Supermatrix phylogeny
We generated the final tree (lnL = −2609551.07) using Maximum Likelihood (ML) in RAxMLv7.2.8. Support was assessed using the non-parametric Shimodaira-Hasegawa-Like (SHL) implementation of the approximate likelihood-ratio test (aLRT; see [94]). The tree and data matrix are available in NEXUS format in DataDryad repository 10.5061/dryad.82h0m and as Additional file 1: Data File S1. A skeletal representation of the tree (excluding several species which are incertae sedis) is shown in Figure 1. The full species-level phylogeny (minus the outgroup Sphenodon) is shown in Figures 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28. The analysis yields a generally well-supported phylogenetic estimate for squamates (i.e. 70% of nodes have SHL values >85, indicating they are strongly supported). There is no relationship between proportional completeness (bp of non-missing data in species / 12896 bp of complete data) and branch length (r = −0.29, P = 0.14) for terminal taxa, strongly suggesting that the estimated branch lengths are not consistently biased by missing data.
Higher-level relationships
Our tree (Figure 1) is broadly congruent with most previous molecular studies of higher-level squamate phylogeny using both nuclear data and combined nuclear and mitochondrial data (e.g. [16–20]), providing important confirmation of previous molecular studies based on more limited taxon sampling. Specifically we support (Figure 1): (i) the placement of dibamids and gekkotans near the base of the tree (Figure 1A); (ii) a sister-group relationship between Scincoidea (scincids, cordylids, gerrhosaurids, and xantusiids; Figure 1B) and a clade (Episquamata; Figure 1C) containing the rest of the squamates excluding dibamids and gekkotans; (iii) Lacertoidea (lacertids, amphisbaenians, teiids, and gymnophthalmids; Figure 1D), and (iv) a clade (Toxicofera; Figure 1E) containing anguimorphs (Figure 1F), iguanians (Figure 1G), and snakes (Figure 1H) as the sister taxon to Lacertoidea.
These relationships are strongly supported in general (Figure 1), but differ sharply from most trees based on morphological data [13–15, 19, 22, 95]. Nevertheless, many clades found in previous morphological taxonomies and phylogenies are also present in this tree in some form, including Amphisbaenia, Anguimorpha, Gekkota, Iguania, Lacertoidea (but including amphisbaenians), Scincoidea, Serpentes, and many families and subfamilies. In contrast, the relationships among these groups differ strongly between molecular analyses [17–20] and morphological analyses [14, 15]. Our results demonstrate that this incongruence is not explained by limited taxon sampling in the molecular data sets. In fact, our species-level sampling is far more extensive than in any morphological analyses (e.g. [14, 15]), by an order of magnitude.
We find that the basal squamate relationships are strongly supported in our tree. The family Dibamidae is the sister group to all other squamates, and Gekkota is the sister group to all squamates excluding Dibamidae (Figure 1), as in some previous studies (e.g. [16, 18]). Other recent molecular analyses have also placed Dibamidae near the squamate root, but differed in placing it as either the sister taxon to all squamates excluding Gekkota [17], or the sister- group of Gekkota [19, 20]. Our results also corroborate that the New World genus Anelytropsis is nested within the Old World genus Dibamus[96], but the associated branches are weakly supported (Figure 2).
Gekkota
Within Gekkota, we corroborate both earlier morphological [97] and recent molecular estimates [55, 56, 59, 98] in supporting a clade containing the Australian radiation of "diplodactylid" geckos (Carphodactylidae and Diplodactylidae) and the snakelike pygopodids (Figures 1, 2). As in previous studies [55], Carphodactylidae is the weakly supported sister group to Pygopodidae, and this clade is the sister group of Diplodactylidae (Figures 1, 2). We recover clades within the former Gekkonidae that correspond to the strongly supported families Eublepharidae, Sphaerodactylidae, Phyllodactylidae, and Gekkonidae as in previous studies, and similar relationships among these groups [55–57, 59, 60, 98–100].
Within Gekkota, we find evidence for non-monophyly of many genera. Many relationships among the New Caledonian diplodactylids are weakly supported (Figure 2), and there is apparent non-monophyly of the genera Rhacodactylus, Bavayia, and Eurydactylodes with respect to each other and Oedodera, Dierogekko, Paniegekko, Correlophus, and Mniarogekko[101]. In the Australian diplodactylids, Strophurus taenicauda is strongly supported as belonging to a clade that is only distantly related to the other sampled Strophurus species (Figure 2). The two species of the North African sphaerodactylid genus Saurodactylus are divided between the two major sphaerodactylid clades (Figure 3), but the associated branches are weakly supported. The South American phyllodactylid genus Homonota is strongly supported as being paraphyletic with respect to Phyllodactylus (Figure 3).
A number of gekkonid genera (Figure 4) also appear to be non-monophyletic, including the Asian genera Cnemaspis (sampled species divided into two non-sister clades), Lepidodactylus (with respect to Pseudogekko and some Luperosaurus), Gekko (with respect to Ptychozoon and Lu. iskandari), Luperosaurus (with respect to Lepidodactylus and Gekko), Mediodactylus (with respect to Pseudoceramodactylus, Tropiocolotes, Stenodactylus, Cyrtopodion, Bunopus, Crossobamon, and Agamura), and Bunopus (with respect to Crossobamon), and the African Afrogecko (with respect to Afroedura, Christinus, Cryptactites, and Matoatoa), Afroedura (with respect to Afrogecko, Blaesodactylus, Christinus, Geckolepis, Pachydactylus, Rhoptropus, and numerous other genera), Chondrodactylus (with respect to Pachydactylus laevigatus), and Pachydactylus (with respect to Chondrodactylus and Colopus). Many of these taxonomic problems in gekkotan families have been identified in previous studies (e.g. [59, 99, 102]), and extensive changes will likely be required to fix them.
Scincoidea
We strongly support (SHL = 100; Figures 1, 5, 6, 7, 8, 9, 10) the monophyly of Scincoidea (Scincidae, Xantusiidae, Gerrhosauridae, and Cordylidae), as in other recent studies [16–20]. All four families are strongly supported (Figures 5, 6, 7, 8, 9, 10). A similar clade is also recognized in morphological phylogenies [14], though without Xantusiidae in some [13].
