New partial dentaries of amphitheriid mammalian Palaeoxonodon ooliticus from Scotland, and posterior dentary morphology in early cladotherians

We describe two partial dentaries of mammals from the Middle Jurassic of Scotland. They belong to the early cladotherian Palaeoxonodon ooliticus . These dentaries comprise the first specimen of P. ooliticus ever found—although its significance was initially unrecognised so it remained undescribed until now—and the most recently discovered specimen, found during fieldwork in 2017. The new specimen preserves part of the coronoid process of the dentary, previously unknown for P. ooliticus , demonstrating the presence of a deep masseteric fossa, with a prominent crest enclosing the fossa anteriorly, and a masseteric foramen, located in the masseteric fossa on the buccal surface of the dentary. On the lingual surface, the mandibular foramen is offset from the Meckel’s sulcus, and positioned below the alveolar plane. These morphologies allow an updated analysis of the phylogenetic position of P. ooliticus , confirming a sister-taxa relationship between Palaeoxonodon and Amphitherium . The position of the mandibular foramen, and the slight extension of the masseteric fossa into the body of the dentary are new autapomorphies for Palaeoxonodon .


Introduction
The Middle Jurassic was a key period in the evolutionary history of mammals.Recent research suggests a global radiation of early mammals beginning in the Early Jurassic, and entailing large increases in diversity and phenotypic disparity (Luo 2007;Close et al. 2016).This gave rise to docodontans, shuotheriids, australosphenidans, multituberculates, and cladotherians in the Early to Middle Jurassic (Luo et al. 2002;Kielan-Jaworowska et al. 2004).Continued fossil discoveries are critical to understanding this radiation.Furthermore, Cladotheria includes Theria, the clade comprising living marsupials and placentals and their stem lineages.Therefore, early cladotherian fossils provide information on the ancestral morphologies of the group, from which the more derived structures of extant therians evolved.
The early cladotherian Palaeoxonodon was originally reported based on isolated teeth from the Middle Jurassic of England (Freeman 1976a(Freeman , b, 1979;;Sigogneau-Russell 2003), and its detailed phylogenetic affinities remained uncertain.An almost complete dentary from the Isle of Skye, Scotland (NMS G. 2015.17.1), described by Close et al. (2016), revealed that the lower tooth row of Palaeoxonodon shows a gradient of variation in several dental characters that were used to delimit species in previous taxonomy.Based on these new observations, a previously erected Palaeoxonodon species, Palaeoxonodon freemani, and the genus Kennetheridium (Sigogneau-Russell 2003), were both synonymised with the type species of Palaeoxonodon, Palaeoxonodon ooliticus.This dentary also provided additional characters for a more informative phylogenetic analysis, supporting previous work that suggested Palaeoxonodon is the sister taxon to

Editors' choice
Amphitherium (Close et al. 2016; but see also Averianov et al. 2015 for an alternative hypothesis).
Here, we report two further specimens of Palaeoxonodon from the Isle of Skye.A newly discovered specimen, NMS G.2017.37.1, was found during fieldwork in 2017.Although it is not as complete as that found by Close et al. (2016), it preserves the base of the coronoid process and adjacent regions of the dentary.These features were missing from previous specimens and allow additional characters to be scored for phylogenetic analysis.Another specimen, NMS G.1992.47.123, was recovered during fieldwork by Michael Waldman and Robert J.G. Savage in 1973.It had previously been mentioned as an undescribed pantotherian by Clemens (1986) and Evans et al. (2006).The discovery of this specimen pre-dates all other known specimens of Palaoxonodon, including the holotype of P. ooliticus (Freeman 1976a).However, for reasons unknown, it was never described.After being transferred posthumously from the late RJG Savage of the University of Bristol to the National Museum of Scotland, it was rediscovered by EP in the collections at NMS in 2016.

