© 2006 Nature Publishing Group

A Cretaceous terrestrial snake with robusthindlimbs and a sacrumSebastian Apesteguıa1,2 & Hussam Zaher3

It has commonly been thought that snakes underwent progressiveloss of their limbs by gradual diminution of their use1. However,recent developmental and palaeontological discoveries suggest amore complex scenario of limb reduction, still poorly documentedin the fossil record2–5. Here we report a fossil snake with a sacrumsupporting a pelvic girdle and robust, functional legs outside theribcage. The new fossil, from the Upper Cretaceous period ofPatagonia, fills an important gap in the evolutionary progressiontowards limblessness because other known fossil snakes withdeveloped hindlimbs, the marine Haasiophis, Pachyrhachis andEupodophis, lack a sacral region. Phylogenetic analysis shows thatthe new fossil is the most primitive (basal) snake known and thatall other limbed fossil snakes are closer to the more advancedmacrostomatan snakes, a group including boas, pythons andcolubroids. The new fossil retains several features associatedwith a subterranean or surface dwelling life that are also presentin primitive extant snake lineages, supporting the hypothesis of aterrestrial rather than marine origin of snakes.Pachyrhachis problematicus3,Haasiophis terrasanctus4 and Eupodo-

phis descouensi5, three marine fossil snakes from the Tethyan coasts ofNorthern Gondwana, were until now the only known snakes withwell-developed hindlimbs. The presence of fully formed hindlimbsenforced the idea that these were the most primitive (basal) snakesand perfect transitional taxa linking extant snakes to an extinct groupof marine lizards, the Mosasauroidea3,6,7. However, the presence ofseveral other features typical of the more advanced macrostomatansnakes such as pythons, boas and colubroids8 supports the competinghypothesis that these fossils were advanced (macrostomatan) snakesinstead, with no special bearing on the origin of snakes4,8,9. Addition-ally, all three taxa resemble modern snakes in lacking a sacral regionand in having a pelvis that is not suspended from the axial skeletonbut rather lies within the ribcage10.The snake reported here was found in the context of a rich early

Upper Cretaceous fossil fauna11,12 from north Patagonia, Argentina,and represents the earliest limbed snake from a fully terrestrialdeposit. It retains several primitive features absent in any knownfossil or recent snake, including a remarkably primitive pelvis.

Squamata Oppel, 1811Serpentes Linnaeus, 1758

Najash rionegrina gen. et sp. nov.Etymology. From Hebrew Najash, the legged biblical snake; rio-negrina, for Rıo Negro Province, Argentina, where the fossil wasfound.Holotype. MPCA (Museo Paleontologico ‘Carlos Ameghino’,Cipolletti, Rıo Negro) 390–398, 400, consists of a large fragmentof the left dentary and anterior portion of the correspondingsplenial, and a nearly complete and articulated postcranial skeletoncomposed of several sections bearing a total of 122 articulated and

associated vertebrae (109 presacrals, 2 sacrals, 11 caudals), pelvicgirdle, two femora, one fibula and the proximal head of the right tibia(Figs 1 and 2).Locality and horizon. Upper section of the Candeleros Formation(Cenomanian–Turonian13) at ‘La Buitrera’, Rıo Negro Province,Argentina.Additional material. The posterior half of a non-associated brain-case with its right otico-occipital region preserved (Fig. 1a) andseveral associated presacral vertebrae (MPCA 385); several dis-articulated cranial and vertebral elements of a larger individual,including an incomplete left dentary, axis, and associated presacraland caudal vertebrae (MPCA 380–383).Diagnosis. A snake with a strongly concave ventral surface of theparasphenoid rostrum, forming a deep and straight gutter; two sacralvertebrae present; single large parazygantral foramen on each side ofneural arch; proximal caudal vertebrae with blunt haemapophyses;robust femora with a large trocanter.The three specimens referred to N. rionegrina are identified as

snakes on the basis of the completely enclosed braincase, fusedparietals, axis with sutured anterior and fused posterior hypapo-physes, large number of presacral vertebrae (109 preserved), zygo-sphenal and zygapophyseal facets separated by a non-articular area,anterior margin of zygosphenal tectum straight or slightly convex,divided synapophyses, three distally forked lymphapophyses, haema-pophyses on posterior tail vertebrae.The preserved posterior portion of the braincase ofNajash (Fig. 1a)

