Boidae

Palci, Alessandro, Hutchinson, Mark N., Caldwell, Michael W., Smith, Krister T. & Lee, Michael S. Y., 2020, The homologies and evolutionary reduction of the pelvis and hindlimbs in snakes, with the first report of ossified pelvic vestiges in an anomalepidid (Liotyphlops beui), Zoological Journal of the Linnean Society 188, pp. 630-652 : 638-647

Boidae (sensu Pyron et al., 2013)

The ossification of the femur and of all three elements of the pelvis appears to be typical for boines (e.g. Boa constrictor Linnaeus, 1758 , Candoia carinata and Eunectes murinus ). The pelves of Candoia carinata ( Fig. 2R), Boa constrictor and Eunectes murinus are similar to that of Cylindrophis , but the anterior element is much longer relative to the others and curves ventrally towards the anterior tip, which is still capped with cartilage in adults ( Fürbringer, 1870). McDowell (1979: 19) states that male specimens of Candoia bibroni (Duméril & Bibron, 1844) always have large ‘pelvic spurs’ (i.e. keratinous claws) that are connected to a femur and one or two ossified pelvic elements (he does not specify which ones), whereas female members of the same species have smaller spurs that are supported only by an irregular cartilage or lack spurs altogether. Male specimens of Candoia carinata always have large pelvic spurs ( McDowell, 1979; present study). Interestingly, McDowell (1979) mentions that males of Candoia carinata have only one or two ossified pelvic bones, but we found all three in the specimens we examined, although the dorsal element is small and therefore might have been missed by McDowell. Most female specimens of Candoia carinata lack spurs, ossified femora and pelvic elements, but some specimens do have small spurs (sometimes only on one side and all lacking any internal skeleton), and this is significantly correlated with geographical distribution ( McDowell, 1979). Males of Candoia aspera (Günther, 1877) also always have large spurs, and these are connected to a femur and a single pelvic element according to McDowell (1979), whereas females lack spurs (with only rare exceptions) and pelvic vestiges.

The condition in Epicrates cenchria (Linnaeus, 1758) appears variable and, perhaps, sexually dimorphic. A dry specimen of unknown sex (ZFMK 86470) reveals the presence of three ossified elements alongside a well-developed femur, whereas radiographs of two other specimens, both females (BMNH 1845.8.25.193 and SMF PH 26), show only the femur and one pelvic element (the most anterior).

Two specimens of Calabaria reinhardtii (Schlegel, 1851) , a dry skeletal specimen ( ZFMNH 89190 ) and a radiographed specimen ( BMNH 1971.344 ), show only two ossified elements on each side of the body, a femur and the anterior pelvic element .

Intra- and interspecific variation affecting size and number of ossified elements has been observed in species of Eryx . For instance, Fürbringer (1870) described Eryx jaculus (Linnaeus, 1758) [referred by the author using the synonym Eryx turcicus (Olivier, 1801) ] as having a bulky cylindrical femur connected to an anterior, a dorsal and a ventral pelvic element. In contrast, the CT scan of a specimen of Eryx jaculus (SAM R36885) shows only an ossified anterior element and no trace of a femur (possibly cartilaginous), whereas a radiograph of Eryx colubrinus (Linnaeus, 1758) (ZFMK 46058) shows an ossified anterior element and a small, calcified nodule in place of the femur (i.e. no dorsal or ventral pelvic bones).

A micro-CT-scanned specimen of Ungaliophis continentalis (USNM 344819) ( Fig. 2T) shows a well-developed femur (hence, it is likely to represent a male; see Bogert, 1968), a long, anterodorsally directed anterior element and a stout, slightly bowed, posteroventrally directed ventral element. There is no evidence of a dorsal element. Another specimen of Ungaliophis (USNM 29215), but belonging to the species Ungaliophis panamensis , a female, shows no ossified limb or pelvic rudiments. With regard to Exiliboa placata , a female (USNM 209414), which was micro-CT scanned, and a juvenile (BMNH 1977.209, sex unknown), which was radiographed, show neither pelvic nor limb rudiments. However, McDowell (1975: 14) states that male Tropidophoidea (which, in his classification, includes Exiliboa and Ungaliophis ) all have pelvic spurs (and presumably femora supporting them) and a vestigial pelvic girdle, the only exception being Tropidophis semicinctus (but see above). Lack of pelvic bones in the specimens we examined could be attributable to sexual dimorphism, because, as mentioned above, these elements are often unossified or missing altogether in females.

