Urrayira whitei gen. et sp. nov.: a dasyuromorphian (Mammalia: Marsupialia) with incipient zalambdodonty from the Middle Pleistocene of Queensland, Australia

Abstract

Urrayira whitei gen. et sp. nov. is described based on dental remains from middle Pleistocene cave sites at Mount Etna, Queensland. Its higher-level systematic affinities are unclear but it appears to be a dasyuromorphian. It is unusual in having a specialized reduced dentition characterized by reduction of the stylar cusps, protocone and talonid, resulting in an incipiently zalambdodont morphology that emphasizes the shearing crests. In addition, only two upper premolars are present, and we assume that it is P3 that has been suppressed, as has occurred multiple times within Dasyuridae. Maximum parsimony and undated Bayesian analyses of a 174 morphological character matrix intended to resolve relationships within Dasyuromorphia, with a molecular scaffold enforced, suggest that Urrayira is a dasyurid. In the maximum parsimony analysis, Urrayira is sister to Planigale gilesi (which also lacks P3), whereas in the undated Bayesian analysis, Urrayira resolves as part of a trichotomy at the base of Dasyuridae, together with Sminthopsinae and Dasyurinae; however, support values are generally low throughout the tree. While the majority of rainforest-adapted taxa in the Mount Etna sites became either extinct or were locally extirpated at, or soon after, 280 ka, there is no evidence that U. whitei gen. et sp. nov. even persisted until that time. Urrayira whitei was likely a rainforest-specialist, thus may have been particularly vulnerable to incipient effects of the Mid-Brunhes climatic shift towards aridity that eventually drove the disappearance of the Mount Etna rainforest and its associated fauna.

Jonathan Cramb* [[email protected]], Queensland Museum, PO Box 3300, South Brisbane BC, Queensland 4101, Australia; Scott Hocknull [[email protected]], Queensland Museum, PO Box 3300, South Brisbane BC, Queensland 4101, Australia; Robin M. D. Beck [[email protected]], School of Science, Engineering and Environment, University of Salford, Manchester M5 4WT, UK; Shimona Kealy [[email protected]], Archaeology and Natural History, College of Asia and the Pacific, The Australian National University, Canberra, ACT, 2601, Australia; Gilbert J. Price [[email protected]], School of Earth and Environmental Sciences, The University of Queensland, Brisbane, Queensland 4072, Australia.

Key words:

• Dasyurid
• Dasyuromorphia
• rainforest
• zalambdodont

DASYUROMORPHIANS are faunivorous marsupials noted for both their diversity and their rather poorly understood fossil record, especially from northern Australia (Wroe 2003, Van Dyck & Strahan 2008, Black et al. 2012, Kealy & Beck 2017). Four families are currently recognized: Thylacinidae (thylacines), Myrmecobiidae (the numbat), Malleodectidae (dentally derived forms that have been identified as possible specialist molluscivores), and Dasyuridae (the most diverse family today, comprising >70 modern species, including quolls, dunnarts, antechinuses, and the Tasmanian Devil). Assignment of isolated teeth to Dasyuridae has proven challenging, in particular those from Oligocene–Miocene deposits, due to an absence of unequivocal dental synapomorphies that may characterize the family (Wroe 1997a, 2003, Long et al. 2002, Archer & Hand 2006, although see Murray & Megirian 2006 for an alternative perspective).

Hocknull (2005) described diverse rainforest assemblages from cave sites at Mount Etna, eastern central Queensland (Fig. 1), that are now known to be mid-Pleistocene in age (Hocknull et al. 2007). Dasyurid diversity in these is unusually high in comparison to other rainforest assemblages, and contains several previously undescribed taxa (Cramb et al. 2009). One undescribed taxon noted by Hocknull (2005) features a highly distinctive dentition. Here, this taxon is fully taxonomically assessed and described for the first time, including new material not available to Hocknull (2005).

Figure 1. Map of Queensland showing fossil localities mentioned in the text. 1: Mount Etna; 2: Russenden Cave; 3: Floraville; 4: Riversleigh World Heritage Area.

In comparison to the understudied northern records, the Pleistocene record of fossil dasyurids is relatively well-documented in southern Australia (e.g., Smith 1972, Wakefield 1972, Dawson & Augee 1997, Fraser & Wells 2006). Despite this rich record, relatively few dasyurid species have been erected on the basis of Pleistocene fossils. Antechinus yammal and A. yuna Cramb & Hocknull, 2010 are exceptions, described from Chibanian-aged (170–>450 ka) deposits at Mount Etna. Only three extinct dasyurids of undoubted Pleistocene age have been described from sites other than Mount Etna: A. puteus Van Dyck, 1982 from Russenden Cave, south-east Queensland, Sminthopsis floravillensis Archer, 1982 from Floraville in north-west Queensland (see Price et al. 2021 for comments on the age of the Floraville deposits), and the widespread Sarcophilus laniarius Owen, 1838 (Long et al. 2002). Most dasyurids recovered from Pleistocene deposits have been assigned to living species. Given that the majority of dasyurid species described solely on the basis of fossils are from Pliocene or older deposits (Long et al. 2002, Wroe 2003), this has created the impression that dasyurid assemblages were relatively stable during the Quaternary. In contrast, middle Pleistocene faunas from Mount Etna challenge that assumption in that they contain high diversity assemblages of dasyurids dominated by prehistorically extinct species, including members of multiple previously unknown genera (Hocknull 2005, Hocknull et al. 2007, Cramb et al. 2009, Cramb & Hocknull 2010).

