The earliest-diverging avemetatarsalian: a new osteoderm-bearing taxon from the Triassic (?Earliest Late Triassic) of Madagascar and the composition of avemetatarsalian assemblages prior to the radiation of dinosaurs

Abstract

Understanding the evolution of the earliest avemetatarsalian (bird-line) archosaurs and inferring the morphology of the last common ancestor of Archosauria are hampered by a poor fossil record in critical temporal intervals. Here we describe an early-diverging avemetatarsalian from the ?Earliest Late Triassic (~235 Ma) ‘basal Isalo II’/Makay Formation of Madagascar, which helps bridge these gaps. This taxon, Mambachiton fiandohana gen. et sp. nov., is represented by well-preserved postcranial material and possibly a postfrontal bone. Features of the neck region include anteroposteriorly elongated vertebrae with laterally expanded dorsal ends of the neural spines with three pairs of osteoderms per cervical vertebra, lying dorsal to those vertebrae. Inclusion of Mambachiton in a phylogenetic analysis of archosauromorphs recovers it at the base of Avemetatarsalia, outside of the aphanosaur + ornithodiran clade. This new specimen indicates that osteoderms were present in the earliest avemetatarsalians, but were lost in more crownward lineages. The plesiomorphic morphology of the taxon also underscores the difficulty of identifying early avemetatarsalians from incomplete skeletons. This early-diverging avemetatarsalian occurring together with a lagerpetid and silesaurid in the ‘basal Isalo II’/Makay Formation of Madagascar documents the co-occurrence of multiple non-dinosaurian avemetatarsalian clades in Gondwana near the Middle–Late Triassic transition. Translated abstract (Malagasy and French) is provided in the Supplementary information.

Pan-Aves, Archosauria, Mesozoic Era, Africa, Dinosauria, phylogeny

INTRODUCTION

Western Madagascar preserves an extensive sequence of vertebrate fossil-bearing Late Paleozoic–Mesozoic strata, including the Permo-Triassic Sakamena and Triassic–Jurassic Isalo ‘groups’ (Wescott and Diggens 1998, Rakotosolofo et al. 1999, Flynn and Wyss 2003, Flynn et al. 2022). Among these, the richest records of continental tetrapods have been recovered from the ‘basal Isalo II’ beds in the Morondava Basin of south-western Madagascar, providing an important window into central Gondwanan faunas in the Mid-to-Late Triassic (Flynn et al. 1999, 2010). The most abundant taxa in the ‘basal Isalo II’/Makay Formation fauna are the traversodontid cynodonts Dadadon isaloi and Menadon besairiei, the rhynchosaur Isalorhynchus genovefae, and the allokotosaurian archosauromorph Azendohsaurus madagaskarensis (Flynn et al. 1999, 2000, 2010, Goswami et al. 2005, Kammerer et al. 2008, 2012, Ranivoharimanana et al. 2011, Nesbitt et al. 2015). Rarer synapsid faunal components include a kannemeyeriiform dicynodont and the predatory cynodont Chiniquodon kalanoro (Kammerer et al. 2010).

Based on the cynodont fauna, particularly the shared presence of Menadon besairiei, the ‘basal Isalo II’ beds/Makay Formation have been correlated with the Carnian Santacruzodon Assemblage Zone (AZ) of the Brazilian Santa Maria Formation (Melo et al. 2010, Schultz and Langer 2010, Schultz et al. 2020). These faunas in turn are also similar to one from the Chañares Formation of Argentina, sharing the probainognathian Chiniquodon and massetognathine traversodontids (Romer 1972a, Abdala and Ribeiro 2010). Curiously, less overlap is observed with the more geographically proximate faunas of eastern Africa (Sidor and Nesbitt 2018), which may be attributable to temporal, environmental, and/or ecological distinctions. Despite the general similarity between the ‘basal Isalo II’/Makay Formation fauna and its South American counterparts, however, the former has produced few representatives of several major clades that are well-known from Mid–Late Triassic Brazilian and Argentine deposits, notably Archosauria. Kammerer et al. (2020) recently described the first crown archosaur from these ‘basal Isalo II’/Makay Formation deposits, the lagerpetid avemetatarsalian Kongonaphon kely.

Nesbitt et al. (2017, 2018) recently recognized a new group of archosaurs, Aphanosauria, including the earliest-diverging members of Avemetatarsalia (the only members of this clade falling outside of the dinosaur–pterosaur subclade Ornithodira). Aphanosaurs differ greatly from all previously known Triassic avemetatarsalians in their retention of a suite of plesiomorphic features, including a semi-sprawling, quadrupedal posture and ‘crocodile-normal’ ankle. Interestingly, all known aphanosaurs (the Tanzanian Teleocrater rhadinus, Brazilian Spondylosoma absconditum, Indian Yarasuchus deccanensis, and Russian Dongusuchus efremovi) were previously referred to disparate other archosaur and archosauriform groups, and their similarity to those other clades may indicate mosaic evolution at the base of Archosauria (which also may be contributing to poor phylogenetic resolution among the crocodile-line subclades, and the uncertain position of Phytosauria; see: Nesbitt 2011, Ezcurra 2016).

Here, we present a distinctive new ‘basal Isalo II’/Makay Formation archosaur, which also exhibits a mixture of features present in disparate early bird- and crocodile-line archosaurs. This taxon differs from known aphanosaurs in having an extensive array of osteoderms dorsal to the cervical vertebrae, a plesiomorphic feature for archosauriforms, which was subsequently lost in previously known avemetatarsalians (apart from various dinosauriform groups, where it reappeared). Discovery of this taxon provides a new point of comparison with the fauna from the Santa Maria Formation; supporting the idea of broadly similar tetrapod faunas across southern Gondwana during the Middle-to-Late Triassic, it also highlights the persistence of at least one taxon retaining a body plan similar to that inferred for the ancestral archosaur long after the clades’s first appearance. This new archosaur derives from the same locality as the lagerpetid Kongonaphon kely (Kammerer et al. 2020), from strata correlative to those from a nearby locality that hosts the remains of a non-dinosaurian dinosauriform (silesaurid) (Nesbitt et al. 2015), showing that successive outgroups to dinosaurs formed an avemetatarsalian fauna prior to the arrival of dinosaurs.

Institutional abbreviations

AMNH, American Museum of Natural History, New York, NY, USA; FMNH, Field Museum of Natural History, Chicago, Illinois, USA; MSM, Arizona Museum of Natural History (formerly Mesa South-west Museum), Mesa, AZ, USA; NHMUK, Natural History Museum, London, United Kingdom; NMT, National Museum of Tanzania, Dar es Salaam, Tanzania; SAM, Iziko South African Museum, Cape Town, South Africa; SMNS, Staatliches Museum für Naturkunde, Stuttgart, Germany; UA, Université d’Antananarivo, Antananarivo, Madagascar.

MATERIAL AND METHODS

Association

The holotype and referred specimens of the new taxon were collected from one locality (=M-13) at the same stratigraphic level over three field seasons. A rhynchosaur premaxilla (FMNH PR 5070) and the distal end of a synapsid humerus (FMNH PR 5071) and synapsid tibia (FMNH PR 5072) were found with the holotype, as were unidentified remains that are not assigned to the holotype because they cannot be confidently assigned to this taxon. The other reptile remains described below were found at the same locality, but there is no clear direct association with the holotype.

Preparation

The specimens were mechanically prepared by staff at the Field Museum of Natural History. Cyanoacrylate (‘Super Glue’) used in specimen consolidation during fieldwork was removed during preparation. Final surface preparation was completed by Vicki Yarborough at Virginia Tech using steel needles to remove matrix and adhesives. Some sand matrix was added back in (e.g. to the right scapulocoracoid) for stability purposes, using Paraloid B-72, so that this is reversible, if necessary.

Phylogenetic analysis

We incorporated the new taxon Mambachiton fiandohana gen. et sp. nov. in the phylogenetic matrix of the Complete Archosauromorph Tree Project (CoArTreeP; Ezcurra 2016), as modified by Ezcurra et al. (2020a), Ezcurra et al. (2020), Ezcurra et al. (2020b), and then by Foffa et al. (2022); these datasets have the largest sampling of taxa and characters at the base of, and immediately outside, Archosauria. A few character scores were changed for Teleocrater rhadinus (character 497 0→1; character 664 0→1) and a score for sacral 1 of Teleocrater rhadinus was added (369-0) based on newly discovered material. As with the analysis of Foffa et al. (2022), character 119 was excluded, and the following operational taxonomic units were excluded based on previous justifications (see: Ezcurra et al. 2018, 2020b, 2020a, Foffa et al. 2022): Dinocephalosaurus orientalis, Fuyuansaurus acutirostris, Pectodens zhenyuensis, Protanystropheus antiquus, Trachelosaurus fischeri, Tanystropheus haasi, Malerisaurus robinsonae, Arctosaurus osborni, ‘Chañares rhynchosaur’, Eorasaurus olsoni, Prolacertoides jimusarensis, ‘Archosaurus complete’, ‘Panchet proterosuchid’, Vonhuenia fredericki, ‘C. rossicus combined’, ‘C. magnus combined’, Chasmatosuchus vjushkovi, Koilamasuchus gonzalezdiazi, ‘Kalisuchus rewanensis holotype’, ‘NMQR 3570’, Shansisuchus kuyeheensis, ‘Uralosaurus combined’, Osmolskina czatkoviensis, ‘Osmolskina complete’, Triopticus primus, Angistorhinus talainti, ‘Otter Sandstone archosaur’, Dagasuchus santacruzensis, Hypselorhachis mirabilis, ‘Waldhaus poposauroid’, Vytshegdosuchus zbeshartensis, Bystrowisuchus flerovi, Bromsgroveia walkeri, ‘Moenkopi poposauroid’, Lutungutali sitwensis, Nyasasaurus parringtoni, and Scleromochlus taylori (i, ii, iii). In total, 823 characters and 159 taxa were analysed in this study. The following characters were treated as ordered based on previous justification (see: Ezcurra 2016): 1, 2, 7, 10, 17, 19–21, 29, 36, 40, 42, 46, 50, 54, 66, 71, 74–76, 122, 127, 146, 153, 156, 157, 171, 176, 177, 187, 202, 221, 227, 263, 266, 278, 279, 283, 324, 327, 331, 337, 345, 351, 352, 354, 361, 365, 370, 377, 379, 386, 387, 398, 410, 414, 424, 430, 435, 446, 448, 454, 455, 458, 460, 463, 470, 472, 478, 482, 483, 485, 489, 490, 502, 504, 510, 516, 521, 529, 537, 546, 552, 556, 557, 567, 569, 571, 574, 581, 582, 588, 636, 648, 652, 662, 701, 731, 735, 737, 738, 743, 749, 766, 784, and 816.

The dataset was run according to the criteria presented in Ezcurra et al. (2020) and Foffa et al. (2022), which are briefly reiterated here. The data matrix was analysed under equally weighted maximum parsimony in TNT 1.5 using the following search strategies: start with a combination of the tree-search algorithms Wagner trees, tree bisection and reconnection (TBR) branch swapping, sectorial searches, phylogenetic ratchet, and tree fusing until 100 hits of the same minimum tree length were achieved. Then the trees with the fewest steps were subjected to a final round of TBR branch swapping. Zero-length branches in any of the recovered most-parsimonious trees were collapsed.

X-Ray computed tomography

The cervical series of the holotype (UA 8-25-97-132) was scanned at the Duke University Shared Materials Instrumentation Facility (SMIF) using the Nikon XT H 225 CT high-resolution micro-computed tomography scanner in 2017. Original scans were taken at 160 kV and 235 uA with a 0.125 mm-thick copper filter. The entire cervical series was scanned while still partially covered in matrix, allowing the fossil to remain articulated. Slices were segmented by hand using Materialise Mimics (v.20) on a Cintiq 24HD tablet. Osteoderms used in the principal component analysis were subsequently isolated in separate masks and exported in PLY files. The model was smoothed by a factor of 2 within Mimics for figures. Specimen images and renderings are on file at the AMNH, FMNH, and UA.