Casarea , Bolyeria and caenophidians show no trace of pelvic vestiges or hindlimbs, regardless of sex ( Bellairs, 1950; Underwood 1967; McDowell, 1975; present study). Wallach & Günther (1998) also reported that the pelvis is absent in Xenophidion , a rare, recently described taxon that has been associated with Bolyeriidae in recent analyses ( Zheng & Wiens, 2016).

SEXUAL DIMORPHISM WITHIN BOAS AND PYTHONS

Sexual dimorphism in pelvic and hindlimb morphology has been most often discussed in large constricting snakes. According to McDowell (1975: 10), in Boidae and Pythonidae (his Booidea) cloacal spurs are always present in males and associated with a vestigial femur, an anterior pelvic element (which McDowell called ‘ilium’), and sometimes also a vestigial ischium (he did not mention a third posterodorsal element, and this might be the basis of his misinterpretation of the anterior element as the ilium; see below under Homology of the Pelvic Elements). Female Pythonidae and Boidae typically also have spurs, but these are usually missing in female Candoia (see above), whereas many female Boidae , despite retaining spurs, have a cartilaginous rather than an ossified pelvis ( McDowell, 1979; a list of examined species of female Boinae is unfortunately missing here).

The condition in Boa constrictor and Python bivittatus Kuhl, 1820 appears variable. Fürbringer (1870) mentioned that in some specimens belonging to these species the pelvis consists of three ossified elements plus a femur, but in others the ventral and dorsal elements are cartilaginous. This seems to be the condition also in a male Boa constrictor in the collections of the British Museum of Natural History (BMNH IV.11.16), because the radiographs show only the femur and the most anterior element. Moreover, radiographs of a juvenile specimen of Boa constrictor (ZFMK 49854) show no trace of a femur or pelvic elements. Such absence might be attributable to immaturity (elements have not yet ossified) or sex (specimen might be female). The fact that pelvic elements ossify early in ontogeny in other species [e.g. in Python sebae (Gmelin, 1789) ; Boughner et al., 2007] suggests that this specimen might be a female and that ossification of the pelvic girdle is sexually dimorphic in this species.

The extent to which ontogeny and sexual dimorphism affect the size and ossification pattern of the pelvic elements in boas and pythons remains to be quantified, because statistical studies involving large numbers of specimens have never been attempted.

Based only on external features (size of the spurs), previous authors have suggested that in some species of Pythonidae and Boidae only the males have externally visible spurs (i.e. claws that protrude from the body wall), whereas in other species where external spurs are present in both sexes, they are larger in the males (e.g. Stickel & Stickel, 1946; McDowell, 1975, 1979; Shine et al., 1998; O’Shea, 2007; Cogger, 2014). As mentioned above, in some species (e.g. Candoia spp. ) the females consistently have smaller spurs or lack any trace of the hindlimbs and pelvic girdle ( Stickel & Stickel, 1946; McDowell, 1979; present study). Despite all this variability, some species do not show any significant variation in the number of ossified elements during ontogeny or between sexes. One example of this is provided by the Australian python Antaresia stimsoni , in which only two ossified elements (the femur and the anterior element) are visible in radiographs of several specimens representative of an ontogenetic series from neonate to adult and inclusive of both males and females. The only difference between the largest specimen (a male) and all others consists of some calcified cartilage between the femur and the pelvic elements ( Fig. 2V; Supporting Information, Data S6, Fig. S1).

HOMOLOGY OF THE PELVIC ELEMENTS

The homology of the elements of the snake pelvis has long been considered problematic, and different authors have used various terms to refer to them (e.g. puboischium, iliopectineum, ischium, pubis and ilium), and sometimes the same term has been used to refer to different bones (e.g. ilium vs. pubis; pubis vs. ischium; Fürbringer, 1870; Perrier, 1928; Bellairs, 1950; List, 1966; McDowell, 1975, 1979; Renous et al., 1976). Figure 2 provides an overview of the main types of pelvic girdles that can be observed in snakes. As might be expected, a complete triradiate pelvis, formed by an ilium, ischium and pubis, is found in the oldest known, and reasonably complete, snake fossils (e.g. Eupodophis , Haasiophis , Najash and Pachyrhachis ), and these taxa also present a complete, or almost complete, hindlimb. However, as we move up in the phylogeny ( Figs. 3–6), we can see that the situation becomes much more complex, and not only do we observe multiple losses (or, more specifically, loss of ossification) of the hindlimbs and pelvic elements, but also new configurations in the geometry (topology) of the pelvis. This great diversity, combined with the inconsistent loss of elements (e.g. some taxa lack the posterodorsal and/or the anteroventral element, whereas others lack the posterodorsal and the posteroventral elements), makes it difficult to homologize the bones, especially when only one element remains (e.g. Liotyphlops beui ).

