Pythons in the Eocene of Europe reveal a much older divergence of the group in sympatry with boas

Abstract

Extant large constrictors, pythons and boas, have a wholly allopatric distribution that has been interpreted largely in terms of vicariance in Gondwana. Here, we describe a stem pythonid based on complete skeletons from the early-middle Eocene of Messel, Germany. The new species is close in age to the divergence of Pythonidae from North American Loxocemus and corroborates a Laurasian origin and dispersal of pythons. Remarkably, it existed in sympatry with the stem boid Eoconstrictor. These occurrences demonstrate that neither dispersal limitation nor strong competitive interactions were decisive in structuring biogeographic patterns early in the history of large, hyper-macrostomatan constrictors and exemplify the synergy between phylogenomic and palaeontological approaches in reconstructing past distributions.

1. Background

Evolutionary convergence on the same ecomorph in different geographic areas is widely recognized in squamate reptiles, that is, lizards and snakes. To take one example, Wiens et al. [1] explored the roles of biogeography and competition in the convergent evolution of burrowing versus cryptic, surface-dwelling limbless forms. Larger constrictors [2], namely New World and Malagasy boas (Boidae and Sanziniidae sensu [3]) and Old World pythons (Pythonidae sensu [4]), furnish a further, unexpected example. These taxa frequently group together in phylogenetic analyses of morphology [5–9], although genetic data demonstrate that they are more distantly related [4,10,11]. The morphological topology arises from numerous shared characters related to what we would call hyper-macrostomy: a suite of features including large gape, thermo-sensory labial pits, strongly projected supratemporals and mandibles, posteroventrally oriented quadrates, mainly parallel-fibred adductor muscles, crista circumfenestralis enclosing the stapes, close apposition of the prefrontal laminae anterior to the frontals and well-developed sagittal crests for muscle attachment on the parietal and basisphenoid (e.g. [5,12,13]). Since boids and pythonids nowhere co-occur on Earth today, modern distributions suggest that they independently evolved in the New World and Old World, respectively [4].

This hypothesis has been corroborated by further evidence. Molecular data suggest that the Oceanian Candoia and the Malagasy Acrantophis and Sanzinia, commonly thought to represent Old World relictual boid lineages [14], are either only distantly related to Neotropical boids or only questionably associated with them [3,15]. Moreover, palaeontological evidence supports the primary allopatry of boas and pythons. The oldest definitive records of the total clade of Boidae are from the late Palaeocene of Cerrejón, Colombia, and early Eocene of Itaboraí, Brazil [16,17], and the oldest definitive records of the total clade of Pythonidae are from the latest Oligocene to early Miocene of Riversleigh, Australia [18]. Still, molecular time estimates suggest that Booidea and Pythonoidea diverged in the mid-Cretaceous [11,15,19], implying a ghost lineage of at least 40 Myr for both groups. The isolated vertebrae from the Campanian–Maastrichtian of South America referred to ‘Boidae’ [20] are too poorly preserved to be confidently assigned to this family.

Two tenable hypotheses exist to explain the biogeographic history of the total clade of Pythonidae, the Gondwanan and Laurasian dispersal models [18]. In the Gondwanan model, Pythonoidea originated in Laurasia, and the pythonid stem dispersed via South America and Antarctica to Oceania, Southeast Asia and Africa [15]. This hypothesis recognizes North American Loxocemus and Southeast Asian Xenopeltis as successive living outgroups of Pythonidae, following molecular evidence. It predicts that stem pythonid fossils should be found in South America, Antarctica and Australia [18]. In the Laurasian model, the total clade of Pythonidae originated in Europe or Asia and dispersed to Africa and Australia. This hypothesis is also consistent with the recognition of Loxocemus and Xenopeltis as pythonid outgroups and is informed by tantalizing hints of python-like fossils in the Eocene of Europe [21–24].

Here, we revisit undescribed python-like fossils from the Eocene of Messel, Germany [25] and demonstrate that they are the oldest known stem pythonids. We further examine their implications for pythonoid evolution and biogeography.

2. Methods

Measurements of body proportions were made using ImageJ v. 1.52 k [26].