Within the New World family Xantusiidae, we corroborate previous analyses [103, 104] that found strong support for a sister-group relationship between Xantusia and Lepidophyma, excluding Cricosaura (Figure 5). These relationships support the subfamily Cricosaurinae for Cricosaura[105]. We also recognize Xantusiinae for the North American genus Xantusia and Lepidophyminae for the Central American genus Lepidophyma[106, 107].
Within the African and Madagascan family Gerrhosauridae (Figure 5), the genus Gerrhosaurus is weakly supported as being paraphyletic with respect to the clade comprising Tetradactylus + Cordylosaurus, with G. major placed as the sister group to all other gerrhosaurids. Within Cordylidae (Figure 5), we use the generic taxonomy from a recent phylogenetic analysis and re-classification based on multiple nuclear and mitochondrial genes [108]. This classification broke up the non-monophyletic Cordylus[109] into several smaller genera, and we corroborate the non-monophyly of the former Cordylus and support the monophyly of the newly recognized genera (Figure 5). We support the distinctiveness of Platysaurus (Figure 5) and recognition of the subfamily Platysaurinae [108].
We strong support (SHL = 100) for the monophyly of Scincidae (Figure 6) as in previous studies (e.g. [20, 50, 51]). We strongly support the basal placement of the monophyletic subfamily Acontiinae (Figure 6), as found in some previous studies (e.g. [20, 51]) but not others (e.g. [50]). Similar to earlier studies, we find that the subfamily Scincinae (sensu [110]) is non-monophyletic, as Feylininae is nested within Scincinae (also found in [20, 50, 51, 111]). Based on these results, synonymizing Feylininae with Scincinae produces a monophyletic Scincinae (SHL = 97), which is then sister to a monophyletic Lygosominae (SHL = 100 excluding Ateuchosaurus; see below) with 94% SHL support (Figures 6, 7, 8, 9, 10). This yields a new classification in which all three subfamilies (Acontiinae, Lygosominae, Scincinae) are strongly supported. Importantly, these definitions approxi\mate the traditional content of the three subfamilies [50, 110], except for recognition of Feylininae.
We note that a recent revision of the New World genus Mabuya introduced a nontraditional family-level classification for Scincidae [112]. These authors divided Scincidae into seven families: Acontiidae, Egerniidae, Eugongylidae, Lygosomidae, Mabuyidae, Scincidae and Sphenomorphidae. However, there was no phylogenetic need for considering these clades as families, since the family Scincidae is clearly monophyletic, based on our results and others (see above). Thus, their new taxonomy changes the long-standing definition of Scincidae unnecessarily (see [113]). Furthermore, these changes were done without defining the full content (beyond a type genus) of any of these families other than Scincidae (the former Scincinae + Feylininae) and Acontiidae (the former Acontiinae).
Most importantly, the new taxonomy proposed by these authors [112] is at odds with the phylogeny estimated here, with respect to the familial and subfamilial classification of >1000 skink species (Figures 6, 7, 8, 9, 10). For instance, Sphenomorphus stellatus is found in a strongly supported clade containing Lygosoma (presumably Lygosomidae; Figure 10), which is separate from the other clade (presumably Sphenomorphidae) containing the other sampled Sphenomorphus species (Figure 7; note that these Sphenomorphus species are divided among several subclades within this latter clade). An additional problem is that Egernia, Lygosoma, and Sphenomorphus are the type genera of Egerniidae, Lygosomidae, and Sphenomorphidae, but are paraphyletic as currently defined (Figures 7, 8, 9, 10), leading to further uncertainty in the content and definition of these putative families.
Furthermore, Mabuyidae apparently refers to the clade (Figure 9) containing Chioninia, Dasia, Mabuya, Trachylepis, with each of these genera placed in its own subfamily (Chioniniinae, Dasiinae, Mabuyinae, and Trachylepidinae). However, several other genera are strongly placed in this group, such as Eumecia, Eutropis, Lankaskincus, and Ristella (Figure 9). These other genera cannot be readily fit into these subfamilial groups (i.e. they are not the sister group of any genera in those subfamilies), and Trachylepis is paraphyletic with respect to Eumecia, Chioninia, and Mabuya (Figure 9). Also, we find that Emoia is divided between clades containing Lygosoma (Lygosomidae) and Eugongylus (Eugongylidae), and many of these relationships have strong support (Figure 10). Finally, Ateuchosaurus is apparently not accounted for in their classification, and here is weakly placed as the sister-group to a clade comprising their Sphenomorphidae, Egerniidae, Mabuyidae, and Lygosomidae (i.e. Lygosominae as recognized here; Figures 7, 8, 9, 10).
These authors [112] argued that a more heavily subdivided classification for skinks may be desirable for facilitating future taxonomic revisions and species descriptions. However, this classification seems likely to only exacerbate existing taxonomic problems (e.g. placing congeneric species in different families without revising the genus-level taxonomy). Here, we retain the previous definition of Mabuya (restricted to the New World clade; sensu [114]), and we support the traditional definitions of Scincidae, Acontiinae, Scincinae (but including Feylininae), and Lygosominae (Figures 6, 7, 8, 9, 10; note that we leave Ateuchosaurus as incertae sedis). The other taxonomic issues in Scincidae identified here and elsewhere should be resolved in future studies. Our phylogeny provides a framework in which these analyses can take place (i.e. identifying major subclades within skinks), which we think may be more useful than a classification lacking clear taxon definitions.
Of the 133 scincid genera [1], we can assign the 113 sampled in our tree to one of the three subfamilies in our classification (Acontiinae, Lygosominae, and Scincinae; Appendix I). We place 19 of the remaining genera into one of the three subfamilies based on previous classifications (e.g. [110]), with Ateuchosaurus as incertae sedis in Scincidae. Below, we review the non-monophyletic genera in our tree. Many of these problems have been reported by previous authors [51, 111, 115–118], and for brevity we do not distinguish between cases reported in previous studies, and potentially new instances found here.
