Skeletal Anatomy of Acaenasuchus geoffreyi Long and Murry, 1995 (Archosauria: Pseudosuchia) and its Implications for the Origin of the Aetosaurian Carapace

ABSTRACT Acaenasuchus geoffreyi is a diminutive armored archosaur from the Upper Triassic Chinle Formation of northern Arizona, U.S.A., with uncertain evolutionary relationships and skeletal maturity. Known only from osteoderms, the taxon has been considered a valid taxon of aetosaur, juvenile specimens synonymous with the aetosaur Desmatosuchus spurensis, or a non-aetosaurian pseudosuchian archosaur. Here, we describe new fossils of Acaenasuchus geoffreyi that represent cranial, vertebral, and appendicular elements as well as previously unknown variations in the dorsal carapace and ventral shield. The skull bones are ornamented with the same anastomosing complex of ridges and grooves found on the paramedian and lateral osteoderms, and the appendicular skeleton resembles that of Revueltosaurus callenderi, Euscolosuchus olseni, aetosaurs, and other armored archosaurs such as erpetosuchids. Histology of osteoderms from the hypodigm of Acaenasuchus geoffreyi shows multiple growth lines, laminar tissue, and low vascularity, evidence that the individuals were close to skeletal maturity and not young juveniles. A revised phylogenetic analysis of early archosaurs recovers Acaenasuchus geoffreyi and Euscolosuchus olseni as sister taxa and members of a new clade that is the sister taxon of Aetosauria. This new phylogeny depicts a broader distribution of osteoderm character states previously thought to only occur in aetosaurs, demonstrating the danger of using only armor character states in aetosaur taxonomy and phylogeny. Acaenasuchus geoffreyi is also a good example of how new fossils can stabilize ‘wild card’ taxa in phylogenetic analyses and contributes to our understanding of the evolution of the aetosaur carapace.


INTRODUCTION
The Aetosauria are a clade of Late Triassic pseudosuchian archosaurs best known for their interlocking dorsal carapace and ventral shield of osteoderms , and as one of the only groups of large-bodied herbivorous tetrapods during the Late Triassic Epoch (Walker, 1961;Small, 2002;Desojo and Vizcaíno, 2009). This diverse group possesses taxonomically informative osteoderms whose abundance in the fossil record has led to them being used for biostratigraphy of Upper Triassic non-marine sedimentary units, especially in the North American Southwest (e.g., the Chinle Formation and Dockum Group), and eastern North America (Newark Supergroup rift basins along Atlantic margin) (see discussions in Long and Murry, 1995;Lucas, 2010;Parker and Martz, 2011;Desojo et al., 2013;. Members of this clade include medium-sized to large (>2 m total length; Desojo et al., 2013), wide-bodied taxa (e.g., Desmatosuchus spurensis, Paratypothorax andressorum, and Typothorax coccinarum) and smaller, narrower taxa (e.g., Aetosaurus ferratus, Neoaetosauroides engaeus, and Stenomyti huangae).
Acaenasuchus geoffreyi, a diminutive taxon named for a small number of aetosaur-like osteoderms from northern Arizona, has been hypothesized to be a small-bodied aetosaur taxon, a juvenile of an existing larger-bodied aetosaur taxon, or a nonaetosaurian pseudosuchian archosaur (Long and Murry, 1995;Heckert and Lucas, 2002a;Parker, 2016a). Here, we redescribe the taxon with new material that includes the first cranial and non-osteoderm postcranial skeletal elements of Acaenasuchus geoffreyi and hypothesize its phylogenetic relationships by placing it in an analysis with aetosaurs, closely related taxa, and other early archosaurs. Charles Camp made a large collection of small vertebrates in 1926 from what he called the "meal pots" (Camp, 1924:756-757) that occur in the upper part of the Blue Mesa Member (sensu Woody, 2003;Martz and Parker, 2010;Martz et al., 2012) of the Chinle Formation in the 'Blue Hills' near St. Johns, Arizona (Fig. 1). The tiny osteoderms in this collection were not described until more specimens were found at the nearby Placerias and Downs quarries in the 1980s, and they were first identified as belonging to a juvenile Desmatosuchus spurensis (Long and Ballew, 1985). Subsequently, they were named a new taxon of diminutive stagonolepidid aetosaur (Murry and Long, 1989;Long and Murry, 1995), Acaenasuchus geoffreyi. Since then, this taxon has again been hypothesized to be a juvenile of Desmatosuchus spurensis (Lucas andHeckert, 1996a, 1996b;Estep et al., 1998, Heckert and Lucas, 1999, 2000, 2002a, a valid small-bodied aetosaur taxon different from Desmatosuchus spurensis (Hunt, 1998;Hunt and Wright, 1999;Irmis, 2005;Parker, 2005aParker, , 2005bDesojo et al., 2013), a possible sister taxon of Desmatosuchus spurensis (Harris et al., 2003;Kischlat, 2000), or perhaps not even an aetosaur at all (Parker, 2008(Parker, , 2016a. Acaenasuchus geoffreyi is not always included in anatomical phylogenetic analyses owing to its highly incomplete nature (e.g., Parker, 2016a), but when it is included it is usually recovered as a member of the stagonolepidid sub-clade Desmatosuchinae (Parker, 2007;Desojo et al., 2012;Heckert et al., 2015;Schoch and Desojo, 2016).
Anatomical modules or regions (e.g., the pectoral girdle and forelimb, the skull, the carapace, etc.) may evolve at different rates in different groups (Thompson, 1917;de Beer, 1954;Takhtajan, 1991;Rae, 1999;Clarke and Middleton, 2008;Mounce, 2013;Parker, 2016a), so taxa named for incomplete specimens that are only represented by one module often act as wild card taxa in phylogenetic analyses (Donoghue et al., 1989;Kearney, 2002;Kearney and Clark, 2003;Norell and Wheeler, 2003). Furthermore, juvenile specimens can also be recovered in phylogenetic positions different from those of adults of the same species (e.g., Fink, 1981;Tykoski, 2005;Wiens et al., 2005;Campione et al., 2013;Tsai and Fordyce, 2014;Griffin, 2018;Gee, 2020). The taxonomic and ontogenetic uncertainty surrounding Acaenasuchus geoffreyi is compounded by these problems. Here, we describe new fossils representing cranial, vertebral, and appendicular elements of Acaenasuchus geoffreyi, as well as additional osteoderms that preserve carapace variation in the taxon and re-evaluate the original hypodigm. We also perform osteohistological analysis to assess skeletal maturity of specimens belonging to the hypodigm and use this new information in a revised phylogenetic analysis of early archosaurs.
Institutional Abbreviations-AMNH, American Museum of Natural History, New York, New York, U.S.A.; CRILAR, Centro Regional de Investigaciones Científicas y Transferencia The shaded circle in C for UCMP V7308 represents the holotype locality and the dashed circle in A for Rincon Basin E represents the lost specimen mentioned in Long and Murry (1995). U-Pb detrital zircon dates modified from Ramezani et al. (2011Ramezani et al. ( , 2014 and Atchley et al. (2013). are far less common than osteoderms of Acaenasuchus geoffreyi and Vancleavea campi. Because the skeletal anatomy of Vancleavea campi is well known (Long and Murry, 1995;Parker and Barton, 2008;Nesbitt et al., 2009), we can easily distinguish its elements from those of Acaenasuchus geoffreyi.
Preparation of PEFO fossils of Acaenasuchus geoffreyi included adhering small fragments with a 1:2 solution (by volume) of Paraloid B-72 in acetone. All work on PEFO specimens was done under a Leica MZ6 binocular microscope with 10× oculars and a 1.0× objective at up to 40× magnification, or a Wild M7A with 10× oculars and a 1× objective at up to 30× magnification. Minimal matrix removal was performed with water on a soft bristled brush or with a 1/32-inch chisel-pointed carbide needle in a pin vise. In addition, minimal air abrasion was performed on some specimens with a Crystal Mark Swam-blaster MV-1 using iron powder screened through a 0.0025 inch, Tyler equivalent 250 mesh screen. The abrasion was done between 1 and 6 psi with a flow of 1-2 set on the dial. The air abrasion was performed under a Wild M65S surgical scope with 10× objectives and f = 200 mm objective at up to 25× magnification. We molded and cast certain PEFO specimens to reconstruct a vertebra; molds were made with Polytek Platsil 73-25 silicone rubber and casts were made with Polytek EasyFlo 60 polyurethane.
