The taxon Mammalia consists of numerous extinct lineages as well as three extant clades: the monotremes (egg-laying mammals), marsupials (“pouched” mammals that give live birth to relatively altricial young), and placentals (mammals that give live birth to precocial young). Here, we follow the crown-group definition of Mammalia, which circumscribes the clade as the most recent common ancestor of living monotremes, marsupials, and placentals, and all its descendants (sensu Rowe 1988). Mammalia is nested within the clade Mammaliaformes (= Mammalia of Kielan-Jaworowska et al. 2004), which arose and diversified beginning in the Late Triassic, within approximately 20 million years of the Permo-Triassic mass extinction (Benton 2005; Fraser and Sues 2010) (Fig. 1). Mammaliaformes is a subset of Cynodontia, which, in turn, is a subset of Synapsida, one of the primary branches of terrestrial vertebrates that arose in the late Paleozoic. The evolution that led from pre-mammalian cynodonts to mammals is an important transition in vertebrate evolution (Kemp 2005; Kielan-Jaworowska et al. 2004; Luo 2007).

Mammaliaformes underwent multiple episodes of diversification through the Mesozoic, resulting in several now extinct lineages (Kemp 2005; Kielan-Jaworowska et al. 2004; Luo 2007). The timing of the origin of Mammalia is poorly constrained. This is largely due to the incomplete and scarce nature of mammalian fossils from the Jurassic and Early Cretaceous and to disagreements regarding the relationships among basal mammals (Cifelli and Davis 2013). There is current debate about whether mammals originated in the Late Triassic, but are poorly sampled in the fossil record at this time, or much later during the Middle Jurassic, when a diversity of unequivocal mammals first appear in the fossil record (Cifelli and Davis 2013; Zheng et al. 2013; Zhou et al. 2013). Resolving this debate in part hinges on resolving the phylogenetic positions of enigmatic Triassic–Jurassic mammaliaforms, such as the Haramiyida, which may or may not belong to crown group Mammalia. For example, if haramiyidans prove to be within the crown clade Mammalia and represent a paraphyletic group that gave rise to Multituberculata, as has been suggested by some workers (e.g., Butler 2000; Zheng et al. 2013; Bi et al. 2014), then Mammalia must have originated deep in the Triassic.

Once Mammalia originated, several fundamental lineages split from each other during the early stages of mammalian evolution. According to several phylogenetic analyses (Kielan-Jaworowska et al. 2004; Luo et al. 2001; Luo et al. 2002; Rougier et al. 2007), monotremes may fall within Australosphenida, which is sister taxon to the Theriimorpha. The Theriimorpha later diversified into a number of now-extinct groups, such as the multituberculates and eutriconodontans, as well as the trechnotherians, which include the living therian mammals (metatherians and eutherians) and their closest fossil relatives (Averianov et al. 2013; Kielan-Jaworowska et al. 2004; Luo et al. 2007a; Luo et al. 2007b). Much of this diversification probably occurred in Asia (Averianov et al. 2013; Martin and Averianov 2010; Martin et al. 2010) before spreading to other continents during the later Jurassic, although this pattern may be a figment of the relatively complete and well-studied Asian fossil record compared to that of other regions. Trechnotherians underwent three major diversifications, all likely confined to Laurasia, before the origin of the tribosphenic molar (teeth in which a lingual upper cusp occludes into a distal basin on the lower molars; Fig. 2) within the Boreosphenida (Averianov et al. 2013). Each evolutionary radiation of trechnotherians involved successive transformation of the molar form, eventually resulting in the tribosphenic pattern found in the immediate ancestors of metatherians and eutherians (stem boreosphenidans).

Figure 1. 

Phylogeny and diversification patterns of major mammal lineages after Luo (2007).

Figure 2. 

Molar terminology and wear facet designation showing upper and lower molar tooth cusp homologies between the “symmetrodont” Kuehneotherium (A) and the metatherian Kokopellia (B) after Davis (2011a).

