Rageryx schmidi, Smith & Scanferla, 2021

Smith, Krister T. & Scanferla, Agustín, 2021, A nearly complete skeleton of the oldest definitive erycine boid (Messel, Germany), Geodiversitas 43 (1), pp. 1-24 : 4-18

Rageryx schmidi n. gen., n. sp.

( Figs 1 ; 2 A-C; 3A-E; 4A-D; 5A-C; 6A-C; 7A, B; 8A-D; 9A-D; 10A-D; 11B-E; Appendix 1: Figs S1-S 8 )

urn:lsid:zoobank.org:act:AEB68284-BF3F-4980-8CD3-960415C732E0

HOLOTYPE AND ONLY KNOWN SPECIMEN. — HLMD-Me 9723, nearly complete skeleton.

TYPE LOCALITY. — Early or middle Eocene ( MP 11, Ypresian or Lutetian) of the Middle Messel Formation, Germany. Known only from type locality.

ETYMOLOGY. — After Dietmar Schmid, past president of the Senckenberg Gesellschaft für Naturforschung, in recognition of his estimable service to the society.

DIAGNOSIS. — Small boid snake with the following unique combination of characters: skull with short snout and moderately extensive braincase; orbits located in front of longitudinal midpoint of skull; anterior margin of premaxilla in line with arch defined by maxillae; nasal process of premaxilla small but distinct, with flat anterior face; maxillary tooth count around 16; maxilla with small posteromedial flange for ectopterygoid; prefrontal with anterolateral lamina and medial and lateral footprocesses; frontal table trapezoidal; parietal with low mid-sagittal ridge on posterior third; parasphenoid rostrum triangular with very broad base and weak, posterior ventral midline ridge; right Vidian canal larger than left; shelf bounding groove for posterior opening of Vidian canal obscures foramen for palatine branch of cranial nerve VII; supraoccipital significantly exposed externally; free end of supratemporal short; quadrate ramus of pterygoid with longitudinal, dorsally open groove; palatine ramus of pterygoid long; ectopterygoid straight with simple anterior end; coronoid present but reduced, lacking a distinct anteromedial process; compound bone with gradual decay of coronoid eminence anteriorly, and low, straight prearticular crest; about 220 precloacal and 38 caudal vertebrae; mid-trunk vertebrae with low and short neural spine; posterior trunk vertebrae without depressed neural arch; caudal neural spines relatively thick but not bifurcated; neural arch in middle of tail vaulted with flat, vertical posterior surface; distal caudal vertebrae short and tall (much taller than long); supernumerary process of caudal vertebrae present at least in rudimentary form, viz., pterapophyses present on majority of caudals, small postzygapophyseal wings and posterior extensions of the prezygapophysis present; and zygosphene-zygantral articulations present on distal caudal vertebrae.

REMARKS

Three erycine taxa have been named from the Eocene of Europe, Calamagras gallicus Rage, 1977 , Cadurceryx filholi Hoffstetter & Rage, 1972 and Cadurceryx pearchi Holman, Harrison & Ward, 2006 . The first is known from much of the Ypresian (early Eocene), from MP 8+9 to MP 10 ( Rage 1977, 2012). The second is known possibly as early as MP 13 and certainly from MP 16 to MP 19+20 ( Hoffstetter & Rage 1972; Rage 1984, 2012, 2013). The third is described only from the late Eocene Headon Hill Formation of England. The Middle Messel Formation, considered to be MP 11 ( Franzen 2005), falls in-between and straddles the Ypresian-Lutetian boundary ( Lenz et al. 2015).

Smith (2013) found that Calamagras weigeli from the late Eocene of North America is closely related to Ungaliophiinae , a result confirmed here. He also found that the type species of Calamagras Cope, 1873 , Cal. murivorus Cope, 1873 from the early Oligocene, shows similarity of proportions to Cal. weigeli , and even if a lack of referred material with diagnostic characters (absence of hemapophyses on all caudal vertebrae; see Smith 2013) currently prevents a critical appraisal of its relationships, it is not unlikely that Cal. murivorus will also turn out to be related to Ungaliophiinae . Thus, the genus name Calamagras cannot be used for HLMD-Me 9723.

Cadurceryx filholi , in contrast, is a highly derived species in which the accessory processes are not only more highly developed but also extend far into the trunk. This unusual morphology is not seen in any living erycine. Furthermore, Cad. filholi shows a very depressed neural arch on posterior trunk vertebrae, a derived character that it shares with extant Eryx (and also Charininae ) and that distinguishes it from the vaulted neural arch of HLMD- Me 9723. Whatever its precise phylogenetic relations, Cad. filholi is clearly not closely related to HLMD-Me 9723, and the name Cadurceryx cannot be applied to the specimen.

Cadurceryx pearchi , finally, is known exclusively from caudal vertebrae ( Holman et al. 2006). The two diagnostic features of Cad. pearchi with respect to Cad. filholi , according to the authors, were the size of the cotyle (larger than neural canal) and the presence of a ‘tubercle’ on the anterior end of the prezygapophysis. The tubercle was not labelled, but if we understand it correctly, it is absent in Rageryx schmidi n. gen., n. sp. The size of the cotyle with respect to the neural canal appears to be variable in the type series of Cad. pearchi ( Holman et al. 2006: figs. 2a, d and 3a, e), potentially as a result of vertebral position or intraspecific variation, calling into question the validity of this diagnostic feature. There are additional differences between Cad. pearchi and R. schmidi n. gen., n.sp, such as caudal vertebrae being taller than wide and having a more vaulted neural arch. Thus, there is no reason to consider that these two species are closely related.

The question now arises whether Rageryx n. gen. is the appropriate generic name for “ Calamagras ” gallicus . Rage (1977) stated that small pterapophyses are present in the holotype caudal vertebra, MNHN GR 7896, but other accessory processes are absent. It is remarkable, therefore, to find more of them present in Rageryx schmidi n. gen., n. sp., a younger European taxon. However, as noted in the description, it is the pterapophyses that occur most proximally in HLMD-Me 9723. Therefore, MNHN GR 7896 might derive from a more proximal part of the tail where the other processes do not occur. On the other hand, its proportions speak for a more distal position in the caudal series. MNHN GR 7896 also appears to have more bulbous processes than HLMD-Me 9723. With regard to the shape of the neural spine, HLMD-Me 9723 and referred trunk vertebrae of “ Cal. ” gallicus appear to be fully comparable. A clear resolution of this taxonomic problem will have to await the discovery and description of further caudal material from the Ypresian of France, preferably from the type locality, i.e., Grauves in the Paris Basin.

