto the process itself (Fig. 4). The extent (posterior third of the process) is similar to that of Doleserpeton annectens, whereas Amphibamus grandiceps also has a large patch anteriorly where the process meets the vomers. The pterygoids are slender and without a posterolaterally flaring transverse flange (in contrast to the plesiomorphic condition). The ectopterygoid is a long and narrow strut, although whether it was edentulous (as in, e.g., D. annectens) is unknown. The basal plate of the parasphenoid is much wider than it is long and expands prominently posterolaterally to closely approach the quadrate ramus of the pterygoid. However, the choana is not expanded anteromedially, forming a plesiomorphic elongate oval, and N. macrorhinus retains at least one pair of vomerine ‘fangs.’

Narial elongation in dissorophoids

The most conspicuous feature of Nanobamus macrorhinus is the elongate external naris, which is matched only by those of some trematopids in its length relative to the overall skull (Fig. 2). Although the external naris is superficially similar to that of trematopids, the construction in UCMP 203686 is fundamentally different. As in the amphibamiform Georgenthalia clavinasica, the posterior elongation excavates only the lacrimal (Anderson et al., Reference Anderson, Henrici, Sumida, Martens and Berman2008a), which widely separates the short prefrontal from the opening. As with G. clavinasica, this particular construction indicates nontrematopid affinities. In trematopids, both the prefrontal and the lacrimal are truncated as a result of the narial elongation, the lacrimal does not extend along the dorsal margin of the naris, and the prefrontal contributes substantially to the posterior narial margin. The prefrontal is highly variable among amphibamiforms in both total length and in relative anterior and posterior extents, but it is always widely separated from the naris, as in most temnospondyls. It should be noted that the prefrontal is narrowly excluded from the naris in a partial trematopid snout (MCZ 2531) described by Dilkes (Reference Dilkes1993), the preserved length of which is already longer than UCMP 203686. The naris extends posterior to the septomaxilla in MCZ 2531, but it does not have the keyhole-shaped profile in lateral view that is formed by the dorsal inflection of the maxilla in larger trematopids, instead being uniformly oval. In UCMP 203686, the inflection into the naris is along the dorsal margin, formed by a ventral inflection of the nasal and the lacrimal. The maxilla is dorsally expanded in UCMP 203686, but this occurs posterior to the naris (Fig. 2). This underscores the point that superficial similarities in anatomy can result from distinct, nonhomologous modifications to the skeleton that could influence phylogenetic analyses based on character construction and coding (Simões et al., Reference Simões, Caldwell, Palci and Nydam2017).

Among nontrematopid dissorophoids, elongate nares are also noted in the slender-headed morph of the micropholid Micropholis stowi from the early Triassic of South Africa (Schoch and Rubidge, Reference Schoch and Rubidge2005) and the branchiosaurid Leptorophus Schönfeld, Reference Schönfeld1911 from the early Permian of Germany (Boy, Reference Boy1985; Schoch, Reference Schoch2014). In neither of these taxa does the naris approach either the full length or the subdivided condition seen in Nanobamus macrorhinus, and only in Georgenthalia clavinasica is the construction similar, where anteriorly extensive processes frame the naris posteroventrally and posterodorsally. As previously noted, the phylogenetic (and geographic) separation of N. macrorhinus and G. clavinasica complicates a fuller understanding of this type of narial elongation. The lacrimal of Leptoropus tener (Schönfeld, Reference Schönfeld1911) is more abruptly truncated by the naris, and it lacks these circumnarial processes (Schoch, Reference Schoch2014). The branchiosaurid Schoendfelderpeton prescheri Boy, Reference Boy1986 is more similar to N. macrorhinus and G. clavinasica in having these processes, but that taxon has a more typical, circular naris. The narial elongation of Micropholis stowi is found only in the slender-headed morph of the taxon, and the lacrimal is either truncated abruptly or only slightly contacts the naris (e.g., Schoch and Rubidge, Reference Schoch and Rubidge2005).