Most Mesozoic mammal fossils consist of fragmentary jaws and teeth, which largely explains the intense emphasis that paleontologists place on the evolution of the mammalian dentition. One of the most distinctive and important features that arose in some mammaliaforms is precise molar occlusion. The upper and lower molars of some basal mammaliaforms (e.g., kuehneotheriids) are nearly triangular and have a loosely interlocking fit. From such teeth, basal cladotherians (e.g., Amphitherium; Fig. 2) evolved more precisely occluding teeth in which the lower molars possess a distally projecting talonid, initially consisting of only a single small distal cusp. This talonid may have initially served as an interlocking mechanism with the succeeding lower molars, but it also added another shearing facet between upper and lower teeth during mastication. Many non-tribosphenic trechnotherians (e.g., “symmetrodontans,” dryolestoids, stem zatherians) retained or slightly modified this early tooth morphology in which the talonid supported only a single cusp that is now regarded as homologous with the hypoconid of tribosphenic molars (see Davis 2011a). More advanced non-tribosphenic trechnotherians called “peramurans” evolved a talonid with a second cusp (hypoconulid). Although it still lacked a basin, the expanded talonid participated more fully in mastication with the opposing upper molar.

The final step towards the first tribosphenic molar was the acquisition of a molar talonid basin, which appears to have happened with the acquisition of the protocone (Lopatin and Averianov 2007). One of the most plesiomorphic examples of a talonid basin and protocone is found in the Early Cretaceous basal boreosphenidan Kielantherium gobiense. The protocone is small and the talonid bears two cusps; the hypoconid and hypoconulid.

This two-cusped talonid morphology was modified further in derived tribosphenic mammals. The appearance of the lingual talonid cusp (entoconid) of the lower molar, which helped to close the lingual side of the talonid to form a basin (Davis 2011a) may have functioned to keep food in place within the talonid basin during the final mortar and pestle phase of the chewing stroke (Crompton and Hiiemae 1970; Crompton and Kielan-Jaworowska 1978). These developments resulted in the basic pattern of molar occlusion of living therian mammals, in which lower molars have a basined talonid (crushing heel) and trigonid (shearing end). The tribosphenic molar of therian mammals is not to be confused with the superficially similar pattern found in taxa grouped as australosphenidans and shuotheriids in some phylogenies (Davis 2011b; Luo et al. 2001). Lower molars of those groups (upper molars are not known in australosphenidans) indicate that the talonid (or pseudo-talonid of pseudo-tribosphenic mammals in the case of shuotheriids, see Luo et al. 2007b) may not be completely homologous or functionally identical to the tribosphenic molars of boreosphenidans (see Davis 2011a).

Recent studies in experimental developmental biology (e.g., Harjunmaa et al. 2014) are shedding light on the evolution of the mammalian dentition by showing how genetic modifications can alter developmental pathways that control tooth shape. Although most of these studies focus on mice (Mus musculus), rodents with highly derived dentitions, some studies (e.g., Moustakas et al. 2011) have examined the effects of genetic modification on tooth development and patterning in the extant metatherian Monodelphis domestica (the short-tailed opposum), an animal that retains a relatively plesiomorphic, heterodont dentition. Studies such as these will undoubtedly lead to a better understanding of the molecular mechanisms underlying the evolution of mammalian teeth. The evolution of the distinctive tightly interlocking occlusal molar pattern of tribosphenic mammals was accompanied by necessary changes in tooth ontogeny and growth. In Mammalia, tooth replacement was suppressed further. This is likely related to the determinate growth pattern of the skull (Luo et al. 2004). Teeth are replaced no more than once (diphyodonty) so that an initial set of deciduous teeth were replaced by a second generation of permanent teeth, or teeth, such as molars, were not replaced at all.

Theria is a clade of boreosphenidan mammals that is defined as the most recent common ancestor of extant marsupials and placentals and all of its descendants. The only extant members of boreosphenidans can be grouped into two sister clades, the Metatheria (living marsupials and their close fossil relatives) and Eutheria (living placentals and their close fossil relatives). The most recent and comprehensive molecular clock divergence estimates place the origin of Theria approximately 170–190 million years ago (e.g., dos Reis et al. 2012, 2014). The oldest confidently identified therian fossil is the eutherian Juramaia from the early Late Jurassic (ca. 160 million years old) of China (Luo et al. 2011). Because Metatheria and Eutheria are sister clades, the age of Juramaia implies that metatherians and eutherians must have diverged from each other by 160 million years ago.

