Reappraisal of the South American Miocene Snakes of the Genus Colombophis, with Description of a New Species

A redescription of the extinct snake genus Colombophis is presented, on the basis of new specimens from the late Miocene of southwestern Brazilian Amazonia, and those previously reported for the middle Miocene of Colombia and Venezuela. The reappraisal of Colombophis allows the recognition of a new species, C. spinosus sp. nov. The revised diagnosis of the genus is based on the midtrunk vertebrae, distinct from those of other snakes mainly in the features of the neural arch, position and shape of the neural spine, inclination of the zygapophyses, shape of the centrum, and development of the haemal keel. The affinities of Colombophis with “Anilioidea” are still unresolved; it is distinguished from all known extinct and extant “anilioids” due to its great vertebral size and the frequent presence of paracotylar foramina. The posterior paired apophyses of the haemal keel in some vertebrae, and the high neural spine of C. spinosus also contrast significantly with the “anilioid” genera, making the allocation of the genus into this probably paraphyletic group not well supported. Here, we recognized Colombophis as a basal alethinophidian of uncertain relationships.


Reappraisal of the south American Miocene snakes of the genus Colombophis, with description of a new species Introduction
Colombophis was a medium−size to large genus of snake, hitherto represented exclusively by the type species Colom− bophis portai Hoffstetter and Rage, 1977, based on about 40 midtrunk vertebrae recovered from the middle Miocene Villavieja Formation (Fish Bed) at the Los Mangos locality, near La Venta, Colombia. In spite of its relatively large size, Colombophis was considered belong to the "Anilioidea" (Hoffstetter and Rage 1977), a probably paraphyletic group of alethinophidian snakes. Later, Hecht and LaDuke (1997) recognized some additional incomplete vertebrae from the same area, but they did not describe or discuss the morphol− ogy of the genus. More recently, Head et al. (2006) referred a single vertebra from the middle Miocene of the Socorro For− mation (Venezuela) to Colombophis cf. C. portai. New spec− imens from the late Miocene of the Solimões Formation, southwestern Brazilian Amazonia, increase the knowledge of the vertebral morphology of Colombophis. Reviewing all the available material assigned to this genus and evaluating the intracolumnar and intrageneric variation, allows us to recognize a new species of Colombophis and to evaluate the taxonomic allocation of the genus into the "Anilioidea".

Material and methods
The material comes from the middle Miocene of Colombia (La Victoria and Villavieja formations) and Venezuela (Up− per Member of Socorro Formation), and from the late Mio− cene, southwestern Brazilian Amazonia (Solimões Forma− tion) (Fig. 1).
The La Venta Fauna of Colombia is one of the most con− spicuous Cenozoic vertebrate paleofaunas in South America. It originates from Honda Group beds including the La Victoria and the Villavieja formations, in the Magdalena River Valley, between the eastern and central Andes Mountains of south− western Colombia (Guerrero 1997). The vertebrate fauna from these formations belongs to the Laventan SALMA (South American Land−Mammal Age), middle Miocene . The holotype of the species Colombophis portai Hoffstetter and Rage, 1977, recognized among the remains from La Venta fauna, was not available for this study because the whereabouts of the specimens at MPNHN are currently un− known (Jean−Claude Rage, personal communication 2008). More recently, new specimens of Colombophis from the La Victoria and the Villavieja formations were collected during the Duke University−INGEOMINAS expeditions to the upper Magdalena River Valley between 1985 and 1991 . Part of this material was published by Hecht and LaDuke (1997) and is now stored at IGM. One of us (AA) had the opportunity to reanalyze the squamate material from this collection and found discrepancies regarding the collection numbers of the specimens and the taxonomic assignments pub− lished by Hecht and LaDuke (1997). Moreover, different boxes generally contain more than one specimen, including remains corresponding to more than one taxon, under the same collec− tion number. Due to the discrepancies mentioned above and the fact that it was not possible to know exactly which speci− mens were studied by Hecht and LaDuke (1997), the collection numbers of IGM are distinguished herein by addition of nu− merals between parentheses. They should not be considered as the same numbers of the published specimens.
The specimen from northwestern Venezuela comes from the Upper Member of the Socorro Formation, which crops out in the Falcón Basin, close to the Urumaco Municipality. It is only one vertebra, referred previously to Colombophis cf. C. portai by Head et al. (2006) and stored at AMU−CURS. Based on numerous previous studies of foraminifera and palynomorphs, a middle Miocene age was proposed for the Socorro Formation (Sánchez−Villagra and . According to the mammal fauna, the localities in the Solimões Formation in southwestern Brazilian Amazonia are referred to the Huayquerian SALMA, which would corre− spond to the late Miocene (Cozzuol 2006;Latrubesse et al. 2007) or even to the Montehermosan SALMA (late Mio− cene/Pliocene) (Latrubesse et al. 1997). In addition, Latru− besse et al. (2007), largely based on palynological data ob− tained in typical mammal fossil localities, proposed a late Miocene age for the fossils from the Solimões Formation.