Material and methods
The two specimens described here are both in the collections at the NMS: NMS G.1992.47.123 andNMS G.2017.37.1.Both were collected from the Kilmaluag Formation of the Great Estuarine Group (Bathonian, Middle Jurassic) of the Strathaird Peninsula, north of Elgol, Isle of Skye, Scotland.NMS G.2017.37.1 was collected at Cladach a' Ghlinne, whereas NMS G.1992.47.123 was collected approximately 0.8 km south of Cladach a' Ghlinne along the coastline.The Kilmaluag Formation comprises dolomitised blue-grey limestone, interbedded with calcareous siltstone and shale.The presence of freshwater gastropods and bivalves (Viviparus, Neomidon, and Unio), and abundant freshwater ostracods, such as Darwinula and Theriosynoecum, support the environmental interpretation that this was a freshwater lagoon environment (Andrews 1985;British Geological Survey 2011).The strata yielding the new specimens are frequently referred to as "the vertebrate beds".Also known from this formation are fossil crocodylomorphs, squamates, choristoderes, salamanders, fishes, mammals, and tritylodontids, as well as rare archosaur remains (Waldman and Savage 1972;Evans et al. 2006;Close et al. 2016;Panciroli et al. 2017a, b).
Micro-CT scan data were obtained for NMS G.1992.47.123 using the PCT scanner built in-house at the University of Edinburgh, School of Geosciences Experimental Geoscience Facility.The scanner comprises a Feinfocus 10-160kV dual transmission/reflection source, Perkin Elmer XRD0822 amorphous silicon x-ray flat panel detection, and terbium doped gadolinium oxy-sulfide scintillator.Data acquisition software was written in-house, and scans were reconstructed using Octopus 8.7 software.The scan resolution for this scan is 20 μm.NMS G.2017.37.1 micro-CT scan data were obtained at Cambridge Biotomography Centre (http:// www.cbc.zoo.cam.ac.uk/) using a Nikon Metrology XT H 225 ST micro-CT scanner.The scan resolution for this scan is 7.64 μm.All microCT scans were digitally reconstructed, image processed, and measured using Mimics 19.0 at the National Museum of Scotland.Digital reconstructions are available in SOM (Supplementary Online Material available at http://app.pan.pl/SOM/app63-Panciroli_etal_SOM.pdf).
Lower molar terminology follows that by Sigogneau-Russell (2003) with the following amendments: the talonid cusp is renamed the hypoconid, and the crest running from the hypoconid to the metaconid is the cristid obliqua (as in Davis 2011); the paraconal sulcus is renamed the hypoflexid (as in Close et al. 2016); cusp e is not present in Palaeoxonodon.
Phylogenetic analysis was carried out using TNT 1.5, with settings as in Close et al. (2016): all analyses were carried out by first using the new technology search with default command settings, and stipulating that the shortest tree was to be found 10 times.This was followed by TBR branch swapping on the recovered most parsimonious trees (MPTs), and then calculating a strict consensus.Jackknife resampling statistics were calculated using 1000 replicates, and values are displayed as absolute frequencies.Bremer support values were obtained by TBR branch swapping on the MPTs, incrementally increasing suboptimal trees from one to eight steps (see SOM for TNT script and amended data matrix).
Description.-NMSG.1992.47.123 is a fragment of left dentary containing m2, m3, and m4 in position, with roots, and the anterior part of the alveolus for m5 (Fig. 1).The molar crowns were broken and detached from their roots, and the latter remained inside the dentary fragment.The crowns have been restored to their original positions digitally to facilitate description and comparison.The dentary fragment measures 2.5 mm in anteroposterior length, 0.62 mm in width at the widest point, and the dentary dorsoventral depth below m3 is 1.3 mm (for tooth measurements see Table 1).
All three crowns of NMS G.1992.47.123 are well preserved, except for the following damage: the protoconid, metaconid, and paraconid cusp tips of m4 are missing; the hypoconid tip and tip of cusp f of m3 are missing; and the hypoconid and buccal apical portion of the protoconid are missing from m2 (Fig. 1).The surface of the paraconid of m3 is damaged, missing enamel on the lingual surface of the paraconid, metaconid, and cingulid, and some of the anterobuccal enamel surface is also missing apically on the protoconid.The cristid obliqua of m3 is well preserved, revealing a distinct incipient talonid basin in occlusal view (Fig. 1C), defined lingually by the entocristid.The incipient talonid basin of m4 is well preserved, with a clear entocristid on the lingual edge.