UCMP specimens from the holotype locality UCMP V7308 were initially prepared sometime between collection in the 1930s and publication in 1995. Select UCMP specimens were further prepared for this study using the same equipment and materials as applied to the PEFO specimens. Some of the elements from this locality displayed fresh breaks from collection or damage during storage, but also sometimes a curious weathering pattern that suggested that the material was pre-depositionally eroded or weathered along existing breaks. A good example of this is observed on the anterior bar of the holotype specimen ( Fig. 2A). Numerous needle marks were evident on the specimens from earlier attempts at preparation. In addition, various unknown adhesives and surface coatings had to be removed or modified because they obscured much of the surface anatomy. Most of the historic adhesive was relatively clear and readily soluble in acetone and smelled like a cellulose nitrate adhesive such as Devcon (previously DuPont) Duco Cement. The yellowing surface consolidant may have been a type of shellac; it was only moderately soluble in ethanol. Both of these adhesives may have been applied soon after collection, as they were commonly used for contemporary paleontological preparation (Camp and Hanna, 1937). Under the shellac, the bone had softened significantly and was very easily scratched. A ca. 5% solution of Butvar B-76 in acetone was diluted again by half and used to consolidate the bone surface before the chisel-tipped needle was used to physically thin the shellac until it could be lifted away. The bone was relatively hard and firm under the needle in areas where no historic surface consolidants had been applied. In some cases, the remaining matrix released easily from the uncoated bone. In other cases, ferrous minerals or clays remained on the surface and little surface preparation occurred in order to not damage the specimen. We did not prepare any of the previously collected MNA or SMU specimens. UMNH specimens were surface collected by the Utah Geological Survey and were not mechanically prepared or treated with any consolidants.
Two osteoderm specimens (UCMP 175103 and UCMP 175114) from the type locality were selected for osteohistological analysis. The bones were photographed, molded using silicone rubber, and cast in epoxy resin prior to sectioning. Histological sampling and slide preparation were conducted at UCMP and followed the methods of Lamm (2013). Both bones were vacuum-embedded in US Composites Silmar-41 clear polyester resin and allowed to cure for at least 72 hours. 1-1.5 mm-thick sections were cut using a Buehler IsoMet 1000 saw and were affixed to amp, anteromedial projection; ap, anterior projection; as, articulating surface; de, dorsal eminence; g, groove; k, keel; ls, lateral spike; pf, posterior fossa; tr, transverse ridge. Scale bars equal 1 cm. petrographic glass slides using two-part two-ton epoxy. Using progressively finer grit sizes, the sections were ground with silicon carbide papers on a Buehler Ecomet 3 grinder-polisher. UCMP 175103, a paramedian osteoderm, was longitudinally sectioned along an anterior-posterior axis through the dorsal eminence. UCMP 175114, a lateral osteoderm, was sectioned both longitudinally through the anterior-posterior axis of the main body of the element, and transversely through the lateral spike. The completed histological slides were imaged in plane and cross-polarized light using a Zeiss Axio Imager.M2m petrographic microscope with accompanying Zen 2 v2.0.0.0 software at the University of Utah.
Referred Specimens-Incomplete lists of specimens referred to Acaenasuchus geoffreyi are found elsewhere (Long and Murry, 1995;Heckert and Lucas, 2002a;Polcyn et al., 2002), and a complete list of specimens is found in Appendix 1 and more detailed information is found in Supplementary Data 1. Surface files of selected PEFO specimens are available in Supplementary Data 2 at www.morphobank.org (MorphoBank project P3406). All referred specimens from PEFO, MNA, and UCMP were observed first-hand.
Removed Specimens-The following specimens have been referred to Acaenasuchus geoffreyi (pp. 212-213 in Heckert and Lucas, 2002a), but either lack the character states that diagnose the taxon (see below) or are lost. When appropriate, we provide re-identification of these elements in parentheses. UCMP 27049 (Metoposauridae); UCMP 175104, UCMP 175136, UCMP 175140, UCMP 175140, UCMP 175142, and UCMP 175143 (Revueltosaurus sp.);UCMP 139584, UCMP 139584, UCMP 139585, UCMP 156046, and UCMP 175121 (lost); the UCMP Rincon Basin E specimens (Long and Ballew, 1985;Long and Murry, 1995) may be two uncatalogued osteoderm fragments from a mis-routed loan at the MNA that are in a box labeled 'Tovar Mesa, Winslow.' Revised Diagnosis-Acaenasuchus geoffreyi can be distinguished from all other pseudosuchian archosaurs by the following autapomorphies: small horn present on the posterior process of the squamosal (Fig. 3E); transverse process of trunk vertebrae ornamented distally (Fig. 3M); rounded tubercle present on the lateral surface of the ilium between the preacetabular process and supraacetabular crest (Fig. 4H); dorsal eminence of the paramedian osteoderms forms a "thorn-like process" (Long and Murry, 1995:114;Fig. 2F) or posteriorly curved spike; triangular boss of lateral osteoderm forms a posteriorly curved spike (Fig. 2J); and presence of external lateral osteoderms (an osteoderm ventrolateral to the 'internal' lateral osteoderm but dorsal to the ventral osteoderms ( Fig. 5H, L). Acaenasuchus can further be distinguished by the following unique combination of character states: anterior edge of anterior margin of some paramedian osteoderms is straight (shared with Longosuchus meadei, Lucasuchus hunti, Sierritasuchus macalpini, and Desmatosuchus), dorsal eminence of paramedian osteoderms does not contact posterior margin of the osteoderm in most rows (shared with Desmatosuchus spurensis), lateral osteoderms in cervical region have a slightly recurved spine rather than an elongated horn like those of Desmatosuchus spurensis and some typothoracines (shared with Lucasuchus hunti).
Several high-precision detrital zircon U-Pb ages from PEFO and the St. Johns region constrain the highest and lowest occurrences of Acaenasuchus geoffreyi in Arizona (Fig. 1). The Placerias Quarry (UCMP A269) in the lower part of the Sonsela Member is constrained by a maximum depositional age 219.39 ± 0.12 Ma and the bone-bearing horizon at PFV 449 (the Salado Site) directly overlies a sandstone dated to a maximum depositional age of 221.6 ± 1.4 Ma (Ramezani et al., 2011(Ramezani et al., , 2014. The specimens from the upper part of the Blue Mesa Member at PEFO are constrained by maximum depositional ages of 223.036 ± 0.059 Ma (TPS sample) and 220.123 ± 0.068 Ma (SS-7 sample) (Ramezani et al., 2011;Atchley et al., 2013). However, recent U-Pb age-calibrated magnetostratigraphy from PEFO suggests that ages from the lower part of the Sonsela Member and upper part of the Blue Mesa Member may be biased by redeposited zircons that are significantly older than the depositional age (Kent et al., 2018(Kent et al., , 2019

Cranium and Mandible
Maxilla-A partial left maxilla is preserved in PEFO 43699 ( Fig. 3I, J). It includes part of the tooth-bearing posterior process and three alveoli, two of which contain nearly complete teeth. The lateral side of the maxilla is ornamented with longitudinal grooves below the antorbital fossa and irregular ridges and pits above it (Fig. 3I). A low longitudinal ridge is present on the lateral surface, but it is not as prominent as that of Revueltosaurus callenderi (PEFO 34561) and some aetosaurs such as Aetosaurus ferratus (Schoch, 2007); Stagonolepis robertsoni (Walker, 1961) and Longosuchus meadei (TMM 31185-98) (Nesbitt, 2011). The anteromedial process of the maxilla is short and projects ventral to the alveolar margin. The ventral margin of the antorbital fenestra is concave up and forms the dorsal edge of the element.