The snake specimens collected from this formation are stored at UFAC−PV.
Among "Anilioidea", the comparative material used here in the review of Colombophis includes postcranial specimens of the extant species Anilius scytale (IB 40251, MZUSP 14572, 14573, and 14574). Data from the literature as well as figures of Cylindrophis ruffus from Ikeda (2007) were also used. The vertebrae of Uropeltidae and Anomochilus are un− known by us due to the difficulties of obtaining comparative material of these taxa, so that data from the literature were used (Lee and Scanlon 2002). The osteological nomenclature and measurements follow Auffenberg (1963), Hoffstetter and Gasc (1969), and Rage (1984Rage ( , 1998. The inclination of the prezygapophyses was taken considering the horizontal plane at the floor of the neural canal. The systematic arrangement follows Lee and Scanlon (2002). The measurements are ex− pressed in millimeters.
Emended diagnosis.-Fossil snake with midtrunk vertebrae characterized by the following combination of character states: medium to large size; clearly depressed neural arch, not vaulted in posterior view; shallow median notch of the posterior border of the neural arch; long dorsal surface of the neural arch, smooth or even concave, extending from the anterior edge of the zygosphene to the neural spine; neural spine reduced to a tubercle or relatively high and circular in outline, always re− stricted to the posterior end of the neural arch; zygapophyses prominent and strongly inclined above the horizontal plane, reaching the level of the zygosphene roof; prezygapophyseal process short; variable presence of paracotylar foramina; paradiapophyses weakly divided or even indistinct; centrum not markedly widened anteriorly; haemal keel broad, indistinct, and often only posteriorly developed, with the usual presence of two small and divergent apophyses more or less differenti− ated; and subcentral foramina placed close to the sagittal plane of the centrum, variably enlarged, reduced or absent.
Stratigraphic and geographic range.-Middle to late Mio− cene, Colombia, Venezuela, and Brazilian Amazonia. Hoffstetter and Rage, 1977 Figs. 2-5; Table 1. 1977 Colombophis portai Rage, 1977: 174-179, fig. 4. 1997 Colombophis portai;Hecht and LaDuke 1997: 95-96. Holotype: MNHN. VIV 6, one midtrunk vertebra.  plete remains; nonetheless, they present the general features described by Hoffstetter and Rage (1977) for this species. In general, the vertebrae are medium to large size, in this respect approximating an extant boa of 177 cm (Boa constrictor, MCN.D. 333) for the holotype of Colombophis portai. Also, the vertebrae are robust and not strongly depressed, although longer and broader than high (pr−po > h, pr−pr > h). The ante− rior and posterior trunk vertebrae are smaller than the mid− trunk vertebrae, but there is also variation in the size of the midtrunk vertebrae among the specimens from Colombia. The neural arch is longer than broad (pr−po > pr−pr) and its roof is depressed, especially in the posterior vertebrae ( Fig. 4), whereas there is a slightly vaulted neural arch in the anterior trunk vertebrae (Fig. 2). The posterodorsal notch of the neural arch is well defined but not very deep; each half of the roof being notably flattened. In lateral view, the neural arch rises posteriorly from about the origin of the anterior border of neural spine, which is restricted to the postero− dorsal end of the neural arch, so far from the zygosphene. The roof of the neural arch between the anterior edge of the zygosphene and the anterior edge of neural spine is slightly concave. The neural spine is very low but relatively robust, similar to an almost imperceptible tubercle, circular in out− line. The zygosphene is thin to moderate, but broader than the cotyle (zw > ctw). The anterodorsal edge of the zygo− sphene is variable between specimens, probably due to intra− specific variation. It can be rectilinear, notched or even slightly convex in dorsal view. The zygantra are small and deep, with a small foramen inside each zygantrum. The roof of the zygantra is almost rectilinear and continuous. The neu− ral canal is large, high, and triangular in outline. The medial borders of the prezygapophyses lie at a high position, at the level of the middle of the neural canal. They are antero− laterally directed and strongly inclined dorsally from the horizontal plane. The prezygapophyseal facet is oval and large (prl > prw). The prezygapophyseal process is short, al− though, in dorsal view, it can be seen exceeding laterally the tip of the prezygapophyseal facet due to the strong inclina− tion of the prezygapophyses. The postzygapophyses are also well inclined dorsally and posterolaterally orientated. The  interzygapophyseal constriction, between pre− and postzyga− pophyses, is deep and anteroposteriorly short. The centrum is longer than the width of the neural arch (cl/naw > 1). It is smooth, not markedly widened anteriorly and rather narrow. The subcentral ridges are not developed or only weakly de− fined. The anterior trunk vertebrae bear a prominent hypa− pophysis on the posterior surface of the centrum, broken in all specimens (Fig. 2). In the midtrunk vertebrae, there is a weakly developed haemal keel, which is anteriorly broad, smooth or convex, and usually narrower and prominent in the most posterior portion of the centrum (Fig. 3). The poste− rior trunk vertebrae have a well developed haemal keel that is  (2) 7.9 4.2 4.7 3.8 4.4 10.6 7.7 4.1 3.4 1.4 10 --4.3 2.8 --IGM 183928 8.7 3.4 4.4 4 4.4 9.6 7.7 3.5 3.2 0.9 - (1) 6.6 2.6 3.4 2.8 3 6.3 5.6 2.6 2.3 -8.4 8.7 7.9 2.8 2.1 1 -IGM 184285 (2) 6.5 2.7 3.5 2.5 3.3 8.1 5.7 -

