The evolution of the lepidosaurian lower temporal bar: new perspectives from the Late Cretaceous of South China

Until recently, it was considered axiomatic that the skull of lizards and snakes arose from that of a diapsid ancestor by loss of the lower temporal bar. The presence of the bar in the living New Zealand Tuatara, Sphenodon, was thus considered primitive, corroborating its status as a ‘living fossil’. A combination of new fossils and rigorous phylogeny has demonstrated unequivocally that the absence of the bar is the primitive lepidosaurian condition, prompting questions as to its function. Here we describe new material of Tianyusaurus, a remarkable lizard from the Late Cretaceous of China that is paradoxical in having a complete lower temporal bar and a fixed quadrate. New material from Jiangxi Province is more complete and less distorted than the original holotype. Tianyusaurus is shown to be a member of the Boreoteiioidea, a successful clade of large herbivorous lizards that were dispersed through eastern Asia, Europe and North America in the Late Cretaceous, but disappeared in the end-Cretaceous extinction. A unique combination of characters suggests that Tianyusaurus took food items requiring a large gape.


INTRODUCTION
The reptilian group Lepidosauria encompasses Squamata (lizards, snakes and amphisbaenians) and Rhynchocephalia (Sphenodon and its fossil relatives). Living members of these two subgroups differ in several respects, but the one most widely cited is the presence in the Sphenodon skull of a complete lower temporal bar. For more than a century, the classic view held that the diapsid skull of the ancestral lepidosaur experienced gradual reduction in the lower temporal bar, thereby 'freeing' the quadrate (streptostyly) in squamates (e.g. Romer 1956;Robinson 1967). In fact, research over the last 30 years has demonstrated unequivocally that lepidosaurs inherited a skull without a lower temporal bar, although the quadrate was firmly fixed to the skull by the pterygoid and squamosal (e.g. Whiteside 1986;Evans 2003Evans , 2008Mü ller 2003). The ancestral lepidosaurian skull architecture was modified differently in each descendant group. The fully diapsid skull of Sphenodon is secondary ( Whiteside 1986); one or more rhynchocephalian lineages developed a posterior jugal process that ultimately contacted an enlarged anterior quadratojugal process. Squamates, on the other hand, reduced the quadrate/pterygoid and quadrate/ squamosal contacts in the evolution of streptostyly. However, so persuasive was the original paradigm with respect to the lower temporal bar that this character in Sphenodon is still widely discussed using polarity-related terms such as 'loss' or 'retention' (e.g. Cleurens et al. 1995;Herrel et al. 1998Herrel et al. , 2007Schwenk 2000;Metzger 2002;Lü et al. 2008). The recent description of a fossil lizard with a complete lower temporal bar, Tianyusaurus zhengi ( Lü et al. 2008), thus seems almost paradoxical, but it highlights the lability of this much-misunderstood character.
Tianyusaurus was described on the basis of a single medium-sized specimen, reportedly from the Late Cretaceous of Henan Province, China. Here, we extend the description with three skulls from contemporaneous deposits in Jiangxi Province ( Nanxiong Formation, ca 66 Myr; see the electronic supplementary material). These skulls are more complete and less distorted than the holotype, and permit this remarkable lizard to be placed more fully into phylogenetic and functional context. Two skulls are significantly larger than the holotype and one is smaller.