The teeth are labiolingually compressed and are mesiodistally widest near their apicobasal midpoint (Fig. 3J). Apically, the teeth taper to a rounded tip. The teeth are similar to the maxillary teeth of Revueltosaurus callenderi (PEFO 34561) in that they contain broad denticles along the mesial and distal carinae; however, the teeth of Acaenasuchus geoffreyi are not as labiolingually wide basally as those of Revueltosaurus callenderi. The anatomy of the maxillary teeth is variable in aetosaurs, from the mediolaterally compressed, curved teeth found in Aetosauroides scagliai (Brust et al., 2018) to the bulbous teeth of Desmatosuchus smalli (Small, 2002). The maxillary teeth of Typothorax coccinarum (PEFO 38001/YPM VP.58121) include a labiolingually compressed leaf-shaped morphology more similar to what is found in Revueltosaurus callenderi and Acaenasuchus geoffreyi except that in T. coccinarum defined denticles are lacking on these teeth (Reyes et al., in review).
Jugal-The lateral surface of the jugal (UCMP 285903) is heavily ornamented with anastomosing ridges and corresponding pits. It is anteroposteriorly long (Fig. 3G Small, 2002). A sharp longitudinal ridge occurs on the lateral side of the jugal, this feature is also present in some early theropod dinosaurs and most suchians except rauisuchids, which possess a rounded ridge (Nesbitt, 2011). The longitudinal ridge on the jugal of Acaenasuchus geoffreyi lies above a shallow, ventral-facing groove.
Squamosal-The squamosal of Acaenasuchus geoffreyi (UCMP 285837) in dorsolateral view is only slightly longer anteroposteriorly than it is wide mediolaterally, and does not taper posterodorsally as it does in erpetosuchids ( Fig. 3E; Nesbitt and Butler, 2013;, Revueltosaurus callenderi (PEFO 34561), and the aetosaurs Desmatosuchus smalli (TTU-P9024; Small, 2002), Stenomyti huangae (DMNH 60708; Small and Martz, 2013:fig. 4), and Aetosaurus ferratus (SMNS 5770 S-16 and SMNS 5770 S-18; Schoch, 2007:fig. 8a). The dorsal surface is ornamented with thin ridges and small pits that together form larger ridges that radiate anteriorly and ventrally from the articulation with the parietal. The articular surfaces for the quadratojugal and the postorbital are slotted on the ventromedial surface of the bone. The articular facet for the paroccipital process of the opisthotic is broad and similar to that of Revueltosaurus callenderi (e.g., PEFO 34561) and aetosaurs (Nesbitt, 2011). The posterior margin of the squamosal is concave, with a small horn that projects posterodorsally.

Vertebral Column
Trunk Vertebrae-The only non-sacral vertebrae that can be confidently referred to Acaenasuchus geoffreyi belong to the dorsal series, which we call 'trunk vertebrae,' that are consistently the same size, and include fragments of centra, neural arches, zygapophyses, and transverse processes (  (UCMP 124548, UCMP 124551, UCMP 192558). These vertebrae preserve the parapophysis and the diapophysis on the neural arch/transverse process and lack chevron facets. The centrum is amphicoelous, and both cotyles are strongly concave (most of the UCMP specimens are worn flat). Shallow longitudinal depressions occur on the lateral sides of the centrum just under the neural arch. The centrum and the neural arch are strongly integrated into one another and there is no external trace of a suture in any of the well-preserved specimens. Trunk vertebrae neurocentral sutures close relatively later in ontogeny in pseudosuchians but timing can vary (Brochu, 1996;Irmis, 2007), so either Acaenasuchus geoffreyi reached skeletal maturity early in ontogeny or all the known specimens represent skeletally mature individuals. The neural arch is extremely flat and table-like (Fig. 3M, S), much like that of Euscolosuchus olseni (USNM 448584; Sues, 1992), and is approximately as mediolaterally wide as it is anteroposteriorly long (excluding the transverse processes). Four discrete laminae extend from the base of the transverse process to the side of the neural arch: the anterior centrodiapophyseal lamina, the posterior centrodiapophyseal lamina, the prezygadiapophyseal lamina, and the postzygadiapophyseal lamina (Wilson, 1999). The anterior centrodiapophyseal and posterior centrodiapophyseal laminae are very close to one another (Fig. 3Q) and form a small, circular centrodiapophyseal fossa (Wilson et al., 2011). The two centrodiapophyseal laminae join to form a prominent ventral strut that extends laterally down the length of the transverse process.
The mediolateral length of the transverse process is much greater than the anteroposterior length of its respective centrum ( Fig. 3M-P), much like the broad transverse processes of some doswelliids and erpetosuchids (Ezcurra, 2016), a condition that may have helped support the armor carapace. This may be functionally similar to the condition in the trunk of stagonolepidid aetosaurs (e.g., Parker, 2008:fig. 11;Parker, 2016b:fig. 14), where there is a large amount of variation in the relative contribution of the transverse process and rib to the support structure (e.g., Parker, 2016a:character 40). The distal end of the transverse process is ornamented with small pits and longitudinal grooves (e.g., PEFO 40687,PEFO 40703,PEFO 40712,and PEFO 40723), similar to but not to the extent observed on the ribs of Euscolosuchus olseni (USNM 448590; Scheyer and Sues, 2016); there are no trunk ribs preserved in PFV 211 that unambiguously belong to Acaenasuchus geoffreyi, but the direct articulation of the distal end of the transverse process and proximal end of the rib suggest that these two structures were of similar anteroposterior width. The transverse process terminates in the offset circular, concave articular facets of the diapophysis and the parapophysis. The presence of the diapophysis and the parapophysis completely on the transverse process of the trunk vertebrae differs significantly from the condition found in most non-crocodylomorph pseudosuchians where the parapophysis is situated on the neural arch close to the base of the process. The pairing on the transverse process is found on in Acaenasuchus geoffreyi, all aetosaurs (e.g., Desmatosuchus spurensis, UMMP 7476), and Alligator (e.g., Chiasson, 1962). The functional significance of this has not been determined, but its distribution is convergent among these three taxa.
In Acaenasuchus geoffreyi the prezygapophyseal, centrodiapophyseal, and postzygapophyseal centrodiapophyseal fossae are subtriangular and deep. The prezygapophyses (and the postzygapophyses) lie very close to one another across the midline of each vertebra, unlike those of Euscolosuchus olseni, in which the zygapophyses extend anteriorly/posteriorly and laterally on tapering processes (USNM 448584;Sues, 1992: fig. 2b). The neural arch contains bilateral accessory articular structures medial to the prezygapophysis and postzygapophysis (best preserved in UCMP 192558; Fig. 3S, T). They are similar to hyposphene/hypantrum articulations found in some aetosaurs (Desmatosuchus spurensis, MNA V9300; Scutarx deltatylus, PEFO 34045) and other archosaurs (see Stefanic and Nesbitt, 2019), but they do not occur on the midline. The anterior component is a dorsal subtriangular fossa found medial to the prezygapophysis and the posterior component is a ventral subtriangular process medial to the postzygapophysis. Small, circular fossae occur on the midline between both pairs of zygapophyses just below the level of the base of the neural spine. The neural spine on the trunk vertebrae projects up from the middle of the flat dorsal surface of the neural arch of Acaenasuchus geoffreyi (similar to that of Doswellia kaltenbachi; USNM 244214; Dilkes and Sues, 2009: fig. 4a; Ezcurra, 2016) and is robust yet dorsoventrally short, much like that of Euscolosuchus olseni (USNM 448584; Sues, 1992: fig. 2a).