Given the distinctiveness and relatively early occurrence of Cadurceryx in Europe ( Rage 2012), one hypothesis is to derive it from “ Calamagras ” gallicus or Rageryx schmidi n. gen., n. sp. (or a closely related form). A second hypothesis is that it represents an early dispersal of a taxon more closely related to Eryx into Europe. Cranial remains of these fossils apart from Messel are meager, and there are no fossils, at present, from MP 12 ( Rage 2012). Consequently the data are insufficient to determine the relationships of Cad. filholi .

DESCRIPTION

HLMD-Me 9723 is a nearly complete skeleton, highly contorted and missing four sections of the axial skeleton ( Fig. 1A ). It is preserved on three plates that had been broken during excavation and rejoined. Because of the missing sections, the total length of the animal can only be given as c. 52 cm. A counterpart is not present in the HLMD collections.

The orbits are located in front of the longitudinal midpoint of the skull ( Fig. 1B ; Appendix 1: Fig. S1 ). The skull on the whole shows a short snout and moderately extensive braincase. Thus the general proportions of the skull are similar to ungaliophiine boids, especially Ungaliophis Müller, 1880 , and to Tropidophis Bibron in Ramon de la Sagra, 1840 ( Cundall & Irish 2008: fig. 2.62 and 2.65). They differ both from the primitive alethinophidian pattern, in which the elongate braincase is narrow anteriorly but widened at the otic capsule, and from most erycines, in which the braincase behind the orbit is shortened ( Cundall & Irish 2008).

Like those of several adult boid skulls examined, the posterior region of the parietal bone of HLMD-Me 9723 is clearly projected posteriorly, especially the supratemporal processes. This posterior projection overlaps the anterodorsal edge of the supraoccipital, like in adult individuals of large and small boids and macrostomatans in general (e.g., Smith & Scanferla 2016). Also, the well-preserved tips of the neural spines of the trunk vertebrae show well-finished caps of bone.Taken together, these anatomical traits present in the parietal bone, and the advanced state of ossification observed in skull and trunk vertebrae, evidence that HLMD-Me 9723 represents an adult postnatal ontogenetic stage. Thus, the species it represents was apparently smaller than typical adult Rosy Boas ( Lichanura trivirgata , length 43-112 cm), and many adult Rubber Boas ( Charina bottae , length 35-84 cm) ( Stebbins 2003). HLMD-Me 9723 is also smaller than many Eryx . Er. johnii and Er. tataricus are the largest of the genus, with total length in large individuals> 1 m, but all other species are smaller ( Seufer 2001).

Premaxilla

The premaxilla is only partially preserved, yet there is every indication that its anterior margin was in line with the curvature of the arch defined by the maxillae ( Fig. 1B, C ). Thus, it was not produced far forward, unlike in extant erycines, Loxocemus

bicolor Cope, 1861 and Calabaria reinhardtii . A small, ovate nasal process is present that projects posterodorsally and has a flat anterior surface that would have been visible externally between the nasal bones. It is similar to that of many boids and totally unlike the low crest completely hidden between the nasal bones in Eryx ( Cundall & Irish 2008) .

Maxilla ( Fig. 2 ; Appendix 1: Fig. S2 )

The maxilla is elongate, slightly dorsally arched element ( Fig. 2 A-C). Its anterior end is most notable for a large dorsal foramen (superior alveolar foramen, s.a.f.) continued anteriorly by a deep groove. This foramen is most comparable to that in Charina bottae , which, however, is disposed more laterally. It presumably transmits the subnarial artery and superior alveolar nerve onto the snout, but the reasons for its large dimensions are unknown. A weak ascending process begins to rise adjacent to the groove and terminates at the level of the superior alveolar foramen and palatine process (pl.pr.); it is more strongly developed than in Lichanura trivirgata ( Fig. 2H, I ), but not as strong as in Ch. bottae . The lateral surface of the bone is pierced by a single, elongate labial foramen (l.f.), which transmits branches of the maxillary artery and superior alveolar nerve. A small facet, probably for the prefrontal, is present medial to the ascending process.

The dorsal surface of the posterior end of the maxilla appears to show several fine, longitudinal striae ( Fig. 2A ), whose origin may lie with the ectopterygoid articulation. The palatine process is poorly preserved, and little can be said of its size, orientation, and morphology; however, it appears to have been asymmetrical, with a steep posterior margin and a probably more gradual anterior margin. Presumably the palatine process was pierced by a foramen, as in other boids. A single foramen is visible on the dorsal surface at the level of the posterior margin of the palatine process, as in other boids ( Fig. 2D, G ).

Nasals ( Fig. 3 ; Appendix 1: Fig. S3 )

The elongate nasals are gently dorsally convex in sagittal and transverse planes ( Fig. 3 C-E). They contact one another on the midline for most of their length. Each comprises a plate posteriorly over the nasal capsule (the horizontal lamina) and a long, slightly thickened anteromedial process ( Fig. 3A, B ). All examined Eryx also have a strong horizontal expansion of the anterior end of the horizontal lamina ( Fig. 3 F-H), which is lacking in Lichanura trivirgata ( Fig. 3 K-N) and other boids. The anteromedial processes curve ventrally to contact the nasal process of the premaxilla, which slightly separates them at their distal end. An anterolateral process, as in Boa Linnaeus, 1758 , is lacking. Posterolaterally the nasal is overlapped by the prefrontal, as in Li. trivirgata and Charina bottae but unlike in Eryx ( Rieppel 1978a; Cundall & Irish 2008). The posterolateral end of the nasal is squared off, but posteromedially there appears to be a posterior expansion; due to the tightly apposition of the frontal in this region, the true morphology is uncertain. A triangular, posteromedial prong extending posteriorly between the frontals was present in all examined Eryx but is lacking in Li. trivirgata and Ch. bottae . (In Boa and Candoia carinata , the nasals taper posteriorly and so they necessarily form a triangular point, but there is no anteromedian notch in the frontal; rather the space is occupied by the prefrontal. Thus, such a prong can be regarded as absent in these taxa.) There is no evidence of a vertical buttress that would have articulated with the lateral (subolfactory process) or medial pillars of the frontal at the prokinetic joint, unlike in extant erycines ( Fig. 3I, J, N, O ) as well as some other fossorial forms ( Rieppel 1978a; b; pers. obs.), but it is likely that the nasal contacted the frontal beneath the olfactory tracts, as in most constrictors. All examined Eryx have one or two foramina through the posterolateral corner of the nasal, although it is uncertain what structures pass through it; these foramina are lacking in HLMD-Me 9723 and other boids.