Morphometrics

PLY files of well-preserved cervical osteoderms were imported from Mimics to quantify their shape variation. Osteoderms were selected on the basis of completeness and minimal inferred distortion to provide the truest possible assessment of shape variation. The 13 selected osteoderms included eight from the left lateral column, and four from the right. Even and odd numbered osteoderms (counting from the anterior part of the series) are represented in the analysis, including paired and overlapping elements from throughout the length of the cervical column (but no complete posterior osteoderms from the right lateral column are included). The models, unsmoothed, were exported for shape analysis using the programs Landmark (2006) and Klingenberg (2011). Eleven non-sliding landmarks were placed in the positions described in Table 1. A principal components analysis was conducted on both left and right lateral osteoderms to quantify variation of the 3D shape. The landmarks captured variation along the rim of the osteoderms, as well as the height and width of their keel, and ventral ridge. Principal components 1 and 2 capture the greatest variation in shape. For left lateral osteoderms, principal component 1 accounted for 35% of the variation, and principal component 2 for 22%. For right lateral osteoderms, principal components 1 and 2 accounted for 55% and 28% of the variation, respectively.

Table 1.

Osteoderms landmarks used in the morphometric analysis

Landmarks	Positions
1	Anterior peg tip
2	Distal angle between peg and body
3	Anterior most lateral edge before curving
4	Distal most curve on the outer curve
5	Posterior most lateral edge before curving back
6	Poster endpoint just before extension of the keel of the osteoderm
7	Posterior most tip of keel
8	Proximal sides posterior most point before curving
9	Anterior most tip of keel
10	Posterior most interior end of the keel
11	Posterior most end of the ventral ridge of the osteoderm (and deepest bit)

Systematic paleontology

 

Archosauromorpha Huene 1946

 

Archosauria Cope 1869 sensu Gauthier and Padian 1985

 

Avemetatarsalia Benton 1999

Mambachiton fiandohana gen. et sp. nov.

Etymology:

Genus name is a combination of the Malagasy mamba, meaning crocodile, and Ancient Greek χιτών (khiton), which can refer to a suit of armour. A secondary reference to the molluscan chiton is also intended, based on the evocative superficial similarity in armour morphology between Mambachiton fiandohana and polyplacophorans. Species name is the Malagasy word meaning source or beginning, in reference to this taxon’s phylogenetic position near the crocodile–bird split, the node to which Archosauria is definitionally linked.

Holotype—

UA 8-25-97-132, an articulated series of anterior to posterior cervical vertebrae and osteoderms, anterior cervical vertebra neural arch fragment, cervical rib, and disarticulated osteoderms (Fig. 1A). A right partial postfrontal was directly associated with the holotype.

Figure 1.

Reconstruction of Mambachiton fiandohana, based on that of Teleocrater rhadinus (Nesbitt et al. 2017). Light blue bones represent elements from the holotype and light red bones are from the referred material which we consider to be from the same individual. Scale bar = 25 cm.

Referred specimen—

FMNH PR 5065, middle trunk vertebrae, posterior trunk vertebrae, sacral vertebra one, anterior caudal vertebra, right scapulocoracoid, left scapulocoracoid, right ilium, and right proximal portion of femur (Fig. 1B).

All of the referred elements were found near the holotype in the same horizon over a number of field seasons. We consider it highly probable that the holotype and the referred bones pertain to a single individual, based on consistency of size, consistency of character states for a taxon near the base of Archosauria, similar preservation, and the fact that none of the elements are duplicated with the holotype or each other. However, considering the lack of direct association with the cervical series and several years over which these elements were collected in the field, we refrain from including them in the holotype. Our hypothesis that all of these bones belong to Mambachiton fiandohana can only be tested with the discovery of another specimen showing association of these and diagnostic cervical elements; for now, justification for referring these non-holotype elements to Mambachiton fiandohana is given in the description.

Locality and age—

The holotype and referred elements of Mambachiton fiandohana were collected within the ‘basal Isalo II’ beds/Makay Formation, in a coarse-grained grey sandstone at Locality M-13, east of Sakaraha in the southern Morondava Basin of south-western Madagascar (precise locality information on file at the AMNH, FMNH, and UA). Other taxa recovered from the holotype’s locality include the traversodontids Menadon besairiei and Dadadon isaloi (Flynn et al. 1999, 2000), the rhynchosaur Isalorhynchus genovefae (Flynn et al. 1999), the lagerpetid Kongonaphon kely (Kammerer et al. 2020), and the other reptile remains described below. The age of the ‘basal Isalo II’ deposits is Ladinian–Carnian (Mid-to-Upper Triassic) based on the correlations and caveats provided in Flynn et al. (2000) and Kammerer et al. (2020).

Diagnosis

—The holotype of Mambachiton fiandohana differs from all other archosauriforms by the presence of the following combination of character states (asterisks indicate autapomorphies): *small tuber present at the dorsal margin of the prezygadiapophyseal lamina on the lateral side of the prezygapophysis in the cervical vertebrae; epipophyses absent on the dorsal surface of the postzygapophysis; laterally expanded dorsal portion of the neural spine; tapering anterior process of the cervical osteoderms articulating with a distinct groove on the ventral surface of the preceding osteoderm; smooth, unsculptured osteoderms; *high number of osteoderms per cervical vertebra (five to eight, depending on position); staggered arrangement of osteoderms across the midline.

Potential further diagnostic character states for this new taxon discernable from the referred material include: neural spines of the trunk vertebrae and first sacral vertebra laterally expanded dorsally; weakly developed hyposphene–hypantrum articulations between the trunk vertebrae; lateral articulation surface of sacral rib anteriorly and posteriorly constricted, resulting in an ‘I’-shape; coracoids with short postglenoid processes; distal end of the scapula expands anteroposteriorly more than the proximal end; rugose tuber occurs just distal to the glenoid of the scapula; ilium with notch on the articular surface for reception of the ischium.

Ontogenetic assessment—

No histological sectioning was performed, but a number of co-ossifications throughout the skeleton suggest that the holotype and referred materials of Mambachiton fiandohana were near skeletal maturity at time of death (see: Griffin et al. 2020). The neurocentral sutures are completely co-ossified throughout the preserved portions of the cervical series (the first preserved neural arch is broken) and the trunk, sacral, and anterior caudal vertebrae are completely or partially co-ossified at the neurocentral sutures; only one side of the neurocentral suture is fully co-ossified in the posterior trunk vertebra. Co-ossification of the scapulocoracoids is complex; anteriorly it is complete, but a clear suture occurs posteriorly on the left element. No co-ossification occurs in this region on the right element.

Nomenclatural acts—

This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank Life Science Identifiers (LSIDs) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix zoobank.org/. The LSIDs for this publication are as follows: urn:lsid:zoobank.org:pub:5CD3A1E5-2882-4004-8875-FCD534CF88AF.

Mambachiton: urn:lsid:zoobank.org:act:A15DFD4D-8C6D-4940-B708-464540433830.

Mambachiton fiandohana: urn:lsid:zoobank.org:act:630DA554-FF99-4046-BBAE-42391392FB00.

Description

Postfrontal

A largely complete left postfrontal (UA 8-25-97-132; Fig. 2), found directly associated with the holotype, represents the only cranial material potentially referrable to Mambachiton fiandohana. It is uncertain whether this bone actually pertains to Mambachiton fiandohana, or to an otherwise unknown reptile from the quarry. However, it is clear based on its small size and other features, that this element is not that of a rhynchosaur, the most common archosauromorph from the locality. Yet, we cannot exclude the possibility that it could pertain to a different avemetatarsalian or other, unknown archosauriform. This bone, which forms the posterodorsal portion of the orbit, bears a complex sutural surface with the frontal that angles anterolaterally. The dorsal portion of the orbital margin is raised slightly relative to the rest of the element and is faintly rugose. A fossa on the posteromedial portion indicates that the postfrontal would have participated in the supratemporal fossa, but a contact surface at the posterior edge (either for the postorbital or parietal) indicates that the postfrontal did not participate in the supratemporal fenestra.

Figure 2.

Right postfrontal of the holotype of Mambachiton fiandohana (UA 8-25-97-132) in dorsal (A) and ventral (B) views. Arrow indicates anterior direction. Scale bar = 1 cm. Abbreviations: a., articulates with; fr, frontal; om, orbit margin; pa, parietal; stf, supratemporal fossa.

Cervical series

The cervical series (UA 8-25-97-132) consists of six complete vertebrae, the first of which articulates with its neural arch; a fragment of the postzygapophyses of an anterior cervical was also recovered (Figs 3–8). The articulated cervical vertebrae encompass most of the cervical series, but their exact positions within the complete series (e.g. vertebrae 3–9) is unclear. The last element of the series is possibly transitional between the cervical and trunk series, but the parapophysis of this element still lies entirely on the centrum.

Figure 3.

Articulated cervical vertebra and osteoderm series of Mambachiton fiandohana (UA 8-25-97-132) in right lateral view as a photograph (A), line drawing (B), and CT rendering model with the osteoderms and matrix removed (C). Grey shading in (B) indicates sandstone matrix. Arrow indicates anterior direction. Scale bar = 5 cm. Abbreviations: 1st, first preserved cervical vertebra; 7th, seventh preserved cervical vertebra; cr, cervical rib; os, osteoderms.

Figure 4.

Articulated cervical vertebra and osteoderm series of Mambachiton fiandohana (UA 8-25-97-132) in left lateral view as a photograph (A), line drawing (B), and CT rendering model with the osteoderms and matrix removed (C). Grey shading in (B) indicates sandstone matrix. Arrow indicates anterior direction. Scale bar = 5 cm. Abbreviations: 1st, first preserved cervical vertebra; 7th, seventh preserved cervical vertebra; cr, cervical rib; os, osteoderms.

Figure 5.

Articulated cervical vertebra and osteoderm series of Mambachiton fiandohana (UA 8-25-97-132) in dorsal view as a photograph (A), line drawing (B), and CT rendering model with the osteoderms and matrix removed (C). Grey shading (B) indicates sandstone matrix. Arrow indicates anterior direction. Scale bar = 5 cm. Abbreviations: 1st, first preserved cervical vertebra; 7th, seventh preserved cervical vertebra; cr, cervical rib; os, osteoderms.

Figure 6.

Articulated cervical vertebra and osteoderm series of Mambachiton fiandohana (UA 8-25-97-132) in ventral view as a photograph (A), line drawing (B), and CT rendering model with the osteoderms and matrix removed (C). Grey shading in (B) indicates sandstone matrix. Arrow indicates anterior direction. Scale bar = 5 cm. Abbreviations: 1st, first preserved cervical vertebra; 7th, seventh preserved cervical vertebra; cr, cervical rib; os, osteoderms.

Figure 7.

Anterior to middle cervical vertebrae of the holotype of Mambachiton fiandohana (UA 8-25-97-132). Second persevered cervical vertebra in left lateral (A), right lateral (B), anterior (C), posterior (D), ventral (E), and dorsal (F) views. Fourth persevered cervical vertebra in left lateral (G), right lateral (H), anterior (I), posterior (J), ventral (K), and dorsal (L) views. Colours of the vertebrae correspond to those in Figures 3–6. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: dia, diapophysis; ns, neural spine; par, parapophysis; pre, prezygapophysis; poz, postzygapophysis.

Figure 8.

Middle to posterior cervical vertebrae of the holotype of Mambachiton fiandohana (UA 8-25-97-132). Fifth persevered cervical vertebra in left lateral (A), right lateral (B), anterior (C), posterior (D), ventral (E), and dorsal (F) views. Seventh preserved cervical vertebra in left lateral (G), right lateral (H), anterior (I), posterior (J), ventral (K), and dorsal (L) views. Colours of the vertebrae correspond to those in Figures 3–6. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: dia, diapophysis; ns, neural spine; par, parapophysis; pre, prezygapophysis; poz, postzygapophysis.