Those cases where all three pelvic elements are still present and retain a plesiomorphic configuration can be interpreted in a straightforward manner, based on a comparison with the lizard condition (e.g. Rena humilis Baird & Girard, 1853 , Trilepida dimidiatum , Typhlops pusillus Barbour, 1914 and the fossil taxa Najash rionegrina , Wonambi , Pachyrhachis , Haasiophis and Eupodophis ): the element directed posterodorsally is identified as the ilium, the one directed anteroventrally is the pubis and the element pointing posteroventrally is the ischium. However, in some situations, because of the altered geometry of the pelvis, the homology of the elements is not so obvious. This is, for example, the case of Indotyphlops braminus (and several other typhlopoid species), where the pelvis is represented by only two rod-like elements lying close to each other and oriented parallel to the midsagittal plane of the body ( Fig. 2M). Their homology was first clarified by Duerden & Essex (1923) after comparison with other typhlopoid species where a triradiate pelvis is retained, as in some specimens of Typhlops jaimaicensis ( Fig. 2K). The pelvis of Typhlops jamaicensis ZFMK 70500 shows a typical configuration, where the ilium and pubis point posterodorsally and anteroventrally, respectively, whereas the ischium, much larger than the other elements, contributes to the long posteroventral process, which is identified as the same bone as the similarly oriented single rod-like element of Indotyphlops braminus . This primary homology interpretation is confirmed by the fact that often the single remaining bone is connected to cartilaginous precursors of the ilium and pubis that are oriented posterodorsally and anteroventrally, respectively, leaving no doubt about their identification [e.g. Afrotyphlops schlegeli (Bianconi, 1849) MCZ 70064; see also List, 1966].

The anomalepidid Liotyphlops beui also has a single rod-like pelvic element, but its orientation is transverse rather than longitudinal as in typhlopoids, leaving its identification and primary homology in doubt. List (1966) reported a similar structure, although cartilaginous, in a large specimen of Liotyphlops albirostris and interpreted it as the result of fusion between the pubis and ilium. However, List’s (1966) interpretation of the ilium was based on the presence of a dorsally directed process, which is absent in our specimen of Liotyphlops beui . Although we agree with List (1966) that the pelvic vestige of Liotyphlops might represent the fusion of two elements, it is far from clear which ones. If we consider that one process is oriented anteromedially (plesiomorphic orientation for the pubis) and the other posteroventrally (plesiomorphic orientation for the ischium), it would make sense to interpret the element as the result of the fusion of the pubis and ischium (rather than the pubis and ilium, as suggested by List in the case of Liotyphlops albirostris ). However, as shown below, the orientation of a single pelvic element cannot always be relied upon to infer its identification and primary homology in snakes, and the presence of other nearby elements becomes essential for a reliable interpretation.

Asmentionedabove, Cylindrophis andmembersofthe Boidae can have three pelvic elements, but their relative size and orientation are unusual when compared with the more plesiomorphic condition observed in lizards, stem fossil snakes and leptotyphlopids. In particular, boids have a large anterior element, the largest of the three, which is directed anterodorsally, unlike any other pelvic element in the mentioned outgroups. However, the small element pointing posterodorsally is most likely to be the primary homologue of the ilium, owing to its similar position to the same element in leptotyphlopids and fossil snakes. If we accept this primary homology assessment, then the ventral element can be either the ischium or the pubis, because it can be directed somewhat anteroventrally (e.g. Cylindrophis ruffus SAMA R 12956) or somewhat posteroventrally (e.g. Cylindrophis maculatus ZFMK 33609). However, the most common condition for this element is to be pointing posteroventrally (e.g. Eunectes murinus , Candoia carinata and Ungaliophis continentalis ), which makes the first interpretation (ischium) more likely. Moreover, if we examine the condition in the fossil snake Najash rionegrina ( Fig. 2C), we can see that the pubis is considerably larger than the ischium in this snake, and this suggests that the large anterior element in Cylindrophis and boines is the primary homologue of the pubis, despite its unusual anterodorsal orientation. Some previous authors have interpreted this element as the primary homologue of the ilium (e.g. Perrier, 1928; McDowell, 1979), possibly based on its relative size and direction approaching the vertebrae. However, Raynaud (1972) and Renous et al. (1976) solved the primary homology conundrum by examining the muscle attachments and path of the obturator nerve, all of which indicate that the anterior pelvic element is indeed the primary homologue of the pubis. In Pythonidae and Boidae , there is a muscle that originates on the anterior pelvic element and inserts on the femoral trochanter ( Fürbringer, 1870; Bellairs, 1950; Renous et al., 1976). Herrel et al. (2008) have shown that in the lizard Anolis carolinensis only two muscles originate on the pelvis and attach to the internal trochanter on the femur: one originates on the ischium (m. ischiofemoralis) and the other on the pubis (m. pubofemoralis). These muscles cannot be differentiated in Iguana iguana (Linnaeus, 1758) , where only a m. puboischiofemoralis is observed ( Russell, 1988), and this appears also to be the condition in Python , as shown by Renous et al. (1976). The m. puboischiofemoralis can, in turn, be subdivided into two parts (internus and externus, both in Iguana and in Python ) ( Renous et al., 1976; Russell, 1988), and these are equivalent to Fürbringer’s (1870 m. ileopectineo-trochantineus longus and brevis.