We follow the stem-based definition of Kielan-Jaworowska et al. (2004), in which Metatheria is defined as all mammals more closely related to living marsupials (such as kangaroos and opossums) than to living placentals (such as humans and hedgehogs) and monotremes. According to most recent phylogenetic analyses (Luo et al. 2003; Luo et al. 2011; Rougier et al. 2004), Metatheria includes basal stem taxa (e.g., Sinodelphys), the Deltatheroida, and Marsupialiformes, with the latter group consisting of the Marsupialia (crown group metatherians) plus stem metatherians more closely related to Marsupialia than to Deltatheroida (Vullo et al. 2009).

Many of the anatomical features that are used to distinguish living Metatheria (marsupials) and Eutheria (placentals) are based on aspects of “soft anatomy” that do not commonly preserve in fossils and do not have reliable osteological correlates (Kielan-Jaworowska et al. 2004). Many of these features relate to differences in the reproductive system. Marsupials give birth to altricial young after a relatively brief gestation period compared to placental mammals, whose young are born in a more advanced developmental state (precocial). This is at least partly because marsupials have not evolved the ability to prevent immunological recognition and rejection of the young by the mother. Marsupials then undergo a prolonged period of lactation, compared to most placentals. In most, but not all, marsupials this occurs in a pouch.

Although the vast majority of features distinguishing metatherians and eutherians relate to soft tissues, there are limited characters of the dentition, cranium, and postcranial skeleton that diagnose Metatheria (or very proximal groups on the phylogeny). The possession of these features is usually taken as strong evidence that a fossil in question belongs to Metatheria.

Metatherians differ from eutherians in their dental formula and tooth replacement pattern. The metatherian pattern of tooth replacement is postulated to be intimately related to the metatherian reproductive process, which includes an extended period of lactation with nipple fixation following birth of extremely altricial young (Cifelli and Muizon 1997; 1998b; Cifelli et al. 1996b; but see van Nievelt and Smith 2005). However, the identification of tooth loci and resulting tooth homologies remains controversial, as does the identification of the plesiomorphic condition shared by the therian ancestor of metatherians and eutherians (Averianov et al. 2010; Cifelli et al. 1996b; Clemens 1979; Luckett 1993; Marshall and Muizon 1988; Owen 1868; Reig et al. 1987; Rich et al. 2002). These issues are complicated by the paucity of specimens preserving relatively complete tooth rows. Therefore, the dental formula of many taxa is based upon composites constructed from numerous specimens (see Wible et al. 2009). Even among specimens with relatively complete dentitions, there are sometimes differences of interpretation because some specimens may retain deciduous teeth along with their permanent successors.

In recent years, there has been a growing consensus that the plesiomorphic dental pattern for Theria consists of seven to eight postcanine teeth, consisting of four to five premolars and three molars. Many Cretaceous eutherian mammals retain this plesiomorphic dental formula, but the ancestral placental mammal has reduced the premolar number to four, based on a loss of the P3/p3 position (the P3/p3 is variably present in stem therians and early eutherians) (Averianov et al. 2010; Novacek 1986; O’Leary et al. 2013; Szalay 1977). The metatherian dentition differs from this in the loss of a permanent ultimate premolar. Under this dental replacement model, the deciduous ultimate premolar (DP5/dp5) is retained (Fig. 4). The DP5/dp5 has traditionally been identified as the M1/m1 due to its position and molariform morphology; the remaining molars were typically identified as M2–M4. Therefore, the metatherian premolar/molar formula is usually given as 3.4. However, based on this model, the remaining molars are correctly identified (in a development sense) as M1–3 and are therefore homologous with the eutherian M1–3. This pattern was certainly established among at least some metatherians in the Late Cretaceous (Cifelli and Muizon 1998a; Cifelli and Muizon 1998b; Cifelli et al. 1996b; Clemens 1966; Fox and Naylor 2006) and is hypothesized to have been shared by all metatherians (Averianov et al. 2010; O’Leary et al. 2013). The metatherian tooth replacement pattern is considered to be the most derived (in terms of most extreme modification from the ancestral condition) for all basal therian mammals (Luo et al. 2004) and diagnostic for Metatheria (Rougier et al. 1998).