The centra of the cervical series shorten posteriorly (Table 2) to the fourth preserved vertebra, before lengthening again slightly. The lateral sides of the centra bear a shallow fossa (Figs 3, 4) between the articular facets; the depth of these fossae is constant throughout the series, although taphonomic crushing has distorted them on the left side of the sixth preserved vertebra and on the right side of the fifth (Fig. 8A–F). The amphicoelous centra have circular facets. A distinct but weakly extended ridge on the midline of the ventral surface (=ventral keel) is present on the first three preserved vertebrae (Figs 6, 7E) and on the anterior half of the fourth (Fig. 7K), but is absent on the fifth (Fig. 8E) and sixth. Paramedian ridges parallel the midline ridge on the first complete vertebra (Fig. 6). On the anterior cervical vertebrae, the posterior surfaces of the parapophyses are oval, the long axis oriented anteroposteriorly, whereas in the more posterior cervical vertebrae, the long axis is oriented posterodorsally. The anteroventral position of the parapophysis creates a concave surface medial to the structure, but lateral to the midline, in the first two preserved vertebrae. The parapophyses and diapophyses are separated throughout the series, though these structures converge posteriorly as the centrum shortens.

Table 2.

Measurements (mm) of the bones of Mambachiton fiandohana. Please note that the vertebra number is not absolute in the holotype presacral column but referred to the number system present in Figures 3–8 and in the description

Specimen	Length (mm)
Holotype (UA 8-25-97-132)
C ‘1’ centrum length	30.6
C ‘2’ centrum length	29.5
C ‘3’ centrum length	24.1
C ‘4’ centrum length	22.4
C ‘5’ centrum length	23.2
C ‘6’ centrum length	23.0
Mid-trunk vertebra (FMNH PR 5065)
Anterior centrum width	18.2
Anterior centrum height	17.2
Posterior centrum width	17.1
Posterior centrum height	15.9
Vertebral height (to top of neural spine)	44.3
Neural spine height	20.8
Length of dorsal edge of neural spine	22.5
Anterior neural canal height	6.2
Posterior neural canal height	6.5
Centrum length	23.8
Mid-trunk vertebra (FMNH PR 5065)
Anterior centrum width	16.5
Anterior centrum height	18.5
Posterior centrum width	18.0
Posterior centrum height	19.1
Vertebral height (to top of neural spine)	49.7
Neural spine height	20.0
Anterior neural canal height	6.2
Posterior neural canal height	7.5
Centrum length	23.9
Posterior trunk vertebra (FMNH PR 5065)
Anterior centrum width	20.3
Anterior centrum height	17.7
Posterior centrum width	17.5
Posterior centrum height	16.3
Vertebral height (to top of neural spine)	51.2
Neural spine height	20.8
Length of dorsal edge of neural spine	18.5
Anterior neural canal height	7.6
Posterior neural canal height	10.1
Centrum length	22.6
Posterior trunk vertebra (FMNH PR 5065)
Anterior centrum width	16.0
Anterior centrum height	20.5
Posterior centrum width	18.3
Posterior centrum height	19.4
Vertebral height (to top of neural spine)	47.1
Anterior neural canal height	5.4
Posterior neural canal height	4.2
Centrum length	23.6
First sacral vertebra (FMNH PR 5065)
Anterior centrum width	22.8
Anterior centrum height	17.9
Posterior centrum width	20.7
Posterior centrum height	16.4
Vertebral height (to top of neural spine)	48.3
Neural spine height	23.6
Length of dorsal edge of neural spine	19.7
Length of sacral rib (mediolateral from suture)	23.2
Anteroposterior length of the lateral edge of sacral rib	17.0
Anterior neural canal height	8.8
Posterior neural canal height	9.3
Centrum length	22.9
Anterior caudal vertebra (FMNH PR 5065)
Anterior centrum width	19.2
Anterior centrum height	17.1
Posterior centrum width	19.5
Posterior centrum height	17.7
Length of left transverse process	21.2
Length of right transverse process	23.3
Anterior neural canal height	5.8
Posterior neural canal height	4.9
Centrum length	20.5
Scapulocoracoid (left) (FMNH PR 5065)
Ventral width	41.3
Least width of scapular blade	19.3
Glenoid height	21.3
Glenoid width	15.0
Height of coracoid foramen	6.6
Scapulocoracoid (right) (FMNH PR 5065)
Ventral width	45.0
Least width of scapular blade	19.6
Glenoid height	20.8
Glenoid width	16.3
Ventral edge of coracoid	58.0
Height of coracoid foramen	5.6
Proximal half of femur (left) (FMNH PR 5065)
Proximal width (maximum)	32.7
Length of preserved portion of femur	67.4
Ilium (right) (FMNH PR 5065)
Length of dorsal edge	91.5
Anterior process length	16.6
Posterior process length	45.0
Height above acetabulum	30.9
Acetabular width	37.3
Length of ventral edge	53.9
Height of iliac contribution to acetabulum	31.3
Length of medial ridge on posterior process	62.8
Width of anterior sacral facet	17.1
Width of posterior sacral facet	26.2

The neural aches preserve a series of distinct laminae and deep fossae as in other long-necked archosaurs (e.g. Arizonasaurus babbitti, Nesbitt 2005; Teleocrater rhadinus, Nesbitt et al., 2018). Laterally, a deep centrodiapophyseal fossa (sensuWilson et al. 2011) lies medial and ventral to the diapophysis, and is roofed by a posterior centrodiapophyseal lamina (sensuWilson 1999). A few small laminae of bone are present within this fossa in the more anterior cervical vertebrae (Fig. 7A–F). A shallow parapophyseal centroprezygapophyseal fossa is present in the first preserved vertebra, deepening posteriorly as the result of a more pronounced prezygadiapophyseal lamina. A small tuber is present at the dorsal margin of the prezygadiapophyseal lamina on the lateral side of the prezygapophysis. A postzygapophyseal centrodiapophyseal fossa is absent in the anteriormost preserved neural arch (Figs 3–5), but appears in the second preserved neural arch (first preserved complete vertebra; Fig. 7A, B) and deepens posteriorly on that vertebra as the posterior centrodiapophyseal lamina develops. The postzygapophyseal centrodiapophyseal fossa is a deep pit, with its anterior extent obscured by the posterior centrodiapophyseal lamina in the second completely preserved vertebra; this condition is similar to that of Teleocrater rhadinus (e.g. NMT RB505) and saurischian dinosaurs (Langer and Benton 2006).

The prezygapophyses project anterolaterally at an angle about 45^o from the anteroposterior plane, and a large gap separates the articular surfaces at the midline (Figs 3–8). The postzygapophyses are angled about 45^o laterally (from horizontal) in posterior view; a wide gap between the articular surfaces widens anteriorly. No epipophyses are present at the dorsal margin in any of the cervical vertebrae (structures present in most avemetatarsalians), but a slight swelling of the neural spine posteriorly occurs in the homologous position. A fossa between the pre- and postzygapophyses excavates the base of the neural spine.

The neural spine is mediolaterally thin and situated over the posterior two-thirds of the centrum. A shallow fossa occurs at the lateral base of the neural spine throughout the preserved series. Most of the neural spines are hidden under osteoderms, but (computed tomography) reconstructions permit a number of observations (Figs 8, 9). Neural spines are expanded laterally at their dorsal margins, while the dorsal margins are nearly flat. The neural spines are slightly shorter (measured from their bases) than the centra are long. The neural spines shorten anteroposteriorly as their corresponding centra shorten posteriorly. The anterior edge of the neural spine slopes slightly anterodorsally in the more anterior elements, but not to the steep degree seen in aphanosaurs (Nesbitt et al. 2018). The posterior margin of all the neural spines are nearly vertical.

Figure 9.

Digital reconstruction of the osteoderms of the holotype of Mambachiton fiandohana (UA 8-25-97-132). Both paramedian rows of osteoderms (alternating purple and red = right side; alternating pink and turquoise = left side) in dorsal (A) and ventral (B) views. Left paramedian row in lateral (C), dorsolateral (D), ventral (E), and medial (F) views. Isolated osteoderm from the left paramedial row in dorsolateral (G), lateral (H), medial (I), and ventral (J) views and isolated osteoderm from the right paramedial row in dorsolateral (K), lateral (L), medial (M), and ventral (N) views. Arrows indicate anterior direction. Scale bars = 5 in A–F and 1 cm in G–N. Abbreviations: ap, anterior process; gap, groove for anterior process.

A single isolated rib, missing much of the shaft, is preserved anteriorly in the cervical series (Figs 3–6). Its disarticulation prevents its position within the cervical series from being assigned with certainty, but it appears to derive from the posterior portion of the series based on the positions of the capitulum and tuberculum. A sheet of bone on one side of the capitulum and the tuberculum may represent an anterior process, as is present in most archosauromorphs. The capitulum is broken, but a thin web of bone links it to the tuberculum.

Cervical osteoderms

Osteoderms cover the dorsal margins of the cervicals throughout the preserved series (in the holotype UA 8-25-97-132; Figs 3–5, 9, 10). Twenty articulated osteoderms are preserved on the right side of the holotype and nine on the left (Fig. 9). The long (mediolateral) axes of the osteoderms are deflected ventrolaterally about 60^o to the horizontal in posterior view. Each osteoderm is thick medially and progressively thins laterally as the osteoderms become dorsoventrally compressed. The dorsal surface of the medial edge sometimes preserves a ridge, with a slight depression adjacent to it. Where exposed, the surface of the medial portions of the osteoderms is slightly rugose, but no interdigitating contact between elements is evident. In anterior view, the lateral portion of the osteoderms flexes slightly dorsally. In dorsal view, the lateral margin is rounded, with a greater lateral extent more posteriorly, and the posterior margin varies from rounded to straight; some osteoderm margins bear small projections, but the pattern is inconsistent. CT data show that each osteoderm possesses a round (in cross-section), anteriorly tapering process (Fig. 9B, C, G–L). This process makes up half the length of the medial edge; although it occurs near the midline, it does not form the midline in articulation. No clear dorsal articular surfaces mark where osteoderms overlap (Figs 9, 10). Dorsal surfaces are smooth and unsculptured, but at least seven non-serial osteoderms on the right side bear a small dimple, which differs in size and location between elements. The osteoderms are consistently shaped throughout the column; they become only slightly longer (anteroposteriorly) posteriorly (see analysis below).

Figure 10.

Osteoderms and possible osteoderms of the holotype of Mambachiton fiandohana (UA 8-25-97-132). A, close up of the right paramedian osteoderms in lateral view. B, close up of the left paramedian osteoderms in lateral view. Three disarticulated left paramedian osteoderms in ventral (C) and medial (D) views. E, possible osteoderms. Isolated osteoderm with a unique morphology compared to the others in dorsal (F) and posterior (G) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: anl, anterior lamina; ap, anterior process; gap, groove for anterior process.

In Mambachiton, the configuration of osteoderms relative to each other and to their associated vertebra is unique. They are arranged in two rows along the midline, but are staggered across the midline, as in some pseudosuchians (e.g. Prestosuchus chiniquensis; Nesbitt 2011; Nundasuchus songeaensis, Nesbitt et al. 2014). The osteoderms are imbricated, such that the posterior edge of the more anterior osteoderm overlaps the anterior edge of the subsequent one. Additionally, the tapering anterior process articulates with a distinct groove (Fig. 9J, N) on the preceding osteoderm, an arrangement unique among archosauriforms. The number of osteoderms per vertebra exceeds the typical ‘one paramedian pair to vertebra’ ratio typical in Archosauriformes (Nesbitt 2011). Nine osteoderms cover the longest cervical vertebra in Mambachiton, whereas six or seven osteoderms occur per vertebra in the posterior portion of the preserved vertebral series. The number of paramedian osteoderms per vertebra in Mambachiton is the highest known among any archosauriform.

Variation-

Our principal components analysis (Fig. 11) show some clustering based on relative location of the osteoderms along the cervical column. PC1 shows the most variance along Landmarks 3, and 1 (respectively) which correlates with the angle between the anterior process and the rugose medial flare. PC2 shows the most variance along Landmark 5, 3, and 4 (respectively) which correlates with the medial edge of the osteoderm.

Figure 11.