DESCRIPTION AND COMPARATIVE ANATOMY
Squamata (Oppel 1811) Boreoteiioidea ( Nydam et al. 2007) T. zhengi (Lü et al. 2008 Diagnosis (emended and extended from Lü et al. 2008). A large squamate, with a maximum midline skull length over 87 mm, which resembles other boreoteiioid squamates (sensu Nydam et al. 2007) in having vomers and pterygoids meeting in the palatal midline; suborbital fenestra reduced by expansion of the ectopterygoid; no palatal dentition; welldeveloped pterygoid lappet on quadrate; large medially directed adductor fossa; deep vertical pterygoid flanges; clavicle greatly expanded with fenestrate medial end. Resembles many Asian boreoteiioids in having postcaniniform teeth labiolingually compressed and multicuspid; resembles Gilmoreteius, Tuberocephalosaurus and Aprisaurus (Alifanov 2000) in having enlarged maxillary caniniform teeth. Differs from other boreoteiioids and all other squamates in the following combination of derived characters: long posterior process of jugal forming complete lower temporal bar; tympanic crest of quadrate rugose with expanded anteroventral margin; deep fixed quadrate-pterygoid overlap; squamosal and supratemporal fused or partially fused and firmly attached to quadrate dorsal condyle; premaxilla excluded from ventral narial margin by expanded anterior maxillary process; upper temporal fenestra small; lower temporal fenestra and post-temporal fenestra large; lower jaw relatively shallow so that the deep vertical pterygoid flange extends below the lower margin of the closed mandible. Lü et al. (2008) described the main features of the skull of Tianyusaurus. Here, we focus on details not reported, or misreported, in the original description and those features of the skull that, together, seem to contribute to a functional explanation of the unusual morphology. More detailed descriptions and illustrations of individual Jiangxi specimens can be found in the electronic supplementary material.

MORPHOLOGY
The skull of Tianyusaurus is fully diapsid with small dorsally placed upper temporal fenestrae and large posttemporal and lower temporal fenestrae (figures 1 and 2). The choana is not subdivided and the suborbital fenestrae of mature specimens are reduced in width by expansion of the ectopterygoid and pterygoid. Lü et al. (2008) figured the holotype skull as anteriorly deep and described the post-temporal fenestrae as dorsoventrally shallow and the jugal bar as bulging. However, that skull has been distorted by lateral compression and breakage. The Jiangxi Tianyusaurus specimens have shallower skulls, not unlike those of Sphenodon in general shape, and, judging from the breakage patterns, they are closer to the original shape.
The skull is robust with strong sutures. Anteriorly, the maxilla forms nearly three-quarters of the narial margin. Its long, dorsally concave premaxillary process expands anteriorly to brace the premaxilla, meeting the contralateral maxilla in the posterior midline. The facial process has a procumbent anterodorsal edge overhanging the naris. In NHMG 9317 and NHMG 9316, the postorbital is separated from the parietal by a slender supratemporal, but in NHMG 8502 the post-orbital . By comparison with extant taxa, this crest marked the anterior limit of origin of the aponeurosis to which superficial fibres of the pterygoideus muscle were attached. However, a groove running posterior and parallel to this crest suggests fibres, perhaps of the deep pterygoideus, passed around the pterygoid neck from a dorsomedial attachment. This would increase their functional length. The sphenoid and basioccipital are fused in mature specimens, but the latter is the shorter element in NHMG 9316. It has a large laterally placed basal tubera. In the sphenoid, a short vidian canal perforates the base of each basipterygoid process from a large posterior foramen. The supraoccipital has a steep posterodorsal margin with a median column that abuts the parietal (figures 1f and 2b; electronic supplementary material). In the oto-occipital, the metotic fissure is divided, as in all squamates, into a small vagus foramen and an elongate lateral opening of the recessus scalae tympani. The latter is framed by a low crista tuberalis and by a narrow occipital recess in the basioccipital. The long, posterolaterally directed paroccipital processes have expanded distal tips meeting the suspensorium. One or two hypoglossal foramina perforate the exoccipital. The prootic has a well-developed crista prootica.
