Mosaic anatomy in an early fossil squamate

Preservation and taphonomy of NMS G.2023.7.1

Disarticulated bones of NMS G.2023.7.1 are unfold over a diameter of roughly 19 cm on an undulating bedding floor. This preservation is just like that of specimens of P. estesi (NHMUK (Natural History Museum, London, UK) PV OR 41388) and cf. P. estesi (NHMUK PV R8851), from the Early Cretaceous Purbeck Limestone Group⁸. Various parts are seen on the floor of the block, together with a proper mandible, braincase, numerous different cranium bones, vertebrae, ribs, a partial proper coracoid, humeri, a proper ilium and proper femur (Fig. 1b). Numerous extra bones are current within the matrix (Fig. 1b). Of the cranium, NMS G.2023.7.1 contains the left jugal, left postfrontal, left and proper parietals, left and proper squamosal, braincase, left vomer, left palatine, proper pterygoid, proper dentary, proper angular, and proper compound bone incorporating the surangular, prearticular and articular. Of the axial skeleton NMS G.2023.7.1 contains 32 vertebrae or partial vertebrae, many dorsal ribs, and a cervical intercentrum (Extended Data Fig. 5 and Supplementary Data 2). Of the appendicular skeleton, NMS G.2023.7.1 contains the best coracoid, proper and left humeri lacking distal ends, proper ilium, fragment of proper pubis, proper femur lacking distal finish, left femur lacking epiphyses, proper tibia lacking epiphyses, one metapodial, and three phalanges, together with one ungual phalanx. Details of all these parts, together with hyperlinks to 3D digital fashions, are included in Supplementary Data 1.

The morphologies and lack of duplication of all squamate parts preserved in NMS G.2023.7.1, are according to task to a single comparatively large-bodied particular person (see fundamental textual content). Although the bones are disarticulated, they’re unfold in a rostral to caudal sample with cranium parts and anterior vertebrae at one aspect of the block, and hindlimb and caudal vertebrae on the reverse edge, with dorsal vertebrae, ribs, and forelimb parts between (Fig. 1b and Extended Data Fig. 5). NMS G.2023.7.1 is remoted from different skeletons recovered from the identical bedding floor by a minimum of two metres distance. Those skeletons belong to mammals, amphibians and fish, however not squamates. Thus, it’s unlikely that any squamate parts reported right here symbolize distinct squamate people or species, transported from adjoining areas of the lagoon ground.

The presence of enormous numbers of simply transported Voorhies Group 1 parts (for instance, vertebrae and ribs⁵¹) signifies that the skeleton was subjected to currents that have been neither robust sufficient to take away these parts, nor robust sufficient to move allochthonous Group 2 and three parts (that’s, limb bones and cranium parts⁵¹) from different people into the bone scatter earlier than it was buried. Other non-squamate vertebrate parts similar to bone crumbs and a tritylodontid tooth are current within the area of the skeleton, however these Group 3 lag parts are almost definitely to have been current on the lagoon ground when the skeleton was deposited, and comparable materials is quite common all through vertebrate-bearing ranges of the sequence and round different skeletons.

These observations, the rarity of squamate stays in comparison with different teams within the Kilmaluag Formation assemblage¹⁴, and the rarity of enormous squamate stays particularly (of virtually 200 specimens collected from the Elgol Coast SSSI from 2014–2024, NMS G.2023.7.1 is the one specimen to protect any large-bodied squamate stays) present robust help for the view that NMS G.2023.7.1 represents a single squamate particular person.

Reconstruction of cranium and physique proportions

The cranium reconstruction proven in Fig. 3 was performed in Blender 3.5.0 by arranging parts of NMS G.2023.7.1 in 3D house, just about extant squamate anatomy. The define of the maxilla was based mostly on the maxilla of NHMUK PV OR 48388, the holotype of P. estesi. The cranium size is estimated at 41.4 mm and primarily makes use of data from the preserved parts of the dentary and compound bone. The reconstructed cranium has lengthy, low proportions. Evidence for this comes from the braincase dimensions, relative to the lengths of the mixed palatal parts and mandibular parts. We allowed extra vertical peak behind the cranium to accommodate slight crushing of the braincase.

The life reconstruction proven in Fig. 1a was produced by Mick Ellison on the American Museum of Natural History, in session with R.B.J.B. and S.E.E., utilizing measurements derived from the specimen. Measurements have been constructed from 3D digital fashions, utilizing Meshlab 2023.12⁵², and are reported in Supplementary Data 2. The size of the presacral vertebral column was estimated based mostly on the summed lengths of the 23 particular presacral vertebrae with measurable lengths (84.5 mm, excluding the condyles), giving an estimated presacral size of 99.2–110.2 mm if 27–30 presacral vertebrae have been initially current (permitting for the lacking atlas and the potential for lacking cervicals or dorsals). Of this, the summed cervical lengths give an estimated neck size of 25.9 mm or extra. The straight-line size of the longest full dorsal rib (19 mm) and ilium size (15 mm) knowledgeable reconstruction of physique depth, and the partial humerus size (10.1 mm), femur size (19 mm) and tibia size (11.5) knowledgeable reconstruction of limb lengths. Other features, similar to hand and foot morphology, are usually not knowledgeable by proof and ought to be thought of as generalized.