Prefrontal (Appendix 1: Fig. S4 )

On the whole, both elements are poorly preserved. A large anterolateral lamina is present ( Fig. 1B ), unlike in Eryx . The triangular dorsomedial projection of this lamina that extends medially behind the nasals toward the contralateral element is comparable in extent to Charininae ; the prefrontals do not meet one another on the midline. The orbital lamina forms the anterior wall of the orbit. The medial foot-process

is well developed and curves slightly laterally. The prefrontal is complete enough to deduce that the lateral foot-process was less well developed, more like Boa constrictor than the larger process of Eryx , Charininae and Calabaria reinhardtii . A small foramen of unknown significance pierces the ventral margin of the bone in the embayment between the two foot-processes.

Frontal ( Fig. 4 ; Appendix 1: Fig. S5 )

The smooth frontal table is trapezoidal, with a long medial, shorter and slightly concave lateral, and oblique anterior and posterior margins ( Fig. 4A ). Consequently there is no deep median notch between the frontals for reception of the nasals. In this it differs from the parallelgram-shaped table of Eryx ( Fig. 4E ). It appears that a small foramen exits dorsally through the posterolateral corner. The anterolateral margin is smooth and lacks a distinct notch for the prefrontal, unlike in Lichanura trivirgata ( Fig. 4I ). A prominent supraorbital shelf, as in Boinae , is absent ( Fig. 4B, C ). The posterolateral corner has a postolateral projection for accomodating the anterolateral corners of the parietal dorsally ( Fig. 4A, D ), like in Eryx (Fig. E, H) but unlike in Boinae and Li. trivirgata (Fig. I, L). A small facet for articulation of the postorbital may be present.

The ventral portion of the frontals is almost certainly present but could not be distinguished due to crushing. Thus, the extent and morphology of the medial and lateral frontal pillars and the posteroventral projection bounding the optic foramen cannot be ascertained.

Postorbital (Appendix 1: Fig. S6 )

The dorsal portion shows an elongate facet where it articulated along the parietal and, anteriorly, a small part of the frontal ( Fig. 1B ). It tapers strongly ventrally, so that the postorbital process is thin, unlike in Boinae and some Eryx (e.g., Er. colubrinus ). The process is broken, so that its ventral extent is uncertain, but comparison of the left and right elements suggests that it was less extensive than in most booids. The preserved portion shows no evidence of a posterior deflection, as is present in Candoia carinata and many Eryx (e.g., Er. colubrinus , Er. conicus , Er. jayakari and Er. tataricus , but not Er. jaculus ).

Parietal

The parietal bone is relatively broad, only slightly longer than wide ( Fig. 1B ). Its widest point is found anterior to midlength, well in front of the otic capsules. The anterior margin is shallowly concave except at the midline, where a small process projects between the frontals. The dorsal surface of the anterior part of the parietal is flat, and a low mid-sagittal ridge is developed only in about the posterior one-third of the bone. In this respect it is similar to Lichanura , Charina and Ungaliophiinae and differs from Eryx and Boinae , in which the sagittal crest is sharper and far more extensive ( Cundall & Irish 2008). The ventral extent of the parietal forming the lateral wall of the braincase is almost certainly present, but could not be distinguished due to crushing.

Parabasisphenoid (“sphenoid” of Cundall & Irish 2008)

( Fig. 5 ; Appendix 1: Fig. S7 )

This is a triangular element with a broad, regularly tapering rostrum ( Fig. 5A, B ). The rostrum is broader relative to the width of the basisphenoid portion of the bone than in any examined constrictor. Its dorsal surface is weakly concave in transverse section ( Fig. 5A ). Its ventral surface is nearly flat proximally, but distally there is a ventral keel formed beyond the terminus of the cristae trabecularis. On the main body of the basisphenoid portion there is a weak, midline ridge, similar to that seen in some constrictors, like Loxocemus bicolor , Lichanura trivirgata , and Candoia carinata , but no indication that the ridge bifurcates anteriorly, as it does in Lo. bicolor (Smith 2013) and Li. trivirgata ( Fig. 5H ).

In dorsal view the sella turcica – dorsal margin of the dorsum sella or pituitary fossa – is approximately in line with the greatest lateral extent of the bone, but crushing has nearly obliterated the fossa.It appears that the badly crushed parasphenoid wings are strong with a well-developed articulation for the parietal articulation, but their exact extent cannot be determined.If our interpretation is correct, these project more strongly than any observed in extant constrictors, except Candoia carinata , where

they are also anteroposteriorly longer. Neurovascular foramina are difficult to distinguish in the CT scan. The right egress for cranial nerve VI, however,appears to be present at approximately the level of the lateral margin of the pituitary fossa. Assuming mirror symmetry for the left egress, the foramina would be widely spaced, like in Lichanura trivirgata ( Fig. 5G ) and most examined snakes but unlike in Loxocemus bicolor (Smith 2013) , some Eryx [e.g., Er. johnii ( Fig. 5D ), Er. tataricus ] and Candoia carinata . The right Vidian canal is distinctly larger than the left one, as in Boidae ( Underwood 1976) .