We compiled an expanded total evidence dataset that combines morphological [27] and molecular [11] information for 90 extinct and extant species of anguimorphs, iguanians and snakes, representing the crown clade Toxicofera [10,11,28], scored for 524 471 characters (electronic supplementary material, S1–S4). The dataset was subjected to total evidence, equally weighted parsimony analysis in TNT [29].

We applied the package BioGeoBEARS [30] for R [31] to data on the geographic ranges of our tree tips, using modified zoogeographic regions (after [32]; electronic supplementary material, figure S2) to better reflect global palaeogeographic history since the Jurassic (electronic supplementary material, S5).

3. Systematic palaeontology

Serpentes Linnaeus, 1758 [33]

Pythonoidea Fitzinger, 1826 [34]

Messelopython freyi gen. et sp. nov.

urn:lsid:zoobank.org:act:E8C2F2DA-1434-4580-8D30-A5EDA71A2CAC

urn:lsid:zoobank.org:act:84AF3F8E-1B68-4807-A5A3-31D95A9ADE84

Holotype: SMNK-PAL 461, nearly complete skeleton with partial skull (figure 1a–c)

Figure 1.

Hypodigm of Messelopython freyi gen. et sp. nov. (a) and (b) Photograph and drawing of skull of holotype, SMNK-PAL 461a. (c) Photograph of holotype. (d) and (e) Photograph and drawing of skull of paratype, SMF-ME 2784a. (f) Drawing of left splenial and angular in medial view in counterpart of paratype. (g) Drawing of right ectopterygoid in dorsal view in main part of paratype. (h) and (i) Photograph and drawing of skull of paratype, HLMD-Be 165. Abbreviations: an, angular; boc, basioccipital; bpt-pr, basipterygoid process; bs, basisphenoid; d, dentary; ec, ectopterygoid; fr, frontal; mx, maxilla; n, nasal; ot, otooccipital; pa, parietal; par, prearticular part of compound bone; pl, palatine; pf, postfrontal; prf, prefrontal; pro, prootic; pt, pterygoid; q, quadrate; sa, surangular part of compound bone; smx, septomaxilla; so, supraorbital; soc, supraoccipital; spl, splenial; st, supratemporal.

Paratypes: SMF-ME 710, nearly complete skeleton with a badly crushed skull (figure 1f); SMF-ME 2784, a nearly complete skeleton with a skull (figure 1d,e,g); HLMD-Be 165, a nearly complete skeleton with a partial skull (figure 1h,i).

Referred specimen: HLMD-Me 10583

Diagnosis: Medium-sized pythonoid (electronic supplementary material, S6) sharing with Pythonidae the following synapomorphies: a toothed premaxilla with no midline diastema; palatine foramen present in palatine; and mid-sagittal crests on parietal and basisphenoid. Autapomorphies: 6 premaxillary teeth; sigmoid lateral margin of maxilla; large, crescentic supraorbital bone nearly equal in length to the frontal; ectopterygoid overlap of pterygoid reduced.

Type locality and horizon: Messel Pit, Germany; Middle Messel Formation, latest early to earliest middle Eocene [35].

Etymology: ‘Messel’, after the type locality [25], and ‘python’, after its close living relatives. The specific epithet honours Eberhard ‘Dino’ Frey, for his detailed studies of fossil and extant reptiles.

4. Results

The premaxilla of M. freyi (figure 1) possesses a short nasal and a pair of long vomerine processes. It is toothed, with six tooth positions, and lacks a median tooth or midline diastema. The long, paired nasals taper anteriorly. The prefrontals lack a lateral foot-process but possess an ascending lamina; these laminae do not meet on the midline. The paired frontals are long and have a concave lateral margin, corresponding to the supraorbital bone, for which an articulation facet is developed (see explanation on homologies of circumorbital bones in electronic supplementary material, characters 73, 74, 77 and 170). The orbital margins of the frontals are upturned. They articulate with the parietal on an anteriorly concave suture. The postfrontal has a bifurcated dorsal articulation on the parietal corner, and it does not reach the frontal. It possesses a long articulation with the supraorbital, which extends some distance ventrally. The large size and crescentric shape of the supraorbital distinguishes M. freyi from all extant Pythonoidea (electronic supplementary material, figure S1).