Within Acontiinae, we find that the two genera are both monophyletic (Figure 6). Within Scincinae, many genera are now strongly monophyletic (thanks in part to the dismantling of Eumeces; [49, 50, 119]), but some problems remain (Figure 6). The genera Scincus and Scincopus are strongly supported as being nested inside of the remaining Eumeces. Among Malagasy scincines (see [49, 120]), Pseudacontias is nested inside Madascincus, and the genera Androngo, Pygomeles, and Voeltzkowia are all nested in Amphiglossus (Figure 6).
We also find numerous taxonomic problems within lygosomines (Figures 7, 8, 9, 10). Species of Sphenomorphus are widely dispersed among other lygosomine genera. The genus Tropidophorus is paraphyletic with respect to a clade containing many other genera (Figure 7). The sampled species of Asymblepharus are only distantly related to each other, including one species (A. sikimmensis) nested inside of Scincella (Figure 7). The genus Lipinia is polyphyletic, with one species (L. vittigera) strongly placed as the sister taxon to Isopachys, and with two other species (L. pulchella and L. noctua) placed in a well-supported clade that also includes Papuascincus (Figure 7).
Among Australian skinks, the genus Eulamprus is polyphyletic with respect to Nangura, Calyptotis, Gnypetoscincus, Coggeria, Coeranoscincus, Ophioscincus, Saiphos, Anomalopus, Eremiascincus, Hemiergis, Glaphyromorphus, Notoscincus, Ctenotus, and Lerista, and most of the relevant nodes are strongly supported (Figure 8). The genera Coeranoscincus and Ophioscincus are polyphyletic with respect to each other and to Saiphos and Coggeria (Figure 8). The genus Glaphyromorphus is paraphyletic with respect to a clade of Eulamprus (Figure 8). The genus Egernia is paraphyletic with respect to Bellatorias (which is paraphyletic with respect to Egernia and Lissolepis) and Lissolepis, although many of the relevant nodes are not strongly supported (Figure 9). The genera Cyclodomorphus and Tiliqua are paraphyletic with respect to each other (Figure 9).
Among other lygosomines, Trachylepis is non-monophyletic [121], with two species (T. aurata and T. vittata) that fall outside the strongly supported clade containing the other species (Figure 9). The latter clade is weakly supported as the sister group to a clade containing Chioninia, Eumecia, and Mabuya. In Mabuya, a few species (M. altamazonica, M. bistriata, and M. nigropuncata) have unorthodox placements within a monophyletic Mabuya, potentially due to uncertain taxonomic assignment of specimens by previous authors [51, 122]. The genus Lygosoma is paraphyletic with respect to Lepidothyris and Mochlus, and many of the relevant nodes are strongly supported (Figure 10). Among New Caledonian skinks, the genus Lioscincus is polyphyletic with respect to Marmorosphax, Celatiscincus, and Tropidoscincus, and both Lioscincus and Tropidoscincus are paraphyletic with respect to, Kanakysaurus, Lacertoides, Phoboscincus, Sigaloseps, Tropidoscincus, Graciliscincus, Simiscincus, and Caledoniscincus, with strong support for most relevant nodes (Figure 10). The genera Emoia and Bassiana are massively polyphyletic and divided across multiple lygosomine clades (Figure 10). The genus Lygisaurus appears to be nested inside of Carlia, although many of the relevant branches are only weakly supported (Figure 10).
Lacertoidea
Within Lacertoidea (Figure 1), we corroborate recent molecular analyses (e.g. [16, 17, 19, 20]) and morphology-based phylogenies and classifications (e.g. [13, 15]) in supporting the clade including the New World families Gymnophthalmidae and Teiidae (Figure 11). Within a weakly supported Teiidae (Figure 11), the subfamilies Tupinambinae and Teiinae are each strongly supported as monophyletic, as in previous studies [61]. In Tupinambinae, Callopistes is the sister group to a clade containing Tupinambis, Dracaena, and Crocodilurus. The clade Dracaena + Crocodilurus is nested within Tupinambis, and the associated clades have strong support (Figure 11). We find that the teiine genera Ameiva and Cnemidophorus are non-monophyletic (Figure 11), interdigitating with each other and the monophyletic genera Aspidoscelis, Dicrodon (monotypic), and Kentropyx, as in previous phylogenies [62].
A recent study [123] proposed a re-classification of the family Teiidae based on analysis of 137 morphological characters for 101 terminal species (with ~150 species in the family [1]). Those authors erected several new genera and subfamilies in an attempt to deal with the apparent non-monophyly of currently recognized taxa in their tree. However, in our tree, some of these new taxa conflict strongly with the phylogeny or are rendered unnecessary. First, they recognize Callopistinae as a distinct subfamily for Callopistes, arguing that failure to do so would produce a taxonomy inconsistent with teiid phylogeny. However, we find strong support for Callopistes in its traditional placement as part of Tupinambinae (Figure 11), and this change is thus not needed based on our results. The genus Ameiva is paraphyletic under traditional definitions [62]. In our tree, their conception of Ameiva is also non-monophyletic, with species found in three distinct clades (Figure 11). We also find non-monophyly of many of their species groups within Ameiva, including the ameiva, bifrontata, dorsalis, and erythrocephala groups (Figure 11). Their genera Aurivela (Cnemidophorus longicaudus) and Contomastyx (Cnemidophorus lacertoides) are strongly supported as sister taxa in our tree, and are nested within Ameiva in their erythrocephala species group, along with Dicrodon (Figure 11).
On the positive side, many of the genera they recognize are monophyletic and are not nested in other genera in our tree, including their Ameivula (Cnemidophorus ocellifer), Aspidoscelis (unchanged from previous definitions), Cnemidophorus (excluding C. ocellifer, C. lacertoides, and C. longicaudus), Holcosus (Ameiva undulata, A. festiva, and A. quadrilineatus), Kentropyx (unchanged from previous definitions), Salvator (Tupinambis rufescens, T. duseni, and T. merianae), and Teius (unchanged from previous definitions). We did not sample Ameiva edracantha (their Medopheos).