Prootic ( Fig. 6 )

The opening for the maxillary ramus of the trigeminal nerve (V2) is presumably situated between the prootic and the parietal, with the prootic deeply notched for the nerve, but the notch is not distinct on either side ( Fig. 6A ). The opening for the mandibular ramus (V3) is located posteriorly. In most boids ( Fig. 6D, G ) these foramina are separated by an ophidiosphenoid (sensuGauthier et al. 2012), but it is lacking in Eryx colubrinus , Er. jaculus , Er. muelleri , and Er. tataricus (among examined Eryx ); the region is too damaged to be certain in HLMD-Me 9723. The hyomandibular branch

of the facial nerve (VIIh) opens within the trigeminofacialis chamber, and well within the margins of the lateral opening of V3. Dorsally the prootic evinces an elongate groove for the reception of the supratemporal.

In ventral view the prootic exhibits an ophidiosphenoid foramen anteriorly ( Fig. 6C ), like in many boids but unlike in Ungaliophis continentalis Müller, 1880 . There is an anteromedially trending groove that would have continued into the posterior opening of the Vidian canal on the basisphenoid. At the base of the groove is a relatively large foramen for the palatine ramus of cranial nerve VII, and the dorsolateral margin of the groove is expanded as a shelf that obscures said foramen in lateral view. A groove as such has a capricious phylogenetic distribution. Lichanura trivirgata ( Fig. 6I ) and Charina bottae , but not Ungaliophiinae or Eryx ( Fig. 6F ), share with HLMD-Me 9723 the shelf that obscures the foramen for the palatine branch of the facial nerve (VIIp). The ventral edge of the prootic in HLMD-Me 9723 is wedge-shaped, fitting into the broad notch between the parabasisphenoid anteriorly and basioccipital posteriorly.

In medial view two foramina, a larger anterior opening for cranial nerve V and a smaller posterior opening for cranial nerve VII, pierce the cranial vault to enter the trigeminofacialis chamber ( Fig. 6B ). The impression of the vestibulum and parts of the relatively large anterior and lateral semicircular canals can be seen. In contrast to Eryx , in particular ( Fig. 6E ), the vestibulum is relatively small. The anterior semicircular canal extends to the anterior margin of the bone and well away from the vestibulum.

Supraoccipital

The supraoccipital achieves significant exposure between the parietal and otoccipitals ( Fig. 1B ), comparable to Loxocemus bicolor , Lichanura trivirgata and Charina bottae , but unlike in Eryx ( Cundall & Irish 2008) . Anteriorly on the midline a

small prong projects into a corresponding notch in the posterior margin of the parietal.

Otooccipital (sensu Maisano & Rieppel 2007)

This paired element is poorly exposed, and segmentation was deemed too subjective due to crushing. The dorsal flanges of the otooccipitals meet broadly on the midline behind the supraoccipital, thereby forming the dorsal margin of the foramen magnum as in other snakes.

Supratemporal (Appendix 1: Fig. S8 )

The supratemporal is an elongate element, weakly convex ventrally and weakly concave dorsally, with rounded anterior and posterior ends ( Fig. 1B ). Dorsolaterally at the posterior end is a facet for the quadrate articulation. Just

anterior to this facet the bone has dorsal bulge that is mirrored by a concavity on the ventral surface. The free end of the supratemporal is very short, extending only slightly beyond the otooccipital.

Quadrate (Appendix 1: Fig. S9 )

The quadrate consists of a transversely oriented ventral condyle that articulates with the mandible and an oblique, triangular dorsal portion that contacts the supratemporal ( Fig. 1B ). The overall triangular shape is comparable to that in Ungaliophiinae , Charininae and Eryx . The ratio of the width of the cephalic condyle of the quadrate to the bone’s height is 0.55, comparable to Ungaliophiinae (c. 0.53 in Exiliboa placata Bogert, 1968 , 0.55 in Ungaliophis continentalis ) and Lichanura trivirgata (0.49) but unlike Boinae (0.29 in Candoia carinata , 0.33 in Boa imperator Daudin, 1803 ), Charina bottae (0.43), and Eryx (0.44 in Er. tataricus ). The quadrate is therefore relatively short and broad. A thick suprastapedial process (not “extrastapedial process” as in Smith 2013) projects ventrally from the medial edge of the bone; as it curves slightly anteriorly, it is hidden in lateral aspect. There is a prominent foramen in the dorsal portion of the bone on both right and left sides, which therefore cannot be attributed to an artifact; a comparable feature was not observed in any extant boid (although a small foramen in roughly this position is found in Exiliboa placata ). Another, smaller foramen pierces the anterior surface of the shaft just above the ventral condyle; such a foramen is present in numerous boids.

Septomaxilla

The dorsolateral process of the left element is preserved, but the posterior spine is broken ( Fig. 1B ). Little more can be discerned.

Vomer

The vomers are presumably present but not clearly identifiable due to crushing.

Palatine

At least the right palatine is almost certainly present, but it cannot be recognized clearly due to crushing.

Pterygoid ( Fig. 7 ; Appendix 1: Figs S10 , S 11 )

The pterygoid is best represented by the left element (anteriorly) and the right element (posteriorly). The anterior portion (that portion anterior to the ectopterygoid articulation) is long, comparable to the posterior portion ( Fig. 7A, B ). In this it is similar to Charina bottae and relatively longer than in Lichanura trivirgata ( Fig. 7G, H ), Eryx conicus and Boa imperator ; in other examined Eryx the anterior portion is much shorter than the posterior portion ( Fig. 7E, F ). The ectopterygoid process evinces a deep, roughly circular ectopterygoid facet, but the ectopterygoid process is not prominent ( Fig. 7A ); it is comparable to Boinae and some Eryx like Er. johnii ( Fig. 7A ) rather than Lichanura trivirgata ( Fig. 7G ) and Charina bottae .