Colombophis portai
8.5 5.5 6.3 5 6.2 16.6 -3 4.4 5.2 -16.9 10.1 -4.3 2.9 6.1 defined by the subcentral grooves (Fig. 4). The subcentral grooves are shallow in the anterior, mid−, and posterior trunk vertebrae, from the ventral margin of the cotyle to the middle of the centrum. They delimit the haemal keel anterolaterally, and they narrow toward the precondylar constriction. The subcentral foramina are variably enlarged, reduced, or ab− sent, and when present, located anterior to the prominent part of the haemal keel, and close to the sagittal plane of the centrum. They are usually located on the broad and flat ante− rior portion of the haemal keel (Fig. 5). Most specimens have a haemal keel with a rounded distal end, slightly projecting below the ventral surface of the centrum. In some specimens (mainly observed in the midtrunk vertebrae), the haemal keel has a bilobed distal end, where there are two small and diver− gent apophyses more or less differentiated (Fig. 5). The subcentral ridges and grooves are also morphologically dis− tinct among specimens. The vertebrae that show bilobed haemal keel usually have relatively deep subcentral grooves. Despite the poor preservation of the vertebrae, we infer that these different morphologies are probably linked to regio− nalization of the column. The cotyle and condyle are almost circular, slightly broader than high. The cotyle is not or scarcely visible in ventral view because it is not inclined and its rim is continuous and prominent. The main axis of the condyle is not notably inclined above the horizontal plane.
Only two specimens (UFAC−PV 3484 and 3478) could rep− resent juveniles, due to the small size, and because the cotyle and condyle are very dorsoventrally depressed. The presence of paracotylar foramina is irregular, indicating probably an intraspecific variation. Some specimens have only one fora− men or a pair of foramina on each side of the cotyle (UFAC−PV 4089, Fig. 3C), but others do not show any fo− ramina. In most specimens, the paradiapophyses are not pre− served; when present they are relatively small, usually sur− passing the ventral margin of the cotyle, and separated from it by well defined notches that become deeper in the posterior trunk vertebrae. The paradiapophyses are undivided. In the anterior and posterior trunk vertebrae, the paradiapophyses are almost vertical in lateral view, and in the midtrunk verte− brae they are posteroventrally inclined. In the posterior trunk vertebrae, the paradiapophyses are more prominent latero− ventrally, although they maintain far from the level of the prezygapophyseal tip (Fig. 4).  Description.-Although some vertebrae are somewhat frag− mented, data association, comparisons and description were possible, mainly based on the holotype. There are variations in vertebral morphology, but in general, the vertebrae are large, robust and high; higher than long (h > pr−po) and broader than high (pr−pr > h), with a centrum that is shorter than the width of the neural arch (cl/naw<1), and a neural arch much shorter than broad (pr−po < pr−pr). In anterior view, the neural arch is broad due to the long prezygapophyses. The zygosphene is rather thick and shows a straight dorsal margin, having small zygosphenal articular facets that are inclined dorsally. In two anterior trunk verte− brae (UFAC−PV 1609 and 2952), the dorsal margin of the zygosphene is slightly elevated in the middle. The width of the zygosphene varies considerably relative to the transverse diameter of the cotyle, being nearly equal as in the holotype (zw~ctw), wider, or even narrower than the cotyle. The prezygapophyses are slender, long and strongly inclined dorsolaterally, around 25°from the horizontal plane, reach− ing the level of the dorsal margin of the zygosphene (Figs. 6 372 ACTA PALAEONTOLOGICA POLONICA 55 (3), 2010  foramen, considered consistent with intraspecific variation as in C. portai. The paradiapophyses, fragmented on the left side in the holotype, are relatively small, not surpassing the ventral margin of the cotyle. In the posterior trunk vertebra (UFAC−PV 3485), the paradiapophyses extend further later− ally, almost reaching the median level of the prezygapo− physes, the parapophyseal facet almost exceeding the ventral limit of the cotyle, probably due to the greater lateral expan− sion and the anteroventral orientation of the parapophyseal facet (Fig. 8).