The dentary is supported ventrally by a large angular and braced medially by a plate-like splenial extending from the symphysis to the coronoid. The latter has a low coronoid process, a small lateral lappet and a posterior crest. The surangular is shallow anteriorly, but deepens posteriorly. Its most notable feature is a strong lateral crest that divides the bone into a shallow ventrolateral surface that meets the angular and a broad dorsal shelf or platform that is angled from dorsomedial to ventrolateral. The platform is most marked in mature specimens and bears a posteriorly deepening concavity such that the posterodorsal surface is almost horizontal (see figures and cross section in fig. S4 in the electronic supplementary material). Its lateral edge is in line with the lower temporal bar (rather than bowed) and marks the ventrolateral edge of the adductor chamber (NHMG 9316). By comparison with modern lizards, this broad surface gave insertion to fibres of the 1b portion of the MAMES and perhaps part of the medialis component (MAMEM). The inner edge of the platform forms the dorsal rim of a large medially directed adductor fossa (MAME posterior), the ventral margin of which is formed by the long pre-articular. A shallow anteromedial to posterolateral groove curves onto the lateral jaw surface and probably marks the course of the main (superficial) part of the pterygoideus muscle, whereas a concavity along the ventromedial margin of the articular, associated with a short medially directed crest, probably accommodated deep pterygoideus fibres (Throckmorton 1976(Throckmorton , 1978. The articular is short, with a transverse joint surface that is inclined at an anterodorsal to posteroventral angle to the long axis of the jaw. The retroarticular process is reduced to a tubercle in adult specimens. The premaxilla bears 6-7 small teeth, the dentary approximately 33 and the maxilla approximately 24. The smaller anterior teeth and the large maxillary caniniforms are monocuspid. The post-caniniforms are deeply pleurodont, labiolingually compressed and have expanded multicuspid crowns. The maxillary teeth are most clearly preserved in NHMG 9316 (figure 1d ) and are asymmetrical, with a large posterior cusp and a series of smaller cusps (usually four) along the oblique anterior edge. The cuspidation on the dentary teeth is more symmetrical.
The smallest skull, NHMG 9316, shows differences that are probably ontogenetic: retention of braincase sutures; weaker development of crests and rugosities; more rounded upper temporal fenestrae; a larger suborbital fenestra; no maxillary caniniforms; a less inclined quadrate; a weaker jugal-quadrate contact; and a small remnant of the retroarticular process. Lesser differences exist between NHMG 8502 and NHMG 9317, and between these large adults and the Henan holotype, but, allowing for preservation and taphonomy, these are most parsimoniously interpreted as individual variation.
We coded Tianyusaurus into the data matrix of Conrad (2008) and ran a preliminary heuristic analysis (PAUP* v. 4.0b10; Swofford 2002) with 148 exemplars (364 characters) representing all major squamate lineages and several fossil taxa. All but 20 characters were run unordered (as in Conrad 2008), multistate taxa were treated as uncertainty, and Kuehneosauridae, Marmoretta and Rhynchocephalia were outgroups. The strict consensus of 1501 equally parsimonious trees (LZ2969, CIZ0.169, RCZ0.111) unequivocally placed Tianyusaurus with Boreoteiioidea (sensu Nydam et al. 2007; Polyglyphanodontidae of Conrad 2008). A branch and bound analysis was then performed on a subset of 19 fossil and living teiioid and boreoteiioid taxa, with Lacertidae as the designated outgroup. It yielded four maximum parsimony trees (LZ354, CIZ0.588, RCZ0.311) in which the Tianyusaurus was consistently placed with the Mongolian genera Tuberocephalosaurus and Aprisaurus, and a second, as-yet unnamed Jiangxi lizard ( Jiangxi 2 in figure 3a), in a subgroup of boreoteiioids. Boreoteiioidea emerged as the monophyletic sister group of TeiidaeCGymnophthalmidaeCChamops (Late Cretaceous, North America), a result closer to that of Nydam et al. (2007) than to Conrad (2008). However, bootstrap and decay analyses provided only weak support for all clades except Gymnophthalmidae and that comprising the tuberocephalosaurus (including Tianyusaurus and the second Jiangxi taxon), and a larger analysis of boreoteiioid relationships is needed.