Living dasyurids have a generalist tribosphenic dentition that is broadly similar to the ancestral state of therian mammals as a whole (Beck et al. 2022). Tribosphenic molars have been retained by numerous lineages of small insectivorous therians, but several therian lineages show a distinctive modification of this ancestral pattern termed ‘zalambdodonty’. Zalambdodonty presents in which either the paracone or metacone are greatly reduced or lost, giving the molar crown a distinctive V-shape, with the protocone and talonid also often greatly reduced or lost (Asher & Sánchez-Villagra 2005). Among eutherians, the zalambdodont condition of the upper molars is typically approached via reduction of the metacone (see for example Seiffert et al. 2007 for discussion of the development of zalambdodonty among afrosoricids). In contrast, metatherians appear to develop zalambdodonty in their upper molars by reduction of the paracone (see for example Archer et al. 2011). Among Australian marsupials, there are two clades that have evolved incipient or full zalambdodonty: notoryctids (species of Notoryctes Stirling, 1891 and Naraboryctes Archer et al., 2011) and yalkaparidontids (species of Yalkaparidon Archer et al., 1988). Here we describe a new genus and species that independently demonstrates the development of incipient zalambdodonty in a third group of Australian marsupials, the dasyuromorphians.

Geological setting

Hocknull (2005) and Hocknull et al. (2007) described the geological, sedimentological and taphonomical settings for the fossil deposits yielding dasyurid remains in the Mount Etna region. Fossil dasyurids have been recovered from all known deposits at Mount Etna spanning approximately the last 500,000 years (500 ka). The exact geological ages of the oldest deposits are currently unknown, due to the limits of U-Th dating undertaken, but are likely older than 500 ka (Hocknull et al. 2007). Fossils pertaining to the new taxon described herein are known from three deposits: QML1284, 1385 and 1384. QML1284 is located on Limestone Ridge to the east of Mount Etna. Hocknull (2005, 2009) interprets QML1284 as a cave chamber deposit due to the repeating travertine flowstone floors, fine sedimentation and mixture of small-bodied and large-bodied fauna. Articulated remains are common. The small-bodied faunal inclusions are most likely the result of predator accumulation, such as an owl roost. U-Th dating of the flowstones associated with the fossil-bearing layers returned ages of >500 ka, as the resulting assessments were near the limits of the technique. QML1384 and 1385 form part of the (now destroyed) Elephant Hole Cave system that occurred on the western flank of the Mount Etna Limestone Mine. QML1385 is the lowest deposit known at Mount Etna and constituted a series of small solution pipes connecting to the higher chamber deposits of QML1384. QML1384 is divided into two discrete deposits, an upper and lower unit (QML1384UU and QML1384LU). These two deposits are likely divided by a thick flowstone, although direct contacts between the two were not uncovered during mining operations. Fossils from QML1385 are primarily from small-bodied fauna and show some degree of bone rounding and reworking. These remains were likely derived from an upper deposit, possibly QML1384LU. QML1384LU was assessed to be greater than ∼300 ka (Hocknull et al. 2007), with QML1384UU being younger than this.

All three deposits show similar degrees of small-bodied fauna preservation, suggesting predator (owl, raptor and/or ghost bat) accumulation was likely a key contributor to all three deposits. QML1284 does differ from QML1384 and 1385 because it represents a deposit accumulating on what is interpreted to be a functional cave floor, whilst the other two deposits occur in vertically deep and massive sediment deposits with limited internal stratigraphy. The dasyurid remains described here have been recovered through a combination of acid processing of indurated blocks and sieving. Remains were picked under microscope magnification. Associated fauna include abundant remains of frogs, lizards, small mammals (marsupial, rodent and bat) and small birds.

Materials and methods

Dental terminology is from Cramb & Hocknull (2010), and is shown in Fig. 2. Numbering follows Luckett (1993). Specific names recognized here follow the Australasian Mammal Taxonomy Consortium (2021) for extant dasyurids, and Travouillon et al. (2020) for extinct species. Body mass for the new species was estimated using methods described by Myers (2001); best estimates were calculated from M² area (crown length × crown width, following the suitability ranking regressions of Myers 2001) based on three of the most complete specimens.

