relationship between snakes and some burrowing squam- ates that are limbless or have reduced limbs, such as amphisbaenians and/or dibamids (Rieppel & Zaher 2000b; Zaher & Rieppel 2000; Kearney 2003), has been con- sidered as support for the terrestrial hypothesis, but this Molecular evidence for a is not required. Specializations shared by snakes, amphis- baenians and dibamids include the loss, reduction and terrestrial origin of snakes consolidation of skull bones; braincase enclosure; loss or reduction of limbs and girdles; and increased uniformity Nicolas Vidal and S. Blair Hedges* along the vertebral column (Coates & Ruta 2000). The NASA Astrobiology Institute and Department of Biology, terrestrial origin hypothesis agrees with data derived from 208 Mueller Laboratory, Pennsylvania State University, the sensory system of living snakes and in particular with University Park, PA 16802, USA the peculiarities of their eyes (Walls 1940). Under this * Author for correspondence ([email protected]). hypothesis, the fossil marine snakes (pachyophiids) are Recd 22.09.03; Accptd 01.12.03; Published online 30.01.04 considered to be derived alethinophidian snakes and therefore to have no particular bearing on snake origins. Biologists have debated the origin of snakes since the This derived position of pachyophiids implies either a re- nineteenth century. One hypothesis suggests that evolution of the limbs or several independent losses of snakes are most closely related to terrestrial lizards, limbs in living snake lineages (Greene & Cundall 2000; and reduced their limbs on land. An alternative Rieppel et al. 2003). The latter has been a common theme hypothesis proposes that snakes are most closely among vertebrates in general and squamates in particular related to Cretaceous marine lizards, such as mosa- (Greer 1991). saurs, and reduced their limbs in water. A presumed According to the marine hypothesis, snakes are derived close relationship between living monitor lizards, varanoids, a group that includes several families of extinct believed to be close relatives of the extinct mosa- marine reptiles (mosasauroids) such as aigialosaurs, dol- saurs, and snakes has bolstered the marine origin ichosaurs and mosasaurs and two living families of terres- hypothesis. Here, we show that DNA sequence evi- trial lizards, the Helodermatidae (genus Heloderma) and dence does not support a close relationship between Varanidae (genera Lanthanotus and Varanus) (Lee 1997, snakes and monitor lizards, and thus supports a ter- 1998, 2000; Lee et al. 1999; Lee & Caldwell 2000). Under restrial origin of snakes. the marine hypothesis, the varanid lizards would be the closest living relatives of snakes. This result has also been Keywords: evolution; phylogeny; reptiles; squamates; suggested in molecular analyses using mitochondrial lizards; mosasaurs genes, although always with limited taxon sampling and usually not with robust statistical support (Forstner et al. 1. INTRODUCTION 1995; Macey & Verma 1997; Rest et al. 2003). When Living snakes (ca. 3000 sp.) are grouped together with mosasauroids are taken into account in morphological lizards (ca. 4700 sp.) and amphisbaenians (ca. 160 sp.) in analyses, they form a paraphyletic branching pattern lead- the reptilian order Squamata, named for their scaly skin ing to snakes, and are thus closer to snakes than are varan- (Zug et al. 2001; Uetz 2003). Snakes are divided into two ids (Lee & Caldwell 2000). The group comprising snakes main groups. The scolecophidians (‘blindsnakes’) are bur- and mosasauroids, to the exclusion of other varanoids, is rowing species with small gape size and feed on small prey called Pythonomorpha (Cope 1869; Lee 1997) and is (mainly ants and termites). The alethinophidians (‘typical based on the sharing of presumably derived characters of snakes’) are more ecologically diverse and most species the skull, the lower jaw and the dentition. According to feed on relatively large prey, primarily vertebrates this scheme, the extinct marine pachyophiids are con- (Cundall & Greene 2000; Vidal & Hedges 2002b). In sidered to be intermediates between limbed marine squa- addition to the more obvious characteristics of body mates (mosasauroids) and living snakes (Lee & Caldwell elongation and loss of limbs, other features of living snakes 2000; Lee & Scanlon 2002). This basal position of both include absence of eyelids and external ears and the pres- marine groups of squamates implies a marine-to-terrestrial ence of deeply forked tongues (Coates & Ruta 2000). transition leading to living snakes, an event otherwise Two main hypotheses have been proposed and debated unknown to have occurred within tetrapods. concerning the ancestral mode of the life of snakes: a ter- restrial (burrowing