Sacral Vertebrae-Most of the morphology of the first primordial sacral vertebra can be reconstructed from portions attached to fragmentary sacral ribs (PEFO 38761, PEFO 40704, and UCMP 285840), partial sacral centra (PEFO 40723), and half of a sacral vertebra (Fig. 3U, V;UCMP 192558). These specimens are identified as the first primordial sacral by the rounded distal end of the sacral rib that does not extend posterolaterally (p. 115 of Nesbitt, 2011). The centrum is amphicoelous and dorsoventrally short. An anteroposteriorly long depression separates the centrum from the sacral rib in ventrolateral view. The anterior face of the centrum does not project anteriorly past the anteriormost extent of the sacral rib. The sacral rib is robust and is situated only on the anterior half of the centrum; it is waisted near the centrum where anterior and posterior fossae are present at its base but is anteroposteriorly expanded distally. In lateral view, most of the sacral rib is round in outline except for the posterior edge, which tapers posteriorly. The neural arch is low, and the zygapophyses are only separated from each other by a thin midline slot. The neural spine is very short dorsoventrally, and it terminates in a slightly concave 'table' that is bifurcated anteriorly and slopes ventrally posteriorly (Fig. 3U, V). Flat neural spine 'tables' are also present on the precaudal vertebrae of a number of reptile lineages, including tanystropheids, phytosaurs, ornithosuchids, Revueltosaurus callenderi, aetosaurs, stem-paracrocodylomorphs, and dinosaurs (Nesbitt, 2011;Pritchard et al., 2015;Marsh and Rowe, 2018;, but they are mediolaterally narrower than those of Acaenasuchus geoffreyi (UCMP 192558) and Euscolosuchus olseni (USNM 448584; Sues, 1992: fig. 2a). A concave dorsal surface of the neural spine is also present in erpetosuchids .
Humerus-Preserved specimens of the humerus include complete proximal and distal ends but lack the mid-diaphysis PEFO 40693,PEFO 40704,and PEFO 40739). Three bulbous structures are observed on the head of the humerus in proximal view (Fig. 4F); the median humeral head is larger than the lateral and medial tuberosities. The medial tuberosity is the largest of these in Revueltosaurus callenderi (e.g., PEFO 34561). The medial corner of the proximal end of the humerus is gently expanded in anterior view like in some aetosaurs (e.g., Aetosauroides scagliai, PVL 2073; Parker, 2016a) but unlike the prominent medial deflection observed in Stagonolepis olenkae (ZPAL AbIII 1175; Lucas et al., 2007:fig. 5e;Parker, 2016a), Longosuchus meadei (TMM 31185-84b; Long and Murry, 1995), and Revueltosaurus callenderi (PEFO 34561). A semicircular fossa is present just beneath the articular surface of the humeral head in anterior view, similar to that present in Revueltosaurus callenderi (PEFO 34561). The deltopectoral crest is fairly short proximodistally and projects anterolaterally to form a rounded subtriangular outline in lateral view. The ectepicondylar flange of Acaenasuchus geoffreyi completely encloses the ectepicondylar foramen like it does in most aetosaurs (except Aetosaurus ferratus; SMNS 5770 S-5; Schoch, 2007:fig. 10f;Parker, 2016a). This feature is found on some individuals of Revueltosaurus callenderi (e.g., PEFO 34561) but in specimens the flange does not completely close and forms an ectepicondylar groove (e.g., PEFO 34269). A subcircular 'cuboid fossa' is present on the distal end of the humerus in anterior view (Langer et al., 2007), and the radial and ulnar condyles are subequal in size in distal view.

Osteoderms
The hypodigm material described by Long and Murry (1995) represents the 'typical' aetosaur-like arrangement of paramedian and lateral osteoderms and all specimens exhibit the characteristic dorsal sculpturing and ornamentation of anastomosing ridges and pits. Overall, the paramedian osteoderms are rectangular and mediolaterally wider than anteroposteriorly long in dorsal view (e.g., the holotype specimen UCMP 139576; Fig. 2A, B; Long and Murry, 1995:figs. 117a, b, 118k, l) and the lateral osteoderms are subrounded or subhexagonal (e.g., Fig. 2J-M; Long and Murry, 1995:figs. 117d, 118a). Long and Murry (1995) inferred how osteoderm shape varied relative to body position in their diagnosis of Acaenasuchus, but did not actually assign specific specimens to specific anatomical regions in the subsequent description.
During our examination of the original hypodigm and comparison of these specimens with the new PEFO fossils, we noted extensive variation in shape, size, and articulation patterns for the osteoderms of Acaenasuchus geoffreyi. Based on anteroposterior carapacial variation trends observed in stagonolepidid aetosaurs (e.g., Parker, 2005bParker, , 2007, this allows us to classify Acaenasuchus osteoderms by region. Ideally, carapacial regions should be determined by which vertebrae are roofed by specific osteoderms (e.g., Long and Ballew, 1985;Desojo et al., 2013), but in the absence of vertebra-osteoderm association in Acaenasuchus geoffreyi, we designate regions using comparisons with similar regionalized osteoderms in aetosaurs and Revueltosaurus callenderi. Further support for hypothesized carapacial regions comes from new PEFO specimens that are either articulated or co-ossified to one another. The cervical region is characterized by shorter paramedian osteoderms that interlock with a pair of co-ossified lateral osteoderms (Fig. 6A, B), the trunk region includes the 'typical' paramedian + lateral osteoderm pair with the widest paramedian osteoderms in the carapace (not co-ossified; Fig. 6A, B, D), and the pelvic/anterior caudal region includes rings of interlocking serial rows (Fig. 6A, B, E).
Some features are shared across osteoderms despite the serial position within the carapace. The dorsal ornamentation of the paramedian and lateral osteoderms is similar (and similar to the ornamented skull bones described above), consisting of subcircular to oblong pits radiating away from the dorsal eminence or lateral spike, with relatively more subcircular pits closer to the dorsal eminence or lateral spike ( Fig. 2A, J, F). Each subcircular pit contains a foramen. The dorsal eminence and lateral spike are covered in foramina. This pattern is very similar to that of Doswellia kaltenbachi (USNM 244214; Dilkes and Sues, 2009: fig. 10) and Euscolosuchus olseni (e.g., USNM 448587), but different from Revueltosaurus callenderi (e.g., PEFO 34561) in which the depressions are more circular and evenly spaced (e.g., PEFO 34561) and aetosaurs, which generally have either more widely spaced circular pits (Typothorax coccinarum) or a pattern of elongate grooves and pits (e.g., Desmatosuchus spurensis, Calyptosuchus wellesi) (Long and Ballew, 1985;Parker, 2008). The ventral surface of the osteoderms of Acaenasuchus geoffreyi is mostly smooth except for the occasional foramen or rugose patch on the posterior margin that articulates with the dorsal surface of the anterior bar of the subsequent osteoderm (Fig.  2G, K). These rugose patches are also observed on the osteoderms of Euscolosuchus olseni (e.g., USNM 448587). The lateral margins of the paramedian osteoderms are sigmoidal. Medially, a paramedian osteoderm articulates with the adjacent paramedian osteoderm, and laterally articulates with at least one lateral osteoderm (see below); both articulation surfaces form a tongue-in-groove system (Fig. 2C, D, J, H, I). The medial edge of a paramedian osteoderm is flat where it articulates with another paramedian osteoderm, but the lateral surface is concave to articulate with the convex medial surface of the lateral osteoderm. Each paramedian osteoderm has a dorsal eminence that projects posterodorsally into a spike or "thorn" (Long and Murry, 1995:114); some dorsal eminences curve posteriorly in medial/lateral view (Fig. 2H), whereas others remain relatively straight and dorsally erect (PEFO 40742). A dorsal ridge is present on the eminence that extends from the point of the spike to the posterior margin of the anterior bar. Unlike stagonolepidid aetosaurs, in which the dorsal eminence is often positioned either near the mediolateral mid-point or closer to the medial margin of the osteoderm (Parker, 2007), in Acaenasuchus geoffreyi it is positioned either near the mid-point (e.g., UCMP 139576) or closer to the lateral edge of the paramedian osteoderm (e.g., PEFO 40000). This observation is confirmed by specimens of paramedian osteoderms articulated with lateral osteoderms (PEFO 40740; Fig. 6A). This is similar to the dorsal eminence of Revueltosaurus callenderi (e.g., PEFO 34561) and Euscolosuchus olseni (USNM 448587 and USNM 448582), in contrast to the condition in paramedian osteoderms of stagonolepidids. A transverse ridge extends medially and laterally away from the posterior margin of the dorsal eminence and terminates before reaching either margin of the element ( Fig. 2A). An elongate posterior fossa is formed behind this ridge and under the dorsal eminence. Each paramedian osteoderm has a raised anterior bar like that of Revueltosaurus callenderi and most stagonolepidid aetosaurs (Parker, 2007(Parker, , 2016a, in contrast to the depressed 'anterior lamina' of Desmatosuchus spurensis (MNA V9300; Parker, 2008), as well as the non-raised anterior bar of Euscolosuchus olseni (e.g., USNM 448587, USNM 448587, and USNM 448582). As mentioned above, the anterior bar of the holotype UCMP 139576 was abraded or prepared away (inset, Fig. 2A), appearing to represent a depressed lamina, which had been used to argue that Acaenasuchus geoffreyi was a juvenile Desmatosuchus spurensis (Heckert and Lucas, 2002a), but does not reflect the true condition observed throughout the carapace. The lateral osteoderms also possess a raised anterior bar (Fig.  2J, M) and a large lateral spike projecting dorsolaterally from the middle or lateral half of the osteoderm (Fig. 2J, L).