Figure 2. Urrayira whitei gen. et sp. nov. compared with Planigale sp., showing terminology of major molar features. A, B, Urrayira whitei, A, left M2; B, left M2. C, D, Planigale sp., C, left M2; D, left M2. Abbreviations: ANC, anterior cingulum; ME, metacone; MTC, metacrista; PA, paracone; PMC, premetacrista; POP, postprotocrista; PPC, postparacrista; PR, protocone; PRC, paracrista; PRP, preprotocrista; PSC, posterior cingulum; STB, stylar cusp B; STD, stylar cusp D; (lower): ACD, anterior cingulid; CRO, cristid obliqua; HCD, hypoconulid; HCR, hypocristid; HYD, hypoconid; MCD, metacristid; MED, metaconid; PAD, paraconid; PCD, paracristid; PRD, protoconid; PSD, posterior cingulid.

Phylogenetic relationships of the new taxon were tested using a morphological character matrix modified from that of Kealy & Beck (2017), which was specifically designed to infer relationships within Dasyuromorphia. We deleted taxa that lacked morphological character scores (i.e., that were represented by molecular data only) from Kealy & Beck’s (2017) original matrix, leaving 53 taxa (40 dasyuromorphians and 13 non-dasyuromorphian outgroup taxa) scored for 173 characters. We added one additional morphological character (Character 174) that we considered potentially useful in resolving relationships among the taxa considered here, namely body mass. We agree with Yates (2014, p. 5) that body mass appears to show considerable phylogenetic signal within Dasyuromorphia; for example, no known thylacinid has a body mass <1 kg (Rovinsky et al., 2019), whilst within Dasyuridae, all known taxa with a body mass >1 kg are within a single tribe, Dasyurini (Weisbecker et al., 2013). We used the following states: 0 = <10 g; 1 = 10–100 g; 2 = 100 g to 1 kg; 3 = 1–10 kg; 4 = >10 kg. Body mass estimates for modern taxa were taken from Weisbecker et al. (2013), and the PanTHERIA (Jones et al. 2009) and PHYLACINE (Faurby et al. 2018) databases. Body mass estimates for fossil taxa were taken from Travouillon et al. (2009), Ladevèze et al. (2011), Beck (2015), Rovinsky et al. (2019), and Muizon & Ladevèze (2020).

Two additional taxa were here added to this 174 character matrix: the new taxon described here, plus the modern Planigale gilesi Aitken, 1972, which was noted by Hocknull (2005) as being closest in morphology to the new taxon, and is the only Planigale species known to have entirely lost the third upper and lower premolars (P³ and P₃; Aitken 1972; Archer 1976a). Only 13 characters (=7.5%) of the new taxon described here could be scored meaningfully, with 4 characters (=2.3%) inapplicable, and the remaining 157 characters (=90.2%) unknown. Planigale gilesi was scored based on cranial specimens QMJM4349, QMJ21973, and QMJM14278 (postcrania were not examined), and could be meaningfully scored for 48 characters (=27.6%), with 5 characters (=2.9%) inapplicable, and the remaining 121 characters (=69.5%) unknown.

The final matrix was analysed using maximum parsimony and undated Bayesian analysis, enforcing a molecular scaffold among the Recent taxa based on recent molecular analyses of Dasyuromorphia (Westerman et al. 2016, Kealy & Beck 2017, Feigin et al. 2018) and Marsupialia (Beck et al. 2022), but with the positions of the fossil taxa free to vary. All 40 characters originally specified as ordered by Kealy & Beck (2017) were also ordered here, as was our new body mass character (character 174), as this represents an obvious morphocline. Preliminary analyses indicated that Sminthopsis floravillensis was an unstable/“wildcard” taxon, and so it was deleted, leaving a total of 54 taxa. Maximum parsimony analysis was implemented in TNT v1.5 (Goloboff & Catalano 2016), with the molecular scaffold enforced using the “force” command. A two-stage search strategy was used, with the first stage comprising a “New Technology” search using combined Sectorial Search, Ratchet, Drift and Tree fusing (default settings in each case), until the same minimum tree length was hit 100 times, followed by a “traditional” search with tree bisection–reconnection branch swapping within the set of trees saved from the first stage. The final set of multiple most parsimonious trees was then summarized using strict consensus. Support values were calculated for nodes present in the strict consensus using 2000 bootstrap replicates, with bootstrap support calculated as absolute frequencies; however, these values should be treated with scepticism because they were calculated using the molecular scaffold, and so do not represent “true”, unconstrained support.