Given our results, major taxonomic rearrangements within Teiidae seem problematic at present, especially with the extensive paraphyly of many traditional and re-defined teiid genera, the lack of strong resolution of many of these relationships based on molecular and morphological data, and incomplete taxon sampling in all studies so far. Thus, we provisionally retain the traditional taxonomy of Teiidae, pending additional data and analyses. However, we note that Ameiva, Cnemidophorus, and Tupinambis are clearly non-monophyletic based on both our results and those of recent authors [123], and will require taxonomic changes in the future. We anticipate that many of these newly proposed genera [123] will be useful in such revisions.
We find strong support (SHL = 98) for monophyly of Gymnophthalmidae (Figure 11). Within Gymnophthalmidae, we find strong support for the monophyly of the previously recognized subfamilies [63, 64, 124], with the exception of Cercosaurinae (Figure 11). Previous researchers considered the genus Bachia a distinct tribe (Bachiini) within Cercosaurinae, based on a poorly supported sister-group relationship with the tribe Cercosaurini [63, 64]. Here, we find a moderately well supported relationship (SHL = 84) between Bachia and Gymnophthalminae + Rhachisaurinae, and we find that this clade is only distantly related to other Cercosaurinae. Therefore, we restrict Cercosaurinae to the tribe Cercosaurini, and elevate the tribe Bachiini [64] to the subfamily level. The subfamily Bachiinae contains only the genus Bachia (Figure 11), identical in content to the previously recognized tribe [64]. Within Cercosaurinae, we find that the genus Petracola is nested within Proctoporus (Figure 11). In Ecpleopinae, Leposoma is divided into two clades, separated by Anotosaura, Colobosauroides, and Arthrosaura (Figure 11), and many of the relevant nodes are very strongly supported. These issues should be addressed in future studies.
Our results show strong support for a clade uniting Lacertidae and Amphisbaenia (Figure 1), as in many previous studies [16–20, 23]. We also find strong support for monophyly of amphisbaenians (SHL = 99), in contrast to some molecular analyses [19, 20]. Relationships among amphisbaenian families are generally strongly supported and similar to those in earlier molecular studies (Figure 12), including the placement of the New World family Rhineuridae as sister group to all other amphisbaenians [70, 71, 125]. The family Cadeidae is placed as the sister-group to Amphisbaenidae + Trogonophiidae (Figures 1 and Figure 12) with weak support, but has been placed with Blanidae in previous studies, with strong support but less- extensive taxon sampling [125, 126].
We find strong support for monophyly of the Old World family Lacertidae (Figure 13). Within Lacertidae, branch support for the monophyly of most genera and for the subfamilies Gallotiinae and Lacertinae is very high (Figure 13). However, we find that relationships among many genera are poorly supported, as in previous studies [65, 67, 68]. Our results (Figure 13) also indicate that several lacertid genera are non-monophyletic with strong support for the associated nodes, including Algyroides (paraphyletic with respect to Dinarolacerta), Ichnotropis (paraphyletic with respect to Meroles), Meroles (paraphyletic with respect to Ichnotropis), Nucras (polyphyletic with respect to several genera, including Pedioplanis, Poromera, Latastia, Philocortus, Pseuderemias, and Heliobolus), and Pedioplanis (paraphyletic with respect to Nucras).
Higher-level phylogeny of Toxicofera
We find strong support (SHL = 96) for monophyly of Toxicofera (Anguimorpha, Iguania, and Serpentes; Figure 1), and moderate support for a sister-group relationship between Iguania and Anguimorpha (SHL = 79). Relationships among Anguimorpha, Iguania, and Serpentes were weakly supported in some Bayesian and likelihood analyses [16–19], but strongly supported in others [20]. We further corroborate previous studies in also placing Anguimorpha with Iguania [16–20]. In contrast, some other studies have placed anguimorphs with snakes as the sister group to iguanians [127, 128].
Anguimorpha
Our hypothesis for family-level anguimorphan relationships (Figures 1, 14) is generally similar to that of other recent studies [17, 19, 20, 129], and is strongly supported. Our results differ from some analyses based only on morphology, which place Anguidae near the base of Anguimorpha [130]. Here, Shinisauridae is strongly supported as the sister taxon to a well-supported clade of Varanidae + Lanthanotidae (Figures 1, 14). Varanid relationships are similar to previous estimates (e.g. [131]). Xenosauridae is here strongly supported as the sister-group to a strongly supported clade containing Helodermatidae and the strongly supported Anniellidae + Anguidae clade (Figures 1, 14). However, previous molecular analyses have placed Helodermatidae as the sister to Xenosauridae + (Anniellidae + Anguidae), typically with strong support [16, 17, 19, 20].
Within Anguidae (Figure 14), our phylogeny indicates non-monophyly of genera within every subfamily, including Diploglossinae (Diploglossus and Celestus are strongly supported as paraphyletic with respect to each other and to Ophiodes), Anguinae (Ophisaurus is strongly supported as paraphyletic with respect to Anguis, Dopasia, and Pseudopus), and Gerrhonotinae (Abronia and Mesaspis are non-monophyletic, and Coloptychon is nested inside Gerrhonotus). Some of these problems were not reported previously (e.g. Coloptychon, Abronia, and Mesaspis), due to incomplete taxon sampling in previous studies [129, 132, 133], but relationships within Gerrhonotinae are under detailed investigation by other researchers, so these issues are likely to be resolved in the near future.
Iguania
We find strong support (SHL = 100) for the monophyly of Iguania (Figure 1). This clade is strongly supported by nuclear data [16, 17, 19, 20], but an apparent episode of convergent molecular evolution in several mitochondrial genes has seemingly misled some analyses of mtDNA, leading to weak support for Iguania [134], or even separation of the acrodonts and pleurodonts [17, 135] in previous studies. Within Iguania (Figure 1), we find strong support (SHL = 100) for a sister-group relationship between Chamaeleonidae and Agamidae (Acrodonta), and for a clade of mostly New World families (Pleurodonta; SHL = 100).
