APPENDIX 1

Phylogenetic Analysis of Microleo attenboroughi: Results and Description of Characters

The phylogenetic relationships of Microleo attenboroughi within the Thylacoleonidae were assessed as well as the relationships of thylacoleonids within the Suborder Vombatiformes, using a branch-and-bound search in PAUP * 4.0b10 (Swofford, 2002) based on a modified data matrix of Black et al. (2012) employing a total of 93 characters. Thirty characters were designated ordered. All characters were equally weighted and parsimony-informative. Five phascolarctid taxa in that matrix were not included in this analysis because many had high levels of missing data and the phascolarctids that were included were deemed informative for that taxon. The Oligo-Miocene peramelid Galadi speciosus and dasyurid Barinya wangala were used as outgroups. Within the thylacoleonid ingroup taxa, Priscileo pitikantensis, Wakaleo alcootaensis and Thylacoleo hilli have relatively high levels of missing data because they are only known from relatively incomplete specimens. The dentary specimen of Thylacoleo sp. cf. T. hilli (AM F63584) was used to code characters for the lower dentition of T. hilli on the basis of the assumption that it is likely to be representative of that species. Sixty five of the 71 characters of the original data matrix (Black et al., 2012) analysis were utilised. Six characters were removed because they either related to phascolarctid taxa not included in the analysis or were uninformative. One character relating to P3 morphology (ch.14) was modified to incorporate states in newly included taxa. A heuristic search was performed using branch and bound analysis. New characters relating to dental morphology (ch. 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17 18, 27, 28, 30, 38, 39, 40, 41, 42, 43, 44, 48, 50, 53, 57, 58, 59) were included in the data matrix. Bootstrap values were calculated using 1000 replicates.

Results

Our phylogenetic analysis recovered nine most-parsimonious trees (see consensus tree in Supplementary Figure 2: tree length = 272 steps, consistency index = 0.580, rescaled consistency index = 0.435, retention index = 0.750) in all of which Microleo attenboroughi is recovered as the sister taxon to all other known thylacoleonid taxa. Unequivocal thylacoleonid apomorphies (node 10) include bunodont molars and features of P3, including its length relative to M1 length (> 1.2), a longitudinal blade that has a long v-shaped (notched) occlusal edge and lack of a basal posterior cingulum. Autapomorphies distinguishing Microleo from all other thylacoleonids are features of the P3; its fusiform shape, medial cuspule, a posterolingual crest and steep posterobuccal crest at the rear of the crown. Bootstrap support was relatively strong for the thylacoleonid clade (91%) and for the Priscileo - Wakaleo - Thylacoleo clade (70%; see Supplementary Figure 2). Synapomorphies for the Priscileo - Wakaleo - Thylacoleo clade are: a posterior longitudinal blade on P3 that is gently-angled and M2 with confluent postprotocrista and premetaconulecrista. Priscileo roskellyae is recovered as the sister taxon to P. pitikantensis which in turn is recovered as the sister taxon to a Wakaleo/Thylacoleo clade. Unequivocal synapomorphies of the Wakaleo - Thylacoleo clade are increased P3 length relative to cheektooth row (> 0.4) and loss of M4/m4. Bootstrap support for this clade was moderate (83%). Species of Wakaleo form a clade (bootstrap support 62%) but their relationships are unresolved probably as a result of the limited data for both W. oldfieldi and W. alcootaensis. Synapomorphies for the Wakaleo clade include; reduced number of upper and lower premolars, development of an anterolingual cuspule and a moderately-developed posterobuccal crest on P3. Monophyly for species of Thylacoleo was well-supported (bootstrap frequency = 97%) and based on synapomorphies of P3 including: width greater anteriorly than posteriorly, a convex posterior longitudinal blade, a posterolingual crest, a significant increase in length, and reduction of the number of lower incisors to one.