PHYLOGENETIC ANALYSES

Parsimony analyses enforcing our first constraint (i.e. constraint 1, with anomalepidids unresolved relative to other scolecophidians) produce six most parsimonious tree topologies [471 steps; consistency index (CI) = 0.344; retention index (RI) = 0.701], the strict consensus of which is shown in Figures 3 and 4, where characters related to the presence of pelvic elements and hindlimbs are optimized onto the topology (under parsimony). The position of Casarea dussumieri , which is variable in recent phylogenetic analyses of squamates (e.g. Hsiang et al., 2015; Figueroa et al., 2016; Streicher & Wiens, 2017; Miralles et al., 2018), is resolved with fairly high bootstrap support (BS = 71.42; see Supporting Information, Data S6, Fig. S2). The bolyeriid Casarea resolves as the sister taxon to Caenophidia (in support of the results of Figueroa et al., 2016; Miralles et al., 2018).

Parsimony analyses enforcing our second constraint (i.e. constraint 2, with anomalepidids as the sister group to Alethinophidia) obtain 18 most parsimonious trees (479 steps; CI = 0.338; RI = 0.693), the consensus tree of which shows a large polytomy at the base of the snake radiation (Supporting Information, Data S6, Fig. S3). This analysis was thus unable to resolve the relationships between scolecophidians, the two alethinophidians Anilius and Tropidophis , and fossil snakes, namely Dinilysia , Najash , Wonambi and the clade Pachyrhachis + Haasiophis + Eupodophis . No character optimization was performed using this consensus topology because of the large polytomy.

Bayesian analysis enforcing constraint 1 produces a majority-rule consensus tree that is identical to the consensus tree obtained through parsimony (minus branch lengths; Supporting Information, Data S6, Figs S4, S 5; see these figures also for character state optimizations. See Supporting Information, Data S6, Fig. S6 for posterior probabilities of clades).

Bayesian analyses that enforce constraint 2 result in a consensus tree that is different from that obtained in the corresponding parsimony analysis. This tree topology and relative character state optimizations for pelvic elements and hindlimbs are shown in Figures 5 and 6 (see Supporting Information, Data S6, Fig. S7 for posterior probabilities of clades). The difference between this tree topology and that resulting from the parsimony analysis resides in the placement of two ‘scolecophidian’ lineages (Typhlopoidea + Leptotyphlopidae and Anomalepididae ) at the base of all snakes (fossil and extant), followed by an alethinophidian clade that includes all fossil snake taxa. Two extant species, Anilius and Tropidophis , are reconstructed at the base of a clade that includes all fossil snakes. Interestingly, a morphological resemblance between Anilius and the fossil snake Dinilysia had been noted before. In fact, Dinilysia was originally classified as an anilioid by Smith-Woodward (1901) (Ilysia is the former name of Anilius ). The relatively close position of Tropidophis to the fossil snakes, in contrast, is likely to be largely attributable to the molecular constraint enforced (i.e. the constraint forces Tropidophis to be close to Anilius ).