or semi-burrowing) origin (Camp 1923; Mahendra 1938; Walls 1940) and a marine origin 2. METHODS (Cope 1869; Nopcsa 1923; Caldwell & Lee 1997; Lee We used sequences from two protein-coding nuclear genes, recom- 1997; Lee & Scanlon 2002). This controversy, which has bination-activating gene 1 (RAG1) and oocyte maturation factor (C- implications for understanding the evolution of locomotor mos), obtained from all 19 families of living lizards and amphisbaeni- and feeding systems in squamate reptiles (Gans 1961; ans (Zug et al. 2001) and 17 out of the 25 families of living snakes (Vidal & Hedges 2002a,b). The usefulness of C-mos for resolving Cundall & Greene 2000), has recently been fuelled by the interfamilial squamate relationships is well known (Saint et al. 1998; discovery or reanalysis of fossils of three marine snake Vidal & Hedges 2002a,b), although RAG1 has until now never, to species (‘pachyophiids’) with small but well-developed our knowledge, been sequenced in squamate reptiles despite its potential for resolving various higher-level vertebrate relationships hindlimbs (genera Pachyrhachis, Haasiophis and (Venkatesh et al. 2001). Maximum-likelihood (ML), Bayesian infer- Eupodophis) (Caldwell & Lee 1997; Rage & Escuillie´ 2000; ence, minimum evolution (ME) and maximum-parsimony (MP) Tchernov et al. 2000; Rieppel et al. 2003). methods were used to analyse the RAG1 and C-mos sequences of 64 According to the terrestrial hypothesis, the character- species, both separately and combined. Details of the samples used, methods for obtaining and sequencing the DNA and methods of istics that define snakes were acquired in ancestors that phylogenetic analysis are detailed in electronic Appendix A, available were burrowing or semi-burrowing (Walls 1940). A close on The Royal Society’s Publications Web site. Proc. R. Soc. Lond. B (Suppl.) 271, S226–S229 (2004) S226 2004 The Royal Society DOI 10.1098/rsbl.2003.0151 Origin of snakes N. Vidal and S. B. Hedges S227 60/96 Elapsoidea 65/94 Laticauda 70/95 Bungarus Elapidae 88/100 Dendroaspis 86/100 Micrurus 57/68 Leioheterodon Psammophylax 70/100 Lamprophiidae Lamprophis 82/99 88/100 Mehelya 81/99 Alsophis < 50/89 Diadophis Dipsadidae 100/100 Leptodeira Caenophidia 74 100 99 Hapsidophrys 83/100 100 Phyllorhynchus Colubridae 50/66 Bothriechis Viperidae Stoliczkaia Xenodermatidae Acrochordus Acrochordidae 100/100 Rhinophis * Uropeltidae Uropeltis Casarea Bolyeriidae snakes 75/100 Charina 72/95 Lichanura 99/100 Boa Gongylophis 74/100 Eryx Boidae Alethinophidia 62/ Candoia (big gape) 65/100 64 Ungaliophis –/57 Calabaria 100/100 Acrantophis 82/96 Liasis snakes 99/100 Apodora Pythonidae 91/100 Python 71 Loxocemus Loxocemidae 100/100 100 Xenopeltis Xenopeltidae 100/100 Tropidophis Tropidophiidae 60/99 Trachyboa Anilius 59/50 Scolecophidia * Aniliidae (small gape) 100/100 Typhlops Typhlopidae 85/96 Ramphotyphlops Leptotyphlops Leptotyphlopidae 60/100 100/100 Agama Agamidae 64/99 Chamaeleo Chamaeleonidae Sceloporus Iguanidae 98/100 Anniella 69/100 Diploglossinae Anguidae Anguimorpha Heloderma –/78 Helodermatidae 98/95/98/100 –/57 Varanus Varanidae Shinisaurus Xenosauridae 89/100 Trogonophidae 100/100 Amphisbaena Amphisbaenidae 85/100 Bipes Bipedidae < 50/80 Rhineura Rhineuridae 77/100 80 89 Lacerta Lacertidae 100 Teiinae Teiidae 100 Gymnophthalmus Gymnophthalmidae 69/68 100/100 Eumeces Scincidae Tiliqua < 50/100 100/100 Xantusia Xantusiidae Cordylus Cordylidae 100/100 Lialis Gekkonidae Gekkoninae Dibamus Dibamidae Sphenodontidae Sphenodontida Pelomedusidae Chelonia (turtle) Figure 1. (Caption overleaf.) Proc. R. Soc. Lond. B (Suppl.) S228 N. Vidal and S. B. Hedges Origin of snakes Figure 1. Phylogenetic relationships of snakes, lizards and amphisbaenians inferred from DNA sequences of the nuclear living snakes genes RAG1 and C-mos. For the critical node (Anguimorpha), values for all four tree-building methods are shown in the following order: bootstrap support values from pachyophiids ML, MP, ME and Bayesian posterior probabilities. For marine other nodes, bootstrap support values from MP are shown, followed by Bayesian posterior probabilities. Support values anguid lizards above 50% are shown at nodes in the Bayesian consensus marine tree; a dash indicates a node not appearing in the bootstrap consensus MP tree. An asterisk identifies a lineage with mosasaurs small gaped species, and the box indicates the taxon (Varanidae; monitor lizards) believed to be the closest relative of snakes under the marine hypothesis. The tree is varanid lizards rooted with a tuatara (Sphenodon) and turtle. Detailed results for all methods are in electronic Appendix A. other squamates (lizards, 3. RESULTS amphisbaenians) The resulting phylogenetic trees show remarkable con- sistency among different methods