The elongate parietal possesses a low, mid-sagittal crest over nearly three-quarters of its length. It is somewhat inflated anteriorly. The posterior margin of the parietal contacts the small, oval supraoccipital in an anteriorly directed V-shaped suture, as in most pythonids. The ophidiosphenoid is broad and the foramen for the maxillary branch of the trigeminal nerve is open anteriorly in the prootic. A low, mid-sagittal crest is present on the parabasisphenoid.

The supratemporal extends beyond the posterior end of the braincase. The quadrate has a triangular, plate-like dorsal portion, whereby the cephalic condyle is at an oblique angle to the mandibular condyle.

The maxilla has a weakly sigmoid lateral margin and a single labial foramen. As in most pythons [18], it retains an anterior foramen located just on the medial side of the bone, and two dorsal foramina on the palatine process for the superior alveolar nerve. The palatine process is broad and rounded, rather than posteriorly directed and club-like as in Booidea. The ectopterygoid expands in width anteriorly, where its distal tip is bifurcated. Posteriorly, it possesses a short articulation on the pterygoid, but no pronounced posterior process as in Pythonidae. The toothed palatine is pierced by the palatine nerve. Details of the articulation with the pterygoid are unknown. Both palatine and pterygoid were toothed, although an exact tooth count cannot be given. The pterygoid terminates posteriorly just short of the mandible.

The dentary has one mental foramen on the right side of SMF-ME 2784 and two on the left. The surangular notch is not particularly deep. The splenial has a posteroventral knob, which would have inserted into a concavity on the angular. The compound bone possesses a weak but distinct ventrolateral crest. The surangular crest rises relatively straight and gradually, whereas the prearticular crest is strongly convex. The anterior surangular foramen is small.

The maxilla possesses approximately 18 tooth spaces. The teeth are long and needle-shaped, diminishing from front to back, but anterior teeth are not notably enlarged.

There appear to be somewhere between 205 and 210 trunk vertebrae in the holotype. HLMD-Be 165 has approximately 209 trunk vertebrae followed by at least two visible cloacals. SMF-ME 2784b preserves the tail, with at least two cloacal vertebrae with bifurcated lymphapophyses followed by approximately 67 vertebrae and tail tip. This yields an approximate total vertebral count of 275. The trunk vertebrae are short and high. Hypapophyses are present only on anterior trunk vertebrae. Mid-trunk vertebrae have an anteroposteriorly long neural spine whose base extends onto the zygosphenal tectum (electronic supplementary material, figure S2). Short accessory processes are present. Broadly rounded zygantral mounds (see [36,37]) are present even in the posterior trunk. Haemapophyses are present in the tail.

Our phylogenetic analysis strongly supports the position of Messelopython freyi on the stem of Pythonidae (figure 2a; electronic supplementary material).

Figure 2.

(a) Phylogenetic relationships of Messelopython freyi gen. et sp. nov. Support values for nodes are Jackknife/Bremer, values of 100% (Jackknife) or greater than 20 (Bremer) are indicated with an asterisk. (b) Palaeogeographic map of continental distributions at 48 Mya based on Gplates model [38] with location (palaeocoordinates for 48 Mya) of Palaeogene records of the total clades of Boidae, Loxocemidae and Pythonidae. Large arrows indicate hypothesized dispersal directions for the total clade of Pythonidae.

Excluding Messelopython freyi, BioGeoBEARS results support a Gondwanan origin of Alethinophidia and dispersal of stem pythonids over Antarctica to Australia and Asia. Including M. freyi, both DEC and DEC + J support a Laurasian, and more specifically European, origin of Pythonoidea and Pythonidae (electronic supplementary material, figures S4–S9), with the dispersal of the total clade of Pythonidae to Africa and Oceania (figure 2b); analyses assuming an African range for our Python terminal do not alter this conclusion.