DIVERGENT PATTERNS OF PELVIC REDUCTION ACROSS SNAKES

The placement of fossil snakes, all of which retain a complete pelvic girdle with the only possible exception of Dinilysia (condition unknown), in a basal position or nested in alethinophidia has some obvious effect on character state optimization. In order to avoid confusion, we will first describe the optimization on the strict consensus of the six most parsimonious trees obtained from parsimony analysis when enforcing constraint 1, and then we will describe the optimization performed on the Bayesian tree obtained under constraint 2. The strict consensus obtained from parsimony analysis when enforcing constraint 2 was too poorly resolved to allow any meaningful character state optimization. In contrast, enforcing constraint 1 in our Bayesian analysis produces the same result as parsimony analysis enforcing the same constraint

AAbbreviations: Al, Alethinophidia; An, Anomalepididae ; Bo, Boidae ; Ca, Caenophidia; Le, Leptotyphlopidae ; Op, Ophidia (i.e. all extant and fossil snakes); Pa, Pachyophiidae ; Py, Pythonidae ; Sc, Scolecophidia; Se, Serpentes (crown snakes); Ty, Typhlopoidea.

(minus branch lengths). Therefore, character state optimization on this last tree will be discussed only briefly (state optimization on the Bayesian tree that enforced constraint 1 is still shown in Supporting Information, Data S6, Figs S4, S 5).

OPTIMIZATION OF CHARACTER STATES ON THE PARSIMONY TREE

When optimizing character states on our parsimony consensus tree (enforcing constraint 1), Pagel’s (1994) statistical test of character correlation returned significant results (at α = 0.05) for all three pairwise comparisons of the three bones (character 114, ilium presence/absence; character 115, ischium presence/ absence; and character 116, pubis presence/absence; Table 1). This means that evolutionary gains and losses of each ossified pelvic element are correlated with each of the other two elements. Although one might think that characters related to limb reduction would be correlated when looking at comparisons that span fully limbed and limbless taxa, it is not obvious that such correlations should exist between (already reduced) individual elements within a highly limb-reduced group.

According to parsimony optimization, an ossified ilium is present in stem snake lineages (basal Ophidia) and, ancestrally, in crown snakes (Serpentes: Scolecophidia and Alethinophidia) ( Fig. 3A). The

The strict consensus of 18 trees obtained from the parsimony analysis while enforcing constraint 2 was too poorly resolved to allow any meaningful statistical testing. Pagel’s (1994) test was used to evaluate a possible correlation between gains and losses of ossified pelvic elements. Abbreviations: C1, constraint 1 (loose constraint); C2, constraint 2 (tight constraint, enforcing paraphyly of Scolecophidia); logL(D), log likelihood of the dependent model; logL(I), log likelihood of the independent model; LR, likelihood ratio.

ilium ossification is lost in some scolecophidians and is lost early in Alethinophidia, i.e. in the most recent common ancestor (MRCA) of Afrophidia (sensu Vidal & Hedges, 2009: alethinophidians above Anilius and tropidophiids). Within the last clade, an ossified ilium has reappeared multiple times, in Cylindrophis , Python , Boinae and Candoia .

An ossified ischium is present in all stem snake lineages, but the ancestral condition in Serpentes and Alethinophidia is uncertain ( Fig. 3B). Again, there have been several independent losses and/ or reappearances within Scolecophidia and Alethinophidia. The reconstructed history of the ischium has more ambiguous branches than is the case for the ilium, but regardless of which equally parsimonious optimization is chosen, the pattern of losses and gains is slightly different from that for the ilium ( Fig. 3B). For instance, Indotyphlops and Tropidophis retain the ischium but have lost the ilium (or at least an ossified ilium), whereas the opposite is true for Ungaliophis , where an ossified ischium is lost (a cartilaginous condensation might still be present) but the ossified ilium is retained. Apart from that, overall, the ilium and ischium have rather similar evolutionary trajectories, although the ancestral state distribution for the ischium is more uncertain, and an ossified ischium might have been present in the MRCA of boas and pythons, whereas the ilium was possibly lost at that point and later reappeared convergently in some lineages (alternatively, the ilium was lost independently multiple times).

The pubis and hindlimb exhibit a very divergent pattern compared with the ilium and ischium. An ossified pubis is present in all stem snake lineages and, ancestrally, in Alethinophidia, but the ancestral condition in Scolecophidia is uncertain ( Fig. 4A). The pubis is widely retained in Alethinophidia basal to Caenophidia, unlike the ilium and ischium. Thus, the pubis (in contrast to the ilium and ischium) is optimized as present in all basal alethinophidian lineages,and is lost late in three alethinophidian lineages ( Uropeltidae , Xenopeltis and the MRCA of Casarea + caenophidians). The probable reasons for this discordant pattern are discussed in the next section.