DISCUSSION (a) Phylogenetic relationships
Tianyusaurus is nested within Boreoteiioidea ( Nydam et al. 2007), a successful group of mainly herbivorous Late Cretaceous lizards encompassing the Euramerican transverse-toothed polyglyphanodontines (e.g. Polyglyphanodon; Gilmore 1942) and a more diverse East Asian assemblage including Adamisaurus (Sulimski 1972), gilmoreteiids (macrocephalosaurines;Langer 1998), and the smaller 'mongolochamopines' (Alifanov 2000), the monophyly of which is untested. Boreoteiioids are first recorded in the Neocomian of Japan (Evans & Manabe 2008) and their roots may be Asian. They represent the largest known radiation of herbivorous lizards, but also the only Late Cretaceous terrestrial lizard clade that did not survive the K-T extinction.

(b) Functional morphology
Tianyusaurus shares many traits with other large Asian boreoteiioids (vomer-pterygoid contact, large jugal, reduced suborbital fenestra, posteriorly extended postorbital, long paired frontals, short parietal, probable absence of meso-and metakinesis), but it has a shallower skull and lower jaw, longer maxilla, less massive jugal, more extensive quadrate-pterygoid overlap and no retroarticular process. With the exception of Polyglyphanodon (Gilmore 1942), no other boreoteiioid is known to elongate the posterior jugal ramus, and none appears to have completely immobilized the quadrate. Many living herbivorous lizards (and presumably boreoteiioids) stabilize their quadrates during powerful static biting using a combination of temporal ligaments and muscle action (Cleurens et al. 1995;Herrel et al. 1998). This is possible because the lizard quadrate generally rotates posteriorly during jaw closure (e.g. Cleurens et al. 1995;Herrel et al. 1998;Schwenk 2000;Moazen et al. 2008), putting the temporal ligaments into tension (Herrel et al. 1998) as the animal bites. By contrast, in crocodiles (Cleurens et al. 1995) and Sphenodon ( Jones 2008;Schaerlaeken et al. 2008), forces on the quadrate tend to be directed anteriorly during biting, albeit for different reasons in each, putting the lower skull margin into compression. Under those conditions, a bony lower temporal bar rather than a ligament provides a better mechanical solution (Cleurens et al. 1995). Lü et al. (2008) concluded that the same argument must apply to Tianyusaurus. We agree that this may be the case, but it raises the question as to why the skull of Tianyusaurus was loaded so differently from that of its immediate relatives. A tendency towards predominantly anterior rather than posterior rotation of the quadrate during biting could result from one of several changes in morphology and /or feeding strategy: (i) a cropping action involving strong backward head movements needed to pull plant material from its source (and thus a strong anteriorly directed food resistance force), (ii) an increase in bite force concomitant with an anteriorly directed food reaction force (as in examples cited by Cleurens et al. 1995 andHerrel et al. 1998), (iii) the development of pro-oral shear (as in Sphenodon: Gorniak et al. 1982;Jones 2008), or (iv) the development of a powerful pterygoideus muscle (Cleurens et al. 1995). A cropping action cannot be ruled out for Tianyusaurus, but the small size of the anterior teeth and the absence of any strong occipital crests suggestive of powerful neck muscles render this unlikely as a primary explanation. The large lower temporal and subtemporal fenestrae, expanded post-temporal fenestrae, large medially directed mandibular adductor fossa and robust anterior quadrate rim indicate that Tianyusaurus had well-developed external adductor muscles, but probably no more so than its boreoteiioid relatives (with deeper, apparently more powerful jaws). The broad subhorizontal surangular platform in Tianyusaurus suggests the superficial (1b) part of the external adductor was thick but, contra Lü et al. (2008), the lower temporal bar ran in line with the edge of the surangular platform rather than bowing laterally. This would have limited extension of MAMES onto the lateral surface of the mandible (Rieppel & Gronowski 1981), restricting fibre lengths and, potentially, bite force (Schaerlaeken et al. 2008), although the latter