The anterior portion is toothed ventrally ( Fig. 7B ). A single row of at least seven – and, if the tooth row continued as far posteriorly as the ectopterygoid articulation, as in other erycines, at least twelve – loci. The posterior portion (quadrate ramus) tapers continuously toward the sharp tip. As in Li. trivirgata ( Fig. 7G ) and Ch. bottae , it evinces a longitudinal dorsal groove ( Fig. 7C ). There is a medially or dorsally open groove in all examined Eryx ( Fig. 7E ), Boinae and Ungaliophiinae in the distal part of the quadrate ramus, but it represents a different surface of the bone; namely the medial edge crosses over dorsally to become the lateral edge of the bone, so the groove actually represents the ventral surface of the bone in Charininae .

Ectopterygoid ( Fig. 8 ; Appendix 1: Fig. S12 )

The left ectopterygoid ( Fig. 8A, B ) is the better preserved one; the portion after the anterior break is easily back-rotated. The bone as a whole is nearly straight, unlike the slightly curved element in many extant erycines ( Fig. 8 C-F) and Xenopeltis unicolor and the more strongly curved element of other constrictors. It shares with erycines the apomorphic absence of a sharp lateral corner. The bone grows in width anteriorly, like in most erycines, and evinces an articulation facet for the maxilla ventrally ( Fig. 8B ); the anterior end is simple and rounded. The bone becomes more robust posteriorly and has an expanded, posteromedially directed facet for the pterygoid; this facet is better developed than in most erycines, particularly Lichanura trivirgata ( Fig. 8E, F ).

Dentary ( Fig. 9 ; Appendix 1: Fig. S13 )

The dentary curves distinctly medially at its anterior end ( Fig. 9A, B ), more broadly so than in any examined boids ( Fig. 9E, H ) except Ungaliophiinae . The anterior end of the bone is also slightly downturned ( Fig. 9C, D ), like in Ungaliophiinae but unlike extant erycines ( Fig. 9F, G, I, J ). There is a single, large, anteriorly opening mental foramen at about the level of the 6th tooth. The subdental shelf is deep. The Meckelian groove is fully open and runs along the ventromedial edge of the bone. The connections with the postdentary bones are damaged and cannot be made out clearly. There is a deep notch for the compound bone, and the posterodorsal dentigerous and posteroventral processes appear to be roughly the same length, as in extant erycines ( Fig. 9F, I ) and Ungaliophis continentalis (but not Exiliboa placata ).

Splenial

The splenial cannot clearly be made out.

Coronoid (Appendix 1: Fig. S14)

A coronoid is closely associated with the medial surface of the anterior end of the compound bone ( Fig. 10D ). It is short and weakly arched, with a concave anterodorsal and convex posteroventral margin. A strong, anteriorly trending process is lacking, so the bone lacks the L-shape seen in Boa imperator and all Eryx where the bone could be examined (i.e., Er. colubrinus , Er. conicus , Er. jaculus , Er. jayakari, Er. muel-

leri) as well as fossil taxa Messelophis variatus Baszio, 2004 and Rieppelophis ermannorum (Schaal & Baszio, 2004) from Messel ( Scanferla et al. 2016). Yet, it is not so highly reduced as in Lichanura trivirgata , much less completely absent, as in Charina bottae ( Kluge 1993) and Ungaliophiinae .

Angular

The angular cannot clearly be made out.

Compound bone ( Fig. 10 ; Appendix 1: Fig. S14)

Part of the large surangular process can be distinguished anteriorly, including a medially directed ridge that inserted beneath the dentary tooth row ( Fig. 10D ). Behind the dentary articulation, there is a fine groove roughly between the surangular and prearticular portions of the bone; however, this groove continues posteriorly, longitudinally transects the glenoid fossa ( Fig. 10A ), and extends ventrally through the compound bone only on the posterior portion, so it must be an artifact rather than an indication of incomplete fusion of surangular and prearticular. The lateral surangular crest rises straight and gradually toward the coronoid eminence ( Fig. 10C, D ); its dorsal extent is approximately equal to that of the coronoid itself, so that both bones contribute to the coronoid eminence ( Fig. 10D ). The anterior margin of the compound bone contribution to this eminence decays more gradually, like in Eryx ( Fig. 10G, H ) and Candoia carinata and unlike in Lichanura trivirgata ( Fig. 10K, L ), Charina bottae , Boa imperator and Ungaliophiinae . The dorsal extent of the coronoid eminence is similar to that in other erycines. The medial, prearticular crest is low and nearly straight ( Fig. 10D ). It is nowhere visible behind the surangular crest in lateral aspect, unlike in examined boids except Ca. carinata and the erycines Er. colubrinus , Er. jayakari , Er. muelleri and Ch. bottae . Between them is the deep adductor fossa ( Fig. 10A ). Just below the coronoid eminence on the lateral surface is the anterior surangular foramen ( Fig. 10C ). Near the ventral margin on the lateral surface is a longitudinal ridge, also seen in some extant erycines. The retroarticular process is slightly longer than in most extant erycines and Ungaliophiinae .

Dentition

It is not the teeth at the very front of the maxilla that are the longest but rather those in the middle of the anterior half ( Fig. 2B ). Moving from front to back, that is, tooth length increases at first slightly before decreasing for most of the length of the jaw.

The maxillary tooth count can only be given approximately due to damage to both elements. On the left side, there is a toothless gap in the middle ( Fig. 2B ); based on the average length of the three preceding and three succeeding tooth bases, we estimate that this gap corresponds to approximately five teeth. There are four teeth at the front, and alveoli for seven teeth behind, for a total tooth count of c. 16. This value is higher than in most erycines, except for Lichanura trivirgata ( Kluge 1993) , but is lower than in most other constrictors ( Underwood 1976). The count is comparable to that seen in Ungaliophiinae .

The dentary teeth appear to be broadly comparable to those of the maxilla, with longer teeth anteriorly than posteriorly ( Fig. 9C, D ). However, a tooth count cannot be given due to damage.