NMS G.2017.37.1 is a fragment of left dentary 7.2 mm long with an approximate dorsoventral depth of 1.6 mm below m3 (Fig. 2) (for tooth measurements see Table 1).The teeth m3 and m4 are preserved in position.Most of m2 is also preserved in position, but was fragmented prior to discovery.A fragment of the m1 root is also present.The alveolus for m5 is present but empty (Fig. 2B 1 ).The cusp tips are missing from both m3 and m4, but the protoconid, paraconid, and metaconid positions are clear on both molars, as are the hypoflexid, hypoconid, and incipient talonid basin, defined lingually by the entocristid.None of these features are easily identified in m2, in which the hypoconid and hypoflexid are missing, and only the base of the anterior half of the crown is present.
Multiple fractures are present on the dentary, but the general osteology is evident.The Meckel's sulcus is preserved on the lingual surface of the dentary, extending an-Table 1. Dental measurements (in mm) for Palaeoxonodon ooliticus Freeman, 1976b, NMS G.1992.47.123 and NMS G.2017.37 teriorly from a point just ventral to the mandibular foramen, up to the fractured region ventral to m2 (Fig. 2C 1 ).The mandibular foramen is present posteriorly on the lingual surface of the dentary, posteroventral to the empty alveolus for m5 and offset from the Meckel's sulcus.The pterygoid shelf is unclear from the preserved morphology.
The masseteric fossa, located posteriorly on the buccal surface of the dentary, is deep and is defined by ridges an-terodorsally and ventrally.A masseteric foramen is located anteroventrally within the masseteric fossa.The anterior margin of the masseteric fossa is well-developed and extends anteriorly, ventral to the alveolus for m5.It extends just slightly into the body of the mandible ventral to m5 (Fig. 2A 1 , B 1 ).
Stratigraphic and geographic range.-Middle to Late Batho nian, England and Scotland.

Phylogenetic analysis
We reran the phylogenetic analysis of Close et al. (2016), which was based on the matrix of Zhou et al. (2013), updating scores for characters of the posterior portion of the dentary present in NMS G.2017.37.1.These included the position of the mandibular foramen, the morphology of the anterior margin of the masseteric fossa, the presence of the masseteric foramen, and the presence of a coronoid scar (see Appendix 1 for newly scored characters and SOM for full matrix and TNT script).We also re-evaluated and amended six characters present in NMS G. 2015.17.1 that were either not scored by, or that we interpreted differently from, Close et al. (2016).These include confirming the absence of Crompton's groove, the absence of a premolar or precanine diastema, confirming the presence of the hypoconid, and the morphology of the postcanine roots.Additionally, we re-evaluated and rescored six characters present in Amphitherium, Peramus, and Arguimus (see Appendix 1 for justification for rescoring, and Zhou et al. 2013 for full character list).
The phylogenetic analysis recovered 5376 most parsimonious trees.Despite the additional character scores, we found no change in the relationships between Palaeoxonodon, Amphitherium, and Peramus from previous analyses, with Palaeoxonodon and Amphitherium placed in a sister-group relationship.These relationships have weak Bremer support values of 1, and this clade (Amphitheriidae) forms an unresolved polytomy with Nanolestes, Arguimus, and the clade including Kielantherium + Aegialodon + Theria, as in Close et al. (2016).
Five autapomorphies were recovered for Palaeoxonodon by our phylogenetic analysis, the first three of which were already identified by Close et al. (2016), and the last two of which are identified for the first time here: (i) the posteriormost mental foramen being positioned in the canine/incisor region of the dentary (more anteriorly than in Amphitherium or Peramus); (ii) the absence of a mesio-lingual cingular cuspule e; and (iii) an absent/weak ectoflexus in the second upper molar; (iv) the slight extension of the masseteric fossa onto the body of the mandible; and (v) the position of the mandibular foramen offset from the Meckel's sulcus.Palaeoxonodon and Amphitherium share synapomorphic characters in the presence of a narrow labial stylar shelf of the penultimate upper molar, and the more labial position of the metacone relative to the paracone.Palaeoxonodon, Amphitherium, and Peramus share the following synapomorphies: the convergence of the Meckel's sulcus with the ventral border of the mandible; and possessing open rooted postcanines.