The presence of narial elongation within dissorophoids thus continues to be an intriguing yet enigmatic feature. Notwithstanding the longstanding interpretations of Mordex calliprepes Steen, Reference Steen1938 and Ecolsonia cutlerensis Vaughn, Reference Vaughn1969 as dissorophids (both are now widely accepted to be trematopids), the aforementioned amphibamiforms represent the presence of an enlarged narial opening outside of Trematopidae. Nanobamus macrorhinus is the most extreme example of narial elongation in a nontrematopid, with the relative proportions closely approaching those of substantially larger trematopids, e.g., Acheloma Cope, Reference Cope1882, which represents an extreme end member on the trematopid spectrum. Regardless of the exact configuration, the functional import of elongation of the external naris in dissorophoids remains unresolved. It has only been explored in trematopids (e.g., Dilkes, Reference Dilkes1993), and the question remains open at present. The fact that narial elongation evidently evolved independently in both large- and small-bodied taxa and in both terrestrial and aquatic settings only confounds the issue further, because previous hypotheses of potential soft tissue structures (e.g., salt gland, Jacobson's organ) would not intuitively predict the absence of any clear ecological pattern or signal.

Ontogenetic maturity and Nanobamus

A consideration that merits brief discussion is whether ontogeny can be implicated in either the architecture of the naris or the broader mosaicism of cranial features in Nanobamus macrorhinus. It must be noted that this discussion is largely cursory because the ontogeny of most amphibamiforms is poorly known. This has produced uncertainty regarding the inferred maturity of other taxa (e.g., Bolt, Reference Bolt1977, Reference Bolt and Nitecki1979) that can only be more conclusively resolved when a large proportion of the skeletal anatomy is known from specimens spanning a broad size range (e.g., Schoch and Rubidge, Reference Schoch and Rubidge2005; Clack and Milner, Reference Clack and Milner2010; Sigurdsen and Bolt, Reference Sigurdsen and Bolt2010). As a result, it is difficult to precisely determine the maturity of UCMP 203686 because of the paucity of information for amphibamiforms and the absence of postcranial material associated with the specimen. It is relatively small (midline skull length of ~2 cm), although size is not a precise determiner of development maturity, and it is not the smallest known amphibamiform. Some aspects of the skull proportions, e.g., the relatively large orbits and narrow interorbital region, could suggest a degree of immaturity, at least in comparison to larger amphibamiform specimens. Conversely, other features, e.g., the presence of the palpebral ossifications, a relatively small pineal foramen, and prominent interdigitation along some sutural contacts (e.g., frontal-parietal suture; Fig. 1A), indicate a degree of maturity. Comparisons with the known ontogenetic sequences of other amphibamiforms are limited to a few taxa and are largely restricted to features that are not preserved in UCMP 203686, e.g., the relative ossification of the neurocranium and otic capsule (e.g., Sigurdsen and Bolt, Reference Sigurdsen and Bolt2010) or the pelvic girdle (e.g., Clack and Milner, Reference Clack and Milner2010). Clack and Milner (Reference Clack and Milner2010) did indicate that posterior expansion of the postfrontal (which is large for an amphibamiform in N. macrorhinus) characterized advanced ontogeny in Platyrhinops lyelli. Features that are applied to the ontogenetic maturity of temnospondyls more broadly (e.g., position of the pineal foramen) must be more carefully considered until they are assessed in amphibamiforms. At present, our interpretation of N. macrorhinus is that it undoubtedly represents a postmetamorphic individual that was probably relatively, but not fully, mature.

Based on our interpretation, it should be considered that incomplete ontogenetic maturation at the time of death could have produced the anatomy of Nanobamus macrorhinus. More precisely, differential timing of ossification and development of the different cranial regions and features, which does occur in temnospondyls more broadly (e.g., Boy and Sues, Reference Boy, Sues, Heatwole and Carroll2000), could have produced the combination of retained plesiomorphies and apomorphies that characterizes the taxon. This could apply to other amphibamiforms, particularly those for which no ontogenetic sequence is available. However, we reiterate that this requires additional material beyond the handful of taxa characterized from many specimens to be more precisely identify ontogenetic patterns among the terrestrial amphibamiforms. This also applies to the unusual narial architecture seen in N. macrorhinus, for which there is no evidence at present to support an ontogenetic influence. The most compelling evidence for an ontogenetic influence on the narial morphology comes from the long-snouted early Permian trematopid Phonerpeton Dilkes, Reference Dilkes1990 in which small individuals have circular, undivided nares that transition to elongate, ‘keyhole-shaped’ openings in larger individuals (Dilkes, Reference Dilkes1993). However, in Micropholis stowi, arguably the amphibamiform with the best-known ontogenetic trajectory (Schoch and Rubidge, Reference Schoch and Rubidge2005), there is no evidence for either pronounced elongation or late-stage subdivision of the naris (or the opposite, transformation toward a circular, undivided naris). Both Georgenthalia clavinasica and N. macrorhinus are known only from the type specimens, so it is not possible to evaluate any hypothesis of ontogenetic transformation in these taxa.