needs confirmation. The only jaw muscle with an alignment capable of exerting powerful anteriorly directed forces on the quadrate-mandible during jaw closure is the pterygoideus (Throckmorton 1978). In Sphenodon, this powerful muscle is primarily responsible for anterior (pro-oral) translation of the mandible, as permitted by an elongated joint surface on the articular (Robinson 1976). The de novo development of a lower temporal bar in Sphenodon is generally linked to the need to stabilize the quadrate during pro-oral shearing (Gorniak et al. 1982;Jones 2008;Schaerlaeken et al. 2008). A similar explanation might seem appropriate for Tianyusaurus, except that its short, posteroventrally inclined quadrate-articular joint was incapable of anteroposterior sliding. Nonetheless, pterygoideus can contribute to both jaw opening and jaw closure ( Frazzetta 1962;Throckmorton 1978;Gorniak et al. 1982;Sinclair & Alexander 1987;Schwenk 2000;Reilly et al. 2001;McBrayer & White 2002), and the Jiangxi Tianyusaurus skulls show clear traces of deep and superficial pterygoideus origin and insertion (anterior pterygoid crest, medial ridge and recess on articular, lateral surangular crest and strong pterygoid flange).
Other unusual features of the Tianyusaurus skull morphology may be relevant. All boreoteiioids have strong pterygoid flanges, but in Tianyusaurus these flanges are narrow and extend below the level of the jaw margin when the mouth is closed. This is an unusual trait in lizards, but it does occur in chameleons (S. E. Evans 2008, personal observations) that have a large gape (Herrel et al. 2001), the flange guiding and stabilizing jaw movements, resisting lateral displacement. Chameleons also reduce the retroarticular process. In Tianyusaurus, the reduced retroarticular process ( Frazzetta 1962), together with the posteroventrally angled jaw joint and fixed quadrate ( Throckmorton 1976), would facilitate a wide gape. In an akinetic non-propalinal skull (e.g. Caiman: Sinclair & Alexander 1987), pterygoideus operates at maximum mechanical advantage when the mouth is wide open.
Biting with an open mouth and a fixed quadrate, with pterygoideus activated, would load the lower margin of the skull in compression. A framework of lateral bracing bars would also stabilize the skull against torsional forces. Instead of relying on a deep robust skull and jaw to crop tough plant material (as in other boreoteiioids), Tianyusaurus may have taken food items that required a large gape and an ability to exert a penetrating force at the early stages of the bite (e.g. turgid fleshy fruits). This would also be consistent with the differentiation of the upper and lower teeth. The asymmetrical upper teeth seem ideally shaped to perforate (apical cusp) and then cut (oblique cusped blade) as the jaws close, while the lower teeth are better shaped to hold food in the mouth. The large caniniforms, also present in other large boreoteiioids, are most likely to be associated with territorial behaviour, but some living lizards use such teeth to grip large food items ( Torres-Carvajal 2007). According to Tinsley (2004), the Late Cretaceous saw an increase in size of angiosperm fruits in addition to the fleshy fruits of ginkgos, and some conifers, cycads and seed ferns. This may be relevant. Clearly, Tianyusaurus had a feeding strategy that set it apart from its boreoteiioid relatives. The hypothesis of the skull function set out above needs to be tested (e.g. with computer modelling), with a broader survey of boreoteiioid skull and dental morphology and phylogeny. Nonetheless, the discovery of a squamate with a complete lower temporal bar emphasizes the lability of functional characters and the need to understand the biomechanical and developmental constraints that operate on skulls as they evolve. Jones (UCL), for discussion; to Marc Jones and Ryoko Matsumoto (UCL) for their help in figure construction; to Julija Krupic (UCL) for help with Russian translation; and to Randall Nydam and Magdalena Borsuk-Białynicka for their comments on an earlier version of the manuscript. This project was also supported by the National Natural Science Foundation of China (to J.-Y.M.) and the Chinese Academy of Sciences (to X.X.).