Discussion
Most previous cladistic analyses support a sister group relationship between Palaeoxondon and Amphitherium (Prothero 1981;Close et al. 2016; but see Averianov et al. 2015 for an alternative view).In these previous analyses this was based on (i) the large talonid (accounting for one-third of molar length), (ii) possessing a buccally positioned principal hypoconid, (iii) roots that are mostly equal in diameter along their length except for a slight narrowing distally, and (iv) the possession of five premolars with (v) moderately well-developed lingual cingulids.Close et al. (2016) identified the variably present cuspule on the cristid obliqua (mesoconid) in their specimen NMS G. 2015.17.1 as a feature shared with the holotype NHMUK PV M36508 (and two other specimens according to Sigogneau-Russell 2003: NHMUK PV MJ530 and NHMUK PV MJ618) from the Forest Marble Formation.The mesoconid is not present in either of our specimens, NMS G.1992G. .47.123 or NMS G.2017.37.37.1.Therefore we support the interpretation of this character as variable among specimens of Palaeoxonodon.Close et al. (2016) also suggested that the different dental characters used to erect the species P. freemani and Kennetheridium leesi on the basis of isolated teeth (Sigogneau-Russell 2003) are the morphological differences between tooth positions in the same tooth row of their specimen, relegating the latter two taxa to junior subjective synonyms of P. ooliticus.This character variation includes differences in talonid length, prominence of the cingulid, size of the paraconid, and height and cusp sharpness of the trigonid, among different molars in the same individual.Similarly, we see some of this variation in NMS G.1992.47.123: for example the paraconid is much smaller in m3 and m4 than in m2, and slopes at a shallower angle in m2; and the cingulid is more prominent in m2 than m3 or m4.We can add that the entocristid is more prominent in m2, and that the protoconid is broader both anteroposteriorly and buccolingually in m4.This adds to the evidence for variation along the tooth row, and emphasises the need for caution in erecting new species based on isolated teeth.
A posterior portion of the dentary including the base of the coronoid process, masseteric fossa, and mandibular foramen are preserved in NMS G.2017.37.1 and have not previously been preserved in any other specimen of Palaeoxonodon.The preserved morphology suggests a sloping coronoid process, as in other early cladotherians.Strong masseteric ridges and a deep masseteric fossa are also present, as in other early cladotherians including other amphitheriids such as Amphitherium (Butler and Clemens 2001), other peramurids, dryolestids, and paurodontids (Martin 1999;Kielan-Jaworowska et al. 2004;Davis 2012).Unlike paurodontids, the dentary is relatively slender.The preserved morphology of NMS G.2017.37.1 suggests that a pterygoid shelf may have been present on the lingual surface of the dentary; however, its extent is not clear from this specimen.The pterygoid shelf is strongly protruding in many other early cladotherians (Martin 1999;Kielan-Jaworowska et al. 2004).The anterior position of the posteriormost mental foramen, in the canine/ incisor region of the dentary, is recovered here as an autapomorphy of Palaeoxonodon.While there is a possibility that a more posterior foramen may have been present in the frac-tured portion of the dentary below p1 in NMS G. 2015.17.1, we find no evidence for this at present, pending the recovery of new specimens.
Close et al. ( 2016) identified a possible coronoid scar in NMS G. 2015.17.1, which we can confirm in NMS G.2017.37.1.This is absent in Peramus according to previous analyses (Clemens and Mills 1971;Davis 2012), but was coded as present by Zhou et al. (2013).We have rescored this character as absent in Peramus in our analysis.The coronoid scar is present in Amphitherium (Butler and Clemens 2001) and many other early cladotherians, including the zatherian Nanolestes (Martin 2002).However, it is absent in another zatherian, Arguimus (Lopatin and Averianov 2006).Close at al. ( 2016) also inferred a lower incisor count of four based on comparison to Amphitherium.Although we cannot verify this feature using NMS G.1992.47.123 or NMS G.2017.37.1, we consider it likely to have been present, as this character state is common among early Cladotheria.As such, we have used it in our reconstruction (Fig. 3C).