A, principal component analysis (PCA) of both left and right lateral cervical osteoderms of the holotype (UA 8-25-97-132). Grouping is based on general location of the osteoderms in the series: anterior, posterior, and central and correspond to the letters (right side) and numbers (left side) of (C). B, lollipop diagram and wireframe graph for left lateral PC1, superimposed over osteoderm 1 in (C). Arrow indicates anterior direction.

The Principal Component 1 vs. Principal Component 2 graph shows a cluster of similar shapes among anterior and posterior osteoderms respectively (Fig. 11). The shapes of more central osteoderms do not cluster as closely, varying more broadly across both principal components. Anterior osteoderms form a cluster between 0.00 and 0.15 along PC1, and 0.05 and -0.05 in PC2. Posterior osteoderms form a cluster between 0.00 and -0.15 along PC1, and show less variability along PC2, in a cluster between 0.00 and -0.05. This indicates that anterior osteoderm shapes typically show more variation along both PC1 and PC2, whereas the shape of the posterior osteoderms tends to vary along only PC1, and are less variable along PC2. Central osteoderms show higher variability along both PC1 and PC2, more so than either the anterior or posterior osteoderms. Central osteoderms form a cluster mainly between 0.05 and -0.10 along PC1, and between 0.05 and -0.05 along PC2. Two major outliers (central osteoderms 5 and B) cause the central osteoderms to show high deviation in shape, occupying more of the graph space along PC1 and PC2.

The PCA graph (Fig. 11) shows little apparent separation between right and left osteoderm variation, but this may be the result of a relatively smaller sample size for the right elements. Interestingly, mirrored paired osteoderms (taking into account the slight staggering of the left and right osteoderm columns) plot relatively distantly from each other in PCA space (by a similar length), despite being similar to one another in anatomy. In this case, the mirrored pairs are between osteoderms 3 and A, 5 and C, and 6 and D.

These variations among the osteoderms are congruent with our morphometric analysis. Anteriorly positioned osteoderms appear more rounded, whereas posterior elements appear more angular and more parallelogram shaped. Central osteoderms lie somewhere between these two end members, though the morphometric analysis results shows no clear pattern that would indicate a gradual transition in shape from one extreme to the other. Rather, anterior and posterior osteoderms show little variability, and plot separately within the PCA space. Central osteoderms are much more variable, and can overlap in morphology with either the anterior or posterior osteoderms. Left and right paramedian osteoderms are nearly identical, but not when paired together.

Other osteoderms

A variety of other osteoderms were recovered with the holotype (UA 8-25-97-132), but their position in the skeleton is unknown. A cluster of three disarticulated, but similarly oriented, osteoderms prepared in ventral view are the most similar to the osteoderms articulated along the cervical vertebral column (Fig. 10). These osteoderms may have belonged to the anterior portion of the neck dorsal to the vertebrae, given that this part of the neck is slightly disarticulated and the morphology of these osteoderms closely matches the more anterior osteoderms in the articulated cervical series. These osteoderms are triangular, and this triangular shape becomes posterolaterally extended anteriorly. These three osteoderms have long and tapered anterior processes near their medial edges; their posterolateral portions broaden, as in the articulated neck osteoderms. A deep, anteroposteriorly oriented groove lies near the medial edge, probably the articulation surface for the long anterior process, reminiscent of the groove on the ventral surface of the cervical osteoderms (observed through CT data) discussed earlier. The medial surface, thickest dorsoventrally, is covered by a system of ridges and grooves, probably marking the midline and contact surface with its antimere osteoderm.

A rectangular osteoderm also was found directly associated with the holotype (Fig. 11F, G). We are tentatively assigning it to the holotype given its close proximity to the cervical series and because no other taxa from the quarry are known to bear osteoderms. Nevertheless, we do recognize the possibility that another, currently unknown reptile taxon (e.g. an aetosauriform) could be the source of this single osteoderm. Given the highly divergent shape of this osteoderm relative to of all others known from this locality, its position and anatomical directions are unknown. Accordingly, we employ the anatomical orientation of a typical paramedian aetosaur osteoderm, given their general similarities (Parker 2007, 2008, Desojo et al. 2013). This rectangular osteoderm is about twice as wide as long, with a distinct bend just off the mediolateral centre. The presumed anterior edge of the dorsal surface bears a flat lamina across its entire edge. The dorsal surface is ornamented; small dimples are present near the apex, and the medial and lateral portions bear more elongated grooves and rounded ridges. This isolated osteoderm is consistently thick dorsoventrally throughout its body. Its ventral surface is smooth and concave. Overall, this osteoderm is aetosaur-like in form, the presence of an anterior articulation surface (=anterior bar, anterior lamina) being rather rare among archosauriforms. The ornamentation of this osteoderm also is similar to small-bodied aetosaurs (e.g. Small and Martz 2013).

Two other fragments found with the holotype also appear to be osteoderms, but this cannot be confirmed definitively (Fig. 10E). Both have compressed, tapered processes that could correspond to the anterior processes of the other osteoderms.

Referred vertebrae

Four trunk vertebrae (part of FMNH PR 5065) found near the holotype are referred to Mambachiton fiandohana based on similar size, consistency of character states, preservation, and the fact that none of the elements are duplicated with the holotype or each other; we consider it possible, and even likely, that all these vertebrae belong to the same individual as the holotype (Figs 12, 13). Two of these vertebrae are from the middle trunk (Fig. 12), and two from the posterior trunk (Fig. 13), judging from centrum shapes as well as diapophysis and parapophysis shapes.

Figure 12.

Anterior or middle trunk vertebrae referred to Mambachiton fiandohana (FMNH PR 5065). Anterior or middle trunk vertebra in left lateral (A), right lateral (B), anterior (C), posterior (D), dorsal (E), and ventral (F) views. Second anterior or middle trunk vertebra in left lateral (G), right lateral (H), anterior (I), posterior (J), dorsal (K), and ventral (L) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: dia, diapophysis; ns, neural spine; par, parapophysis; pre, prezygapophysis; poz, postzygapophysis.

Figure 13.

Posterior trunk vertebrae referred to Mambachiton fiandohana (FMNH PR 5065). Posterior trunk vertebra in left lateral (A), right lateral (B), anterior (C), posterior (D), dorsal (E), and ventral (F) views. Another posterior trunk vertebra in left lateral (G), right lateral (H), anterior (I), posterior (J), dorsal (K), and ventral (L) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: dia, diapophysis; ns, neural spine; par, parapophysis; pre, prezygapophysis; poz, postzygapophysis; syp, synpophysis.

The well-preserved middle trunk vertebrae (Fig. 12) both are essentially complete. The centra are strongly waisted between the anterior and posterior articular facets, and a slight fossa is present on the lateral sides, just ventral to the closed neurocentral suture. The centra have a length (25 mm) to articular facet height (19 mm) ratio of 1.32, measured from the anterior articular facet. The amphicoelous articular facets are oval, with a slightly longer dorsoventral than mediolateral axis. Both centra lack a ridge on the midline ventrally.

The neural arches bear laterally extended transverse processes composed of the combined stalks of the parapophysis and diapophysis (Fig. 12). The more anteroventrally located parapophysis has a concave articular surface whereas the posterodorsally and laterally located diapophysis has a convex articular surface. A deep centrodiapophyseal fossa (sensuWilson et al. 2011) lies between and ventral to the parapophysis and diapophysis, and is framed posteriorly by a posterior centrodiapophyseal lamina (sensuWilson 1999). More posteriorly, a postzygapophyseal centrodiapophyseal fossa is bordered anteriorly by the posterior centrodiapophyseal lamina and posterodorsally by the postzygadiapophyseal lamina. An anterior, short, and thick centrodiapophyseal lamina is present. In dorsal view, the transverse process is triangular, with the most lateral point at the posterior edge. No parapophyseal centroprezygapophyseal fossa is present in these vertebrae, unlike in Teleocrater rhadinus (NMT RB 500).

The prezygapophyses are angled ~30^o from the mediolateral plane in anterior view, and a wide gap (2.5 mm) separates them at the midline (Fig. 12). The lateral edge of the prezygapophysis bears a thin lamina that terminates posteriorly at the anterior edge of the base of the neural spine. Medially, the prezygapophyses are well separated, a gap that is interpreted as a hypantrum. The bases of the prezygapophyses meet in a depression at the midline, slightly excavating the anterior part of the base of the neural spine. The postzygapophyses have slightly concave articular facets; a thin lamina lies at the medial edge of the facet. The postzygapophyses are separated by a ventrally flat piece of bone, similar to that of Teleocrater rhadinus (NMT RB 516; Nesbitt et al. 2018 text fig. 9); hence, we interpret this structure as a hyposphene, similar to those of other archosaurs (see below). A deep fossa lies between the postzygapophyses, and excavates the posterior portion of the base of the neural spine.

The neural spine lies dorsal to the posterior half of the centrum (Fig. 12). No dorsally facing fossa is present on the lateral side of the neural spine, in contrast to the one that occurs in Teleocrater rhadinus (NMT RB500). The anterior edge of the neural spine is dorsoventrally straight in lateral view, whereas the posterior edge is slanted posterodorsally at the point where the dorsal tip lies posterior to the postzygapophyses. The dorsal portion of the neural spine is expanded on the lateral sides, rounded but rugose on the dorsal surface, and slightly arched in lateral view. The posterodorsal tip is more rounded than the anterodorsal tip.

The posterior part of the trunk is represented by two vertebrae (Fig. 13). Of the two, one (Fig. 13A–F) is considered more anterior, because the diapophysis and parapophysis are still slightly separated, whereas the two structures are combined into a synpophysis in the second posterior trunk vertebra (Fig. 13G–L). The amphicoelous centrum of the posterior trunk vertebrae bears rounded anterior and posterior articular surfaces. The centrum surfaces in the more posterior vertebra are mediolaterally wider than dorsoventrally tall. The centra of both posterior trunk vertebrae are shorter than those of the more anterior trunk vertebrae, yet a clear fossa is still present in each posterior trunk vertebra on the lateral side of the centrum just ventral to the neurocentral suture. The ventral surfaces of the centra are smooth at the midline.

The neurocentral sutures are closed on the right side but open on the left in both posterior trunk vertebrae (Fig. 13). In the more posterior of these, the parapophysis and diapophysis both lie dorsal to the neurocentral suture on the neural arch. As in the more anterior trunk vertebrae, the diapophysis and parapophysis are both located on a transverse process. The articular surfaces of the diapophysis and parapophysis are convex and continuous in both (Fig. 13), the articular surfaces of the diapophysis and parapophysis not being differentiated. The homologous lamina surrounding the diapophysis and parapophysis in the more anterior trunk vertebrae (e.g. anterior and posterior centrodiapophyseal lamina) are not as laterally expanded and are distinctly rounded in these posterior trunk vertebrae. However, the fossae (centrodiapophyseal fossa, postzygapophyseal centrodiapophyseal fossa) are similarly deep. The more posterior vertebra (Fig. 13G–L) has a deep prezygapophyseal centrodiapophyseal fossa, unlike any of the other trunk vertebrae.

In lateral view, the prezygapophyses and postzygapophyses lie in the same horizontal plane, and slope about 35^o relative to the transverse plane in posterior view. Anteriorly, the articular surfaces of the prezygapophyses are separated by a clear gap that widens anteriorly; this gap develops posteriorly into a depression at the midline, where it excavates the base of the neural spine. This gap is similar to that of the hypantrum of the other trunk vertebrae of Mambachiton fiandohana and of other archosaurs (Stefanic and Nesbitt 2018, 2019). The thin lamina of bone on the lateral side of the postzygapophysis in more anterior vertebrae is absent in the posterior trunk vertebrae. Posteriorly, the articular surfaces of the postzygapophyses are slightly concave and are separated by a thin wedge of bone, like that of the more anterior trunk vertebrae. A deep fossa excavates the posterior edge of the neural spine and separates the postzygapophyses on the midline.