We find strong support for the monophyly of Chamaeleonidae and the subfamily Chamaeleoninae, and weak support for the paraphyly of Brookesiinae (Figures 1, 15). The sampled species of the Brookesia nasus group appear as the sister group to all other chamaeleonids (the latter clade weakly supported) as found by some previous authors [136], though other studies have recovered a monophyletic Brookesia[137, 138]. Within Chamaeleoninae (Figure 15), we find strong support for the monophyly of most genera and species-level relationships. However, we find strong support for the non-monophyly of Calumma, with some species strongly placed with Chamaeleo, others strongly placed with Rieppeleon, and a third set weakly placed with Nadzikambia + Rhampoleon. While non-monophly of Calumma has also been found in previous studies [138], a recent study strongly supports monophyly of Brookesia and weakly supports monophyly of Calumma[139].
Monophyly of Agamidae is strongly supported (Figure 16; SHL = 100), contrary to some previous estimates [15, 31]. Most relationships among agamid subfamilies and genera are strongly supported (Figure 16), and largely congruent with earlier studies [17, 29, 34, 140]. There are some differences with earlier studies. For example, previous studies based on 29–44 loci [20, 29, 34] placed Hydrosaurinae as sister to Amphibolurinae + (Agaminae + Draconinae) with strong support, whereas we place Hydrosaurinae as the sister-group to Amphibolurinae with weak support. Other authors [140] placed Leiolepiedinae with Uromastycinae, but we (and most other studies) place Uromastycinae as the sister group to all other agamids.
Our phylogeny indicates several taxonomic problems within amphibolurine agamids (Figure 16). The genera Moloch and Chelosania render Hypsilurus paraphyletic, although the support for the relevant clades is weak. The species Lophognathus gilberti is placed in a strongly supported clade with Chlamydosaurus and Amphibolurus, a clade that is not closely related to the other Lophognathus. Many of these taxonomic problems were also noted by previous authors [141].
Within agamine agamids (Figure 16), most relationships are well supported and monophyly of all sampled genera is strongly supported. In contrast, within draconine agamids (Figure 16), many intergeneric relationships are weakly supported, and some genera are non-monophyletic (Figure 16; see also [142]), including Gonocephalus (G. robinsonii is only distantly related to other Gonocephalus) and Japalura (with species distributed among three distantly related clades, including one allied with Ptyctolaemus, another with G. robinsonii, and a third with Pseudocalotes).
Recent authors suggested dividing Laudakia into three genera (Stellagama, Paralaudakia, and Laudakia) based on a non-phylogenetic analysis of morphology [143]. Here, Laudakia (as previously defined) is strongly supported as monophyletic (Figure 16), and this change is not necessitated by the phylogeny. Similarly, based on genetic and morphological data, recent authors [144] suggested resurrecting the genus Saara for the basal clade of Uromastyx (U. asmussi, U. hardwickii, and U. loricata). However, Uromastyx (as previously defined) is strongly supported as monophyletic in our results (Figure 16) and in those of the recent revision [144], and this change is not needed. We therefore retain Laudakia and Uromastyx as previously defined, to preserve taxonomic stability in these groups [113]. We note that recent studies have also begun to revise species limits in other groups such as Trapelus[145], and taxa such as T. pallidus (Figure 16) may represent populations within other species.
Within Pleurodonta we generally confirm the monophyly and composition of the clades that were ranked as families (or subfamilies) within the group (e.g. Phrynosomatidae, Opluridae, Leiosauridae, Leiocephalidae, and Corytophanidae; Figures 1, 17, 18, 19) based on previous molecular studies [31, 33, 34] and earlier morphological analyses [146, 147].
One important exception is the previously recognized Polychrotidae. Our results confirm that Anolis and Polychrus are not sister taxa (Figures 1, 18, 19), as also found in some previous molecular studies [31, 33, 34], but not others [20, 29]. Our results provide strong support for non-monophyly of Polychrotidae, placing Polychrus with Hoplocercidae (SHL = 99) and Anolis with Corytophanidae (SHL = 99; the latter also found by [34]). Recent analyses placing Anolis with Polychrus showed only weak support for this relationship [20, 29], despite many loci (30–44). We support continued recognition of Dactyloidae for Anolis and Polychrotidae for Polychrus[34], based on a limited number of loci but extensive taxon sampling. We note that these families are still monophlyetic, even if they prove to be sister taxa.
Interestingly, our results for relationships among pleurodont families differ from most previous studies, and are surprisingly well-supported in some cases by SHL values (but see below). In previous studies, many relationships among pleurodont families were poorly supported by Bayesian posterior probabilities and by parsimony and likelihood bootstrap values, though typically sampling fewer taxa or characters [17, 31, 33, 148–151]. Studies including 29 nuclear loci found strong concatenated Bayesian support for many relationships but weak support from ML bootstrap analyses for many of the same relationships [34]. The latter pattern (typically weak ML support) was also found in an analysis including those same 29 loci and mitochondrial data for >150 species [29]. We also find a mixture of strongly and weakly supported clades, but with many relationships that are incongruent with these previous studies. First, we find that Tropiduridae is weakly supported as the sister group to all other pleurodonts (also found by [29]), followed successively (Figures 1, 17, 18, 19) by Iguanidae, Leiocephalidae, Crotaphytidae + Phrynosomatidae, Polychrotidae + Hoplocercidae, and Corytophanidae + Dactyloidae.
The relatively strong support for the clades Crotaphytidae + Phrynosomatidae (SHL = 87), Polychrotidae + Hoplocercidae (SHL = 99), and Corytophanidae + Dactyloidae (SHL = 99) is largely unprecedented in previous studies (although Corytophanidae + Dactyloidae is strongly supported in some Bayesian analyses [34]). As in many previous analyses, deeper relationships among the families remain weakly supported. We also find a strongly supported clade containing Liolaemidae, Opluridae, and Leiosauridae (SHL = 95), with Opluridae + Leiosauridae also strongly supported (SHL = 99). Both clades have also been found in previous studies [149, 151], including studies based on 29 or more nuclear loci [20, 29, 34].
We note that previous studies have shown strong support for some pleurodont relationships (e.g. basal placement of phrynosomatids; see [34]), only to be strongly overturned with additional data [20, 29]. Therefore, the relationships found here should be taken with some caution (even if strongly supported), with the possible exception of the recurring clade of Liolaemidae + (Opluridae + Leiosauridae).