The presence of a Meckel's sulcus is common among stem mammals.In NMS G.2017.37.1, the mandibular foramen is preserved, indicating that Palaeoxonodon had a Meckel's sulcus that was offset from the foramen, a feature not shared with either Amphitherium (Butler and Clemens 2001) or Peramus (Clemens and Mills 1971;Davis 2012).The mandibular foramen in Palaeoxonodon also does not appear to be confluent with the pterygoid shelf, unlike in Amphitherium and Peramus, although damage to this part of the dentary in NMS G.2017.37.1 makes this interpretation uncertain.The position of the mandibular foramen ventral to the alveolar plane, seen in Palaeoxonodon and Amphitherium, is also shared with other early cladotherians, and the basal metatherian Sinodelphys.The clade comprising Palaeoxonodon + Amphitherium + Peramus is in a polytomy with Nanolestes, the clade comprising Kielantherium + Aegialodon + Theria, and Arguimus.Arguimus shares few dentary character states with Palaeoxonodon.Arguimus has no Meckel's sulcus (the absence of Meckel's sulcus is an autapomorphy of Arguimus; see Lopatin and Averianov 2006) and no coronoid scar.Vincelestes also has no Meckel's sulcus or coronoid scar (Bonaparte and Rougier 1987).While other analyses have recovered the relationship of Vincelestes as closer to Theria than either Peramus or Amphitherium ing Palaeoxonodon + Amphitherium + Peramus, Nanolestes, Arguimus, and Kielantherium + Aegialodon + Theria.The phylogenetic affinities of Palaeoxondon have been contested due to the presence of an "incipient basin" (Sigogneau-Russell 1999, 2003) on the talonid, referred to as the incipient talonid basin herein, and the variable presence of a cuspule on the cristid obliqua.In an alternative hypothesis, Averianov et al. (2015) recovered a close relationship between Amphibetulimus and Palaeoxonodon, placing them in a clade with Nanolestes.This clade was found to be phylogenetically distinct from, and crownward to, Amphitherium (Averianov et al. 2015).Support for this relationship with Nanolestes, rather than with Amphitherium and Peramus was based on the absence of a "retromolar space", the proportional size of the stylocone relative to the metacone in the upper molars, and absence of a coronoid scar.We now confirm the coronoid scar is in fact present in Palaeoxonodon.The analysis by Averianov et al. (2015) also included Mozomus, which was found to form a polytomy with Zatheria (including Peramus and Arguimus), and the clade comprising Palaeoxonodon + Nanolestes + Amphibetulimus.Neither Mozomus nor Amphibetuliumus were included in our analysis and so we did not test the relationships between these taxa and other early cladotherians.
Characters of the talonid basin have previously been used to argue for a close affinity between Palaeoxonodon and Peramus, to the exclusion of Amphitherium (Freeman 1976b(Freeman , 1979;;Kielan-Jaworowska et al. 2004).However, the presence of plesiomorphies such as the principal talonid cusp (hypoconid) with no embrasure for occlusal contact with the metacone of the upper molar may suggest Palaeoxonodon is more stemward than Peramus, and indeed several authors have suggested that Palaeoxonodon and the morphologically similar Amphitherium are sister-taxa, occuring stemward of Peramus (Sigogneau-Russell 2003;Davis 2011Davis , 2012)).Close et al. (2016) suggested that the "incipient" talonid basin of Palaeoxonondon and the talonid basin of Peramus could either be homoplastic, or a synapomorphy of the clade Palaeoxonodon + Amphitherium + Peramus, and this currently cannot be resolved.We agree with Close et al. (2016) that the phylogenetic analyses carried out in their paper and herein cannot resolve this unambiguously.