The neural spine of the posterior trunk vertebrae lies dorsal to the posterior half of the centrum as in the other trunk vertebrae. The anteroposterior length of the neural spine is just short of the anteroposterior length of the centrum in the posterior trunk vertebrae. The mediolaterally thin neural spine expands laterally, anteriorly, and posteriorly at its dorsal margin. In dorsal view, the dorsal surface of the neural spine is nearly flat, and slightly rugose. The anterior edge of the neural spine is dorsoventrally straight, whereas the posterior margin is posterodorsally oriented. In the more posterior trunk vertebrae, the neural spine is taller and more anteroposteriorly restricted than in its anterior counterpart. In the more posterior element (Fig. 13G–L), a thin lamina of bone at the midline projects posteriorly.

Sacral vertebra.

A single sacral vertebra (FMNH PR 5065) is known, here identified as the first sacral (Fig. 14). Much of the left sacral rib is broken away, but this vertebra is otherwise complete. The ‘C’-shaped articular surface of the sacral rib articulates with the medial side of the anterior process of the ilium (see below), indicating that this is primordial sacral vertebra 1, and the precise articulation indicates that these two elements are likely from the same individual. The centrum is waisted between the anterior and posterior articular facets, and a fossa is present on the lateral side of the centrum just ventral to the neurocentral suture, similar to that of Teleocrater rhadinus (NMT RB1395). The neurocentral suture is closed. The anterior articular surface is concave, whereas the posterior surface is flat.

Figure 14.

Sacral vertebra 1 referred to Mambachiton fiandohana (FMNH PR 5065) in left lateral (A), right lateral (B), ventrolateral to highlight the articular surface of sacral rib 1 (C), anterior (D), posterior (E), dorsal (F), and ventral (G) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: ns, neural spine; pre, prezygapophysis; poz, postzygapophysis; sr, sacral rib.

The sacral ribs extend directly laterally, having a small ventral component. In ventrolateral view, the sacral rib has an ‘I-beam’ construction, with flat dorsal and ventral faces and concave anterior and posterior surfaces. The posterior concave surface is deep relative to its counterpart in Teleocrater rhadinus (NMT RB1395). The anterodorsal tip of the rib is expanded and curved anteriorly to produce a ‘C’-shaped articulation surface with the ilium. The anterior extension of the rib in Mambachiton fiandohana is greater than that of Teleocrater rhadinus (NMT RB1395) and more similar to that of early dinosauriforms (e.g. Silesaurus opolensis, Dzik and Sulej 2007). The posterolateral and ventral portions of the sacral rib are posteroventrally deflected, but not as posteriorly expanded as in Teleocrater rhadinus (NMT RB1395). The contact between the sacral rib and vertebra forms a strong ridge dorsally, posteriorly, and ventrally, but no clear suture is present between the two elements; this indicates that co-ossification is complete, similar to that of the only known sacral 1 of Teleocrater rhadinus (NMT RB1395). Furthermore, the sacral rib attaches to the anterior half of the centrum, a diagnostic feature of the first primordial sacral vertebra in archosaurs (Nesbitt 2011). Sacral ribs are not shared between sacral vertebrae in Mambachiton fiandohana.

The neural arch is co-ossified to the centrum, with no sign of a neurocentral suture (Fig. 14). In anterior view, mediolaterally thin laminae frame the lateral sides of the neural canal; just lateral to these laminae, there are small but deep fossae bordered laterally by the raised contact with the sacral rib. The prezygapophyses are angled ~45^o to the mediolateral horizontal plane and are separated by a noticeable gap (5 mm in length). The processes meet medially at a fossa on the midline. Posteriorly, the postzygapophyses are angled ~45^o to the mediolateral horizontal plane and meet at the midline at a point; a deep fossa between the postzygapophyses penetrates the base of the neural spine. Like the anterior surface of the vertebra, mediolaterally thin laminae forming the lateral side of the neural canal border a fossa that extends to the sacral rib. The neural canal, the largest of all the vertebrae, is oval in outline, with a long, dorsoventrally-oriented axis. The ventral portion of the neural canal excavates the body of the centrum.

The neural spine lies dorsal to the posterior half of the centrum. The anterior and posterior edges of the neural spine bear thin laminae at the midline. The dorsal end of the neural spine is more prominently expanded (in all directions) and dorsally rounded than in other preserved vertebrae. The surface of this expansion is rugose.

Caudal vertebra.

There is a single caudal vertebra (part of FMNH PR 5065; Fig. 15), from the anterior portion of the series, as suggested by its large, well-developed transverse processes and the lack of chevron facets on the posteroventral portion of the centrum. The neural spine is broken at its base, but this vertebra is otherwise complete. This centrum is amphicoelous and longer than tall, with distinct fossae on the lateral surface ventral to the transverse processes. Ventrally, a small depression occurs on the midline near the posterior portion, but no depression is present anteriorly. A fused caudal rib/transverse process links the centrum with the fused neural arch. Here, a distinct rim circumscribes the transverse process, but no suture is evident; this rim marks where the caudal rib has completely fused to the rest of the vertebra. The wing-like transverse processes project laterally, having a small posterior component; in anterior view they curve very slightly ventrally towards their tips. The right transverse process appears to expand anteroposteriorly at its lateral margin. A rugose surface is present on the dorsal surface of both transverse processes about halfway to their lateral terminations. The rugose surface is larger anteriorly and decreases in size posteriorly.

Figure 15.

Anterior caudal vertebra referred to Mambachiton fiandohana (FMNH PR 5065) in left lateral (A), right lateral (B), anterior (C), posterior (D), dorsal (E), and ventral (F) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: ns, neural spine; pre, prezygapophysis; poz, postzygapophysis; tp, transverse process.

The base of the neural spine is positioned over the posterior half of the centrum, and deep fossae lie between the prezygapophyses and postzygapophyses. The prezygapophyses project ~30^o from horizontal, and angle ~45^o anteriorly. There is a clear gap at the midline between the two articular surfaces. A broad, flat lamina lies between the postzygapophyses at the midline.

Pectoral girdle

Scapulocoracoid:

The scapula and coracoid (parts of FMNH PR 5065) are co-ossified, with only remnants of a suture in both the left (Fig. 16E–H) and right (Fig. 16A–D) elements; a raised rim marks the junction of the scapula and coracoid. The main bodies of the scapulocoracoids retain much of their original three-dimensional form, but the distal portion of the right scapulocoracoid is crushed and slightly displaced. The scapular blade is incompletely preserved distally, and parts of the anteroventral edge of the coracoid are missing. The scapula is wide proximally, constricted anteroposteriorly dorsal to the glenoid, and it expands anteroposteriorly distally. The glenoid is formed 50% by the scapula and 50% by the coracoid. As in early-diverging avemetatarsalians (e.g. Teleocrater rhadinus NMT RB480; Asilisaurus kongwe NMT RB159) and some paracrocodylomorphs (Poposaurus gracilis; Schachner et al. 2020; Postosuchus spp., Peyer et al. 2008), the glenoid projects posteroventrally and the medial edge of the glenoid is only barely visible in lateral view (Fig. 16A, E). A slight rim circumscribing the glenoid separates the humeral articulation surface from the body of the scapulocoracoid. A distinct, large, and posterolaterally expanded tuber is confluent with the dorsal margin of the glenoid. This tuber is mediolaterally compressed and covered in a rugose surface stretching from the dorsal margin to the glenoid. This tuber is the attachment site for the scapular head of the M. triceps, as in Crocodylia (Mook 1921, Meers 2003). Most archosauriforms have a scar or tuber in the same location, but the large size of this feature in Mambachiton fiandohana is similar to that of the loricatan (pseudosuchian archosaur) Batrachotomus kupferzellensis (Gower and Schoch 2009, text fig. 3). However, the position of this tuber in Mambachiton fiandohana differs from that of the more distally shifted feature of Batrachotomus kupferzellensis and other pseudosuchians, and instead more resembles that of some avemetatarsalians (e.g. Asilisaurus kongwe NMT RB159), or taxa with a slightly more distally shifted scar (e.g. Teleocrater rhadinus, NMT RB478). The proximal portion of the scapula, anterior to the glenoid, is concave distal to the anteroposteriorly-oriented acromion ridge that is located on the anterodorsal edge of the proximal portion of the scapula, a character state present in proterochampsids and crown group archosaurs (Nesbitt 2011). The acromion ridge encircles the anterodorsal margin of the proximal portion of the scapula.

Figure 16.

Scapulocoracoids referred to Mambachiton fiandohana (FMNH PR 5065). Right scapulocoracoid (FMNH PR 5065, collected as field number 9-7-03-788) in lateral (A), ventrolateral (B), posterior (C), and medial (D) views. Left scapulocoracoid (FMNH PR 5065, collected as field number 9-1-98-395) in lateral (E), ventrolateral (F), posterior (G), and medial (H) views. Arrows indicate anterior direction. Scale bars = 1 cm. Abbreviations: acp, acromion process; cf, coracoid foramen; co, coracoid; gl, glenoid; sc, scapula; tu, tuber.

In lateral view, both the anterior and posterior edges of the scapular blade are distinctly concave, whereas the anterior edge of the left scapulocoracoid is much more concave. The anterior and posterior edges are rounded from the lateral to the medial surfaces. The posterior edge is thicker than the anterior one. No proximodistally-oriented ridge occurs on the posterior edge, in contrast with some early avemetatarsalians (e.g. Teleocrater rhadinus, NMT RB480; Asilisaurus kongwe NMT RB159) which have a clear ridge. The distal end of the scapular blade is greatly expanded relative to its midshaft, much more so than in other early avemetatarsalians (e.g. Teleocrater rhadinus, NMT RB480; Asilisaurus kongwe, NMT RB159). The distal surface is incompletely preserved, but appears thicker mediolaterally than the mid-portion (as determined from the right element). The medial surface is concave; a mediolaterally thicker portion extends distally from the glenoid.

In ventrolateral view, the oval-shaped coracoid is largely complete (Fig. 16A–D). Its suture with the scapula is anteroposteriorly straight, and nearly continuous (as determined from the right side), lacking a clear notch. The orientation of the glenoid matches that of scapula, but the coracoid portion of the glenoid extends more posteriorly than the scapular component, but not as asymmetrically as in Asilisaurus kongwe (NMT RB159). The posteroventral portion of the coracoid expands posteriorly and ventrally to the glenoid, creating a clear postglenoid process. A cleft separates the ventral rim of the glenoid from the ventral margin of the coracoid. A similar expansion and cleft is present in early crown group archosaurs (e.g. Asilisaurus kongwe NMT RB159; Batrachotomus kupferzellensis, Gower and Schoch 2009, text fig. 3). The posteroventral edge of the coracoid of Mambachiton fiandohana is thickened mediolaterally relative to the anterior half, as in other archosaurs, and the ventral edge is rounded in ventrolateral view. In ventral view, the thickened ventral edge bears a weakly defined groove that parallels the ventral edge. A similar groove is present in Postosuchus alisonae (Peyer et al. 2008) and crocodylomorphs (Nesbitt 2011). The cleft between the glenoid and the ventral edge terminates anteriorly as a slight depression on the lateral surface. The coracoid foramen is located anterior to the glenoid, just posterior to the center of the coracoid. The anteroventral portion of the coracoid is mediolaterally thin.