All pleurodont families are strongly supported as monophyletic (SHL > 85). Within the pleurodont families, our results generally support the current generic-level taxonomy (Figures 17, 18, 19). However, there are some exceptions. Within Tropiduridae (Figure 17), Tropidurus is paraphyletic with respect to Eurolophosaurus, Strobilurus, Uracentron, and Plica. Within Opluridae (Figure 18), the monotypic genus Chalarodon renders Oplurus paraphyletic. Two leiosaurid genera are also problematic (Figure 18). In Enyaliinae, Anisolepis is paraphyletic with respect to Urostrophus, and this clade is nested within Enyalius. In Leiosaurinae, Pristidactylus is rendered paraphyletic by Leiosaurus and Diplolaemus (Figure 18).
Within Dactyloidae, a recent study re-introduced a more subdivided classification of anoles [152], an issue that has been debated extensively in the past [153–156]. Our results support the monophyly of all the genera recognized by recent authors [152], including Anolis, Audantia, Chamaelinorops, Dactyloa, Deiroptyx, Norops, and Xiphosurus (see Figure 19). However, since Anolis is monophyletic as traditionally defined, we retain that definition here (including the seven listed genera) for continuity with the recent literature [113, 157].
Serpentes
Relationships among the major serpent groups (Figure 1) are generally similar to other recent studies [20, 35, 36, 38, 41, 44, 47, 158–160]. We find that the blindsnakes, Scolecophidia (Figures 1, 20) are not monophyletic, as in previous studies [19, 20, 36, 44, 159, 160]. Similar to some previous studies [44, 159], our data weakly place Anomalepididae as the sister taxon to all snakes, and the scolecophidian families Gerrhopilidae, Leptotyphlopidae, Typhlopidae, and Xenotyphlopidae as the sister-group to all other snakes excluding Anomalepididae (Figures 1, 20). Previous studies have also placed Anomalepididae as the sister-group to all non-scolecophidian snakes [19, 20, 35, 36, 158, 160], in some cases with strong support [20]. Although it might appear that recent analyses of scolecophidian relationships [38] support monophyly of Scolecophidia (e.g. Figure 1 of [38]), the tree including non-snake outgroups from that study shows weak support for placing anomalepidids with alethinophidians, as in other studies [19, 20, 36, 160].
We follow recent authors [38] in recognizing Xenotyphlopidae (strongly placed as the sister taxon of Typhlopidae) and Gerrhopilidae (strongly placed as the sister group of Xenotyphlopidae + Typhlopidae) as distinct families (Figure 20). Leptotyphlopidae is strongly supported as the sister group of a clade comprising Gerrhopilidae, Xenotyphlopidae, and Typhlopidae (Figure 20). As in previous studies [38, 44], we find strong support for non-monophyly of several typhlopid genera (Afrotyphlops, Austrotyphlops, Ramphotyphlops, Letheobia, and Typhlops; Figure 20). There are also undescribed taxa (e.g. Typhlopidae sp. from Sri Lanka; [44]) of uncertain placement within this group (Figure 20). The systematics of typhlopoid snakes will thus require extensive revision in the future, with additional taxon and character sampling.
Within Alethinophidia (SHL = 100), Aniliidae is strongly supported (SHL = 98) as the sister taxon of Tropidophiidae (together comprising Anilioidea), and all other alethinophidians form a strongly supported sister group to this clade (SHL = 97; Figures 1, 21). The enigmatic family Xenophidiidae is weakly placed as the sister-group to all alethinophidians exclusive of Anilioidea (Figures 1, 21). The family Bolyeriidae is weakly placed as the sister-group to pythons, boas, and relatives (Booidea), which are strongly supported (SHL = 88). Relationships in this group are generally consistent with other recent molecular studies [20, 35–37, 47, 159].
Relationships among other alethinophidians are a mixture of strongly and weakly supported nodes (Figures 1, 21). We find strong support (SHL = 100) for a clade containing Anomochilidae + Cylindrophiidae + Uropeltidae. This clade of three families is strongly supported (SHL = 89) as the sister taxon to Xenopeltidae + (Loxocemidae + Pythonidae). Together, these six families form a strongly supported clade (SHL = 89; Figures 1, 21) that is weakly supported as the sister group to the strongly supported clade of Boidae + Calabariidae. Within the clade of Anomochilidae, Cylindrophiidae, and Uropeltidae (Figure 21), we weakly place Anomochilus as the sister group to Cylindrophiidae [44], in contrast to previous studies which placed Anomochilus within Cylindrophis[161]. However, support for monophyly of Cylindrophis excluding Anomochilus is weak (Figure 21). As in previous studies [44, 162], we find several taxonomic problems within Uropeltidae (Figure 21). Specifically, Rhinophis and Uropeltis are paraphyletic with respect to each other and to Pseudotyphlops. The problematic taxa are primarily Sri Lankan [44], and forthcoming analyses will address these issues.
Within Pythonidae (Figure 21), the genus Python is the sister group to all other genera. Some species that were traditionally referred to as Python (P. reticulatus and P. timoriensis) are instead sister to an Australasian clade consisting of Antaresia, Apodora, Aspidites, Bothrochilus, Leiopython, Liasis, and Morelia (Figure 21). These taxa (P. reticulatus and P. timoriensis) have been referred to as Broghammerus, a name originating from an act of "taxonomic vandalism" (i.e. an apparently intentional attempt to disrupt stable taxonomy) in a non-peer reviewed organ without data or analyses [163, 164]. However, this name was, perhaps inadvertently, subsequently used by researchers in peer-reviewed work [165] and has entered into somewhat widespread usage [1]. This name should be ignored and replaced with a suitable substitute. Within the Australasian clade (Figure 21), Morelia is paraphyletic with respect to all other genera, and Liasis is non-monophyletic with respect to Apodora, although many of the relevant relationships are weakly supported.
Within Boidae (Figure 21), our results and those of other recent studies [20, 36, 47, 48, 150, 166] have converged on estimated relationships that are generally similar to each other but which differ from traditional taxonomy [167]. However, the classification has yet to be modified to reflect this, and we rectify this situation here. We find that Calabariidae is nested within Boidae [150], but this is poorly supported, and contrary to most previous studies [47, 48]. While Calabaria has been classified as an erycine boid in the past, this placement is strongly rejected here and in other studies [47, 48]. If the current placement of Calabaria is supported in the future, it would require recognition as the subfamily Calabariinae in Boidae.