Undated Bayesian analysis of the same matrix was implemented in MrBayes v3.2.7 (Ronquist et al. 2012). This program does not allow the use of a single tree as a “backbone” constraint; instead, the program PRAP2 (Müller 2004) was used to create a series of “partial” constraints corresponding to the molecular scaffold. A single Mk model (Lewis 2001) was applied to the morphological data, with the assumption that variable characters have been scored (this is justified as 9 of the 174 characters, i.e., 4.9%, are variable but not parsimony informative), and an eight-category log-normal distribution was used to model rate heterogeneity (following Harrison & Larsson 2015). The MrBayes analysis comprised two independent runs of four chains each, with default settings, run for 2 million generations, sampling trees and other parameters every 5000 generations. Examination of parameters in Tracer (Rambaut et al. 2018) indicated that stationarity and convergence between runs was achieved within the first 10% of samples (i.e., 200 thousand generations); the post-burnin trees from the remaining 90% (i.e., 1.8 million generations) were summarized using 50% majority rule consensus. This gives support values as Bayesian posterior probabilities but, similarly to the bootstrap values from the maximum parsimony analysis (see above), these should be viewed sceptically as they are not the result of an unconstrained analysis. The maximum parsimony and undated Bayesian phylogenies were plotted using the ggtree package (Yu et al. 2016) in R (R Core team 2020).

Institutional abbreviations

QMF, Queensland Museum Fossil specimen; QML, Queensland Museum Fossil Locality; QMJM, QMJ, Queensland Museum mammal specimen.

Systematic palaeontology

Order DASYUROMORPHIA Gill, 1872 sensu Aplin & Archer, 1987

Family cf. DASYURIDAE Goldfuss, 1820 sensu Waterhouse, 1838

Urrayira gen. nov.

Diagnosis

A tiny (Planigale-sized, with estimated body mass range 6.4–8.5g, mean 7.5 g; see Table 1) dasyuromorphian characterized by the following combination of apomorphic characters: (1) Reduced protocone on M^1–3; (2) Reduced paracone approximated to the metacone on M^1–3; (3) Reduced stylar cusp D contributing to a large buccal indentation and basin-like stylar shelf on M^2–3; 4. P³ absent; extremely reduced talonids on lower molars; entoconids absent.

Table 1. Measurements of the dentition of Urrayira whitei.

	QMF51743	QMF53663	QMF53664	QMF53665	QMF53666	QMF53667	QMF55122	QMF55121
	P^2 length (alveolar)	0.8
M^1–3 length	3.4
M^1 length	1.3				1.4
M^1 width	1.1				0.9
M^2 length	1.2	1.2		1.3		1.2
M^2 width	1.3	1.4		1.4		1.1
M^3 length	1	1.1	1	1.1
M^3 width	1.2	1.5	1.4	1.4
M^4 length		0.6
M2 length							1.2
M2 trigonid width							0.8
M2 talonid width							0.7
M3 length								1.3
M3 trigonid width								0.9
M3 talonid width								0.8

All measurements in millimetres. Note that the measurements of QMF53667 are likely affected by heavy wear on that specimen.

Etymology

Combination of the Darumbal (First Nations language of the Mount Etna area) words ‘urra’ meaning small and ‘yira’ meaning tooth/teeth. Pronounced “Oo-rah-year-rah.”

Urrayira whitei sp. nov

(Fig. 3)

Figure 3. Urrayira whitei gen. et sp. nov. A, B, QMF51743 holotype left maxilla with M1–3 in occlusal and lingual views; C, QMF53663 paratype left maxilla with M2–4; D, E, QMF55121 left M3 in occlusolateral and buccal views; F, G, QMF55122 right M2 in occlusolateral and occlusal views. Scale bars: A–C = 2 mm; D–G = 1 mm.

Diagnosis

As for genus.

Etymology

Named for Mr Chris White, for his contribution, as Cement Australia Mt. Etna mine manager, to the protection and preservation of the fossils from the Mount Etna Mine, now within Mount Etna Caves National Park.

Holotype

QMF51743, left maxilla with M^1–3. Paratype: QMF53663 left maxilla with M^2–4

Referred material

QML1385: QMF53664 left M³, QMF53665 left maxilla with M^2–3, QMF53666 right M¹, QMF55122 right M₂, QMF55121 left M₃. QML1284: QMF53667 left M¹.

Type locality, unit and age

QML1385, Bench 4, Mount Etna Limestone Mine, approximately 25 km north of Rockhampton. Chibanian age, dated to ∼500 ka (Hocknull et al., 2007).

Description

Maxilla

No incisors, canines or premolars known. QMF51743 (Figure 1(1)) preserves alveoli of C¹ and two premolars, interpreted to be P^1–2 as P³ is lost in several dasyurids (Tate 1947, Beck et al. 2022; see discussion). Length of M^1–3 (based on QMF51743): 3.48 mm. Numerous minute interdental foraminae present in palate between molars and immediately lingual of premolars. Large infraorbital foramen directly above posterior root of M¹. Anterior end of zygomatic arch emerges from maxilla buccal of M³. Maxilla contacts jugal buccal of M³, marked by swelling of maxilla on ventral edge of zygomatic arch.