Discovery, preparation and imaging of NMS G.2023.7.1

NMS G.2023.7.1 was found by S.A.W. in March 2015, on the Elgol Coast SSSI, throughout fieldwork led by R.B.J.B. and S.A.W., assisted by A. Wolniewicz, with permission of the landowner, the John Muir Trust, beneath allow from NatureScot (then, Scottish Natural Heritage). It was extracted as a block of micritic limestone roughly 220 mm lengthy, 180 mm vast and 150 mm deep. This block was embedded in silicone and ready from behind utilizing acetic acid by S. Moore-Fay of Wavecut Platforms, to its present thickness of 15–30 mm.

We scanned the complete slab, utilizing the Nikon Metrology XT H 225 ST X-ray μCT scanner on the School of Earth Sciences X-ray Tomography Facility, University of Bristol, UK, offering a pilot scan of the entire specimen with all bones of their unique positions. Segmentation of this and different scans within the present work was performed utilizing the software program Mimics 19.0 (Materialise) primarily by E.F.G. The scan and parameters for this and all different μCT scans described within the present work can be found on MorphoSource (hyperlinks supplied in Supplementary Data 1).

The pilot CT scan is the premise of the digital map of the specimen proven in Fig. 1b and Extended Data Fig. 5 and was additionally used to determine areas of matrix freed from preserved bones that have been then eliminated with a table-mounted disc cutter. The ensuing discount of slab diameter allowed a greater signal-to-noise ratio in subsequent episodes of CT scanning. We additionally used the pilot scan to separate some parts of the specimen in smaller blocks for high-resolution μCT scanning on the University of Bristol facility. These blocks, and the remaining portion of the specimen, got subpart numbers that comply with from NMS G.2023.7.1: (1) NMS G.2023.7.1.1, the primary a part of the slab, excluding the next sections; (2) NMS G.2023.7.1.2, a small portion together with the braincase, left jugal, proper parietal, left postfrontal and a cervical vertebra; (3) NMS G.2023.7.1.3, small portion together with the left palatine and a dorsal rib; (4) NMS G.2023.7.1.4, small portion together with the left vomer; and (5) NMS G.2023.7.1.5, small portion together with a tritylodontid tooth and unidentified bone fragments (gray parts Fig. 1b, backside proper).

We additionally focused areas of the remaining slab (NMS G.2023.7.1.1) for phase-contrast synchrotron X-ray tomography on beamline ID19 of the European Synchrotron Radiation Facility (ESRF), Grenoble, France (described beneath): (1) the best dentary, angular and compound bone, proper pterygoid, proper squamosal, left parietal and left humerus; (2) the best ilium, partial pubis and femur, phalanges together with an ungual phalanx, dorsal ribs, and eight vertebrae, together with dorsals, caudals and a cervical intercentrum; and (3) the best humerus and scapulocoracoid, plus a phalanx and a cervical and two dorsal vertebrae.

Finally, the primary slab (NMS G.2023.7.1.1) was break up into 4 parts that have been scanned individually at School of Earth Sciences, University of Bristol. This allowed greater high quality 3D fashions of some parts for which fashions from our different scans weren’t of adequate high quality, together with the tibia, atlas and another vertebrae.

We additionally accomplished CT scans of the holotype (NHMUK PV OR 48388) and referred (NHMUK PV R8851) specimens of P. estesi utilizing a Nikon XTEK H 225 ST MicroCT scanner at Cambridge Biotomography Centre, University of Cambridge, UK, additionally out there through MorphoSource (Supplementary Data 1).