Postcranial skeleton ( Figs. 11 ; Appendix 1: S15-S17)

Hind limb rudiments are lacking, so the postcranial skeleton comprises only vertebrae and ribs. With measurements of the average length of four preceding and four succeeding vertebrae that bracket the gaps, and an estimate of the trend of the vertebral column, sometimes supported by the presered ribs, we arrive at an estimate of 258 total vertebrae, of which 220 are precloacal and 38 are cloacal or postcloacal (Appendix 1: Fig. S17). Accordingly 14.7% of the individual vertebrae are caudals (cloacals or postcloacals), comparable to the proportion observed in the three extant ungaliophiine

species (14-17%) and that estimated for fossil Calamagras weigeli from the late Eocene of North America (14%, with 95% confidence interval 8-22%; Smith 2013). Lichanura trivirgata SMF-PH 21 has 236 precloacal vertebrae and 45 caudals (plus tail tip; see below), yielding 16%. A tail tip, apparently composed in part of fused vertebrae (Smith 2013), is present in HLMD-Me 9723 as well.

Anterior trunk vertebrae are defined as those vertebrae after the skull possessing a distinct hypapophysis. It forms a posteroventrally directed spine (Appendix 1: Fig. S15), with the fore margin becoming straighter and more horizontal on more posterior vertebrae. Anterior trunk vertebrae number 52-68 (the boundary between anterior and mid-trunk vertebrae falls in the first gap of approximately 16 missing vertebrae). Thus, the fraction of the trunk vertebrae belonging to the anterior trunk is 0.24-0.31, compared with 0.18 (95% confidence interval 0.11-0.29) in fossil Calamagras weigeli , 0.19 in Charina bottae , 0.25-0.28 in Ungaliophis and 0.35 in Exiliboa placata (Smith 2013) . The neural arches (“tecta” of Jandzík & Bartík 2004) of the atlas are broad (i.e., anteroposteriorly long). The axis and succeeding three vertebrae have rodlike and strongly posteriorly inclined neural arches that project beyond the posterior margin of the neural arch. On the seventh precloacal vertebra the neural spine bears an anterior expansion, which rapidly grows in strength on more posterior vertebrae. By the 10 th the neural spine is strong and more dorsally directed, so that it no longer (or just barely) projects beyond the posterior margin of the neural arch. On these more anterior vertebrae the postzygapophyses are angular, suggesting small or narrow zygapophyseal articulations; additionally, they are very (anteroposteriorly) short. On more posterior anterior trunk vertebrae the neural spine appears to become recumbent again, overhanging the notch in the posterior margin of the neural arch. The neural spine becomes less prominent and is anteroposteriorly short, as judged by the finished cap at its distal end, although it is prolonged by a thin, sharp ridge of bone with an oblique anterior margin that runs as far as the base of the zygosphene. A low, oval zygantral mound that creates space in the zygantrum for accomodation of the zygosphene (Hecht in McGrew et al. 1959) is present on all anterior trunk vertebrae. Vertebral length increases through the preserved portion of the anterior trunk vertebrae.

The mid-trunk vertebral series, as far as it is preserved, commences after the first gap and continues past the second gap (c. 14 vertebrae) and third gap (c. 4 vertebrae). Almost all of the neural spines are broken, but they appear to have been low and short, to judge by the broken edge. The zygantral mounds become more prominent, and on the anterior mid-trunk vertebrae they occur in conjunction with a flaring of the posterior margin of the neural arch beyond the level defined by the postzygapophyses. This flaring disappears by about the middle of the vertebral column; thereafter, the posterior margin of the neural arch is transverse except for the triangular notch on the midline. Vertebral length reaches a maximum in the middle of the mid-trunk vertebrae. A mid-trunk vertebra was not segmented, as a CT scan is not available.

An exact boundary between the mid- and posterior (Appendix 1: Fig. S16) trunk vertebrae cannot be drawn, but vertebral length gradually decreases toward the posterior end of the precloacal vertebral column. The vertebrae also appear to become relatively shorter. The notch on the posterior margin of the neural arch becomes broader, and on the posterior-most posterior trunk vertebrae it is so broad that the transversality of the margin is obliterated and the postzygapophyses are once again angular. The zygantral mound disappears. The neural spine, where preserved, is very low but slightly longer. The neural arch is also more vaulted than in Lichanura trivirgata ( Fig. 11A , posterior view) or Eryx . The anterior end of the zygosphene on a segmented posterior trunk vertebra appears to have a bulge ( Fig. 11B , dorsal view), giving it a crenate shape ( Auffenberg 1963).

The subcentral (lymphatic) grooves are very prominent on posterior trunk vertebrae, forming in the more posterior part of the series a slightly projecting keel or hypapophysis, which projects more strongly than in Lichanura trivirgata ( Fig. 11A , lateral view), where it is ventrally flattened. However, the lateral expansions of the ventral keel seen in Li. trivirgata ( Fig. 11A , ventral view) are absent. Prezygapophyseal accessory processes appear to be completely absent in HLMD-Me 9723 ( Fig. 11B , ventral view), unlike the prominent processes in Li. trivirgata ( Fig. 11A , dorsal view) and some Eryx .

Caudal vertebrae comprise cloacal and postcloacal vertebrae. At least two cloacal vertebrae were present; they are overlapped by the tail tip ( Fig. 1D ) and visible only through the epoxy plate and in the CT scan ( Fig. 1E ). They are identified by the fused, bifurcated, anteroventrally directed lymphapophyses, of which the dorsal spine appears to be the longer one (left side of first cloacal, broken on right).

The postcloacal vertebrae, lacking bifurcated lymphapophyses, are relatively short. However, the neural spine is relatively longer and mediolaterally thicker ( Fig. 11 C-E, dorsal and anterior views). Strictly speaking, it evinces no bifurcation anywhere, unlike in Charina bottae , Lichanura trivirgata ( Fig. 11F ) or Eryx spp., where the spine strongly expands in width distally and the tip presents a deep dorsal, midline groove. However, starting near mid-tail its anterior end is drawn out into a pair of short spines. A short zygosphene appears to be present on postcloacal vertebrae much closer to the tip than in Lichanura trivirgata ( Fig. 11F , dorsal view) or Eryx spp.

A distinctive feature of the postcloacals is the height of the posterior margin of the neural arch. Even in the middle of the tail the neural arch is quite vaulted, and a flat, vertical posterior surface arises. The height of this surface diminishes distally in the tail. Such a surface is not observed in Lichanura trivirgata , but in some Eryx spp. a similar surface is present on anterior caudals.