M¹. Description based on QMF51743. Triangular in occlusal outline, with buccal edge longer than anterior edge. Ectoflexus in buccal margin broad and shallow, not deep as on M^2–3. Metastylar corner angular. Metacone the tallest cusp, followed by paracone and protocone, respectively. Paracone heavily reduced, immediately anterior of and approximated to metacone. Protocone reduced, directly lingual of paracone. All stylar cusps except D obliterated by masticatory wear. Stylar cusp D positioned posterobuccally of metacone. Metacrista long, paracrista much less than half as long as metacrista. Apparently very little curvature in centrocrista (postparacrista and premetacrista). Anterior cingulum broad and complete, indented at base of paracone. Posterior cingulum well developed; no buccal cingulum apparent.

M². Tooth not as anteroposteriorly elongated as M¹. Roughly triangular in occlusal outline, with large buccal indent approximately half way between paracrista and metacrista and immediately anterior of stylar cusp D. Depression on stylar shelf, forming a basin. Metastylar corner angular. Metacone the tallest cusp, followed by paracone and protocone, respectively. Stylar cusp B and metastylar corner both taller than stylar cusp D. Paracone heavily reduced, anterior and slightly buccal of metacone, to which it is approximated. Protocone reduced, directly lingual of paracone. Stylar cusp D reduced, buccal and slightly posterior of metacone. Paracrista slightly longer than half of metacrista length. Anterior cingulum broad and complete, indented at base of paracone. Posterior cingulum robust lingually, narrowing buccally, terminates below approximate mid-point of metacrista. Posterior cingulum partially crenulated. No buccal cingulum.

M³. Approximately triangular in occlusal outline. Buccal ectoflexus deeper than that of M². Metastylar corner rounded. Depression on stylar shelf, forming a basin. Metacone the tallest cusp, followed by paracone and protocone, respectively. Stylar cusp B and metastylar corner both taller than stylar cusp D. Paracone reduced, but larger and more distinct than that on M². Paracone anterobuccal of metacone, directly buccal of protocone. Protocone more reduced than that of M². All stylar cusps reduced. Stylar cusp D present as small protuberance on posterior side of buccal ectoflexus. Paracrista much more than half length of metacrista. Anterior cingulum broad at buccal end, narrow but complete near base of paracone. No sharp indent in anterior cingulum. Posterior cingulum crenulated, curved near base of metacone to reach reduced protocone. Posterior cingulum robust lingually, narrowing buccally, terminates below approximate mid-point of metacrista.

M⁴. Description based on QMF53663. Approximately triangular in occlusal outline, although full shape not known due to damaged only specimen. Paracone the tallest cusp (larger and more robust than that on M³), then probably the protocone (absent due to damage). Metacone and stylar cusp D absent. Anterior cingulum narrow but complete, posterior cingulum narrow. No metastylar indent, metastylar corner of M³ is buccal of stylar cusp B of M⁴.

Lower dentition

No dentary, incisors, canines, or premolars are known. The interpreted absence of P³ may imply that P₃ was also absent. Two isolated lower molars are interpreted as an M₂ and an M₃.

M₂. Based on QMF55122. Cusps and cristids slightly worn. Talonid heavily reduced, approximately two thirds of trigonid breadth. Trigonid shaped like an equilateral triangle. Protoconid the tallest cusp, followed by metaconid, paraconid, hypoconulid and hypoconid (hypoconid probably taller than hypoconulid when unworn). Entoconid absent. Paraconid the most anterior cusp, anterior and slightly lingual of metaconid. Protoconid the most buccal cusp. Metaconid buccal of lingual margin of crown, directly anterior of hypoconulid. Hypoconid posterolingual of protoconid and slightly posterobuccal of metaconid. Paracristid slightly longer than metacristid. Hypocristid parallel to metacristid. Cristid obliqua very short, contacting the trigonid slightly buccally of the metacristid fissure. Anterior cingulid bow-shaped, narrowing lingually. Parastylid small. Buccal cingulid well developed, weakly present at base of protoconid and strongly present at base of hypoconid. Posterior cingulid broad from base of hypoconid to hypoconulid. Narrow lingual cingulid present at base of metaconid.

M₃. Based on QMF55121. Cusps and cristids slightly worn. Talonid heavily reduced, approximately one quarter of trigonid breadth. Trigonid shaped like an equilateral triangle. Protoconid the tallest cusp, closely followed by metaconid, then paraconid, hypoconulid and hypoconid respectively (hypoconid probably taller than hypoconulid when unworn). Paraconid the most anterior cusp. Protoconid the most buccal cusp, posterobuccal of paraconid and anterolingual of metaconid. Hypoconulid a short distance directly posterior of metaconid. Entoconid absent. Hypoconid small, posterior and slightly lingual of protoconid. Paracristid heavily worn, but deep paracristid notch evident. Hypocristid parallel to metacristid. Cristid obliqua very short, contacting the trigonid below metacristid notch. Strong anterior cingulid, weakly joined to buccal cingulid at base of protocone. Parastylid small. Buccal cingulid narrow but deep. Posterior cingulid well developed but less so than that of M₂.