is similar in several respects to that ofDinilysia patagonica and that ofthe fossorial anilioid snakes (pipesnakes). As in Dinilysia and theanilioids, the otico-occipital portion of the braincase is transverselyexpanded. The right posterodorsal portion of the prootic andposterolateral portion of the parietal form a deep and narrow recessthat receives the anterior portion of the missing supratemporal,which was incorporated into the cranial wall as in Dinilysia and theanilioids Cylindrophis and Anilius. The recess is located laterally tothe contact between the prootic and the supraoccipital, suggesting adorsal exposure of the prootic between the supratemporal, exocci-pital and supraoccipital, a characteristic also present in Dinilysia,Cylindrophis and Anilius. The laterosphenoid is lacking, a plesio-morphic condition found in Dinilysia and scolecophidians (worms-nakes). The contact between the posteriorly expanded edge of thesupraoccipital and the right exoccipital suggests that the exoccipitalsdid not meet dorsally, a plesiomorphic lizard trait also known inHaasiophis and in some boine snakes14. Although present, the cristacircumfenestralis of Najash (here represented mostly by the cristaprootica) is the least developed among all known snakes. Thecrista prootica projects weakly laterally to the stapedial footplate,around its anterodorsal portion, although without overlapping thelatter as in all modern snakes. The ventrolateral expansions of both

LETTERS

1Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’, A. Gallardo 470, Buenos Aires (1405), Argentina. 2Fundacion de Historia Natural ‘Felix de Azara’ (CEBBAD),Universidad Maimonides, V. Virasoro 732, Buenos Aires (1405), Argentina. 3Museu de Zoologia da Universidade de Sao Paulo, Av. Nazare 481, Ipiranga, Sao Paulo 04263-000,Brazil.

Vol 440|20 April 2006|doi:10.1038/nature04413

1037

© 2006 Nature Publishing Group

crista interfenestralis and crista tuberalis are broken, but the juxta-stapedial recess remains widely open posteriorly because of a poorlydeveloped posterodorsal margin of the crista tuberalis, a plesio-morphic feature also found in scolecophidians, Dinilysia and basalalethinophidian snakes. The stapedial footplate is broad, as inDinilysia and the fossorial macrostomatan Xenopeltis. As in lizards,Dinilysia and Wonambi, the basipterygoid processes are prominent,rather longitudinally oriented structures that fit in an articularfacet of the pterygoid instead of contacting only the latter as in themore derived snakes. The left dentary retains two mental foramina(Fig. 1b, c). Although no teeth are preserved, their alveoli aretransversely expanded instead of anteroposteriorly wide as in macro-stomatans. A discrete basal plate is present, but the dentaries lack alingual ridge (or subdental shelf) medial to the tooth-bearing region,a primitive condition absent from all known snakes.The cotyle of the procoelous vertebrae form a rounded to slightly

oval surface that receives a rounded condyle (Fig. 1d, e), a derivedcondition shared with Dinilysia and alethinophidian snakes, includ-ing Pachyrhachis and Haasiophis. The neural arch is low, as is typicalin secretive or fossorial forms, and shares with Dinilysia a prominentridge on each side and above the interzygantral ridge. The zygo-sphene is thick and well developed, as in Dinilysia and macrostoma-tans; however, unlike the latter, the interzygapophyseal constriction isshallow and a posterior neural arch notch is absent. Accessoryprocesses of the prezygapophyses are lacking and parazygantralforamina are present in all trunk vertebrae. The anterior trunk

vertebrae bear relatively high and narrow neural spines, developedhypapophyses, and synapophyses directed posteroventrally. Mid-trunk vertebrae are broader, with lower and longer, blade-like neuralspines, shallow and thin haemal keels extending along the entireventral surface of the centrum, and synapophyses reaching the levelof the prezygapophyseal tip. Synapophyses are divided into para-pophyses and diapophyses, a derived condition shared with allalethinophidian snakes. Conversely, synapophyses project laterallybeyond the level of the prezygapophysial tip on the slightly smallerposterior trunk vertebrae (Fig. 1d, e) and are no longer divided in themore posterior ones. Neural spines become reduced to mere ridgesand haemal keels broaden significantly.Unlike any other snake, Najash retains two sacral vertebrae that