The cotyle and condyle are small and round ( Fig. 11D , anterior and posterior views). Paracotylar foramina appear to be present on the distalmost segmented caudal vertebra ( Fig. 11E ), but they are apparently absent anteriorly in the tail and on the segmented posterior trunk vertebra; it is possible that their apparent presence distally in the tail is an artifact (but see Zerova 1989; Georgalis 2019).

The dimensions of distal postcloacal vertebrae – short and tall – are comparable to those in extant erycines ( Fig. 11F , lateral view). Moreover, at least some supernumerary vertebral processes are present on all preserved postcloacal vertebrae ( Fig. 1D, E ). We follow the terminology of Szyndlar (1994) in describing them. Pterapophyses are present as distinct bumps already on the first postcloacal preserved after the gap of c. 11 vertebrae; they grow rapidly in prominence dis-dorsal ventral left lateral anterior posterior

tally in the tail, forming long, anteroposterior ridges with strong, anteriorly projecting spines ( Fig. 11 C-E). Curiously, similar spines are also present in Pterygoboa Holman, 1976 from the North American late Oligocene and Miocene (e.g., Holman 2000; Mead & Schubert 2013). Postzygapophyseal wings and posterior extensions of the prezygapophyses are not present anteriorly but become distinct by about the 16th caudal vertebra (after the gap); they never become prominent. Pleurapophyses were presumably present on all postcloacal vertebrae, but they are everywhere preserved; posteriorly in the tail they have a knoblike appearance distally, but a distinct posterior extension of the pleurapophyses appears to be absent. In addition to these main, named supernumerary processes, the distal-most caudal vertebrae in Charina bottae and Eryx spp. exhibit fine-scale elaborations of small, generally anteroposteriorly arranged ridges ( Szyndlar 1994: fig. 2). Such elaborations are present on the distal-most caudals of HLMD-Me 9723 as well ( Fig. 1D ).

Hemapophyses are apparently present on all preserved postcloacal vertebrae ( Fig. 11 C-E, ventral view), but they are often broken. Thus, their length cannot be ascertained, nor is it known whether “tubercular prominences” of the hemapophyses ( Sood 1941) were present.

The ribs are for the most part broadly rounded ( Fig. 1A ). In some cases the ribs appear to evince distinct bends, but because these appear sporadically and are bounded front and back by ribs of normal curvature, we ascribe the bends to plastic deformation. The ribs suggested a broadly rounded body, or at least one that was not markedly compressed or depressed.

PHYLOGENETIC RELATIONS (SEE Appendix 2)

Phylogenetic analysis using MP produced 126 equally mostparsimonious trees, the strict consensus of which is shown in Fig. 12A . The relationships of basal alethinophidians are fully unresolved. Xenopeltis unicolor and Loxocemus bicolor are successive outgroups to a clade comprising the Cretaceous marine snakes and Henophidia sensu Gauthier et al. (2012), and a sister-group relationship between Lo. bicolor and fossil Ogmophis compactus (Smith 2013) is confirmed (although bootstrap support is low). Ungaliophiinae forms a clade with Calamagras weigeli , also confirming the conclusion of Smith (2013); as above, bootstrap support is low, and whether Cal. weigeli is inside the crown is unresolved (Smith 2013; Fig. 12A ). Booidea sensu Pyron et al. (2014) is not monophyletic; instead, as is typical in phylogenetic analyses of morphology ( Lee & Scanlon 2002; Gauthier et al. 2012; Hsiang et al. 2015), the large boas, or Boidae sensu Pyron et al. (2014) , and pythons, or Pythonidae sensu Reynolds et al. (2014) , cluster together. Unlike in molecular analyses, erycines including Eryx and Charininae form a clade that also includes Rageryx schmidi n. gen., n. sp., which forms the sister-taxon to Charininae . Enforcing a molecular constraint on the tree topology did not influence the placement of any of these three fossil taxa.

Synapomorphies supporting the clade Rageryx schmidi n. gen., n. sp. and Charininae are as follows: ventral extent of postorbital reduced, quadrate ramus of pterygoid horizontally bladelike with longitudinal dorsal groove, exit foramen for hyomandibular branch of cranial nerve VII obscured in lateral view by projecting flange of prootic, and coronoid reduced. Autapomorphies of Rageryx schmidi n. gen., n. sp. according to this analysis are generally reversals, particularly of synapomorphies common to the broader group of erycines, viz. loss of protruding premaxilla and gain of zygosphene-zygantral articulations distally in the tail. The exception is the presence of a posteromedial flange for the ectopterygoid. Given the strong molecular evidence of erycine polyphyly, the putative reversals should be viewed with caution.

Similar relations for the booid fossil taxa are inferred using standard BI ( Fig. 12B ). The posterior probability (PP) for the clade Rageryx schmidi n. gen., n. sp. + Charininae is 0.98. Calamagras weigeli comes out in the crown of Ungaliophiinae , albeit with low support (PP = 0.59). Ogmophis compactus is the sister-taxon to Loxocemus bicolor , similarly with low support (PP = 0.56).

Insofar as it incorporates the age of fossil taxa, BI using the fossilized birth-death model gave results that differed from those of the MP analysis in expected ways. Without topological constraints, Ogmophis compactus (at c. 35 Ma) falls to the stem of a clade Loxocemus + Xenopeltis . Calamagras weigeli (at c. 48 Ma) falls to the stem of Ungaliophiinae ; and Rageryx schmidi n. gen., n. sp. (also at c. 48 Ma) falls to the stem of Charininae . Support for these relationships is low. When minimum molecular topological constraints are enforced, Rageryx schmidi n. gen., n. sp. is strongly supported (PP> 0.99) as a stem representative of Charininae (PP> 0.99), and Cal. weigeli strongly supported (PP> 0.99) as a stem representative of Ungaliophiinae (PP> 0.99). The position of Ogmophis compactus with respect to Xenopeltis , Loxocemus , and Pythonidae , however, is poorly resolved.