Remarks

The very small size and dental apomorphies of this new genus make it unlikely to be confused with any other dasyuromorphian. It is separable from other dasyuromorphians of similar size (Planigale spp., Ningaui spp., Sminthopsis ooldea Troughton, 1965, S. hirtipes Thomas, 1898 and Mayigriphus orbus Wroe, 1997b) by the combination of the greater reduction of the protocone, paracone, stylar cusp D on M^1–3, and the greater reduction of the talonids and associated loss of entoconids on the lower molars. The lack of a P3 separates it from most other tiny dasyuromorphians (Planigale gilesi Aitken, 1972 lacks P3, but differs from the new genus in the other characters already listed).

Phylogenetic analysis

The results of maximum parsimony and undated Bayesian analysis of our 174 morphological character matrix, with a molecular scaffold enforced, are shown in Fig. 3. Both analyses place Urrayira within Dasyuridae. Our maximum parsimony analysis recovered 108 most parsimonious trees of 871 steps; Urrayira is sister to Planigale gilesi in the strict consensus of these (Fig. 4A). In our undated Bayesian analysis, meanwhile, Urrayira forms a trichotomy with Dasyurinae and Sminthopsinae (Fig. 4B). Support values are consistently low in both analyses, likely due to a combination of high levels of morphological homoplasy within Dasyuromorphia, and the relative incompleteness of several fossil taxa, including Urrayira. There is <50% bootstrap support for Urrayira+Planigale gilesi, Sminthopsinae, and Dasyuridae all receive <50% bootstrap support in the maximum parsimony analysis, whilst monophyly of Dasyuridae in the undated Bayesian analysis has a Bayesian posterior probability of 0.52. Regardless, these support values should be treated with scepticism due to our use of a molecular scaffold that constrained relationships among our Recent taxa.

Figure 4. Results from phylogenetic analyses of a 174 morphological character matrix developed for resolving relationships within Dasyuromorphia, with a molecular scaffold enforced as a backbone constraint. A, Strict consensus of 108 most parsimonious trees (length = 871 steps) from maximum parsimony analysis using TNT v. 1.5; numbers to the left of the nodes are support values calculated as bootstrap percentages (based on 2000 bootstrap replicates). B, 50% majority rule consensus of post-burn-in trees from undated Bayesian analysis using MrBayes 3.2.7; numbers to the left of nodes are support values calculated as Bayesian posterior probabilities (expressed as percentages). Note that, in both cases, the support values should be treated with scepticism due to the use of a molecular scaffold.

Discussion and conclusions

Comparison with other Australian zalambdodont marsupials and evidence for dasyurid affinities

Given the incipient zalambdodont molar morphology of Urrayira, it is appropriate to compare it to the two Australian marsupial clades known to have evolved zalambdodonty: yalkaparidontians and notoryctemorphians. The yalkaparidontids Yalkaparidon coheni Archer et al., 1988 and Y. jonesi Archer et al., 1988 (known from late Oligocene to middle Miocene-aged sites at Riversleigh World Heritage Area; Beck et al. 2014) are characterized by an extreme zalambdodont molar morphology in which the upper molars lack any trace of the protocone or the paracone (assuming that it is this cusp, and not the metacone that has been lost, which is not certain), unlike the condition in Urrayira. The lower molars of species of Yalkaparidon are also peculiar in that the posterior half forms a posterolingually elongate “tail” that is unlike the morphology of any other zalambdodont mammal, and which may in fact represent a remnant of the talonid. The non-molar morphology of species of Yalkaparidon is also radically different from that of U. whitei: Yalkaparidon coheni retains a single upper premolar, which Beck et al. (2014) identified as P³, and which is separated from the tiny upper canine by a distinct diastema. With an estimated body mass of 125–250 g, species of Yalkaparidon are also much larger than U. whitei.