separate the trunk region anatomically from the caudal region(Fig. 2). Preserved elements of the appendicular skeleton includethe pelvic girdle, both femora, the proximal articular head of the righttibia, and the right fibula. Only the right pelvic elements arewell preserved and mostly in place. In contrast with Haasiophis,Pachyrhachis and Eupodophis, the last rib in Najash is ventral to theright femur, reflecting the external position of the hindlimbs withrespect to the ribcage (Fig. 2b). Sacral pleurapophyses are long andslightly curved, and their pointed tips are separated, suggesting aloose suspension of the pelvis. Pubis, ilium and ischium are notsutured or fused together proximally, and the medioventral pubo-ischiac symphysis is lacking. Both ilium and pubis are similar in sizeand show a rounded, expanded proximal head and a long rod-like

Figure 1 | Najash rionegrina. a, Posterior half of the braincase in rightventro-lateral view (MPCA 385). b, c, Left dentary and splenial of theholotype (MPCA 390–398) in medial (b) and lateral (c) views. d, e, Posteriortrunk vertebra of the holotype (MPCA 390–398) in anterior (d) andposterior (e) views. f, Two articulated caudal vertebrae of the holotype(MPCA 390–398) in ventral view. Scale bars, 5 mm (a), 1 mm (b–f).Abbreviations: apl, lateral opening of recessus scalae tympani; bpt,

basipterygoid process; bsp, basisphenoid; crp, crista prootica; crt, cristatuberalis; dia, diapophysis; hae, haemapophysis; jug, jugular foramen; met,mental foramen; par, parapophysis; pa, parietal; paz, parazygantral foramen;pro, prootic; plz, caudal pleurapophysis; psp, parasphenoid; pte, pterygoid;spl, splenial; stf, stapedial footplate; zgo, zygosphene; V, foramen for themaxillary and mandibular branches of the trigeminal nerve; VII, foramenfor hyomandibular branch of facial nerve.

LETTERS NATURE|Vol 440|20 April 2006

1038

© 2006 Nature Publishing Group

body that tapers distally. The right ischium is broken in two pieces.However, the proximal part retained its contact with the two otherpelvic elements. As in Pachyrhachis, it is half the size of the latter twopelvic bones and is spatula-shaped proximally. The first three caudalvertebrae bear well-defined, distally bifurcated lymphapophyses thatproject laterally. Free lymphapophyses (bifurcated ribs) and chevronbones are lacking. Haemapophyses are present as small button-likenodules on the posterior edge of the centrum of the more posteriorcaudal vertebrae (Fig. 1f), resembling the condition found inanilioids and some macrostomatan snakes.A phylogenetic analysis, including all relevant fossil snakes, shows

Najash as the most basal snake (Fig. 3), lying outside the cladeconsisting of all living snakes. Statistical support for this hypothesis isstrong. All other mid-Cretaceous snakes are nested within theclade formed by living snakes, with the terrestrial Dinilysia as thesister-group of alethinophidians, whereas the marine Haasiophis,Pachyrhachis and Eupodophis as well as the Pleistocene Wonambinaracoortensis are nested within a poorly resolved macrostomatanclade, supporting a macrostomatan affinity4,8–10,15, instead of a basalposition as the most primitive snakes3,6,7.Both cranial (a transversely expanded occipital region and a broad

stapedial footplate) and vertebral (a low neural arch) morphologicaltraits of Najash show adaptations to a subterranean life, perhaps asa surface-dwelling species that would occasionally use tunnels

produced by burrowers. Najash, scolecophidians, Dinilysia and ani-lioids represent the four successive outgroups to the macrostomatanclade in which the first marine snakes were documented. Thisscenario unequivocally supports the hypothesis of a subterraneanor surface-dwelling origin of snakes.