In posterior view, the two halves of the neural arch are con− siderably flattened. The neural spine is robust and cylindrical, remarkably high and columnar. The posterodorsal notch of the neural arch is relatively well marked. The neural arch is more depressed in the posterior trunk vertebra (UFAC−PV 3485). The postzygapophyses are elongated and strongly inclined dorsolaterally. The zygantra are large and deep, with a small foramen inside. The articular surfaces are well developed, and the roof of each zygantrum constitutes a continuous and straight dorsal margin in the holotype. The condyle is nearly circular. Ventral to the condyle, the haemal keel can be seen sometimes as a posterior prominence (mid− and posterior trunk vertebrae), or as a well developed hypapophysis in the anterior trunk vertebrae.
In lateral view, the neural spine is robust and well devel− oped, being considerably higher in some specimens (UFAC− PV 1609, 2952, 2956, and has an epi− physeal articular facet in the distal end. It is very short antero− posteriorly and its anterior margin is slightly concave, distant from the zygosphene. It is restricted to the posterior extrem− ity of the neural arch, and is vertical in orientation. On the posterolateral margin of the neural spine, a crest follows up on each side, as the continuation of the posterior margin of the neural arch. The side walls of the neural arch are short. The paradiapophyses are robust and are located ventrally far from the prezygapophyseal articular surfaces. The dia− and parapophysial surfaces are weakly separated; the diapo− physis is slightly convex and the parapophysis is rather con− cave. The cotyle is strongly prominent in some specimens, where the anterolateral edge surpasses the level of the ante− rior edge of the zygosphene. Small lateral foramina are visi− ble on the lateral walls of the neural arch, more or less posi− tioned at the diapophysial level (holotype) or just above it (other specimens). The length of the centrum is smaller than the width of the neural arch (cl/naw < 1), and clearly inclined posteroventrally in the holotype and other specimens, where it distally bears a relatively prominent haemal keel that is limited laterally by relatively well marked and deep sub− central grooves.
In dorsal view, the neural arch is much shorter than broad (pr−po < pr−pr). The posterodorsal notch of the neural arch is well−marked but not deep, and the broad and robust base of the neural spine grows up in its midline. The surface between the anterior edge of zygosphene and the neural spine is hori− zontally oriented and smooth, where the distance between the two structures is relatively large, due to fact that the neu− ral spine is situated well posteriorly. The articular facets of the prezygapophyses are comparatively slender, longer than broad (prl > prw), and the main axis is strongly laterally ori− entated. A small and sharp−edged prezygapophyseal process projects beyond the articular facet of the prezygapophysis. In the posterior trunk vertebra, the prezygapophyses are antero− laterally directed. The postzygapophyses are strongly ori− ented laterally. The interzygapophyseal constriction is well− marked and very short, between the pre− and postzygapo− physis on each side. The anterior margin of the zygosphene is straight or concave.
In ventral view, the centrum is triangular, its ventral face being broadly rounded anteriorly, very short (cl < naw), and wide. In the holotype, UFAC−PV 3485, 4027, and 5716E (midtrunk vertebrae), the subcentral grooves are deep from the ventrolateral margin of cotyle until mid−length of the centrum, limiting anterolaterally the haemal keel, which nar− rows posteriorly. In the UFAC−PV 1609, 2952, and 2956 (anterior vertebrae), the subcentral grooves are limited and more evident in the middle portion of the centrum, and there is a hypapophysis in the most posterior portion. The haemal keel is conspicuous, although not very prominent in the midtrunk vertebrae. Usually, it has two divergent margins in its posterior rim that produce a bilobed aspect, attaining the precondylar constriction. Near mid−length of the centrum, on each side of the haemal keel, there are small subcentral fo− ramina, anterolaterally situated and very close together. The subcentral ridges are relatively well marked, extending ap− proximately from the level between the dia−and parapo− physes to the condyle. In the holotype and in UFAC−PV 2952, 3485, 4027, and 5716E, the paradiapophyses are sepa− rated from the ventrolateral edge of the cotyle by a small and shallow notch. In other specimens, this constriction is dis− creet and subtle, probably in part due to the high degree of fragmentation in this region. Much of the condylar surface is exposed in ventral view, where the precondylar constriction is moderately marked.