Urrayira whitei shows closer dental resemblances to notoryctemorphians, and in particular the incipiently zalambdodont Naraboryctes philcreaseri. In the modern notoryctemorphians Notoryctes typhlops (Stirling, 1889) and No. caurinus Thomas, 1920, the paracone has been lost (or may be present as a tiny vestige), as has the talonid, but Na. philcreaseri retains a reduced but still distinct paracone and talonid; both species of Notoryctes and Na. philcreaseri also retain a prominent protocone. However, there are several notable differences between notoryctemorphians and U. whitei. The protocone and associated crests (preprotocrista and postprotocrista) are larger and more prominent in species of Notoryctes and Naraboryctes, and they show a distinct “waisting” where they connect to the buccal half of the tooth, whereas this is not present in U. whitei. The paracone of Na. philcreaseri is positioned further buccally, and hence the centrocrista between the paracone and metacone is straight, than in U. whitei, which retains a weakly v-shaped centrocrista. In the lower molars, Na. philcreaseri retains a prominent entoconid, the trigonid is mesiodistally more compressed and, perhaps most significantly, there is no trace of a posterior cingulid, whereas this cingulid is prominent in U. whitei, which strongly supports dasyuromorphian affinities (see below; Beck et al. 2022). In summary, qualitative interpretation of available dental evidence therefore supports the hypothesis that U. whitei represents a third Australian clade that has evolved zalambdodont specializations, to a similar degree to Na. philcreaseri (although differing in numerous details) but less so than in species of Notoryctes or Yalkaparidon.

Based on their analysis, Beck et al. (2022, p. 217) stated that three dental features are synapomorphies of Dasyuromorphia: stylar cusp D taller than stylar cusp B on M²; M₃ hypoconid lingual to salient protoconid; and lower molars with distinct posterior cingulid. The M₃ hypoconid is lingual to the protoconid in U. whitei, but this feature shows considerable homoplasy within Marsupialia (Beck et al. 2022). Presence of a posterior cingulid shows somewhat less homoplasy among marsupials, and its presence in U. whitei seems reasonably strong evidence that this taxon is indeed a dasyuromorphian. Among Australian marsupials, a clade comprising species of Djarthia, Keeuna and Ankotarinja appears to have independently acquired a posterior cingulid (Kealy & Beck 2017, Beck et al. 2022), but members of this clade are not known after the late Oligocene. The third dental synapomorphy of Dasyuromorphia identified by Beck et al. (2022), namely stylar cusp D taller than stylar cusp B on M², does not appear to be present in U. whitei (although this is not completely certain, due to the worn nature of available specimens), but this feature shows some homoplasy within Dasyuromorphia (Beck et al. 2022, p. 131). In summary, available evidence, particularly the presence of a posterior cingulid, supports U. whitei as a dasyuromorphian. In turn, this suggests that the premolars present in U. whitei are P1–2, as in all known dasyuromorphians with two premolars (all of which are within the family Dasyuridae) it is P3 that appears to be lacking (Tate 1947, Wroe & Mackness 2000, Beck et al. 2022).

Our phylogenetic analyses place U. whitei within Dasyuridae, a result likely driven primarily by the absence of P3, as within Dasyuromorphia only some dasyurids are known to have lost this tooth. A close relationship with species of Planigale, as found in our maximum parsimony analysis, seems plausible: U. whitei shares with Planigale gilesi the absence of P3, and Planigale includes the smallest known living dasyuromorphians and the only ones to achieve body masses of <10 g, as also estimated for U. whitei. Among living dasyurids, species of Planigale also have particularly small paracones and talonids (Archer 1976), although not to the extent of U. whitei. However, loss of P3 has occurred multiple times within Dasyuridae (Tate 1947, Wroe & Mackness 2000, Beck et al. 2022), and support for a close relationship between U. whitei and species of Planigale will remain weak in the absence of more complete remains of the fossil taxon. On current evidence, U. whitei is therefore best considered as cf. Dasyuridae incertae sedis.

Palaeoecology

The apparently derived state of the dentition of U. whitei suggests a specialized diet. This is unusual among dasyurids, most of which are generalist predators of arthropods and small vertebrates (Dickman 2014). Ecological studies of extant small-bodied dasyurids suggest that their broad diet of invertebrates (supplemented by occasional small vertebrates) allows them to switch prey in response to environmental conditions (Gray et al. 2016). The most specialized extant dasyurid is the Tasmanian devil,Sarcophilus harrisii (Boitard, 1841), a scavenger with dentition adapted to crushing bone. Ganbulanyi djadjinguli Wroe, 1998 was described as a specialist bone cracking dasyurid, partially on the basis of a large P2 or P3 that has since been assigned to the ‘marsupial skink’ Malleodectes moenia Arena et al., 2011. Among very small modern dasyurids, such as species of Planigale and Ningaui which share similar body sizes to U. whitei, there appears to be little dental specialization. This lack of specialization is accompanied by a broad diet; for example, Woolnough & Carthew’s (1996) study of the diet of a species of Ningaui found that while their preference was for smaller insects, they were also capable of consuming small vertebrates (skinks), and hard-shelled prey such as beetles.