Ilium:

A right ilium (part of FMNH PR 5065; Fig. 17) is referred to Mambachiton fiandohana because the articular surface for the first sacral vertebra on the medial side of this element corresponds nearly perfectly with the articular surface of the first sacral vertebra found at the same locality. In lateral view, the ilium has a short anterior process and a posteriorly elongated posterior process, similar to that of archosauriforms (e.g. Euparkeria capensis; Ewer 1965) and a variety of archosaurs (e.g. ornithosuchids; von Baczko and Ezcurra 2013). The anterior process (=preacetabular process) is short and does not extend anterior to the pubic peduncle. The surface dorsal to the supra-acetabular crest is slightly convex and smooth; Mambachiton fiandohana does not bear a vertical crest dorsal to the supra-acetabular crest, in contrast to its presence in some pseudosuchians (Arizonasaurus babbitti, MSM P4590; Batrachotomus kupferzellensis, SMNS 80268) or avemetatarsalians (e.g. Teleocrater rhadinus, NHMUK PV R6795; Asilisaurus kongwe, NMT RB159). The anteroventral margin of the anterior process is slightly thickened mediolaterally relative to the rest of the iliac blade, as in Teleocrater rhadinus; however, the anterior tip of the process is not curved medially as it is in Teleocrater rhadinus (NHMUK PV R6795). The dorsal margin of the iliac blade, from the anterior to the posterior process, is slightly convex in lateral view and the dorsal margin is tapered. In lateral view, there is a slight depression dorsal to the supra-acetabular crest, between the anterior and posterior processes. The posterior process tapers to a rounded posterior termination. The ventral edge of the posterior process is expanded medially into a shelf on the medial surface. In ventral view, this shelf is slightly concave and is reminiscent of the brevis fossa or shelf present in avemetatarsalians (e.g. Teleocrater rhadinus, NMT RB1039; Asilisaurus kongwe, NMT RB159), but also is not that dissimilar to morphology of the same area in Batrachotomus kupferzellensis (SMNS 80268) and several other eucrocopod archosauriforms, such as Chanaresuchus bonapartei and Gracilisuchus stipanicicorum (Ezcurra pers. comm.). In posterior view, the ventral termination of the ilium is triangular, with a slightly concave ventral side.

Figure 17.

Right ilium referred to Mambachiton fiandohana (FMNH PR 5065) in right lateral (A), anterior (B), medial (C), dorsal (D), and ventral (E) views. Arrows indicate anterior direction. Scale bar = 1 cm. Abbreviations: ace, acetabulum; ap, anterior process; ip, ischial peduncle; mr, medial ridge; no, notch; pup, pubic peduncle; sac, supra-acetabulum crest; sr1, sacral rib 1 scar; sr2, sacral rib 2 scar.

The acetabulum is relatively deep compared to that of archosauriforms primitively; a well-developed supra-acetabular crest forms its dorsal portion. The sharp portion of the supra-acetabular crest fails to reach the articulation surface with the pubis. A weakly rimmed depression lies within the acetabulum (Fig. 17A), a feature also present in Teleocrater rhadinus (NMT RB1039). Just ventral to it lies an antitrochanter surface, which reaches the articular surface with the ischium. The acetabulum likely is closed, given that its ventral margin is ‘v’-shaped in lateral view. The ischial peduncle, thickest posteriorly, thins anteriorly; a slight rim marks the posterior boundary of its articulation with the ischium. As in other early-diverging avemetatarsalians (Nesbitt et al. 2017; character 414), a small notch occurs on the articulation surface with the ischium, slightly anterior to the main body of the ischial peduncle and the ventral point of the acetabulum (Fig. 17). The articulation between the ischium and ilium is longer than between it and the pubis. The articulation surface of the ischium meets the pubis at a point at an angle of ~50^o.

The medial surface of the ilium bears two distinct sacral rib scars corresponding to primordial sacrals one and two, judging from comparisons with other archosauriforms (Erythrosuchus africanus; Gower 2003) and archosaurs (see: Nesbitt 2011). The first sacral rib articulates in the junction between the main body of the ilium and its anterior process. A distinct scar, as confirmed by the preserved first primordial sacral vertebra (see above), occurs on the anterior process of the ilium, nearly reaching the anterior end of this process. The posterior edge of the first sacral rib scar meets the anterior edge of the second sacral rib scar. The latter tapers posteriorly and meets the anteroposteriorly oriented medial ridge. Based on this scarring, the second sacral rib appears to have articulated with both the dorsal and ventral surfaces of the medial ridge, the articulation extending to within 2 cm of the posterior end. This ridge meets the ventral edge of the posterior process, forming the brevis shelf. The medial ridge thins posteriorly. A large rim separates the sacral rib scars from the smooth medial face of the acetabular portion of the ilium.

Femur:

The proximal portion of a left femur (part of FMNH PR 5065; Fig. 18) was recovered from the locality; the size of its head closely matches that of the acetabulum (FMNH PR 5065). Where present, the poorly preserved surface of the femur is striated proximodistally. Proximally, a straight groove follows the long axis of the femoral head, as in some early avemetatarsalians (e.g. Teleocrater rhadinus, Silesaurus opolensis) and some pseudosuchians (e.g. poposauroids; Nesbitt 2011). Three tubera are present proximally, a prominent anterolateral one, a weakly developed anteromedial one, and a rounded and broad posteromedial one. The medial edge of the posteromedial tuber bears a proximodistally oriented ridge. A faint line between smooth and fibrous bone divides the anterior portion of the anterolateral tuber; the proximal portion of this line marks the medial portion of the femoral head, which transitions smoothly from the shaft as in Teleocrater rhadinus (Nesbitt et al. 2018, text fig. 21) and other aphanosaurs. Medially, much of the fourth trochanter is eroded off, but the remaining portion indicates that it was positioned low, more similar to Teleocrater rhadinus than lagerpetids (Dromomeron gregorii; Nesbitt et al. 2009), silesaurids (e.g. Silesaurus opolensis; Dzik 2003) or dinosaurs (e.g. Herrerasaurus ischigualastensis, Coelophysis bauri). A shallow depression lies anteromedial to the fourth trochanter. Anterolaterally, the proximal surface is slightly concave and bounded distally by a swelling (Fig. 18). This swelling appears to be homologous with the scar for the inferred M. iliotrochantericus caudalis scar in Teleocrater rhadinus (NHMUK PV R6795), and the anterior trochanter in ornithodirans (e.g. Dromomeron gregorii, silesaurids, and dinosaurs). The femoral shaft is hollow, as indicated by the presence of a diagenetic calcite infilling of the core; the cortex is ~2.3 mm wide, and the maximum cross section width is 16.4 mm.

Figure 18.

Referred proximal half of left femur referred to Mambachiton fiandohana (FMNH PR 5065) in proximal (A), posteromedial (B), posterolateral (C), anterolateral (D), and anteromedial (E) views. Arrow indicates anterior direction. Scale bars = 1 cm. Abbreviations: alt, anterolateral tuber; amt, anteromedial tuber; at, anterior trochanter; ft, fourth trochanter (mostly eroded); g, groove; pmt, posteromedial tuber.

OTHER ARCHOSAUR MEMBERS OF THE ‘BASAL ISALO II’ ASSEMBLAGE

Dinosauriformes:

A small, complete, left tibia (FMNH PR 5066; Fig. 19C–F) also was recovered at the locality (M-13) from which the holotype of Mambachiton derives. Its straight shaft is nearly circular in cross section for much of its length. The element is more expanded proximally than distally. Proximally, a straight cnemial crest is separated from the rest of the proximal surface by a small gap laterally. Posterolaterally, the medial and posterior condyles are separated by a slight gap. The lateral, proximodistally oriented crest observed in Silesaurus opolensis (Dzik 2003) is absent. Distally, a groove separates a distinct posterolateral process from the remainder of the distal articular surface. Anterior to this groove, the distal surface is slightly expanded proximally, suggesting that an ascending process of the astragalus had been present. Like UA-8-24-98-199 (see below), this tibia exhibits a number of dinosauriform character states, including: a posterolateral process on the distal end, an articulation facet for the anterior ascending process of the astragalus, and a prominent, anteriorly directed cnemial crest. This tibia and UA-8-24-98-199 (below) are possibly from the same species.

Figure 19.

Dinosauriform remains from the type locality of Mambachiton fiandohana (‘M-13’) (C–K) and another contemporaneous ‘basal Isalo II’/Makay Formation locality with a similar fauna (A, B). Left maxilla fragment (FMNH PR 5068) in lateral (A) and medial (B) views. Smaller left tibia (FMNH PR 5066) in proximal (C), distal (D), lateral (E), and medial (F) views. Larger left tibia (UA-8-24-98-199) in anterior (G), medial (H), posterior (I), and lateral (J), and distal views. Arrow indicates anterior direction. Scales = 1 cm. Abbreviations: a., articulates with; aas, ascending process of the astragalus; anf, antorbital fenestra; anfo, antorbital fossa; cc, cnemial crest; g, groove; lc, lateral condyle; lp, lateral process; mc, medial condyle; pp, palatal process; t, tooth.

A second, larger, partial left tibia (estimated length = 150 mm) (UA-8-24-98-199; Fig. 19G–K) represents one of the largest reptilian remains recovered from the M-13 locality. This element’s shaft is hollow, as observed at breaks. Proximally, the shaft is oval, the anteroposterior axis longer than the mediolateral, perhaps suggesting that a prominent cnemial crest was present. More distally, the shaft is circular in cross-section. The distal end expands relative to the mid-shaft, and a posterolateral process is present. The posterior surface of the posterolateral process is rugose, whereas the other surfaces of the distal end are smooth; the distal end is nearly circular in axial view. A proximodistally oriented groove separates the posterolateral process from the anterior half of the rest of the tibia laterally. The anterolateral portion of the distal surface forms the element’s most proximal portion, suggesting that it articulated with an anterior ascending process of the astragalus.

Two features indicate the dinosauriform affinities of FMNH PR 5066 and UA 8-24-98-199: the distal posterolateral process, and an articulation facet for an anterior ascending process of the astragalus (Nesbitt 2011). The morphology of these tibiae is consistent with that of silesaurids (see: Dzik 2003).

A fragment of a small left maxilla bearing four teeth (FMNH PR 5068; Fig. 19A, B) was recovered from another nearby ‘basal Isalo II’ locality host a similar fauna (including some of the same cynodont and rhynchosaur species). The preservation of this maxillary fragment is poor, but a number of important details can be discerned. In lateral view, much of the main body is flat, bearing a few small foramina. A well-defined ridge marks the anteroventral edge of an antorbital fossa, a small portion of the edge of which is preserved. This ridge does not project laterally beyond the antorbital fossa to the rest of the body of the maxilla. The portion of the maxilla anterior of the antorbital fossa is similar to that of the dinosauriform Lewisuchus admixtus (Romer 1972b; Ezcurra et al. 2020b). Although broken, the anterior portion of the maxilla seems to be nearly complete; this suggests that the anterior end of the maxilla was low and tapered, as in Lewisuchus admixtus. The element’s ventral margin is straight. Medially, a dorsoventrally wide ridge of bone is present along the length of the maxilla; the anteriormost portion expands slightly medially to form the palatal process (broken). Dorsal to this ridge of bone, the maxilla is considerably thinner mediolaterally at the base of the dorsal (=ascending) process.

Four teeth are preserved ventral to the medial ridge, as is at least one empty alveolus. The bases of the teeth appear to be ankylosed to their sockets, like those of silesaurids (Nesbitt et al. 2010, Mestriner et al. 2021) and other avemetatarsalians (pterosaurs, lagerpetids; Ezcurra et al. 2020). Interdental plates can be seen in medial view as triangular structures; a reabsorption pit lies at the base of the third tooth (Fig. 19B).

The surfaces of the teeth in FMNH PR 5068 are cracked and no crown tips are preserved (Fig. 19B). The crowns are bulbous relative to their roots; the crowns swell mesially, distally, lingually, and labially. Carinae are present mesially and distally, and fine serrations are present wherever carinae are clearly preserved. The fine serrations (~10/mm) taper, as in silesaurids (e.g. Silesaurus opolensis, Dzik 2003; Sacisaurus agudoensis, Ferigolo and Langer 2007) and ornithischians (e.g. Lesothosaurus diagnosticus, Galton 1978). The teeth seem to have been mesially and distally symmetrical, and unrecurved.

The combination of the shape of the maxilla, tooth attachment, and tooth shape indicate that the maxilla of FMNH PR 5068 belongs to a silesaurid dinosauriform. Given the currently uncertain relationships of silesaurids (see e.g. Martz and Small 2019, Müller and Garcia 2020), a more precise phylogenetic position of this specimen within the clade cannot be determined with the available data.