5. Discussion

(a). Constraints

The divergence of Pythonidae from its sister taxon, Loxocemus, was previously constrained by a loxocemid from North America, at least 35.2 Ma [39,40]. Messelopython freyi provides a much older hard minimum, used tentatively in two recent works to calibrate the divergence between the total group of pythonids and Loxocemus [11,41].

As a former maar lake, Messel has a well-dated volcanic origin at 48.2 Ma [35]. Essentially all vertebrate body fossils derive from the Middle Messel Formation, which shows varve-like lamination [42], allowing good estimates of duration. Taking the youngest age model (La2010a) of Lenz et al. [35] and the oil-shale accumulation rate, we estimate a hard minimum age of this fossil as 47.57 Ma (electronic supplementary material).

(b). Biogeography

Our study confirms Szyndlar and Böhme's [22] supposition that the total clade of Pythonidae was present in Europe in the Palaeogene. Remarkably, the age of Messelopython freyi—at least 47.57 Ma—places it close in time to the inferred divergence of Loxocemus and Pythonidae in the late Palaeocene or early Eocene [11,40].

Messelopython freyi has a marked influence on inferred historical biogeography. Without it, early snake evolution (including Pythonoidea and Pythonidae) is anchored firmly in Gondwana, with connections across Antarctica to Oceania. With M. freyi, early pythonoid history is anchored in Laurasia, thus rejecting the Gondwanan model of pythonoid biogeography [18] (electronic supplementary material, figures S4–S9). All three major pythonoid lineages are rooted in Laurasia: extant Xenopeltis, stem and recent Loxocemus and now stem Pythonidae (symbols, figure 2b). Future fossil discoveries in the Arctic and North American provinces may help test whether the specific inference of a European, rather than other Laurasian, origin of Pythonoidea might be unduly influenced by the occurrence in Europe of this single, early representative of the clade.

(c). Sympatry and evolution

Complete allopatry is one of the most striking aspects of the distribution of larger boas and pythons today. Larger boas inhabit the Neotropics (Boidae), Madagascar (Sanziniidae) and northern Oceania (Candoia), whereas pythons inhabit Africa, Southeast Asia and Australia. Given a molecular topology (e.g. [15]), their structural similarities, including the presence of pit organs for the perception of radiant heat [43,44], posteriorly expanded supratemporals and mandibles, and extended parallel-fibred adductor muscles [12] must be ascribed to convergence on a hyper-macrostomatan ecomorph. Allopatry today, in turn, suggests that either (i) dispersal limitation maintains lack of contact, or (ii) competition prevents one group from acquiring a foothold in a region already inhabited by the other group.

Remarkably, our results suggest that boids and pythonids coexisted in Europe during the early-middle Eocene (figure 2b) despite their convergent hyper-macrostomatan ecomorphology, implying that their modern allopatry cannot be attributed solely to Gondwanan vicariant events [4,14]. Pythonids more likely originated in Laurasian than Gondwanan landmasses [4,5,41] and subsequently dispersed to Africa, Southeast Asia and Oceania [41]. Furthermore, the total lineages of both Pythonidae and Boidae were present in Europe during the Eocene [44]. This discovery is wholly unexpected given the extant distributions of Booidea and Pythonidae and shows that neither dispersal limitation nor strong competitive interactions are good explanations for hyper-macrostomatan biogeographic patterns in the early Cenozoic.

Data accessibility

The electronic supplementary material includes additional information on the specimens studied, explanation of the age estimation for the Messel Formation, institutional abbreviations, specimens examined, list of characters and states with changes performed. Data matrices, R scripts and parameter files for the BioGeoBEARS analyses are available in the Dryad Digital Repository at https://doi.org/10.5061/dryad.8pk0p2nkx [45].

Authors' contributions

H.Z. and K.T.S. conceived the study, conducted the analyses and wrote the manuscript. H.Z. scored the matrix. Authors approved the final version and agree to be held accountable for the content.

Competing interests

We have no competing interests.

Funding

This work was supported by the Senckenberg Gesellschaft für Naturforschung (K.T.S.) and Fundação de Amparo à Pesquisa do Estado de São Paulo grant no. 2018/11902-9 (H.Z.).