Remarks.-After comparison of Colombophis vertebrae, it became clear that some differences cannot be attributed to intraspecific or intracolumnar variation, and hence warrant the erection of a new species. These differences are mainly the proportions of the vertebrae, the height of the neural spine, the morphology of the paradiapophyses, and the robustness of the zygosphene. The neural arch and centrum of the midtrunk ver− tebrae of C. spinosus are shorter than in C. portai. This is a re− sult of the zygapophyses being laterally oriented in C. spinosus, producing a short neural arch, and values of the centrum length much lower than the width of the neural arch in the middle (cl/naw < 1). In contrast, the zygapophyses are more anterolaterally directed in C. portai and the centrum length is subequal to or greater than the width of the neural arch in the middle (cl/naw ³ 1). In addition, the paracotylar notches and subcentral grooves seem to be relatively more marked in C. spinosus than in C. portai. The neural spine is high, clearly distinctive, very robust, with the main axis verti− cal in C. spinosus, but it is very low and reduced to a small tu− bercle in C. portai. The dia− and parapophyseal articular sur− faces are weakly distinguishable in C. spinosus, but they are undistinguishable in C. portai. Furthermore, the zygosphene of C. portai is thin to moderate, whereas it is usually thicker in C. spinosus. Based on these characters, it is possible to support the recognition of two species of Colombophis.
Recently, Head et al. (2006: fig. 1A) assigned one pre− cloacal vertebra from the middle Miocene of Venezuela to Colombophis cf. C. portai (AMU−CURS 154). According to the authors, the specimen is morphologically indistinguish− able from the specimens of C. portai from the middle Mio− cene of the La Venta Fauna. Nevertheless, the description of this specimen and the direct observation of its features are consistent with the vertebral morphology of C. spinosus. Ac− cording to the description of Head et al. (2006), this speci− men shows no paracotylar foramina (congruent with the intracolumnar variation of Colombophis) and has paradiapo− physes strongly divided. The latter is a character that con− trasts with the diagnosis of the genus, but according to our observations, AMU−CURS 154 displays paradiapophyses weakly divided into two articular facets, where the diapo− physis is slightly convex and the parapophysis is rather con− cave, which support its reference to C. spinosus. In addition, this vertebra is evidently short (neural arch and centrum), the zygapophyses are laterally oriented and define a very short interzygapophyseal constriction, the neural spine looks higher than in C. portai, and the zygosphene is thick, all characters observed in C. spinosus.

Discussion
Traditionally recognized as "Anilioidea", this group of basal alethinophidians is comprised of taxa that retain certain liz− ard−like features and are as well specialized to fossorial habits (Greene, 1997). "Anilioidea" is considered by some authors to be a paraphyletic group (Rieppel 1988;Rage 1998;Lee and Scanlon 2002;Hedges 2002, 2004;Wilcox et al. 2002;Gower et al. 2005). Recent molecular evidence is now quite strong in favor of splitting Anilius (as close relative of tropidophiids s.s.) and Uropeltidae s.l. (Cylindrophis, Anomo− chilus, uropeltines), as closer relatives of booids, pythons and advanced snakes (Wiens et al. 2008). Traditionally, the "ani− lioids" include the South American genus Anilius (red pipe snake or false coral snake), the Asian Anomochilus (dwarf pipe snakes) and Cylindrophis (Asian pipe snake), and the Uropeltidae family (shield−tailed snakes) (Greene 1997). The three former genera are generally included in the Aniliidae, although there is no consensus about its monophyly (Rage 1998;Lee and Scanlon 2002). Six fossil genera have been de− scribed for the group, and nearly all are tentatively referred to the family Aniliidae (Australophis Gómez, Báez, and Rou− gier, 2008;Colombophis Hoffstetter and Rage, 1977;Conio− phis Marsh, 1892;Eoanilius Rage, 1974;Hoffstetterella Rage, 1998;and Michauxophis Bailon, 1988). Although placement among the "Anilioidea" is well supported for most genera, the set of snakes allocated to Coniophis shows a large range of variation and represents probably a paraphyletic or polyphy− letic grouping of pre−macrostomatan snakes (Rage 1998).