Other possible avemetatarsalian remains

A small, toothless, right dentary fragment (UA 9-6-03-781d; Fig. 20A–C) is among the reptile material from the M-13 locality. The lateral side is gently rounded, bearing anteroposteriorly elongated nutrient foramina. The dentary bears three complete and two partial alveoli, separated by clear interdental plates expressed as triangles between the alveoli in medial view. This alveolar configuration suggests that UA 9-6-03-781d had thecodont teeth and thus is assignable to at least Archosauriformes (Gauthier 1986, Nesbitt 2011, Ezcurra 2016). The medial surface is flat just ventral to the alveoli. A clear Meckelian groove lies even more ventrally. This dentary cannot be identified more precisely than as an archosauriform.

Figure 20.

Reptile remains from the type locality (M-13, ‘basal Isalo II’/Makay Formation) of Mambachiton fiandohana. Right dentary fragment (UA 9-6-03-781d) in lateral (A), medial (B), and dorsal/occlusal (C) views. Incomplete mid to posterior caudal vertebra (UA 9-11-03-803b) in anterior or posterior (D), lateral (E), and ventral (F) views. Cervical centrum (UA 9-11-03-803a) in lateral (G) and ventral (H) views. Right ilium (UA 9-10-98-538a) in lateral (I), medial (J), dorsal (K), ventral (L), and anterior (M) views. Partial?radius (UA 9-10-98-538b) in two (N, O) and distal (P) views. Distal end of a right femur (UA 9-1-98-391) in anterior (Q), medial (R), posterior (S), lateral (T), and distal (U) views. Arrows indicate anterior direction. Scales = 1 cm in D-F, I-U and = 5 mm in A-C, G, H. Abbreviations: ace, acetabulum; al, alveolus; ap, anterior process; ctf, crista tibiofibularis; dia, diapophysis; ip, ischial peduncle; lc, lateral condyle; mc, medial condyle; Mg, Meckelian groove; mr, medial ridge; no, notch; par, parapophysis; pup, pubic peduncle; r, ridge; sr1, sacral rib 1 scar; sr2, sacral rib 2 scar.

Two vertebrae were surface collected (not in situ) at the M-13 locality. UA 9-11-03-803A (Fig. 20G, H) is an anterior cervical, and the smaller UA 9-11-03-803B is a distal caudal (Fig. 20D–F).

The cervical vertebra is represented solely by a centrum, including the neurocentral suture. This cervical centrum (14 mm in length) bears a ventrally extended ridge on the ventral midline. The circular anterior and posterior articular facets of this centrum are strongly offset, lending the centrum a parallelogram shape in lateral view; a feature common in avemetatarsalians and pseudosuchians with more elongated cervical vertebrae (e.g. poposauroids; Nesbitt 2011). The lateral sides of this centrum lack fossae. Identification of the cervical centrum at a less inclusive level than Archosauromorpha is difficult given the independent acquisition of elongate cervical within archosaurs and close relatives (Nesbitt 2011, Ezcurra 2016). UA 9-11-03-803A is similar to corresponding elements of Mambachiton fiandohana; it may represent a smaller individual of this taxon. However, the size and shape also is consistent with the conditions inferred for lagerpetids (Ezcurra et al. 2020) and dinosauriforms, both also known from the same locality (see above).

The caudal vertebra (UA 9-11-03-803B; Fig. 20D–F) is slightly weathered and its neural arch is largely broken off. Proportions of this vertebra (length = 21.2 mm, centrum height = 7.8 mm) suggest that it is part of the distal caudal series, and these proportions are typical of archosaurs (e.g. the poposauroid Qianosuchus mixtus, Li et al. 2006; the dinosauriform Asilisaurus kongwe, Nesbitt et al. 2019). The size of this element indicates that it could pertain to the same individual as the holotype of Mambachiton fiandohana, but there is no direct corroborating evidence for this potential association.

A complete right ilium (UA 9-10-98-538a; Fig. 20I–M) was also recovered from the M-13 locality. It has a short anterior process and a longer posterior process, like other archosauriforms (Ezcurra 2016), and the acetabulum is deep relative to the laterally pronounced supra-acetabular crest. The supra-acetabular crest does not contact the anterior end of the pubic peduncle, and the pubic and ischial peduncles meet at a point ventral to the anteroposterior midlength of the acetabulum. The ischial articulation surface is concave in lateral view. An antitrochanter and slight rounded depression are located within the acetabulum. The anterior process of the ilium projects slightly dorsally at its tip, and the posterior portion of the ilium extends well lateral to the rest of the iliac blade. A flat shelf is present at the base of the ischial peduncle, and extends ventral to the posterior process, an area termed the brevis shelf in some archosauriforms (see above). Medially, there are two prominent scars for the attachment with sacral ribs one and two. The scar for sacral rib one has a thin anterodorsal process that contacts the medial side of the anterior process; the scar for the second sacral rib extends posteriorly, just ventral to a medially expanded ridge.

This ilium (UA 9-10-98-538a) is archosauriform, given its clear anterior process, deep acetabulum with a prominent supra-acetabular rim, and brevis shelf. Beyond archosauriform, a narrower phylogenetic placement is difficult to support, given the generally conservative ilium morphology of early-diverging pseudosuchians, avemetatarsalians, and their close allies. UA 9-10-98-538 is far too large to refer the minute lagerpetid Kongonaphon kely from these deposits, and it lacks a number of features typical of silesaurids (see below). UA 9-10-98-538a conceivably represents a much smaller individual of Mambachiton fiandohana, the blade being less than half the length (maximum = 42 mm) of the ilium (FMNH PR 5065) assigned to this taxon above. UA 9-10-98-538a and the ilium of FMNH PR 5065 are similarly proportioned and share a distinctive set of unique features, including: a ventrally flat brevis shelf, concave margin of the ischial articulation, antitrochanter plus slight rounded depression located within the acetabulum, and articulation of sacral rib one lapping onto the medial surface of the anterior process. The anterior process of UA 9-10-98-538a is more pointed than in FMNH PR 5065; however, if these two ilia represent the same taxon, this difference could be ontogenetic given their marked size difference.

The end of a limb bone (UA 9-10-98-538b; Fig. 20N–P), was found in association with and is compatible in size with UA 9-10-98-538a (Fig. 20I–M). The limb fragment lacks clearly diagnostic features, but it plausibly represents an archosaurian distal radius or ulna. Thin walled, its medullary cavity is calcite-filled; the distal end, circular in end view, bears a centralized depression.

The distal end of a small right femur (UA 9-1-98-391) also was recovered at the M-13 locality (Fig. 20Q–U). The 2.9-mm-wide shaft of this element is extremely thin-walled (~0.3 mm). The distal end widens asymmetrically relative to the shaft, the lateral side of the distal end is more expanded than the medial side. A proximodistally oriented fossa occurs on the anterior surface of the distal end. The distal end bears two condyles; the crista tibiofibularis and the lateral condyle are not differentiated. The posteromedial portion of the distal end is slightly excavated just proximal to the medial condyle, as in early-diverging avemetatarsalians (Nesbitt et al. 2017). The distal surface is distinctly rounded.

The thinness of the bone wall and slight excavation of the posteromedial portion of the distal end of UA 9-1-98-391suggest avemetatarsalian affinities, but the lack of additional synapomorphies precludes a more circumscribed placement. For example, this specimen lacks the well differentiated crista tibiofibularis seen in lagerpetids (Irmis et al. 2007, Müller et al. 2018). The specimen is far too small to belong to the holotype of Mambachiton fiandohana; the possibility that it pertains to a juvenile individual of this species cannot be ruled out, however.

DISCUSSION

Phylogenetic position

Adding Mambachiton fiandohana to the Foffa et al. (2022) phylogenetic dataset, we find the new taxon as the earliest diverging member of Avemetatarsalia (Fig. 21) (880 MPTS; 5037 steps) (Supplementary Information Figure S1), with archosaur and stem archosaur relationships identical to that of the results of Foffa et al. (2022). Placement of Mambachiton fiandohana as the earliest-diverging member of Avemetatarsalia is weakly supported (Bremer = 1). Nevertheless, several important archosaur and avemetatarsalian character states support this view. First, the general morphology of Mambachiton fiandohana is entirely consistent with that expected of an early archosaur, even though it is supported by a single apomorphy [496→1, a posteromedial tuber (=anteromedial tuber of Nesbitt 2011) on the femoral head], among all of the diagnostic attributes of Archosauria identified in the analysis (54 1→2; 187 2→3; 188 0→1; 240 0→2; 496 0→1; 529 0→1; 532 0→1; 552→1), because of its incompleteness relative to more complete craniodental and skeletal material of other taxa in the matrix.

Figure 21.

Phylogenetic relationships of Mambachiton fiandohana in a summary tree (880 MPTS; 5037 steps), see Supplementary Information Figure S1 for full tree. Filled circles = nodes, chevron = stem.

The early divergence of Mambachiton fiandohana within Avemetatarsalia is supported by a number of diagnostic character states that have been identified recently (Nesbitt et al. 2017, Ezcurra et al. 2020). Unambiguous synapomorphies supporting avemetatarsalian monophyly, and membership of Mambachiton within it, include: the presence of a spine table with the distal (=dorsal) surface of the expansion of the neural spine is convex (322→0); ratio between transverse width of diapophysis and length of the centrum in anterior trunk vertebrae < 0.70 (357 1→0); glenoid fossa orientation (of the scapulocoracoid) posteroventral (386 1→2); preacetabular process of the ilium present, being longer than two thirds of its height but not extending beyond the level of the anterior margin of the pubic peduncle (460 1→2); anterior trochanter (=lesser or minor trochanter) (=iliofemoralis cranialis muscle insertion) of the femur present and forms a steep margin with the shaft, but is completely connected to it (502 0→1); absence of a dorsal prominence on paramedian osteoderms (591 1→0); and distinct notch (=dorsal expansion) at the ventral portion of the ischial peduncle between the posterior and anterior ends of the ilium (603 0→1).

Within Avemetatarsalia, Mambachiton fiandohana certainly falls outside of the minimally inclusive clade encompassing aphanosaurs and birds, as this new Malagasy taxon lacks three apomorphies of that clade: a laterally extended crest dorsal to the supra-acetabular rim/crest is present in other avemetatarsalians (462→1; state 0 in Mambachiton fiandohana), the dorsal margin of the ilium is straight in lateral view in other avemetatarsalians (466 →1; state 0 in Mambachiton fiandohana) but clearly convex in Mambachiton fiandohana, and the absence of a longitudinal sharp ridge on the posterior edge of the scapular blade in other avemetatarsalians (602→1; state 0 in Mambachiton fiandohana).

Before aphanosaurs were recognized as early-diverging avemetatarsalians, our preliminary analyses placed Mambachiton fiandohana among the earliest diverging poposauroids. This initial result was not surprising, given that this was the placement originally determined for generally similar anatomy of Yarasuchus deccanensis and the finding of Yarasuchus deccanensis as an early-diverging poposauroid (Brusatte et al. 2008, 2010a). However, currently Yarasuchus deccanensis is considered a member of Aphanosauria at the base of Avemetatarsalia (Nesbitt et al. 2017). The position of Mambachiton fiandohana at the base of Poposauroidea is not supported in the current analysis, as no definitive poposauroid character states indicate such a potential position.

Diagnostic attributes of early-diverging avemetatarsalians

The skeleton of Mambachiton fiandohana appears to record the initial steps in the evolution of the avemetatarsalian bauplan. With the evolution of aphanosaurs, this subsequently transformed into the dinosaur bauplan, with early-diverging ornithodirans and dinosauriforms sequentially acquiring more dinosaurian-like character states. The features exhibited by Mambachiton fiandohana (detailed next) clarify the transition of important features between stem archosaurs and pseudosuchians to early-diverging bird-line archosaurs.

Hyposphene–hypantrum articulations:

The trunk vertebrae of Mambachiton fiandohana possess large, nearly parallel medial edges of the prezygapophyses, here interpreted as hypantra, and a slight ventral expansion of a bony lamina at the midline between the postzygapophyses, interpreted as a hyposphene. The shapes of these vertebral structures comport with the recently revised definitions of the hyposphene–hypantrum articulations in archosaurs (Stefanic and Nesbitt 2018, 2019). The hypantrum of Mambachiton fiandohana is identical to that of Teleocrater rhadinus (Nesbitt et al. 2018), whereas its hyposphene is simpler and lacks the lateral pockets present in Teleocrater rhadinus. Still, the hyposphene of Mambachiton fiandohana fits into the hypantrum of the proceeding vertebra, as in other archosaurs.