The lateral spike is triangular in dorsal view (Fig. 2J), hexagonal in cross section (Fig. 2M, F), and has a longitudinal groove posteriorly that is similar to the fossa behind the dorsal eminence and transverse ridge of the paramedian osteoderms (Fig. 2L). The overall similarity of the lateral osteoderms to those of stagonolepidids is the major reason why Acaenasuchus geoffreyi was originally assigned to the Aetosauria, because prior to this study lateral osteoderms of that type in pseudosuchians were only known from aetosaurs (e.g., Long and Ballew, 1985). Our new material demonstrates that although the lateral osteoderms from some regions do appear very similar to those of stagonolepidid aetosaurs in style of articulation and 1:1 correspondence with the paramedian osteoderms, in other areas of the carapace they have different articulations, including in the cervical and caudal regions forming 'armor bands.' Cervical Region-Both paramedian and lateral osteoderms are present in the cervical region, and when they are articulated, the width of the rows tapers anteriorly (Figs. 5A, 6A). In articulated osteoderms from the cervical region (Fig. 5A-D) the lateral osteoderms are co-ossified to one another (Fig. 5B). These lateral elements articulate closely with the respective paramedian osteoderms along an interdigitating mediolateral articular surface, as well as interdigitating articular surfaces on the anterior bar of the paramedian osteoderm and ventral surface of the lateral osteoderms (Fig. 5A). In dorsal view, the lateral margin of the cervical paramedian osteoderms is angled anteromedially and forms a subtrapezoidal outline (unlike the subrectangular outline of trunk paramedians), causing the anterior taper of the lateral margin of the osteoderm rows in the cervical region (Fig.  5A). This is more similar to what is present in Revueltosaurus callenderi (PEFO 34561), and very different to what is present in aetosaurs. Non-desmatosuchine aetosaurs have cervical osteoderms that are mediolaterally wider than anteroposteriorly long with a posteriorly tapering lateral edge. This is because there is an anterolateral extension of the anterior bar that overlies the dorsomedial corner of the corresponding lateral osteoderms (Parker, 2007). The trapezoidal shape in Acaenasuchus geoffreyi more closely resembles that of desmatosuchine aetosaurs such as Desmatosuchus (Parker, 2005b(Parker, , 2008 and Sierritasuchus macalpini  in which the osteoderms taper anteriorly; however, they are different in that these osteoderms in aetosaurs are anteroposteriorly longer than they are mediolaterally wide. The cervical lateral osteoderms of Acaenasuchus geoffreyi are slightly convex and do not form strong angles like those in the trunk region. Instead, the paramedian/lateral osteoderm pair is gently dorsally arced over the neck (Fig. 5C, D). The ventrolateral margin of the cervical lateral osteoderms is tapered and does not articulate with another osteoderm. The dorsal surfaces of the anterior bars are lightly ornamented with low bumps and small, shallow pits, differing from the smooth bars found in Revueltosaurus callenderi and stagonolepidid aetosaurs.
Trunk Region-The paramedian osteoderms in the trunk vary in mediolateral width and include the widest paramedians of the carapace (Figs. 2A, E, F, 6A, B). For example, UCMP 139574 is 1.75 times mediolaterally wider than it is anteroposteriorly long, and for PEFO 40694 the ratio is 2.3. The ornamentation on the dorsal surface of the anterior bar is not as prominent as that of the cervical paramedians and the rows of osteoderms in the trunk are not co-ossified. The anterior margin of the trunk paramedian osteoderms is straight in some specimens (e.g., UCMP 139576) and scalloped in others (e.g., PEFO 40000; Parker, 2018a). Two small projections extend anteriorly from the anterior bar medial to the midpoint; one projection is halfway between midpoint of the osteoderm and the medial margin (similar to aetosaurs, Parker, 2016a), and the other projection occurs on the anteromedial corner of the element ( Fig. 2A, F). The former projection is located more laterally in Revueltosaurus callenderi (e.g., PEFO 34561). The trunk paramedian osteoderms of Acaenasuchus geoffreyi lack the anterolateral projection observed in non-desmatosuchine stagonolepidid aetosaurs, which articulates with the lateral osteoderm (Parker, 2016a). The anterolateral corner of Acaenasuchus geoffreyi paramedian osteoderms is square, owing to the interdigitating suture with the lateral osteoderm. In contrast, the anterolateral corner of the holotype paramedian osteoderm of Euscolosuchus olseni is a long, tapered spine that projects anteriorly (USNM 448587). Both taxa differ from the condition in desmatosuchine aetosaurs where the anterolateral corner of the paramedian osteoderm is slightly excavated to receive a small process from the corresponding lateral osteoderm (Parker, 2007(Parker, , 2016a. The lateral osteoderms in the trunk region of Acaenasuchus geoffreyi are not co-ossified medially with a paramedian osteoderm but articulate tightly with it across an interdigitating suture (Fig. 2J), and do not deviate from the general description above. In posterior view, the lateral osteoderm curves ventrally to form a right angle (Fig. 2L). The angle that the lateral spike projects from the lateral osteoderm varies serially along the trunk among well-preserved specimens, but without more articulated specimens with unambiguous positional data, we cannot describe how the angle changes anteroposteriorly. The ventral margin of the lateral osteoderm tapers to a non-articulating edge. Based on the condition observed in Acaenasuchus, we interpret Euscolosuchus specimen USNM 448587 as co-ossified trunk paramedian and lateral osteoderms, in which the lateral spine of Euscolosuchus olseni is equivalent to the lateral spike on lateral osteoderms of A. geoffreyi, the dorsal eminences are homologous, and the 'dorsal keel' of Euscolosuchus olseni is autapomorphic for that taxon (Sues, 1992). This hypothesis needs further testing via a transverse histological section or computed tomography.