Vombatiform Relationships. The strict consensus tree (Supplementary Figure 2) places thylacoleonids as the sister taxon to all other vombatiform families. This phylogeny differs from previous phylogenies (e.g., Aplin and Archer, 1987; Marshall et al., 1990; Gillespie, 2007; Black et al., 2012) which place thylacoleonids within the vombatomorphian clade. In our analysis, synapomorphies that united phascolarctomorphians and vombatomorphian families (excluding thylacoleonids, node 2) include presence of a single upper and lower premolar, presence of a lingual cusp on P3, a rectangular M2, lower incisors reduced to one, mandibular symphysis that extends to below m1 and a diastema that is < 30% the length of the horizontal ramus. Bootstrap support for this clade was weak (59%). Black et al. (2012) found phascolarctids were united by characters of the lower molars and were the sister group to all other vombatiform families. Bootstrap support for the phascolarctomorphian clade was strong (100%) but was weaker (63%) for the vombatomorphian clade (excluding thylacoleonids). With the exception of thylacoleonids and Marada arcanum, the relationships of all the other vombatomorphian families agree with those presented by Black et al. (2012). Our analysis was unable to resolve the position of Marada with respect to vombatids, diprotodontids and palorchestids.In previous analyses (Aplin and Archer, 1987; Black et al., 2012) thylacoleonids have been united with other vombatomorphian families predominantly on the basis of cranial characters and their repositioning in this analysis may be due to the large number of dental characters in the data matrix. The results of this analysis suggest that thylacoleonids possibly form a clade within Vombatiformes that has similar taxonomic status to phascolarctomorphians. However, in contrast to thylacoleonids, phascolarctomorphians show a greater diversity at the generic level and, at present, stronger support for the clade is required before erecting a possible new thylacoleonid infraorder.

Character Description

Description of 93 characters used in the phylogenetic analysis. Reference is given to characters used in previous analyses and whether they are modified.* Indicates ordered characters. Abbreviations for references: A84b, Archer (1984b); AD82, Archer and Dawson (1982); B07, Black (2007); B12, Black et al. (2012); BA97a, Black and Archer (1997a); BA97b, Black and Archer (1997b); G07, Gillespie (2007); H93, Hand et al. (1993); HSV03, Horovitz and Sanchez-Villagra (2003); HW82, Hope and Wilkinson (1982); L02, Luo et al. (2002); L04, Louys (2004); M86, Murray (1986); M90, Marshall et al. (1990); M94, Myers (1994); MA97, Myers and Archer (1997); R87, Reig et al. (1987), R98, Rougier et al. (1998); S97, Springer et al. (1997); SVW02, Sánchez-Villagra and Wible, (2002); T10, Travouillon et al. (2010); WD 87, Woodburne et al. (1987); W98, Wroe et al. (1998); W00, Wroe et al. (2000).

APPENDIX 2

Data Matrix

Data matrix used for phylogenetic analysis of Vombatiformes. Abbreviations: '?' signifies 'missing data'; '-' signifies 'inapplicable'. Polymorphic states indicated by: A, (0,1).

List of Synapomorphies

List of nodal apomorphies resulting from the phylogenetic analysis. Node numbers refer to Supplementary Figure 2.

APPENDIX 3

Body Mass Estimate

An estimate of the body mass of Microleo attenboroughi, Priscileo roskellyae and Wakaleo sp. nov. was made using a regression equation formulated by Myers (2001) from correlations of marsupial body mass with cranio-dental measurements. The equation used for the analysis was the highest possible ranked regression from the ‘Diprotodontians data set’ of Myers (2001, table 5, p.106) and employed the measurement for upper molar row length (UMRL, i.e., M1-M4). This equation (log y = -0.418 + 3.011 [log x], where x equals UMRL) was chosen because it was based on alveolar measurements rather than molar measurements; the posterior teeth are missing in both fossil taxa. The UMRL measurements (average of the left and right lengths) for M. attenboroughi (11.3 mm) P. roskellyae (QM F23453, 16.4 mm) and Wakaleo sp. nov. (QM F45200, 23.3 mm) resulted in body weight estimates of 590 g, 1813 g, and 5221 g, respectively. As recommended by Myers (2001), a smearing estimate (4.4%) was applied to the equation. Myers (2001) indicates that the regression equations are not appropriate for the larger species of marsupial lions (based on their unusual reduced molar numbers that result in large underestimates of body mass), hence, an estimate was not calculated for Wakaleo vanderleueri.