METHODSThe data matrix used in the phylogenetic analysis is based on two recentlypublished character lists4,15. Twenty-one new characters were added to includepost-cranial morphology, totalling 119 characters coded for 18 snake taxa.Character codings for the fossil snakes Wonambi naracoortensis, Pachyrhachisproblematicus and Eupodophis descouensi, considered by some authors as beingthe most basal snakes7, were reviewed in accordance with recently publisheddescriptions16–18. Codings for Dinilysia patagonica are based on observationsmade by H.Z. on the holotype and new specimens recently discovered19.Characters included in this matrix are intended to address the more inclusivebasal snake interrelationships and the affinities of the relevant fossils Najash,Pachyrhachis, Haasiophis, Eupodophis and Wonambi; and does not attempt toelucidate macrostomatan interrelationships. Recent attempts to reconstructphylogenetic affinities within the macrostomatan clade show extensive conflict-ing character delimitations3,4,7–10,14,15 that should be addressed more thoroughlybefore any new proposal of macrostomatan interrelationships. We rooted theanalysis with a hypothetical varanoid ancestor (coded according to the con-ditions found in the terrestrial Varanus, Heloderma and Lanthanotus, and themarine Mosasauroidea)7. We alternatively used a dibamid ancestor as an out-group9 with no effects relevant to this study. Analyses were performed with

Figure 2 | Sacral region of the holotype of Najash rionegrina (MPCA390–398). a, Dorsal view. b, Ventral view. The left pelvic and limb elementsshow signs of healed traumatisms, with a large callus formation on thefractured femur. The disarticulated fibula is not visible in these views. Scale

bar, 50 mm. Abbreviations: cav, first caudal vertebra; fem, femur; ili, ilium;isc, ischium; ly1–ly3, first, second, and third lymphapophyses; plz, sacralpleurapophysis; pub, pubis; r, rib; sav, sacral vertebrae; tib, tibia; tro,trocanter; psv, last presacral vertebra.

NATURE|Vol 440|20 April 2006 LETTERS

1039

© 2006 Nature Publishing Group

PAUP* version 4.0b10 (ref. 20), with the branch-and-bound search optionimplemented. Bremer support and bootstrap percentages based on 10,000heuristic replications are given in Fig. 3. Bremer support was calculated withthe McClade interface with PAUP. All multistate characters were treated asunordered. Character descriptions, the data matrix and a list of apomorphiesdiagnosing each relevant clade are given in Supplementary Information.

Received 22 August; accepted 9 November 2005.

1. Underwood, G. Simplification and regeneration in the course of evolution ofSquamate reptiles. Coll. Int. Cent. Natl Rech. Sci. 266, 341–-352 (1977).

2. Cohn, M. J. & Tickle, C. Developmental basis of limblessness and axialpatterning in snakes. Nature 399, 474–-479 (1999).

3. Caldwell, M. W. & Lee, M. S. Y. A snake with legs from the marine Cretaceousof the Middle East. Nature 386, 705–-709 (1997).

4. Tchernov, E., Rieppel, O., Zaher, H., Polcyn, M. J. & Jacobs, L. L. A fossil snakewith limbs. Science 287, 2010–-2012 (2000).

5. Rage, J. C. & Escuillie, F. Un nouveau serpent bipede du Cenomanien (Cretace).Implications phyletiques. C. R. Acad. Sci. Paris Earth Sci. 330, 513–-520 (2000).

6. Lee, M. S. Y., Bell, G. L. & Caldwell, M. W. The origin of snake feeding. Nature400, 655–-659 (1999).

7. Lee, M. S. Y. & Scanlon, J. D. Snake phylogeny based on osteology, softanatomy and ecology. Biol. Rev. 77, 333–-401 (2004).

8. Zaher, H. The phylogenetic position of Pachyrhachis within snakes (Squamata,Lepidosauria). J. Vertebr. Paleontol. 18, 1–-3 (1998).

9. Rieppel, O. & Zaher, H. The intramandibular joint in squamates, and thephylogenetic relationships of the fossil snake Pachyrhachis problematicus Haas.Fieldiana Geol. 43, 1–-69 (2000).