Palaeoxonodon is the only taxon in our analysis in which the mandibular foramen is located in the pterygoid fossa and offset from the Meckel's sulcus.In Amphitherium, the foramen is aligned with the posterior end of the Meckel's sulcus, as it is for the most closely related taxa among early Cladotheria (Butler and Clemens 2001;Kielan-Jaworowska et al. 2004).This is a condition seen in crown Marsupialia and some basal metatherians, such as Didelphodon (Fox and Naylor 1986;Wilson et al. 2016), but not in more crownward Metatheria.However, there is some damage to the posterolingual portion of the dentary in NMS G.2017.37.1.More complete, undamaged specimens may shed further light on the placement of the mandibular foramen.The very slight extension of the anterior crest of the masseteric fossa into the body of the dentary is not shared with any other taxon in this analysis, except the extant Oryctolagus (European rabbit).However the more extreme extension of the anterior crest of the masseteric fossa, where it extends below the ultimate pre-molar or first molar, is commonly seen in multituberculates (e.g., Gambaryan and Kielan-Jaworowska 1995).This may suggest Palaeoxonodon had a more extensively developed superficial masseteric muscle than other early cladotherians.Without knowing the position and morphology of the condylar process, which is crucial for determining the in-lever for the dentary (Gill et al. 2014), it is not possible to make informed inferences about the biomechanics of mastication in this genus.This would require more complete mandibles that are amenable for muscle reconstruction, as shown in other studies (Lautenschlager et al. 2017).
The presence of a masseteric foramen (also referred to as the labial mandibular foramen), located in the masseteric fossa on the buccal side of the dentary, is a newly identified feature in Palaeoxonodon.A deep masseteric fossa with a distinct masseteric foramen is a feature shared with peramurans, but is also present in numerous taxa across different lineages, including other zatherians, triconodontids, and basal metatherians and eutherians (Davis 2012).In Palaeoxonodon the foramen is not as large as in some specimens of Peramus, but it quickly merges with the mandibular canal inside the dentary, posteroventrally to the position of m5.The morphology of this foramen in Palaeoxonodon is similar to that described by Davis (2012) for the specimens of Peramus that lack an enlarged masseteric foramen: the mandibular canal is enlarged mesial to the masseteric fossa, and is separated by only a thin wall of bone (Davis 2012: 813-814).Davis suggests the variability in the size of the masseteric foramen in specimens of Peramus may be attributed to post-depositional crushing and infill.Until further specimens of Palaeoxondon are found possessing the posterior portion of the dentary, it is not possible to say if the size of this feature similarly varies in this taxon.

Conclusions
Newly described specimens of Palaeoxonodon ooliticus from the Isle of Skye add valuable new information for morphological comparison, and new data for phylogenetic analysis.The additional features of the masseteric fossa, and the position of the mandibular foramen identified in NMS G.2017.37.1, while not altering the phylogenetic relationships between Palaeoxonodon and its nearest relatives among cladotherians, are new autapomorphies for this genus.They provide additional character scores for future analysis, reducing the amount of missing information for this taxon.
The "re-discovery" of NMS G.1992.47.123 provides a cautionary tale for palaeontologists who delay in the publication of specimens.By not pursuing identification and description of this specimen, an opportunity was missed to name a second Mesozoic mammal from Scotland, an important taxon for understanding the morphology of early Cladotheria.This more complete specimen would also have provided a holotype with some variation in the tooth row, perhaps preventing the erection of taxa that are now suggested to be junior synonyms.This is an informative lesson from the palaeontological history of Mesozoic mammals.NMS G.1992.47.123 also contributes to the description of rich Scottish Mesozoic mammal specimens that have, until now, remained undescribed, and therefore unrecognised scientifically.
Although a relatively large number of upper teeth and lower dentitions from Palaeoxonodon are now known, they have not yet been found in association, and postcrania have yet to be identified.This provides a goal for further discoveries in the rich microvertebrate beds of the Isle of Skye in the coming years.

Fig. 3 .
Fig. 3. Parsimony analysis and reconstructions of dentary of Palaeoxonodon ooliticus (B, C), with Amphitherium (D) for comparison.A. Equally weighted parsimony analysis with additional mandibular characters for the early cladotherian Palaeoxonodon ooliticus, and rescored and updated matrix from Close et al. 2016 (adapted from Zhou et al. 2013).Jackknife resampling statistics and Bremer values (bold) are shown adjacent to individual nodes.B. Composite of NMS G. 2015.17.1 (grey) and coronoid of NMS G.2017.37.1 (red) in buccal (B 1 ) and lingual (B 2 ) views (arrows indicate anterior direction).Data for NMS G. 2015.17.1 from Close et al. (2016).C. Reconstruction of jaw in buccal (C 1 ) and lingual (C 2 ) views.D. Reconstruction based on composite created by EP from NHMUK PV M.36822 and OUMNH J.20075 in buccal view.Scale bars 1 mm.