The distribution of hindlimbs and their rudiments (femora) has a similar distribution and optimization to the ossified pubis ( Fig. 4B). A femur is retained in the MRCA of crown snakes (Serpentes) and in the MRCA of all alethinophidians and is lost late in this clade, in the same three lineages that also lost the pubis (uropeltids, Xenopeltis and in the MRCA of Casarea + caenophidians). The only differences in the distribution and optimization of the pubis and hindlimb are seen in some scolecophidians, where a pubis is present but no trace of hindlimbs can be observed (e.g. Typhlops jamaicensis and Typhlops pusillus ) or, vice versa, a femur is present but an ossified pubis is absent (e.g. Trilepida dimidiatum ). Importantly, according to our evolutionary scenario, the most parsimonious optimization for the loss of the hindlimbs (femora) suggests that these were lost independently at least four times: at the MRCA of Scolecophidia (or independently in Typhlopoidea and Anomalepididae ), in uropeltids, in Xenopeltis and in the MRCA of Casarea and Caenophidia. Although it is possible that some henophidians might still retain the genetic machinery (HoxD genes) required to develop epipodials and autopodia, as suggested by Leal & Cohn (2016), our results do not support a scenario where hindlimbs ever re-evolved within this clade, although it is possible that hindlimbs might have re-evolved within Scolecophidia (i.e. in Leptotyphlopidae ).

OPTIMIZATION OF CHARACTER STATES ON THE BAYESIAN TREES

For Bayesian trees under both constraints 1 and 2, Pagel’s (1994) statistical test of character correlation returned significant results (at α = 0.05) for all three pairwise comparisons of the three bones (character 114, presence of ilium; character 115, presence of ischium; and character 116, presence of pubis; Table 1).

As mentioned above, the Bayesian analysis that enforced constraint 1 resulted in a consensus tree that was the same (minus branch lengths) as that obtained from the parsimony analysis that enforced the same constraint. Consequently, maximum likelihood optimization of character states across this tree topology is similar to its parsimony equivalent. The differences between the two optimizations are in the ambiguous nodes; where parsimony simply provides unresolved state reconstructions, maximum likelihood provides estimates of which of two ambiguous states is more likely. Notable differences are as follows: likelihood optimization finds it more likely for the MRCA of a monophyletic Scolecophidia to have lost the pelvis (all three elements) and hindlimbs, with subsequent independent re-acquisition of some of these elements in the three main scolecophidian lineages; the MRCA of Serpentes most probably had a complete pelvis and femoral spurs; and the MRCA of Alethinophidia most probably retained the pubis and femoral spurs but had lost the ischium and the ilium, although the absence of these last two elements is only marginally more likely than their presence (compare Figs 3, 4 with Supporting Information, Data S6, Figs S4, S 5).

In the Bayesian analysis that enforced constraint 2, a paraphyletic ‘Scolecophidia’ is placed at the base of the snake radiation (all other fossil and recent snakes), with Anomalepididae reconstructed as the sister group to Alethinophidia, a clade that here includes all fossil forms. Interestingly, maximum likelihood optimization now indicates that the ancestral condition in snakes is that of lacking an ossified ilium, an element that appeared convergently in the MRCA of Leptotyphlopidae , within Typhlopoidea, in Tropidophis , within the MRCA of Anilius and Eupodophis and, as in the parsimony tree, also in Cylindrophis , in pythons, in Candoia and in the MRCA of Boa and Eunectes ( Fig. 5A).

An ossified ischium is also reconstructed as absent in the MRCA of all snakes and appears to have evolved convergently in the MRCA of Typhlopidae and Leptotyphlopidae , in the MRCA of Anilius and Eupodophis and, as in the parsimony tree, in Cylindrophis , pythons, Candoia , Ungaliophis and in the MRCA of Boa and Eunectes ( Fig. 5B) An ossified pubis is reconstructed as absent in the MRCA of all snakes and appeared independently in Rena , some typhlopoids and in the MRCA of Alethinophidia (which here includes fossil snakes) ( Fig. 6A). Ossified limbs are reconstructed as absent in the MRCA of all snakes and appeared convergently in Leptotyphlopidae and Alethinophidia. Interestingly, this scenario also implies that well-developed hindlimbs re-evolved from femoral spurs in the MRCA of Najash and Eupodophis (and, possibly, earlier if Wonambi also had well-developed limbs; Fig. 6B). These optimizations are potentially problematic, as discussed below.