Sacral/Caudal Region-Each half of a row of sacral/anterior caudal osteoderms includes a paramedian osteoderm, an internal lateral osteoderm, and an external lateral osteoderm that articulates with the ventrolateral margin of the internal lateral osteoderm ( Fig. 5E-H, L, M). In fact, two rows of osteoderms articulate with one another to form rings similar to those found in Pleistocene glyptodonts (Lydekker, 1894;Carlini et al., 2008: fig. 3f;Zurita et al., 2013:fig. 2j;Arbour and Zanno, 2018: fig. 1) and the anterior caudal region of some stem turtles (Gaffney, 1985:figs. 19, 21;Gaffney, 1990:figs. 85-88, 131-132;Sterli and de la Fuente, 2011:fig. 13). When articulated, these rings taper posteriorly over the sacrum/anterior caudal region (Fig. 5A). Each ring includes two paramedian osteoderms that articulate with one another anteroposteriorly in an interdigitating articular surface (Fig. 5M, Q), unlike the cervical and trunk regions in which the anterior bar articulates with the ventral surface of the preceding osteoderm. In each sacral/anterior caudal ring, the anterior margin of the anterior paramedian osteoderm and the posterior margin of the posterior paramedian are non-articulating surfaces (Fig. 5P). The lateral side of the paramedian osteoderm is beveled to articulate with two lateral osteoderms (Fig. 5N, O, S), unlike the paramedians in the cervical and trunk regions that only articulate laterally with a single lateral osteoderm. In posterior view (Fig. 5H, M), the sacral/anterior caudal paramedian osteoderms are curved ventrolaterally over the body more so than those in the cervical and trunk regions. The lateral osteoderms in the sacral/anterior caudal region have tall, hexagonal lateral spikes that curve slightly anteriorly (Figs. 5F, K, 6A, B), similar to the spines on the sacral/caudal paramedian osteoderms of Rioarribasuchus chamaensis (Parker, 2007). Only partial external lateral osteoderms exist, but they are smaller than the internal lateral osteoderms, have the characteristic ornamentation found on the other osteoderms of Acaenasuchus geoffreyi, and have a small spike extending from near the lateral edge of the osteoderm (Fig. 5F, L). One specimen, UCMP 139582 ( Fig. 6I-K), is a pair of fused anterior and posterior internal lateral osteoderms. The entire complex of paramedian, internal lateral, and external lateral osteoderms in the sacral/anterior caudal region is more robust than the osteoderms in the cervical and trunk regions, in which the osteoderms are dorsoventrally thinner. In some specimens (e.g., PEFO 40702) these rings are tightly articulated, and in others (e.g., PEFO 38737 UCMP 139582, and UCMP 175112) each osteoderm in the two-row ring is co-ossified to neighboring osteoderms.

Additional Records
Our redescription of Acaenasuchus geoffreyi focuses on material from the type locality and referred specimens from nearby sites in northeastern Arizona, U.S.A. (Fig. 1). From collections made in the 1980s in the lower Monitor Butte Member (Chinle Formation) near the Blue Lizard Mine (UCM loc. 88067) in Red Canyon, southeastern Utah, U.S.A. (Parrish and Good, 1987;Dubiel, 1987). Parrish (1999: fig. 4) figured and briefly described four small osteoderms (UCM 76194) that he assigned to an indeterminate archosauriform, but compared closely with the early archosauriform Doswellia. These specimens possess the distinctive ornamentation of Acaenasuchus geoffreyi, with at least one of them preserving a dorsal eminence that does not reach the posterior margin, and a smooth anterior bar (Parrish, 1999: fig. 4). Additional material was collected from nearby sites at the same stratigraphic level by Harvard University in 1986, 1997, and 2003; these uncatalogued specimens were observed by one of us (R.B.I.) in 2007 to include additional Acaenasuchus osteoderm material.

Osteohistology
Our main goal in examining the bone histology of Acaenasuchus is to assess its skeletal maturity, given that some previous authors have explicitly hypothesized that this material represented juvenile specimens of the stagonlepidid aetosaur Desmatosuchus (Heckert and Lucas, 2002a). We selected osteoderms (Fig. 8) for our analysis because: (1) they are the most plentiful element in the known hypodigm; and (2) previous studies demonstrate that osteoderms from both extant and extinct pseudosuchian archosaurs record valuable skeletochronological data (e.g., Hutton, 1986;Games, 1990;Woodward and Moore, 1992;Tucker, 1997;Erickson and Brochu, 1999;Cerda and Desojo, 2011;Cerda et al., 2013Cerda et al., , 2018Taborda et al., 2013;Scheyer et al., 2014). This previous work also showed that longitudinal sections through the dorsal eminence (or equivalent) provide one of the most complete records of growth (Hutton, 1986;Tucker 1997;Cerda and Desojo, 2011;Taborda et al., 2013;Cerda et al., 2018), so we followed this strategy in our work (Fig. 8A-D) in sampling one paramedian osteoderm (UCMP 175103) and one lateral osteoderm (UCMP 175114).
Both osteoderms are dominated by compact cortical bone, with a small number of large vascular endosteal spaces (Fig. 8B, D, F), in contrast with the large number of small vascular spaces in stagonolepidid aetosaurs Cerda and Desojo, 2011;Scheyer et al., 2014;Cerda et al., 2018). This cortical bone comprises parallel-fibered tissue organized into thin laminae or layers (Fig. 8B, D, F), but does not possess the 'crossed' parallel-fibered variant observed in other aetosaurs (Cerda et al., 2018). Overall, the cortex is largely avascular, with only a few large simple canals and primary osteons oriented perpendicular to the plane of section (mediolaterally within the osteoderm). The ornamented surface of the osteoderms is underlain by undulating sets of dense laminae; though this is similar to the condition in other aetosaurs, Acaenasuchus lacks any evidence of the 'cut and fill' structure caused by resorption and deposition between sets of laminae that is observed in stagonolepidid aetosaurs (e.g., Cerda and Desojo, 2011:figs. 3a, b, 4c;Scheyer et al., 2014: figs. 6d, 7, 9, 10;Cerda et al., 2018) andcrocodyliforms (e.g., de Buffrénil, 1982;Hua and de Buffrénil, 1996). As with other pseudosuchians (Hutton, 1986;Games, 1990;Cerda and Desojo, 2011;Scheyer et al., 2014;Cerda et al., 2018), growth marks are particularly apparent in the cortex of the ventral portion of the osteoderm (Fig. 8B, D). These are most apparent in the longitudinal section of UCMP 175114, which displays a minimum of seven growth marks (Fig. 8D); if interpreted as lines of arrested growth (LAGs), they would indicate a minimum age of seven years. In UCMP 175103, growth marks are also present, but it is more difficult to distinguish them from prominent laminae (Fig. 8B).
Overall, the combination of parallel-fibered bone with laminae and sparse simple vascular canals is an indicator of relatively slow skeletal growth (e.g., Francillon-Vieillot et al., 1990;de Margerie et al., 2002de Margerie et al., , 2004Cubo et al., 2008). This is not entirely unexpected, as growth rate scales positively with body size (e.g., Case, 1978), and so one would expect a small-bodied animal (i.e., Acaenasuchus) to grow slower than a larger one (e.g., many stagonolepidid aetosaurs). The relatively avascular compact structure with a small number of large endosteal spaces may also be a function of body size, as this condition is also present in the small-bodied pseudosuchian Revueltosaurus callenderi (Scheyer et al., 2014), though Cerda et al. (2018) hypothesized it reflected the plesiomorphic character state for Aetosauria and its sister groups. Therefore, histological indicators of slow growth cannot be used alone to infer skeletal maturity stage. That said, several observed osteohistological characters do provide evidence for the ontogenetic stage of these FIGURE 7. Acaenasuchus geoffreyi, representative specimens from the Red Canyon area of southeastern Utah, U.S.A. A, UMNH VP 30185, partial paramedian osteoderm in dorsal view. B, C, UMNH VP 30186, partial paramedian osteoderm in B, dorsal and C, lateral views. D-F, UMNH VP 30184, partial lateral osteoderm in D, ventral, E, anterior, and F, posterior views. G-K, UMNH VP 30183, caudal vertebra in G, dorsal, H, ventral, I, anterior, J, posterior, and K, right lateral views. Abbreviations: as, articulating surface; cf, chevron facets; de, dorsal eminence; g, groove; ls, lateral spike; ns, neural spine; poz, postzygapophysis; tp, transverse process. Scale bars equal 5 mm.