The exact function of U. whitei’s specialized, incipiently zalambdodont dentition is difficult to determine, as few detailed studies of the functional morphology of the dasyurid dentition have been published (but see Sanson 1985 for one example). Although clearly specialized, the exact functional implications of zalambdodont molars are unclear (Asher & Sánchez-Villagra 2005). Modern zalambdodont mammals with known diets are broadly faunivorous, most of them preying predominantly on insects. Zalambdodont mammals range from obligate burrowers (chrysochlorids, notoryctids) to long-snouted predators of earthworms (species of Hemicentetes; Eisenberg & Gould 1970) to probable ‘mammalian woodpeckers’ (yalkaparidontids; Beck 2009). In body size, they range from tiny shrew-like forms weighing less than 10 g (geogaline tenrecids) to 1000 g or more (solenodontids and tenrecine tenrecids). Dentition is clearly not the only factor that determines the ecology of such disparate groups.

The prominence of the postmetacrista implies an emphasis on shearing, the result of occlusion with the paracristid, while the reduction of the protocone implies that grinding, produced by occlusion of the protocone and the talonid, is of less importance (Sanson 1985). Beck (2009) and Archer et al. (2011) have suggested that this emphasis on the shearing action of the molars (as opposed to their crushing function) might imply a focus on soft-bodied invertebrate prey. In a rainforest palaeoenvironment, potential prey could include small insect larvae, termites, onychophorans, earthworms or terrestrial flatworms. The small body size of U. whitei may be an important factor in prey selection, as observations of living dasyurids demonstrate that larger-bodied species target a broader variety of prey sizes (Hall 1980, Dickman 2014).

Specialist taxa are especially vulnerable to environmental change (McKinney 1997). Most of the rainforest-adapted species found in sites at Mount Etna appear to have become extinct sometime between 280 and 205 ka. This event was seemingly driven by the climatic changes during the Mid-Brunhes Event, which manifested as greater climatic variability and intensified aridity in comparison to earlier in the Quaternary (Hocknull et al. 2007). That U. whitei is known only from the older middle Pleistocene sites at Mount Etna may suggest that it went extinct before the major faunal turnover event of 205–280 ka (e.g., Hocknull et al. 2007). Alternatively, it is possible that the very small body size and potential rarity of U. whitei in life means that it was a less-common prey species in the diets of the faunivorous Pleistocene taxa, which contributed to the accumulation of vertebrate remains in the local deposits. Thus, it is possible that the absence of this potentially specialist taxon from geologically younger deposits is due to taphonomic biases. Future excavations, coupled with detailed dating and taphonomic analyses, would be necessary to test this hypothesis.

All dasyurid genera erected to date on the basis of fossils have come from Neogene sites (Long et al. 2002). However, there are at least two and probably three novel genera in the Pleistocene deposits at Mount Etna that remain undescribed (Cramb et al. 2009). The same sites also contain representatives of lineages previously known from Pliocene (and in some cases Miocene) deposits, indicating the complexity of the biochronology of Australian mammal faunas (e.g., thylacoleonids and Kurrabi sp. Hocknull 2005, 2009, Hocknull et al. 2007, Invictokoala monticola, Price & Hocknull 2011).

Extant dasyurid tribes are thought to have begun radiating prior to the Pliocene because the oldest records of modern genera date to the early Pliocene (Long et al. 2002, Wroe 2003, Black et al. 2012); this inference is supported by recent molecular and total evidence clock analyses (Westerman et al. 2016, Kealy & Beck 2017, García-Navas et al. 2020, Beck et al. 2022). By the middle Pleistocene, different dasyurid species occupied both dry open habitats and closed rainforests (Cramb et al. 2009). Subsequent extinctions in closed rainforest environments during the Pleistocene, however, resulted in a dasyurid fauna in which greater diversity is seen in arid regions than in more mesic environments, as is the case today. The Pleistocene dasyurid faunas show both comparatively large and small-bodied specialist species that occurred in both arid (e.g., Sarcophilus harrisii and species of Ningaui) and mesic environments (Sarcophilus harrisii and U. whitei). The middle Pleistocene can therefore be seen as a high point of dasyurid diversity.

Acknowledgments

We acknowledge the Darumbal people, the Traditional Owners of the Mount Etna fossil sites, and pay respect to their Elders past and present. Particular thanks go to Nhaya Nicky Hatfield and LeLarnie Hatfield for sharing their language and granting permission to use the name Urrayira. We thank Cement Australia, and C. White in particular, for their proactive management of the significant fossil deposits discovered at Mt. Etna Mine, that subsequently became part of the Mt. Etna Caves National Park, protected from loss for generations to come. Special thanks to M. Archer, whose decades of work has inspired, mentored and supported all of the authors of this paper. We also thank: the Sands family for vital guidance and assistance in the field; G. E. Webb and A.M. Baker for comments on this manuscript; K. Spring and P. Wilson for collection management services; H. Janetzki and W. Goulding for access to comparative specimens, and the volunteers at QM Geosciences for assistance with specimen preparation.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental material

Supplemental data for this article is available online at https://doi.org/10.1080/03115518.2023.2169351.

Additional information

Funding

Research was funded by Australian Research Council grant ARC LP0453664 and ARC LP0989969.