The genus Colombophis was reported from the middle Miocene of Colombia and Venezuela, in northern South America (Hoffstetter and Rage 1977;Hecht and LaDuke 1997;Head et al. 2006). Hence, the new material described in this paper extends the record of the genus to the late Mio− cene of southwestern Brazilian Amazonia.
All previous descriptive works on Colombophis agree in including this genus in the "Anilioidea" (Hoffstetter and Rage 1977;Hecht and LaDuke 1997;Head et al. 2006), although comparisons with other snakes have not been reported. The new vertebral remains of Colombophis from the middle Mio− cene of Colombia and Venezuela, and the late Miocene of southwestern Brazilian Amazonia, provide some basis for re− vision of the genus and consideration of its affinities.
The diagnosis of the genus Colombophis was originally based on around 40 midtrunk vertebrae from the middle Mio− cene of the Villavieja Formation, Colombia (Hoffstetter and Rage 1977). According to the authors, the vertebral morphol− ogy of Colombophis is similar to that of the extant "anilioid" Cylindrophis, differing in their size and their undivided para− diapophyses (Hoffstetter and Rage 1977). Later, Hecht and LaDuke (1997), based on new material from the same forma− tion, added a new character to the diagnosis of Colombophis: the unusual placement of the subcentral foramina, which oc− curs close to the sagittal plane and just posterior to the level of the paradiapophyses; however, this condition is also observed in most "Anilioidea" (our personal observation). Hecht and LaDuke (1997) made a mistake during the English translation of the French diagnosis of the genus as provided by Hoffstetter and Rage (1977), because according to the last authors, the articular facets of the zygapophyses are notice− ably inclined above the horizontal, whereas Hecht and LaDuke (1997) considered that they are slightly inclined.
Some features of Colombophis are shared with Dinilysia patagonica Woodward, 1901, a Late Cretaceous basal snake from Patagonian Argentina. Both are of medium to large size and have vertebrae with the following characteristics: depressed neural arch, long and strongly inclined zygapo− physes, short prezygapophyseal process, and a variable pres− ence of paracotylar foramina. However, Dinilysia shows a straight (not notched) posterior border of the neural arch (Rage and Albino 1989;Scanferla and Canale 2007). Despite variable neural spine height in Colombophis (see below), C. spinosus displays a well developed neural spine as in Dinilysia. In the latter taxon, however, the neural spine is blade−like and posteriorly inclined, with an elongated base, rising close to the dorsal edge of the zygosphene; thus, it is different from the neural spine of both C. spinosus and C. portai, which is restricted to the posterior end of the neural arch. Dinilysia also differs from Colombophis in having: a better developed haemal keel in midtrunk vertebrae, the ante− rior edge of the zygosphene strongly notched, and an anteri− orly widened vertebral centrum. According to Apesteguía and Zaher (2006), Najash rionegrina, the earliest limbed snake from Patagonian Argentina, shows the neural arch flat− tened without posterodorsal notch, but the vertebrae of this genus are characterized by the presence of parazygantral fo− ramina on each side of the zygantrum and the absence of prezygapophyseal process, as in the extinct Madtsoiidae, dif− fering considerably from Colombophis.
Some authors indicate a probable relationship between Colombophis and the extant uropeltids (McDowell 1987;Szyndlar 1994); nevertheless, the large size of Colombophis and the presence of neural spine and haemal keel, especially in C. spinosus, contrast markedly with uropeltid vertebrae, which are small and strongly modified for fossorial habits, losing the neural spine and haemal keel. These and other characters, such as the long prezygapophyseal process, and condyles and cotyles markedly oval, differentiate the primi− tive Scolecophidia from Colombophis , although the frequent presence of large subcentral foramina is reminiscent of this group (Hoffstetter and Rage 1977).
In spite of numerous records in most continents, the fossils assigned to the "Anilioidea" usually consist of isolated verte− brae, and the characters that support the identifications are thought to be mostly primitive. The vertebrae of the extant "Anilioidea" Anilius and Cylindrophis share the following characters also present in Colombophis: a clearly depressed neural arch; prominent and strongly inclined zygapophyses; short prezygapophyseal process; a shallow median notch in the posterior border of the neural arch; and a centrum not markedly widened anteriorly. In the comparison of Colom− bophis with extinct and extant "anilioids", the inclination of the prezygapophyses at more than 20°is a character−state shared with Australophis, Hoffstetterella, Anilius, and Cylin− drophis (Rage 1998;Gómez et al. 2008;our personal observa− tion). The exceptions are Eoanilius and Michauxophis, which display almost horizontal prezygapophyses (Rage 1974;Bailon 1988;Szyndlar 1994); and Coniophis, in which the condition is variable (Hecht 1959;Rage 1984Rage , 1998Albino 1990). Ac− cording to Lee and Scanlon (2002), an inclination between 15°a nd 30°is interpreted as an intermediate condition in modern snakes. Our observations support this statement.