Acaenasuchus specimens. The three growth marks closest to the external surface of UCMP 175114 (Fig. 8D) are much closer together than the preceding internal marks, suggesting that skeletal growth is slowing. The anterior and posterior margins of the osteoderms, where bone appositional rate is greatest (Cerda and Desojo, 2011:fig. 5;Taborda et al., 2013:fig. 4;Cerda et al., 2018: fig. 9), lack the condition of dense longitudinal simple canals observed in juvenile stagonolepidid aetosaurs R.B.I., pers. observ.). Though the sampled Acaenasuchus osteoderms possess relatively little internal cancellous bone compared to stagonlepidids, the larger endosteal spaces do show evidence of remodeling, with resorption lines external to secondary bone lining the circumference of each opening (Fig. 8B, D, F). This remodeling of endosteal vascular spaces is common in older archosaur osteoderms (e.g., de Buffrénil, 1982;Cerda and Desojo, 2011;Cerda et al., 2013Cerda et al., , 2018Scheyer et al., 2014).
Thus, the balance of histological evidence suggests these osteoderms do not belong to a fast-growing young juvenile of a larger-bodied taxon (e.g., Desmatosuchus). Rather, they appear to be from a sub-adult ontogenetic stage where growth is beginning to slow. All of the Acaenasuchus material reported in this paper is within the same general size class as the sampled osteoderms, suggesting that our ontogenetic interpretations are representative for the sample.

PHYLOGENETIC ANALYSIS
We added 16 characters (nos. 420-435) from Parker (2016a) and ten novel characters (nos. 436-445) to the morphological character-taxon data matrix constructed by Nesbitt (2011) as modified by Butler et al. (2014) and Nesbitt et al. (2017). We also added additional states to five existing characters (422, 424, 427, 432, 408), rescored Revueltosaurus callenderi for select characters (171,219,268,383,406), rescored Aetosaurus ferratus for a character (375), and added Desmatosuchus spurensis and Aetosauroides scagliai to the matrix. We scored Acaenasuchus geoffreyi and Euscolosuchus olseni into the character-taxon matrix using Mesquite v3.10 (Maddison and Maddison, 2018) such that the final matrix includes 91 taxa and 445 characters (only the holotype specimens of Teleocrater rhadinus, Poposaurus gracilis, Prestosuchus chiniquensis, Lewisuchus admixtus, and Asilisaurus kongwe were used; Supplementary Data 3, 4; Mor-phoBank P3406). Characters 32,52,75,121,137,139,156,168,188,223,247,258,269,271,291,297,328,356,399,and 413 were ordered in the analysis (as they were in the parent data matrices), which was conducted using a heuristic tree search in TNT v1.5 (Goloboff et al., 2008) with 1,000 replications, random sequence addition, and tree bisection reconnection swapping while keeping ten trees per replication and condensing zerolength branches. A strict consensus tree was computed in TNT from all recovered most parsimonious trees (MPTs). Bootstrap resampling analyses were run using 1,000 replications with replacement. Parsimony-based ancestral state reconstruction was performed on the strict consensus tree using Mesquite to examine character state evolution.
The heuristic search recovered 227 MPTs with a length of 1,414 steps, a consistency index (CI) of 0.368, and a retention index (RI) of 0.769. The strict consensus tree of all MPTs (Fig. 9) recovers largely the same relationships as Nesbitt et al. (2017) and Butler et al. (2017) among non-archosaur archosauromorphs, avemetatarsalians, and non-aetosaurian suchians. Figure 9 displays an abbreviated version of the strict consensus tree (with larger clades collapsed into single branches for brevity) that also displays the GC bootstrap and Bremer values for each node. Within Pseudosuchia, Ornithosuchidae is the earliest diverging clade, and sister taxon to Suchia, in contrast with the topology recovered by Ezcurra et al. (2017) and Müller et al. (2020), in which ornithosuchids are the sister taxon of Erpetosuchidae and along with aetosaurs comprise a clade within Suchia (see von Baczko et al. [2020] for a synthesis of hypothesized ornithosuchid and erpetosuchid relationships within Pseudosuchia). Though poorly supported, our analysis recovers a monophyletic group of non-paracrocodylomorph suchians in which Erpetosuchidae (sensu Nesbitt and Butler, 2013) is the sister taxon to a well-supported clade that comprises Revueltosaurus callenderi Hunt, 1989, Acaenasuchus geoffreyi Long and Murry, 1995, Euscolosuchus olseni Sues, 1992, and Aetosauria (sensu Parker, 2007. The sister taxon to the stem-based group Aetosauria is the clade comprising Revueltosaurus callenderi + (Acaenasuchus geoffreyi + Euscolosuchus olseni) (Fig. 9). The representative aetosaur taxa in our analysis are recovered in the same relative position as found by Parker (2016a), with Aetosauroides scagliai as the most basal aetosaur. Enforcing the constraint of including ornithosuchids + erpetosuchids along with Aetosauria as a clade of suchians (sensu Ezcurra et al., 2017, Müller et al., 2020 requires an additional seven steps. Pulling Revueltosaurus callenderi just outside of the Acaenasuchus geoffreyi + Euscolosuchus olseni clade (making Revueltosaurus callenderi the sister taxon of Aetosauria, sensu Nesbitt, 2011 andParker, 2016a) requires one additional step.
The clade Acaenasuchus geoffreyi + Euscolosuchus olseni is diagnosed by four unambiguous apomorphies: ornamented dorsal surface of the anterior bar of paramedian and lateral osteoderms; anteroposteriorly broadened transverse processes of trunk vertebrae; dorsal surface of neural arch of trunk vertebrae is nearly as mediolaterally wide and anteroposteriorly long; and anteroposteriorly broadened proximal ends of trunk ribs. The clade Revueltosaurus callenderi + (Acaenasuchus geoffreyi + Euscolosuchus olseni) is diagnosed by two unambiguous apomorphies (although neither can be scored for Euscolosuchus olseni): frontal tapers anteriorly along the midline; presence of enlarged denticles on the maxillary teeth.
Using our phylogenetic hypothesis (Fig. 9) as a framework to understand character transformations along the lineage leading to Aetosauria, we can reconstruct the evolution of the aetosaur body plan from more cursorial carnivorous pseudosuchian ancestors. The common ancestor of erpetosuchids and aetosaurs was still likely a small-bodied cursorial carnivore but possessed two paramedian rows of rectangular osteoderms with anastomosing ornamentation (Benton and Walker, 2002;Nesbitt and Butler, 2013;Nesbitt et al., 2018b;Ezcurra et al., 2017). The common ancestor of Revueltosaurus callenderi and aetosaurs evolved cervical and dorsal paramedian osteoderms that were mediolaterally wider than anteroposteriorly long as well as the presence of ventral osteoderms (Parker et al., 2005). This lineage also began to evolve a less carnivorous diet, as reflected in dental and jaw character states (e.g., un-curved teeth with enlarged denticles, shortened tooth row, dorsoventrally enlarged surangular; Parker et al., 2005). Other changes towards the aetosaurian condition included a laterally rotated squamosal, and a more robust forelimb, with an enlarged scapulacoracoid and a mediolaterally wide but proximodistally short humerus. The common ancestor of Revueltosaurus callenderi and aetosaurs also evolved one or more extensive rows of lateral osteoderms that articulate with the lateral edge of the paramedian osteoderms, enlarged transverse processes and dorsal ribs to support this expanded carapace, a dorsoventrally taller but anteroposteriorly shorter ilium, and more robust hind limb elements. To the exclusion of Revueltosaurus callenderi and its closest relatives (Acaenasuchus geoffreyi and Euscolosuchus olseni), aetosaurs further modified the feeding apparatus by losing teeth in the anterior portion of the dentary and modifying the snout and dentary into distinct shapes . Given the presence of ankylothecodonty in ornithodirans and non-archosaur archosauromorphs, and now in Acaenasuchus geoffreyi, an early diverging pseudosuchian, it suggests that either ankylothecodonty is ancestral for archosaurs and later lost independently in both pseudosuchians and ornithodirans, or gomphosis is ancestral for Archosauria and ankylothecodonty is regained independently in lineages such as silesaurids and Acaenasuchus geoffreyi. We note that Revueltosaurus callenderi (e.g., PEFO 34561) also exhibits ankylothecodont tooth implantation, possibly demonstrating this as a plesiomorphic state for the clade (Revueltosaurus callenderi (Euscolosuchus olseni + Acaenasuchus geoffreyi)).