Mambachiton fiandohana and Teleocrater rhadinus are similar in body size, judging from the vertebrae, pectoral elements, and partial femur. Mambachiton fiandohana falls within the size range of small-bodied avemetatarsalians having hyposphene–hypantrum articulations; Mambachiton fiandohana is smaller than any pseudosuchians with such articulations (Stefanic and Nesbitt 2018). Thus, in combination, the small body size and hyposphene–hypantrum articulations further corroborate our proposed phylogenetic position of Mambachiton fiandohana among avemetatarsalians.

Loss of osteoderms:

Osteoderms occur dorsal to the cervical vertebra series, and at least the anterior trunk regions in Mambachiton. Among early-diverging avemetatarsalians, osteoderms may be present in the dinosauriform Lewisuchus admixtus (Romer 1972b, Bittencourt et al. 2015, Agnolín et al. 2022), but no other avemetatarsalian taxon has the extensive covering of osteoderms seen in Mambachiton fiandohana, except within Dinosauria, where osteoderms occur in ornithischians (e.g. Butler et al. 2008), sauropodomorphs (Currie-Rogers et al. 2011), and rarely in theropods (Hendrickx et al. 2022). Although the osteoderm shape, number per vertebra, and arrangement are unique for Mambachiton fiandohana among archosauriforms, they more closely resemble those of stem archosaurs (e.g. Euparkeria capensis, Ewer 1965) and pseudosuchians (e.g. Gracilisuchus, Lecuona et al. 2017; aetosaurs, Parker 2007) than those of other avemetatarsalians. The presence of stem archosaur-like and pseudosuchian-like osteoderms in an early-diverging avemetatarsalian was not unexpected based on prior phylogenetic hypotheses (Sereno 1991, Benton 1999, Brusatte et al. 2010a, Nesbitt 2011, Ezcurra 2016, Ezcurra et al. 2020), but Mambachiton fiandohana confirms that osteoderms characterized the earliest avemetatarsalians, but were lost immediately prior to the divergence of the lineage including aphanosaurs and other avemetatarsalians.

Sacral rib morphology:

The phylogenetic position of Mambachiton fiandohana helps decipher the evolution of sacral rib morphology among stem-archosaurs, pseudosuchians and early dinosaurs. Sacral vertebral structure has been instrumental in resolving the relationships of archosaurs and their closest relatives, (Gauthier 1986, Sereno 1991, Benton 1999, Brusatte et al. 2010a, Nesbitt 2011, Ezcurra 2016). The sacral region has provided a fruitful avenue for parsing out the evolution of the pelvic girdle across Archosauria and establishing the distinction between crocodylian and avian-line archosaurs. Specifically, the plesiomorphic condition for Archosauria, and the condition retained in most pseudosuchians, is for the sacral ribs to be large and have oval to circular articulation surfaces that contact the ilium. By contrast, in dinosaurs (e.g. Herrerasaurus ischigualastensis, Novas 1994, Langer and Benton 2006) the sacral ribs are smaller and dorsoventrally thinner; they also form divided dorsal and ventral contacts with the ilium. The breadth and shape of these contacts are well documented in early dinosaurs (Langer and Benton 2006, Nesbitt 2011), but their transformation sequence from stem archosaurs and pseudosuchians to dinosaurs are little known, owing to the lack of relevant material. The condition in dinosauriforms such as Silesaurus opolensis is more similar to that in dinosaurs than in stem archosaurs and pseudosuchians. Sacral 1 of Mambachiton fiandohana, and newly discovered material from Teleocrater rhadinus, provide a suite of new data that shed light on this transition.

The stem-archosaur condition of sacral rib 1 is exemplified by Euparkeria capensis (SAM-PK-5867). In lateral view, the rib’s articulation with the ilium is oval; its dorsal margin is flat mediolaterally. The antero- and posterodorsal edges bear ridges that originate on the neural arch and flank the rib laterally to the articulation surface with the ilium. In Mambachiton, the articular surface of sacral rib 1 is less massive than in Euparkeria; ‘I’-shaped, its anterior and posterior margins are distinctly concave. In Euparkeria capensis (SAM-PK-5867) the antero- and posterodorsal ridges of sacral rib 1 exhibit more exaggerated features than in Mambachiton fiandohana; in the former, the anterodorsal ridge extends far more anteriorly than the rest of the sacral rib, but the posterodorsal ridge lies well anterior to the more ventral portion of this rib (Fig. 14). The overall shape of the sacral rib 1 articulation surface of Mambachiton is more similar to the condition in dinosaurs (where the sacral ribs are divided dorsally and ventrally) than to thin Euparkeria capensis. However, the dorsal and ventral parts of the sacral rib are still largely connected; the condition in Mambachiton is thus intermediate between those of stem-archosaurs and other avemetatarsalians.

A number of character states of the first sacral rib of Teleocrater rhadinus (NMT RB1395) also occur in Mambachiton fiandohana and dinosaurs, but its ilial articular surface remains large and robust, like that of stem-archosaurs and pseudosuchians. Mambachiton fiandohana and Teleocrater rhadinus both have an anterior extension of the anterodorsal ridge and an anteriorly positioned posterodorsal ridge. Teleocrater, however, lacks the more I-shaped articular surface of Mambachiton, largely because it lacks the large concave surface on the posterior side of the sacral rib. Given that Teleocrater rhadinus is more closely related to dinosaurs than is Mambachiton, in our phylogenetic analysis, the morphology of sacral rib 1 in early-diverging avemetatarsalians are not perfect intermediates between stem-archosaurs plus pseudosuchians relative to dinosaurs and their closest avemetatarsalian relatives.

Avemetatarsalian assemblages in the Triassic

Although dinosaurs appeared in the Middle–Late part of the Carnian Age of the Late Triassic (Rogers et al. 1993, Brusatte et al. 2010b, Langer et al. 2010, Martínez et al. 2011, Langer et al. 2018), it is now clear that non-dinosaurian avemetatarsalians also were present in the same assemblages throughout much of the Late Triassic (Fig. 22) (Ezcurra 2006, Irmis et al. 2007). These assemblages, including lagerpetids, non-dinosaurian dinosauriforms (possibly including silesaurids), and members of Dinosauria, flourished during the Late Triassic in the higher latitudes of southern Pangea, as shown by the fauna from Ischigualasto (Martínez et al. 2013), and in the low latitudes of Pangea, as demonstrated by a number of assemblages from the Chinle Formation (e.g. Hayden Quarry, Irmis et al. 2007, Marsh and Parker 2020) and the Dockum Group (e.g. Otis Chalk Quarries, Stocker 2013, Nesbitt and Ezcurra 2015). A number of assemblages including non-dinosaurian dinosauriforms and a variety of dinosaurs occur in the higher latitudes of northern Pangea, but lagerpetids are either absent or rare in those (Marsh and Parker 2020, Foffa et al. 2022). Prior to the origin of dinosaurs, avemetatarsalian-containing faunas included lagerpetids and non-dinosaurian dinosauriforms, as recorded in the Chañares Formation of southern Pangea (Marsicano et al. 2015, Ezcurra et al. 2017). Only one type of avemetatarsalian, aphanosaurs, appears to occur in the Middle Triassic. Aphanosaurs are known from southern Pangea [lower portion of the Manda Beds (Teleocrater rhadinus, Nesbitt et al. 2017) and the lower part of the Santa Maria sequence (Spondylosoma absconditum, Nesbitt et al. 2017)], and northern Pangea (Donguz Formation, Dongusuchus efremovi, Sennikov 1988, Niedźwiedzki et al. 2016).

Figure 22.

Time calibrated phylogeny of early-diverging avemetatarsalians highlighting the co-occurrence of Mambachiton fiandohana, a lagerpetid (Kongonaphon kely), and a member of Silesauridae from the M-13 locality and nearby localities from the ‘basal Isalo II’/Makay Formation (light red) of Madagascar prior to the origin of dinosaurs. The staggered beginnings/ends of the boxes represent uncertainty in the duration of the taxon. Filled circles = nodes, chevron = stem.

Sampling elsewhere in the southern Morondava Basin, within the same geological unit as the M-13 locality (‘basal Isalo II’/Makay Formation), is still sparse, but the only crown archosaurs currently known are avemetatarsalians. No pseudosuchians have been recovered, in contrast with most other Middle to Early-Late Triassic archosauromorph-containing assemblages, making it unclear whether the southern Morondava Basin preserves a unique pseudosuchian-free archosaur assemblage, or instead that pseudosuchians were rare there and simply not yet sampled.

With co-occurrence of an early-diverging avemetatarsalian, Mambachiton fiandohana, a lagerpetid (Kongonaphon kely), a non-dinosaurian dinosauriform (likely a silesaurid, see above), and a lack of dinosaurs, the avemetatarsalian assemblage of the M-13 locality and contemporaneous strata nearby (‘basal Isalo II’/Makay Formation, southern Morondava Basin of south-western Madagascar) bridges the gap between the low diversity of avemetatarsalians in currently known Middle Triassic assemblages, to the much richer avemetatarsalian assemblages of the Early-Late Triassic (Fig. 22). Furthermore, the avemetatarsalian assemblage from the Madagascar M-13 locality demonstrates that the earliest diverging avemetatarsalians persisted in assemblages also containing non-dinosaurian dinosauriforms.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article on the publisher's website.

Figure S1. Consensus of the relationships of Mambachiton fiandohana (880 MPTS; 5037 steps) from our additions to the analysis of Foffa et al. (2022). Filled circles = nodes, chevron = stem.

ACKNOWLEDGEMENTS

We thank the Ministère de L’Enérgie et des Mines for authorizing field efforts and for facilitating our joint U.S.-Madagascar exploration, research and education programme, through our long-term collaborations between the AMNH and FMNH, Université d’Antananarivo, Ministe`re de L’Enérgie et des Mines, and ICTE/MICET (Madagascar), including at the Université d’Antananarivo: the Mention Bassins sédimentaires, Evolution, Conservation (BEC), and previously the Département de Paléontologie et d’Anthropologie Biologique. This research was supported by the Division of Paleontology of the American Museum of Natural History; National Geographic Society Grants 5957-97, 6271-98 and 7052-01, for expeditionary support (to J.J.F. and A.R.W.); and the WWF (World Wide Fund for Nature/ World Wildlife Fund), Madagascar, for expedition logistical support. S.J.N.’s research at the Field Museum of Natural History was supported by a Meeker Family Fellowship. We are grateful to Steve Goodman and Asmina Gandhi for their hospitality and generous assistance in innumerable ways. We also thank the valued members of our field crews during the years in which these specimens were collected (1997, 1998, and 2003): R. Andriatompohavana, D. Croft, J. Finarelli, A. Goswami, J.A. Rabarison, T. Rajoelisolo, H. Rakotomalala, P. Ranaivo Vavisaro, B. Randriamanpionona, H. Raveloson, M. Ravokotra, W. Simpson, G. Wesley, R. Whatley, A. Yoder, and R. Zaonarivelo; our intrepid drivers Mssrs Kose, Jean, Honore (Niri), M. Andriamahefa, J.C. Rasoanaivo, and G. Razafindrakoto; the camp chef for our large 2003 expedition, Mssr R. ‘Ledada’ Razafindrasoa; and local assistants Mena, Dada (F.D’A. ‘Dada’ Randriamanendrika), and Hari (Bemiharisoa ‘Hari’ Zafimahefa). Access to the free version of TNT 1.5 was possible because of the Willi Henning Society. We thank John Hutchinson and Oliver Demuth for sharing Euparkeria segmented material for comparison. Finally, we thank Martin Ezcurra and Hans-Dieter Sues for their constructive reviews.

Conflict of interest: None declared.

Author notes

Emily Patellos Current address: University of Southern California, Zumberge Hall of Science, Los Angeles, CA 90089-0740, USA