Amphibamiform biogeography

Amphibamiforms have long been of interest in the context of lissamphibian origins, resulting in a substantial amount of work regarding their morphology, ontogeny, phylogeny, taxonomy, and ecology. One key outcome of previous work is the growing consensus that branchiosaurids are a neotenic branch within Amphibamiformes that secondarily transitioned to an aquatic lifestyle (with occasional metamorphosis to terrestrial adults; e.g., Schoch and Milner, Reference Schoch and Milner2008; Fröbisch and Schoch, Reference Fröbisch and Schoch2009a; Maddin et al., Reference Maddin, Fröbisch, Evans and Milner2013; Schoch, Reference Schoch2019). An interesting aspect of this is the noticeable disparity in the relative abundance and geographical distribution of terrestrial amphibamiforms (predominantly found in North America) and branchiosaurids (predominantly found in Europe) in approximately coeval early Permian environments. There are only two terrestrial amphibamiforms found in Europe: Georgenthalia clavinasica from the upland early Permian Bromacker locality and Platyrhinops fritschi from the Carboniferous of Nýřany (see Clack and Milner, Reference Clack and Milner2010 for discussion of the ecology of Platyrhinops). Conversely, only two branchiosaurids have been found in North America (Eumicrerpeton parvum Moodie, Reference Moodie1909, Milnererpeton huberi Hunt, Lucas, and Berman, Reference Hunt, Lucas and Berman1996), and both are from the Carboniferous. Previous workers (e.g., Schoch and Fröbisch, Reference Schoch and Fröbisch2006; Schoch and Milner, Reference Schoch and Milner2008) have suggested that elevational disparity between environments inhabited by dissorophoids produced this pattern. Branchiosaurids are often found in high-elevation lake deposits of the Variscian orogen, which have been interpreted to be more seasonally disparate (e.g., Clausing and Boy, Reference Clausing and Boy2000). Lakes would have offered a relatively stable environment compared to lowland fluvial systems that would have produced strong selection pressures for alternative life strategies such as neoteny (e.g., Fröbisch and Schoch, Reference Fröbisch and Schoch2009b; Sanchez et al., Reference Sanchez, Steyer, Schoch and De Ricqlès2010). Conversely, dissorophoids in lowland environments (e.g., the Texas red beds) would not have experienced the same physiological constraints of pronounced elevation and thus radiated in terrestrial settings. A similar lack of constraints probably applied to so-called ‘upland localities,’ e.g., the Richards Spur (Oklahoma) and Bromacker (Germany), which might have been better drained than lowland sites but not at high elevations.

Recent work by Pardo et al. (Reference Pardo, Small, Milner and Huttenlocker2019) on global patterns of the Carboniferous-Permian transition (CPT) sheds further light on the geographic disparity. The divergence between branchiosaurids and terrestrial amphibamiforms can be constrained to the late Carboniferous. The fossil record of this time is dominated by wetland environments, including coal swamps, and members of both groups appeared in North America and Europe (e.g., Eoscopus locklardi, Platyrhinops lyelli, Apateon pedestris von Meyer, Reference Meyer1844). The late Carboniferous also marks the onset of a geographically diachronous climate transition across Pangea: the Carboniferous-Permian Transition (CPT), in which dryland environments replaced coal swamps. An onset of the spread of dryland environments in North America during the CPT (Pardo et al., Reference Pardo, Small, Milner and Huttenlocker2019) would have favored the radiation of terrestrial amphibamiforms relative to neotenic branchiosaurids, which thus explains the total absence of the latter in the Permian of North America. In contrast to, but concurrent with, this development in North America, the delayed replacement of wetland environments and the presence of high-elevation aquatic habitats in Europe would have favored the radiation of branchiosaurids. This pattern likely accounts for similar patterns seen in the larger, fully terrestrial olsoniforms, in particular the dissorophids, of which virtually all of the early Permian members (> 20 species) are found in North America. Further exploration of the environmental factors that influenced dissorophoid evolution will be necessary for a more complete appreciation of the ecological context in which crown lissamphibians evolved.