Thus a sister-group relationship between Rageryx schmidi n. gen., n. sp. and crown Charininae is well supported, making the former the oldest confirmed fossil erycine. Because no new data were adduced concerning Messelophis variatus and Rieppelophis ermannorum , it is not surprising that their relationships to other booids are unresolved, as in Scanferla et al. (2016).

Figures

FIG. 1. — HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp.: A, photograph of whole specimen;B, photograph of skull (coated with ammonium chloride) in dorsal view; C, 3D rendering of skull, based on CT scan, in ventral view; D, photograph of tail (coated with ammonium chloride) in roughly dorsal view; E, 3D rendering of tail, based on CT scan, in roughly ventral view. Scale bars: A, 2 cm; B-E, 1 cm.

FIG. 2. — Maxilla: A-C, left maxilla of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal, lateral and medial views,respectively; D-F, left maxilla of Eryx jaculus (Tü-VI.1935) in dorsal, lateral and medial views, respectively; G-I, left maxilla of Lichanura trivirgata (CM 145332) in dorsal, lateral and medial views, respectively. Scale bar: A-C, 1 mm; D-I, 2 mm.

FIG. 8. — Ectopterygoid: A-D, left ectopterygoid of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal, ventral, lateral, and medial views, respectively;E, F, right ectopterygoid (mirrored) of Eryx jayakari BM 1909.10.15.8 in dorsal and ventral views, respectively; G-I, left ectopterygoid of Lichanura trivirgata CM 145332 in dorsal, ventral, and medial views, respectively. Scale bar: A-D, 1 mm; E-I, 2 mm.

FIG. 3. — Nasal:A-E, left nasal of HLMD-Me 9723,holotype of Rageryx schmidi n. gen., n. sp.,in dorsal, ventral,anterior, posterior,and lateral views,respectively; F-J, left nasal of Eryx johnii BM 1930.5.8.31 in dorsal, ventral, anterior, posterior, and lateral views, respectively; K-O, united left and right nasals of Lichanura trivirgata CM 145332 in dorsal, ventral, anterior, posterior, and lateral views, respectively. Scale bar: A-E, 1 mm; F-O, 2 mm.

FIG. 4. — Frontal:A-D, left frontal of HLMD-Me 9723,holotype of Rageryx schmidi n. gen.,n. sp., in dorsal,ventral,lateral,and anterior views,respectively;E-H, left frontal of Eryx johnii BM 1930.5.8.31 in dorsal, ventral, lateral, and anterior views, respectively;I-L, left frontal of Lichanura trivirgata CM 145332 in dorsal, ventral, lateral, and anterior views, respectively. Scale bar: A-D, 1 mm; E-L, 2 mm.

FIG. 5. — Parabasisphenoid: A-C, parabasisphenoid of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal, ventral, and left lateral views, respectively.Portions of the parietal that articulated with the basisphenoid wings are probably artifactually associated here (blurred),but a more precise separation is not possible; D-F, parabasisphenoid of Eryx johnii BM 1930.5.8.31 in dorsal, ventral, and left lateral views, respectively; G-I, parabasisphenoid of Lichanura trivirgata CM 145332 in dorsal, ventral, and left lateral views, respectively. Scale bar: A-C, 1 mm; D-I, 2 mm.

FIG. 6. — Prootic: A-C, left prootic of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in lateral, medial, and ventral views, respectively. A small portion of the parietal is probably artifactually associated here (blurred), but a more precise separation is not possible; D-F, left prootic of Eryx johnii BM 1930.5.8.31 in lateral, medial, and ventral views, respectively; G-I, left prootic of Lichanura trivirgata CM 145332 in lateral, medial, and ventral views, respectively. Scale bar: A-C, 1 mm; D-I, 2 mm.

FIG. 7. — Pterygoid:A, B, left pterygoid of HLMD-Me 9723,holotype of Rageryx schmidi n. gen., n. sp., in dorsal and ventral views, respectively; C, D, right pterygoid of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal and ventral views, respectively; E, F, left pterygoid of Eryx johnii BM 1930.5.8.31 in dorsal and ventral views, respectively; G, H, left pterygoid of Lichanura trivirgata CM 145332 in dorsal and ventral views, respectively. Scale bar: A-D, 1 mm; E-H, 2 mm.

FIG. 9. — Dentary: A-D, left dentary of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal, ventral, lateral, and medial views, respectively; E-G, left dentary of Eryx johnii BM 1930.5.8.31 in dorsal, lateral, and medial views, respectively; H-J, left dentary of Lichanura trivirgata CM 145332 in dorsal, lateral, and medial views, respectively. Scale bar: A-D, 1 mm; E-J, 2 mm.

FIG. 10. — Compound bone: A-D, left compound bone of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp., in dorsal, ventral, lateral, and medial views, respectively; E-H, left compound bone of Eryx johnii BM 1930.5.8.31 in dorsal, ventral, lateral, and medial views, respectively; I-L, left compound bone of Lichanura trivirgata CM 145332 in dorsal, ventral, lateral, and medial views, respectively. Scale bar: A-D, 1 mm; E-L, 2 mm.

FIG. 11. — Vertebrae: A, posterior trunk vertebra (precloacal vertebra number 200) of Lichanura trivirgata SMF-PH 21; B, posterior trunk vertebra of HLMD-Me 9723,holotype of Rageryx schmidi n. gen., n. sp.; C-E, anterior,middle and distal caudal vertebrae of HLMD-Me 9723,holotype of Rageryx schmidi n. gen., n. sp.; F, distal caudal vertebra of Lichanura trivirgata CM 145332. Views: dorsal, ventral, left lateral, anterior, posterior. Scale bar: A, F, 2 mm; B-E, 1 mm.

FIG. 12. — Phylogenetic relationships of Rageryx schmidi n. gen., n. sp.: A, strict consensus of 176 equally most-parsimonious trees; bootstrap percentages>50% are shown above branches;B, majority-rule consensus of 15000 trees from standard Bayesian analysis.Posterior probabilities are shown above branches.

FIG. 13. — Inner ear of HLMD-Me 9723, holotype of Rageryx schmidi n. gen., n. sp. Semitransparent skull for orientation, showing location of left and right bony labyrinth, with inset showing enlarged right inner ear. Scale bar: 1 mm.