10. Zaher, H. & Rieppel, O. The phylogenetic relationships of Pachyrhachisproblematicus, and the evolution of limblessness in snakes (Lepidosauria,Squamata). C. R. Acad. Sci. Paris Earth Sci. 329, 831–-837 (1999).

11. Apesteguıa, S. & Novas, F. E. Large Cretaceous sphenodontian from Patagoniaprovides insight into lepidosaur evolution in Gondwana. Nature 425, 609–-612(2003).

12. Makovicky, P. J., Apesteguıa, S. & Agnolın, F. L. The earliest dromaeosauridtheropod from South America. Nature 437, 1007–-1011 (2005).

13. Corbella, H., Novas, F. E., Apesteguıa, S. & Leanza, H. A. First fission-track agefor the dinosaur-bearing Neuquen Group (Upper Cretaceous), Neuquen basin,Argentina. Rev. Mus. Argentino Cienc. Nat. n.s. 6, 227–-232 (2004).

14. Zaher, H. & Rieppel, O. On the phylogenetic relationships of the Cretaceoussnakes with legs, with special reference to Pachyrhachis problematicus(Squamata, Serpentes). J. Vertebr. Paleontol. 22, 104–-109 (2002).

15. Rieppel, O., Kluge, A. G. & Zaher, H. Testing the phylogenetic relationships ofthe Pleistocene snake Wonambi naracoortensis Smith. J. Vertebr. Paleontol. 22,812–-829 (2002).

16. Scanlon, J. D. Cranial morphology of the Plio-Pleistocene giant madtsoiid snakeWonambi naracoortensis. Acta Palaeontol. 50, 139–-180 (2005).

17. Polcyn, M. J., Jacobs, L. L. & Haber, A. A morphological model and CTassessment of the skull of Pachyrhachis problematicus (Squamata, Serpentes), a98 million year old snake with legs from the Middle East. Palaeontol. Electr. 8,1–-24 (2005).

18. Rieppel, O. & Head, J. J. New specimens of the fossil snake genus EupodophisRage & Escuillie, from Cenomanian (Late Cretaceous) of Lebanon. Mem. Soc.Ital. Sci. Nat. Mus. Civ. Stor. Nat. Milano 32, 1–-26 (2004).

19. Caldwell, M. W. & Albino, A. Exceptionally preserved skeletons of theCretaceous snake Dinilysia patagonica Woodward, 1901. J. Vertebr. Paleontol.22, 861–-866 (2002).

20. Swofford, D. L. PAUP*. Phylogenetic Analysis Using Parsimony (*and OtherMethods). Version 4 (Sinauer, Sunderland, Massachusetts, 2003).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank O. Rieppel, H. W. Greene, J. C. Rage, F. Novasand A. Scanferla for discussion and review of earlier drafts; M. Reguero, S. Bargo,J. Bonaparte and A. Kramarz for access to material; P. Gallina for finding theholotype; A. Scanferla, A. B. Carvalho and M. Isasi for preparation of thespecimens; and L. Lobo for the illustrations. This research was supported by theJurassic Foundation (to S.A.) and the Fundacao de Amparo a Pesquisa doEstado de Sao Paulo BIOTA/FAPESP (to H.Z.). The Agencia Cultura of RıoNegro Province provided the exploration permits.

Author Contributions S.A. was Chief Investigator and Head of the excavationcampaigns in La Buitrera. H.Z. is responsible for the elaboration of the datamatrix and phylogenetic analysis.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to S.A. (paleoninja@yahoo.com.ar) or H.Z. (hzaher@ib.usp.br).

Figure 3 | Phylogenetic relationships of snakes showing the basal positionof Najash rionegrina. The result is expressed in a strict consensus of twoequally parsimonious trees (tree length of 270 steps, ensemble consistencyindex of 0.526, and retention index of 0.654). Bremer support and bootstrappercentages are given in the nodes (see Methods and SupplementaryInformation). Reconstructions of the pelvis and hindlimb elements ofNajash, Pachyrhachis and a boine snake are illustrated for comparison.

LETTERS NATURE|Vol 440|20 April 2006

1040