The recognition that the rectangular ornamented osteoderms that characterized aetosaurs are also found outside of this clade calls into question the ability to use this type of osteoderm to assign fossils that consist of only isolated osteoderms to the Aetosauria. This was already suspected to some degree because aetosaur-like osteoderms are known from Revueltosaurus callenderi (Parker et al., 2005) and is supported further with the determination of Acaenasuchus geoffreyi and Euscolosuchus olseni as non-aetosaur suchians. Thus, isolated osteoderms with character states like the presence of an anterior bar from Upper Triassic deposits are best assigned only to Suchia, unless they possess the unambiguous synapomorphies or unique combination of character states of a known aetosaur taxon (see Parker, 2016a and the references therein). The same recommendation applies to isolated lateral osteoderms, as well. Even though the ornamented anterior bars of non-stagonolepidid aetosaurs could help differentiate them from stagonolepidids, this ornamentation is sometimes lost through erosion (e.g., the holotype specimen of Acaenasuchus geoffreyi, UCMP 139576; Fig. 2A).
In the cases of both Revueltosaurus callenderi and Acaenasuchus geoffreyi it is also important to note that the material was thought to belong to juvenile specimens of previously known taxa (Long and Ballew, 1985;Lucas, 2002a, 2002b). These identifications had been contested (e.g., Irmis, 2005;Parker, 2005aParker, , 2008 based on apomorphies of the osteoderms; however, it required the discovery of associated non-osteoderm remains to provide a final conclusion (Parker et al., 2005; this study), especially when all of the taxa are recovered as disarticulated elements in the same deposits (Heckert and Lucas, 2002a). Histological evidence has since supported the presence of a number of juvenile aetosaur specimens, including those of Aetosauroides scagliai (e.g., MCP 13; Taborda et al., 2013), Aetosaurus ferratus (e.g., SMNS 5770-S16; Taborda et al., 2013), and Coahomasuchus chathamensis (e.g., NCSM 23618; Hoffman et al., 2019).
Other than the overall morphology of the osteoderms, the character states primarily used to identify Acaenasuchus geoffreyi as a juvenile of Desmatosuchus spurensis were the presence of thickened 'tongue-and-groove' lateral and medial articulation surfaces with adjacent osteoderms, and the presence of a depressed anterior lamina rather than a raised anterior 'bar' (both characters following the definition of Long and Ballew, 1985). An 'anterior lamina' was only observed in the holotype specimen whereas all referred Acaenasuchus geoffreyi specimens have a clearly raised anterior bar, and we can now demonstrate the holotype is badly weathered and the beveled anterior surface is an alteration from this weathering and over-preparation. The interdigitating 'tongue and groove' articulations appear to be a convergence with desmatosuchine stagonolepidids; however, they are also fundamentally different between Acaenasuchus geoffreyi and Desmatosuchus spurensis. In both lineages the surfaces are dorsoventrally thickened relative to the rest of the osteoderm, but in Acaenasuchus geoffreyi the lateral and medial faces are often vertical and flat, whereas in Desmatosuchus spurensis they are very complex with more elongate lateral projections and deeper hollows creating a more interlocking suture. Furthermore, in Desmatosuchus spurensis these articulation surfaces are only observed in the cervical and anterior trunk regions (Parker, 2005b(Parker, , 2008, whereas in Acaenasuchus geoffreyi they occur throughout the carapace. In addition, Acaenasuchus geoffreyi also possesses sets of anterior and posterior articular surfaces differing significantly from what is present in aetosaurs. In fact, it was recognition of these different sutural patterns in the osteoderms that originally led to the hypothesis that Acaenasuchus geoffreyi was not a stagonolepidid aetosaur, supported by subsequent discovery of associated nonosteoderm material. Recent work is refining our understanding of the phylogenetic relationships among early archosaurian groups (e.g., Butler et al., 2014;Stocker et al., 2016;Nesbitt et al., 2017), further emphasizing the importance of apomorphy-based identifications in vertebrate paleontology (e.g., Nesbitt and Stocker, 2008;Lessner et al., 2018;Pritchard and Sues, 2019). Because different skeletal modules (e.g., cranium, carapace) can possess conflicting phylogenetic signals (see discussion in Parker, 2016a), taxa known from only a single module can act as wildcard taxa in a phylogenetic analysis (e.g., Kearney and Clark, 2003) grouping with taxa with similar modules. For example, when Acaenasuchus geoffreyi is scored for only osteoderm characters, it is recovered within the clade Stagonolepididae (e.g., Parker, 2007Parker, , 2016aDesojo et al., 2012;Heckert et al., 2015). Pritchard and Sues (2019) recently discussed this phenomenon in regard to the archosauromorph Teraterpeton hrynewichorum. If Teraterpeton is scored for character states observed in the ilium it would be recovered as a rhynchosaurian, and the fifth metatarsal as a tanystropheid; however, a total evidence scoring from a relatively complete specimen provides its current placement within Allokotosauria (Pritchard and Sues, 2019). Thus, new fossils and character state determinations can stabilize these wild card taxa, allowing them to contribute more fully to analyses. In the present case the discovery of additional material of Acaenasuchus geoffreyi clarifies the relationships of that taxon as a member of a newly recognized sister clade of aetosaurs that includes the pseudosuchians Revueltosaurus callenderi and Euscolosuchus olseni.

CONCLUSIONS
The relationships of Acaenasuchus geoffreyi and its taxonomic validity have long been debated due to a lack of non-osteoderm anatomical data and an evaluation of its ontogenetic stage. Our study uses cranial and non-osteoderm postcranial elements as well as osteohistological analysis to revise the skeletal anatomy of Acaenasuchus geoffreyi, better constrain its ontogeny status, and demonstrate that it is a valid small-bodied taxon belonging to a clade that is the sister taxon to Aetosauria. The analysis of non-osteoderm skeletal modules stabilizes the phylogenetic relationships of Acaenasuchus geoffreyi, and our inclusion of newly coded characters for Revueltosaurus callenderi and Euscolosuchus olseni recovers a close relationship to Acaenasuchus and aetosaurs. Our analysis of pseudosuchian relationships shows that erpetosuchids, Revueltosaurus callenderi, Acaenasuchus geoffreyi, Euscolosuchus olseni, and aetosaurs form a monophyletic clade diverse in diet, body size, and osteoderm morphology. The recognition of a new clade of non-aetosaurian armored suchians forms the phylogenetic framework for future studies in the temporal, biogeographic, and trophic evolution of this clade during the Late Triassic Epoch and provide new insights into the evolution of the aetosaur body plan.

ACKNOWLEDGMENTS
We thank D. and J. Gillette (MNA), P. Holroyd (UCMP), and C. Levitt-Bussian (UMNH) for providing collections and archive access, and for facilitating loans from their institutions. M. Polcyn (SMU) graciously sent us locality information and photographs of SMU 75403. H.-D. Sues provided photographs of and helpful comments on Euscolosuchus olseni, and J. Strotman facilitated the molding, casting, and transfer of casts of USNM 448587 and USNM 448582. P. Holroyd (UCMP) graciously provided permission to histologically section two osteoderms, and A. Lee (Midwestern University) kindly assisted with sectioning. K. Ritterbush (University of Utah) generously provided access to her petrographic microscope imaging system, and the histological slides were imaged by N. Ong. R. Long's field notes are on file in the archives at PEFO and the UCMP and those of Charles Camp are available at the UCMP. Thanks to T. Olson for finding MDM specimens at PFV 211. We thank J. Kirkland and D. DeBlieux (Utah Geological Survey) for permission to include in this study material they collected. The collection and preparation of the specimens from PEFO was funded by the Petrified Forest Museum Association and Friends of Petrified Forest National Park. Silhouettes from phylopic.org in