A posterior margin of the neural arch not well−notched in dorsal view is observed in all genera of "Anilioidea" (Rage 1998;Gómez et al. 2008), although Coniophis has an almost rectilinear posterior edge (Albino 1990) and Eoanilius, Hof− fstetterela, and Colombophis have a relatively deeper median notch (Hoffstetter and Rage 1977;Rage 1998;our personal observation). The absence of a strong notch of the neural arch is considered a plesiomorphic condition in snakes (Lee and Scanlon 2002).
As said above, the depressed neural arch of Colombophis and "anilioids" is frequently present in other primitive snakes such as Dinilysia, Najash, Scolecophidia, and Uropeltidae. The centrum not markedly widened anteriorly is found in the two last groups (Rage 1984;Rage and Albino 1989). Thus, the combination of character states that Colombophis shares with "anilioids" are mostly present in primitive snakes.
Other characters of Colombophis are less broadly distrib− uted. Concerning the neural spine, Colombophis differs from Australophis, Hoffstetterella, some species of Eoanilius, and Anilius, because these taxa have a thin, blade−like neural spine with an elongated base that rises close to or in the midline of the neural arch, being anteroposteriorly inclined. Colombophis spinosus shares with Hoffstetterella a neural spine relatively better developed than in the other mentioned genera, but it is higher in C. spinosus than Hoffstetterella. Other "anilioid" genera (the extant Cylindrophis and the ex− tinct Coniophis, Eoanilius, and Michauxophis) have a low and posteriorly restricted neural spine as in C. portai. Among snakes, the presence of a low neural spine is considered a de− rived condition by Lee and Scanlon (2002). This feature is common in extinct and extant "anilioids", and implies that, if C. spinosus is considered an "anilioid", it would be an excep− tion within this group. In conjunction with the position and shape of the neural spine, Colombophis shows a large smooth or slightly concave area between the dorsal margin of zygosphene and the neural spine. This character is also pres− ent in Cylindrophis, Coniophis, Michauxophis, and some Eoanilius specimens.
The dia− and parapophysial surfaces of the paradiapo− physes are slightly distinguishable in C. spinosus, whereas C. portai has indistinguishable paradiapophyses (Hoffstetter and Rage 1977;Hecht and LaDuke 1997). Rage (1998) comments that the dia− and parapophysis are slightly different from one another in Hoffstetterella, Cylindrophis, and Michauxophis, whereas they are not distinguishable in Colombophis portai, Anilius, and some species of Coniophis, and variably distin− guishable in Eoanilius, Coniophis platycarinatus, and C. pre− cedens. The presence of slightly divided paradiapophyses is also observed in some specimens of Anilius (our personal ob− servation) and Australophis (Gómez et al. 2008). According to some authors, distinguishable dia− and parapophysis is a de− rived condition found in all alethinophidian snakes (Rieppel et al. 2002;Apesteguía and Zaher 2006), although it should be considered present in many but not all alethinophidians all (e.g., all macrostomatans and various "anilioids", Jean−Claude Rage, personal communication 2009). Thus, the presence of undivided paradiapophyses in C. portai is primitive.
Although the haemal keel is variable along the column, it is more developed and prominent in the posterior portion of the vertebral centrum, which is in part different from some "ani− lioids". In Hoffstetterella, the haemal keel is a low blade poorly delimited laterally by subcentral grooves (Rage 1998). In Coniophis, it is broad and flat, somewhat delimited laterally by subcentral grooves (see Albino 1990;Rage 1998), but has a very convex surface (also in Eoanilius, which displays a large convex ventral surface). Colombophis differs from Anilius in which the projecting part of the haemal keel extends further anteriorly. It differs from Cylindrophis because in this genus the centrum is extremely convex, although somewhat rounded anteriorly like in Colombophis. Although showing a flattened haemal keel, Australophis somewhat resembles the condition observed in Colombophis, in which the keel is slightly promi− nent in the median portion of the vertebra, being delimited by the subcentral grooves (Gómez et al. 2008). In the most poste− rior portion of the haemal keel of Australophis there are two rounded depressions, one on each side of the distal margin of the haemal keel (Gómez et al. 2008). This character contrasts with the presence of the laterally paired projections that Co− lombophis shows in the same place, reminiscent of the poste− rior apophyses of some madtsoiid snakes (Rage 1998;Scanlon 1997Scanlon , 2005 and not reported in other "anilioid" genus. Scan− lon (1997Scan− lon ( , 2005 interpreted these projections as an autapo− morphic condition of madtsoiids, which could be correlated with intracolumnar variation; however, its presence in Colom− bophis is probably an independent acquisition. In addition, some characters of Colombophis are not pres− ent in any extant or extinct "Anilioidea". The most conspicu− ous of these characters are the large size and the presence of paracotylar foramina in many vertebrae. The vertebrae of Colombophis are larger than those of all other fossil and ex− tant "anilioids" (Table 2); the considerable disparity in verte− bral size between specimens is consistent with intracolumnar variation. The presence of paracotylar foramina is irregular in Colombophis. Some specimens have one or more foram− ina on each side of the cotyle (Hoffstetter and Rage 1977; our personal observation), whereas others do not show any fo− ramina (our personal observation). Anilius and Cylindrophis, as well as the extinct "anilioid" taxa, do not exhibit para− cotylar foramina (Rage 1974(Rage , 1984Bailon 1988;Albino 1990;Gómez et al. 2008). According to Lee and Scanlon (2002), the presence of paracotylar foramina on most or all vertebrae, as seems to be the case in Colombophis, is a plesiomorphic condition.