Phase-contrast synchrotron X-ray tomography

For synchrotron X-ray tomography, the beamline was arrange for filtered white beam (W150-B wiggler hole 39 mm; filtered with 10 mm Cu and 0.5 mm W) leading to a complete built-in detected power of roughly 170 keV. Images have been recorded with an oblique detector comprising a 500 µm LuAG scintillator, a set of two Hasselblad lenses (100 and 150 mm; Victor Hasselblad) set for a 0.67× magnification, and a PCO.edge 4.2 sCMOS digital camera (PCO), leading to a measured pixel measurement of 8.96 µm. The sample-detector distance was set to 13.2 m for propagation part distinction. We used 5,000 projections over a 360° rotation, with an publicity time of 0.06 s per projection (taking the common of three frames of 0.02 s every), 41 flat-field photos and 40 dark-field photos have been used as calibration. The recorded area of view on this configuration was 8.89 mm vertically and 18.35 mm horizontally (992 × 2,048 pixels). The space scanned throughout every rotation was elevated by shifting the centre of rotation by round 8 mm horizontally (similar to 900 pixels on the detector), permitting us to reconstruct tomograms throughout a area if view spanning 34.59 × 34.59 mm (3,861 × 3,861 pixels). We then mixed scans of a number of fields of view to picture wider areas. Data have been processed utilizing PyHST2 with the one distance part retrieval strategy^53,54. Post-processing included a hoop artefact correction⁵⁵, change of dynamic vary from 32-bits to 16-bits utilizing the utmost and minimal 0.001% histogram clipping values of all of the datasets of the collection, and weighted averaging of duplicated tomograms ensuing from the overlap on the vertical axis.

Osteohistology

Three parts have been skinny sectioned from the related partial skeleton NMS G.2023.7.1. The midshafts of a partial humerus and partial femur (Fig. 2c,f), and unknown area of the physique of a rib have been manually ready from the specimen block with a carbide needle. The ensuing items of bone have been embedded in EpoThin 2 low viscosity epoxy resin. Embedded specimens have been minimize transversely at or nearest the mid-diaphysis utilizing an Isomet 1000 precision noticed. After being mounted onto frosted glass slides with clear Gorilla superglue gel, specimens have been floor to optical readability (~30 µm) on a lapidary wheel and photographed in plane-polarized mild on a Nikon Eclipse LV100POL microscope fitted with a Prior ProScan III automated stage adaptor. Composite photos have been processed utilizing Nikon NIS-Elements BR (model 5.24.03) imaging software program utilizing the Extended Depth Focus operate to autofocus and z-stack photomicrographs to enhance picture high quality of microstructural particulars. Two skinny sections have been fabricated from the humerus and femur, and one skinny part was constructed from the rib. High-resolution photomicrographs can be found from MorphoSource (Supplementary Data 1).

Phylogenetic analyses

We included NMS G.2023.7.1 and two different parviraptorid specimens in three phylogenetic datasets. Parviraptorids have been represented in these datasets by three operational taxonomic models (OTUs): (1) the holotype of Breugnathair elgolensis (NMS G.2023.7.1); (2) the holotype of P. estesi (NHMUK PV OR 48388); and (3) the referred specimen of cf. P. estesi (NHMUK PV R8551). Parviraptorid specimens from different localities weren’t included within the analyses as they’re much less full (or incompletely described) however are mentioned within the Supplementary Discussion. Whiteside et al.⁴⁸ just lately reported the Late Triassic Cryptovaranoides microlanius as an anguimorph, and subsequently deeply nested inside the squamate crown group. However, Brownstein et al.⁵⁰ introduced extremely differing anatomical observations and recommended that Cryptovaranoides might as a substitute be an archosauromorph. Given the anomaly of printed anatomical knowledge^48,49,50, we didn’t take a look at the affinities of Cryptovaranoides right here and await extra well-resolved anatomical knowledge. We additionally didn’t embrace the proposed stem lepidosaur Taytalura⁵⁶, additionally due to phylogenetic uncertainties³. Note that we intently thought of the phylogenetic scores proposed by Caldwell et al.⁷ for parviraptorids when scoring these matrices (defined in Supplementary Discussion).

Phylogenetic inference was carried out utilizing Bayesian inference in MrBayes 3.2.7a⁵⁷, utilizing a fossilized delivery–loss of life tree prior^58,59 with a proportion of extant species sampled of 0.038 beneath diversified sampling, and default priors for speciation fee, extinction fee and fossil sampling fee. We used an offset exponential tree age prior with a minimal age of 240 Ma (Middle Triassic) and imply age of 250 Ma (Early Triassic), and a relaxed clock transition fee mannequin with unbiased gamma charges with a default variance enhance prior of 10 and a log-normal clock fee with a imply of −2.56 log models and variance of 1.08. The ages of all operational taxonomic models have been specified utilizing a uniform distribution between their minimal and most potential stratigraphic ages, prolonged from ref. ³ to embody the broader set of fossil taxa included within the analyses performed right here. We constrained the ingroup relationships of squamates to mirror present consensus based mostly on molecular phylogenetic analyses², by specifying a collection of partial constraints matching these proven by ref. ³. The ensuing spine constraint is according to the phylogeny of ref. ², with trichotomies representing areas of uncertainty amongst latest analyses (that’s, implementing no particular relationships between the three constituent teams). For instance, Gekkota and Dibamidae in an unresolved trichotomy with the clade together with all different squamates. The toxicoferan clade (Iguania, Anguimorpha and Serpentes) was left as an unresolved trichotomy in analyses of datasets 1 and a pair of, however constrained to symbolize three totally different phylogenetic hypotheses of toxicoferan ingroup relationships for 3 separate analyses of dataset 3 (defined beneath).