The Malagasy boine genera Acrantophis and Sanzinia are placed as the sister taxa to a weakly-supported clade containing Calabariidae and a strongly supported clade (SHL = 99) comprising the currently recognized subfamilies Erycinae, Ungaliophiinae, and other boines (Figure 21). Regardless of the position of Calabariidae, this placement of Malagasy boines renders Boinae paraphyletic. We therefore resurrect the subfamily Sanziniinae [168] for Acrantophis and Sanzinia. This subfamily could be recognized as a distinct family if future studies also support placement of this clade as distinct from other Boidae + Calabariidae.
The genera Lichanura and Charina are currently classified as erycines [1], but are strongly supported as the sister group to Ungaliophiinae, as in previous studies [20, 36, 47, 166]. We expand Ungaliophiinae to include these two genera (Figure 21), rather than erect a new subfamily for these taxa. The subfamily Ungaliophiinae is placed as the sister group to a well-supported clade (SHL = 87) containing the rest of the traditionally recognized Erycinae and Boinae. We restrict Erycinae to the Old World genus Eryx.
The genus Candoia (Boinae) from Oceania and New Guinea [1], is placed as the sister taxon to a moderately supported clade (SHL = 83) consisting of Erycinae (Eryx) and the remaining genera of Boinae (Boa, Corallus, Epicrates, and Eunectes). To solve the non-monophyly of Boinae with respect to Erycinae (due to Candoia), we place Candoia in a new subfamily (Candoiinae, subfam. nov.; see Appendix I). Boinae then comprises the four Neotropical genera that have traditionally been classified in this group (Boa, Corallus, Epicrates, and Eunectes). We acknowledge that non-monophyly of Boinae could be resolved in other ways (e.g. expanding it to include Erycinae). However, our taxonomy maintains the traditionally used subfamilies Boinae, Erycinae, and Ungaliophiinae, modifies them to reflect the phylogeny, and recognizes the phylogenetically distinct boine clades as separate subfamilies (Candoiinae, Sanziniinae). Within Boinae, Eunectes renders Epicrates paraphyletic, but this is not strongly supported (see also [48]).
Our results for advanced snakes (Caenophidia) are generally similar to those of other recent studies [41, 42, 169], and will only be briefly described. However, in contrast to most recent studies [20, 36, 41, 42, 81, 159, 160], Acrochordidae is here strongly placed (SHL = 95) as the sister group to Xenodermatidae. This clade is then the sister group to the remaining Colubroidea, which form a strongly supported clade (SHL = 100; Figures 1, 22). This relationship has been found in some previous studies [169, 170], and was hypothesized by early authors [171]. Further evidence will be required to resolve this conclusively. Analyses based on concatenation of 20–44 loci do not support this grouping [20, 36], though preliminary species-tree analyses of >400 loci do (Pyron et al., in prep.). Relationships in Pareatidae are similar to recent studies [172], and the group is strongly placed as the sister taxon to colubroids excluding xenodermatids (SHL = 100; Figures 1 , 22), as in most recent analyses (e.g. [41, 43, 44]).
The family Viperidae is the sister group to all colubroids excluding xenodermatids and pareatids (Figure 1), as in other recent studies. The family Viperidae is strongly supported (Figure 22), as is the subfamily Viperinae, and the sister-group relationship between Azemiopinae and Crotalinae (SHL = 100). Our results generally support the existing generic-level taxonomy within Viperinae (Figure 22). However, we recover a strongly supported clade within Viperinae consisting of Daboia russelii, D. palaestinae, Macrovipera mauritanica, and M. deserti (Figure 22), as in previous studies [173]. We corroborate previous suggestions that these taxa be included in Daboia[174], though this has not been widely adopted [1]. The other Macrovipera species (including the type species) remain in that genus (Figure 22).
Within Crotalinae (Figure 22), a number of genera appear to be non-monophyletic. The species Trimeresurus gracilis is strongly supported as the sister taxon to Ovophis okinavensis and distantly related to other Trimeresurus, whereas the other Ovophis are strongly placed as the sister group to Protobothrops. A well-supported clade (SHL = 90) containing Atropoides picadoi, Cerrophidion, and Porthidium renders Atropoides paraphyletic (see also [175]). The species Bothrops pictus, considered incertae sedis in previous studies [176], is here strongly supported as the sister taxon to a clade containing Rhinocerophis, Bothropoides, Bothriopsis, and Bothrops (Figure 22). Most of these relationships are strongly supported.
Viperidae is strongly placed (SHL = 95) as the sister taxon to a well-supported clade (SHL = 100) containing Colubridae, Elapidae, Homalopsidae, and Lamprophiidae (Figure 1). Monophyly of Homalopsidae is also strongly supported (Figure 23). Within Homalopsidae, non-monophyly of the genus Enhydris is strongly supported (Figure 23), and it should likely be split into multiple genera. Homalopsidae is weakly supported (SHL = 58) as the sister group of Elapidae + Lamprophiidae (Figure 1). This same relationship was also weakly supported by previous analyses [41, 44], but other studies have found strong support for placing Homalopsidae as the sister group of a strongly supported clade including Elapidae, Lamprophiidae, and Colubridae [20, 36], including data from >400 loci (Pyron et al., in prep.).
Support for the monophyly of Lamprophiidae is strong (but excluding Micrelaps; see below), and most of its subfamilies are well-supported [40, 41, 177, 178] including Atractaspidinae, Aparallactinae, Lamprophiinae, Prosymninae (weakly placed as the sister-group to Oxyrhabdium), Pseudaspidinae, Psammophiinae, and Pseudoxyrhophiinae (Figure 23). In Lamprophiidae, most genera are monophyletic based on our sampling (Figure 23). However, within Aparallactinae, Xenocalamus is strongly placed within Amblyodipsas, and in Atractaspidinae, Homoroselaps is weakly placed in Atractaspis. In Lamprophiinae, Lamprophis is paraphyletic with respect to Lycodonomorphus but support for the relevant clades is weak.