Based on the detailed comparisons made above, the affini− ties of Colombophis with "Anilioidea" still cannot be re− solved, because many characters are plesiomorphies, shared with other primitive snakes. Also, Colombophis is distin− guished from all known extinct and extant "anilioids" due to its great vertebral size and the frequent presence of paracotylar foramina. The posterior paired apophyses of the haemal keel in some vertebrae, and the high neural spine of C. spinosus, also contrast significantly with all extinct and extant "anilioid" genera. The allocation of the genus into this probably para− phyletic group is not well supported at present.
The combination of characters found in Colombophis is not present in any other extant or fossil snake, supporting its identity as a distinct genus; however, the fact that many of these features are observed in primitive snakes suggests that Colombophis belongs among the broad array of basal ale− thinophidian snakes.

Conclusions
In this paper we report the first record of Colombophis from the Solimões Formation, late Miocene of Southwestern Bra− zilian Amazonia. This record extends the distribution of the genus to the southeast during the Miocene, and implies its survival until the late Miocene. Reassessment of the genus permits the recognition of the new species Colombophis spinosus. The allocation of the genus into the probably para− phyletic "Anilioidea" cannot be resolved for the moment and it should be considered a probable basal alethinophidian of uncertain affinities.
The presence of Colombophis in the middle Miocene of Colombia and Venezuela, and the late Miocene of Brazil (Hoffstetter and Rage 1977;Hecht and LaDuke 1997;Head et al. 2006; this paper) suggests the possibility of general eco− logical similarity among these Miocene faunas. This resem− blance is also supported by the presence of the boid snake Eunectes and the teiid lizard Paradracaena in La Venta fauna and Brazilian Amazonia ). Some authors report similarities among these faunas based on mammalian fossils, but finding more affinities of the southwestern Brazilian Amazonia with Uru− maco than with La Venta (Cozzuol 2006).
The Solimões Formation, southwestern Brazilian Ama− zonia, includes a freshwater vertebrate fauna (rodents, croco− diles, turtles, and freshwater fishes) and, together with paly− nological data, indicates open areas and forest galleries along rivers, swamps, and shallow lakes. It would have been sub− ject to variation in the water level in a seasonal dry−humid tropical climate (Latrubesse et al. 2007). For the Urumaco Fauna, in the Socorro Formation, there is a scenario that in− cludes deltaic and fluvial deposits (Hambalek et al. 1994) with crocodiles, freshwater turtles, and catfishes which in− habited swamps, associated with other kinds of catfishes, sharks, and sirenians frequent in estuarine environments and in large freshwater rivers (Aguilera 2004;Sánchez−Villagra 2006). The La Venta Fauna is a continental deposit, with di− verse and abundant freshwater fishes, turtles, and crocodili− ans indicative of aquatic habitats that developed in a tropical rainforest, mixed with forest/grassland mosaics and open grasslands . The presence of similar fossil snakes in southwestern Brazilian Amazonia, Urumaco, and La Venta is consistent with these restorations.
The vertebral morphology of Colombophis, especially that of C. spinosus, is in part compatible with the lifestyle proposed for Dinilysia patagonica due to the combination of medium− large size, depressed neural arch, and high neural spine. Ac− cording to Albino and Caldwell (2003), the vertebral morphol− ogy of Dinilysia indicates a semi−burrowing or semi−aquatic lifestyle. Considering the proposed paleoenvironment for the Solimões Formation, the habits of Colombophis are well com− patible with a semi−aquatic lifestyle.