Dataset 1

The early reptile dataset of Talanda et al.³ (modified from refs. ^32,45,60), which incorporates an intensive pattern of early members of the reptile crown group, in addition to dwelling and fossil rhynchocephalians and squamates, focussing on deep lepidosaur and deep squamate divergences, however contains comparatively few snakes: three extant species plus the early fossil snakes Najash, Pachyrachis and Dinilysia. To this we added our three parviraptorid OTUs, the hypothesized early anguimorph Dorsetisaurus, and the Late Jurassic squamate Eoscincus⁵, leading to a dataset of 125 taxa and 382 characters.

Dataset 2

An prolonged model of the squamate dataset of Meyer et al.⁶ (modified from Gauthier et al.²⁷). Meyer et al.⁶ prolonged the taxon and character pattern of Gauthier et al.²⁷ by including characters and taxa related to Jurassic squamate divergences, together with new fossil taxa. The ensuing matrix features a robust pattern of early squamates and related characters, in addition to a robust pattern of snakes. To this we added fifteen OTUs: the potential stem lepidosaurs Velbergia bartholomaei, Fraxinisaura rozynekae, Paliguana whitei, Sophineta cracoviensis and Marmoretta oxoniensis, the early rhynchocephalians Gephyrosaurus bridensis and Kallimodon pulchellus, the stem squamates Bellairsia gracilis, Oculudentavis naja, Oculudentavis khaungraae, the potential early anguimorph Dorsetisaurus purbeckensis, and our three parviraptorid OTUs. This resulted in a dataset of 637 characters and 168 suggestions, in comparison with the 155 suggestions included by Meyer et al.⁶.

Dataset 3

An prolonged model of the dataset of Zaher & Smith⁴⁷ (modified from Hsiang et al.⁶¹; after ref. ²⁷), to which we added numerous early squamate and non-squamate fossils. The matrix of Zaher & Smith⁴⁷ included a big pattern of dwelling and fossil snakes and related characters, in addition to 11 iguanians, 21 anguimorphs and two rhynchocephalians. This pattern was meant to concentrate on toxicoferans, with relevance to snake ingroup relationships.

We modified the matrix of Zaher & Smith⁴⁷ by including many species of non-toxicoferan squamates, adopting the scores of ref. ⁶² (modified from ref. ²⁷ based mostly on revisions proposed in more moderen works similar to refs. ^63,64) when, as normally, the equivalent characters have been utilized in these research. We additionally added fossil taxa meant to extend illustration of the stem teams of Lepidosauria, Squamata, and squamate subgroups: the potential stem lepidosaurs V. bartholomaei, Taytalura alcoberi, F. rozynekae, P. whitei, S. cracoviensis and M. oxoniensis, the stem squamates B. gracilis, Huehuecuetzpalli mixtecus, O. naja and O. khaungraae, the early-diverging Jurassic and Cretaceous squamates Hongshanxi xiei, Yabeinosaurus tenuis, Dalinghosaurus longidigitus, Liushusaurus acanthocaudata, Scandensia ciervensis, Meyasaurus faurai, Jucaraseps grandipes, Chometokadmon fitzingeri, Ardeosaurus brevipes, Eichstaettisaurus schroederi, D. purbeckensis, and our three parviraptorid OTUs. We omitted the candidate stem snake Tetrapodopis amplectus from analyses because of substantial doubts about its anatomy and phylogenetic affinities^65,66.

In addition to this, we added two characters. Character 789: gastralia current (0); absent (1). Character 790: entepicondylar foramen current (0); absent (1). This resulted in a dataset of 790 characters and 189 suggestions, in comparison with 788 characters and 90 suggestions by Zaher & Smith⁴⁷. Because analyses of this matrix returned parviraptorids as toxicoferans we ran three separate analyses of this dataset, utilizing topological constraints that symbolize totally different phylogenetic hypotheses for toxicoferan ingroup relationships: (1) snakes constrained as sister to anguimorphs; (2) snakes constrained as sister to iguanians; and (3) anguimorphs and iguanians constrained as sister taxa, to the exclusion of snakes.

Evaluating convergence

Analysis of dataset 1 was run for 133 million generations, dataset 2 was run for 160 million generations, and dataset 3 was run for as much as 180 million generations previous to convergence (at a mean commonplace deviation of break up frequencies of 0.01 or much less). Chains have been sampled each 10,000th era, with a burn-in of fifty%. The efficient pattern measurement was better than 200 for all examined parameters, and a mean potential scale discount issue was 1.01 or much less on all parameters, indicating convergence.

Reporting abstract

Further data on analysis design is on the market within the Nature Portfolio Reporting Summary linked to this text.