The enigmatic genera Buhoma from Africa and Psammodynastes from Asia were both previously considered incertae sedis within Lamprophiidae [41]. Here they are weakly placed as sister taxa, and more importantly, they form a strongly supported clade with the African genus Pseudaspis (Pseudaspidinae; SHL = 95; Figure 23). Therefore, we expand Pseudaspidinae to include these two genera.
The genus Micrelaps (putatively an aparallactine; [1]) is weakly placed as the sister taxon to Lamprophiidae + Elapidae. Along with Oxyrhabdium (see above) and Montaspis[40], this genus is treated as incertae sedis in our classification (Appendix I). If future studies strongly support these relationships, they may require a new family for Micrelaps and possibly a new subfamily for Oxyrhabdium, though placement of these taxa has been highly variable in previous studies [40, 41, 44, 81].
Monophyly of Elapidae is strongly supported (Figure 24), and Calliophis melanurus is strongly supported as the sister group to all other elapids (see also [44]). Within Elapidae (Figure 24), relationships are generally concordant with previous taxonomy, with some exceptions. The genera Toxicocalamus, Simoselaps, and Echiopsis are all divided across multiple clades, with strong support for many of the relevant branches. A recent study [46] has provided a generic re-classification of the sea snakes (Hydrophis group) to resolve the extensive paraphyly of genera found in previous studies (e.g. [179, 180]). Our results support this classification.
Monophyly of Colubridae and most of its subfamilies (sensu [41, 44]) are strongly supported (Figures 1, 25, 26, 27, 28). However, relationships among many of these subfamilies are weakly supported (Figure 1), as in most previous studies [41, 43, 45, 181]. The subfamilies Calamariinae and Pseudoxenodontinae are strongly supported as sister taxa, and weakly placed as the sister-group to the rest of Colubridae (Figure 25). There is a weakly supported clade (Figure 1) comprising Natricinae + (Dipsadinae + Thermophis), but the clade uniting the New World Dipsadinae with the Asian genus Thermophis is strongly supported (SHL = 100; Figure 28). Here, we place Thermophis (recently in Pseudoxenodontinae [41]) in Dipsadinae (following [182]), making it the first and only Asian member of this otherwise exclusively New World subfamily. However, despite the strong support for its placement here, placement of this taxon has been variable in previous analyses [41, 182, 183], and we acknowledge that future analyses may support recognition of a distinct subfamily (Thermophiinae).
The clade of Natricinae and Dipsadinae is weakly supported as the sister group (Figures 1, 25, 26, 27, 28) to a clade containing Sibynophiinae [181] + (Colubrinae + Grayiinae). The subfamily Colubrinae is weakly supported; we find that the colubrine genera Ahaetulla, Chrysopelea, and Dendrelaphis form a strongly supported clade that is weakly placed as the sister group to the rest of Colubrinae, which form a strongly supported clade (Figure 25). This clade was also placed with Grayiinae or Sibynophiinae in many preliminary analyses, rendering Colubrinae paraphyletic. This group of three genera has been strongly supported in the past, and only weakly placed with Colubrinae [41, 44]. It is possible that future analyses will reveal that the clade of Ahaetulla, Chrysopelea, and Dendrelaphis is placed elsewhere in Colubridae with strong support, and thus merit recognition as a distinct subfamily (Ahaetuliinae). A notable feature of this clade is the presence of gliding flight in most species of Chrysopelea, less-developed non-flight jumping with similar locomotor origins in Dendrelaphis, and homologous glide-related traits in Ahaetulla[184].
Numerous colubroid genera are not included in our tree and are not clearly placed in subfamilies based on previous morphological evidence. In our classification, these genera are also considered incertae sedis within Colubridae, including Blythia, Cyclocorus, Elapoidis, Gongylosoma, Helophis, Myersophis, Oreocalamus, Poecilopholis, Rhabdops, and Tetralepis, as in previous classifications [1].
Our phylogeny reveals numerous taxonomic problems within Colubrinae (Figures 25, 26). The genus Boiga is paraphyletic with respect to Crotaphopeltis, Dipsadoboa, Telescopus, Toxicodryas, and Dasypeltis, with strong support (Figure 25). The genus Philothamnus is paraphyletic with respect to Hapsidophrys (Figure 25). The genus Coluber is split between Old World and New World clades (Figures 25, 26). The species Hierophis spinalis is sister to Eirenis to the exclusion of the other Hierophis species (Figure 25). The genus Dryocalamus is nested within Lycodon (Figure 26). The species Chironius carinatus and C. quadricarinatus are weakly placed in a clade of Neotropical colubrines only distantly related to the other Chironius species (Figure 26). The genus Drymobius renders Dendrophidion paraphyletic (Figure 26). The monotypic genus Rhynchophis renders the two species of Rhadinophis paraphyletic (Figure 26). The genus Coronella is rendered paraphyletic (Figure 26) by Oocatochus with weak support (see also [185]). Finally, the genus Rhinechis is nested within Zamenis (Figure 26).
We find numerous non-monophyletic genera within Natricinae (Figure 27), as in previous studies [41, 44, 78, 186]. These non-monophyletic genera include the Asian genera Amphiesma, Atretium, and Xenocrophis. Among New World genera, we find Regina to be non-monophyletic with respect to most other genera, as in previous phylogenetic studies (e.g. [41, 186]). Also, as in previous studies (e.g. [41, 187]), we find that Adelophis[188] is nested deep within Thamnophis[189].
Finally, within a weakly supported Dipsadinae (Figure 28), we find non-monophyly of numerous genera, as in many earlier studies (e.g. [41–43, 190]). These problems of non-monophyly include Leptodeira (with respect to Imantodes), Geophis (with respect to Atractus), Atractus (with respect to Geophis), Sibynomorphus (with respect to Dipsas), Dipsas (with respect to Sibynomorphus), Taeniophallus (with respect to Echinanthera), and Echinanthera (with respect to Taeniophallus). Recent revisions have begun to tackle these problems [42, 43, 190], but additional taxon and character sampling will be crucial to resolve relationships and taxonomy.