The dentition mesial to P4 is not functionally replaced in Marsupialia. Some workers have argued that vestigial deciduous incisors and canines develop, but are resorbed before eruption of permanent teeth in some marsupials (Luckett 1993; van Nievelt and Smith 2005). The mesial premolars of marsupials are not replaced and some workers (Cifelli and Muizon 1998a; Cifelli et al. 1996b; Luckett 1993) have argued that these teeth should be considered deciduous teeth by default, but it may also be possible that the deciduous teeth were lost and only the successional teeth, homologous with the permanent teeth of eutherians, were present (van Nievelt and Smith 2005). McKenna (1975) suggested that the first premolar position was occupied by a retained deciduous tooth in both metatherians and eutherians. Averianov et al. (2010) suggested that the P1 position is lost in metatherians and the mesial two premolars represent retained P2 and P3. More recently, O’Leary et al. (2013) proposed that the most parsimonious scenario is that the 3rd premolar position was lost in the therian ancestor of both metatherians and eutherians. Accordingly, the first two premolar teeth in metatherians would be P1–P2, whereas the third premolar tooth would be homologous with P4 in non-therian mammals (Fig. 4).

Discrete synapomorphies of Metatheria. A list of metatherian synapomorphies recovered by recent higher-level phylogenetic analyses of mammals is given in Table 1. The list includes synapomorphies identified by Wible et al. (2009), with amendments and corrections from O’Leary et al. (2013). The O’Leary et al. (2013) analysis did not include any Cretaceous metatherians and so the soft-tissue characters included in their analysis (some of which can be assessed based on hard-tissue evidence) are based on two extant marsupial taxa.

Potential synapomorphies of Metatheria. Wible et al. (2009) identified a handful of other features as synapomorphies of Metatheria (Table 2). These were not recovered as explicit synapomorphies of Metatheria by O’Leary et al. (2013), but they acknowledged that these optimizations may change when additional taxa are added to their dataset. It is possible, therefore, that some or all of these features may diagnose Metatheria, or if not, then a very proximal portion of mammalian phylogeny.

Figure 3. 

Molar tooth nomenclature of basal metatherian right M2 (upper) and left m2 (lower). Mesial is to the right, buccal is up. Teeth are based on the Late Cretaceous metatherian Glasbius.

Figure 4. 

Homologies between postcanine teeth of adult basal therian, basal metatherian, and basal eutherian mammals after O’Leary et al. (2013).

Sinodelphys szalayi is based on a partial skeleton compressed on a shale slab. It was originally described as an Early Cretaceous metatherian by Luo et al. (2003). This referral was based largely on the molar formula, which was interpreted to consist of four upper premolars and four upper molars and three lower premolars and four lower molars (P4M4/p3m4). The metatherian affinities of this taxon were corroborated by a phylogenetic analysis in the original paper, as well as subsequent analyses by the same authors (e.g., Luo et al. 2011). Szalay and Sargis (2006) later provided additional evidence for the metatherian position of Sinodelphys based on tarsal morphology, arguing that the exceptionally large and distally extending peroneal processes, and small and steeply angled calcaneocuboid articulations facing mediodistally, supported an assignment to Metatheria. However, the diagnostic utility of these characters has not been established using cladistic analysis.

The metatherian identification of Sinodelphys has not been confidently corroborated by subsequent phylogenetic analyses, often because it is excluded. Sinodelphys was not included in some recent phylogenetic analyses of metatherians because the researchers had not examined the holotype specimen (Averianov et al. 2010; O’Leary et al. 2013; Wible et al. 2009; Williamson et al. 2012). In another recent analysis based on a limited sample of dental characters, Sinodelphys was recovered in a basal polytomy with Metatheria and Eutheria (Vullo et al. 2009), meaning that it is equally most parsimoniously a basal metatherian or a basal eutherian. Other doubts about the metatherian affinities of Sinodelphys have been raised based on anatomical considerations. Averianov et al. (2010) noted that no therians have four upper and three lower premolars, raising the possibility that Sinodelphys may be a more basal, non-therian mammal. Averianov et al. (2010) further suggested that a diastema between the first and second preserved premolars represents a missing tooth, so that the specimen would have had one additional lower premolar, and therefore not the characteristic “three premolars” of metatherians (using the traditional terminology where the homolog of the final premolar is considered a molar, see above). In addition, they suggested that the teeth identified by Luo et al. (2003) as M1/m1 are correctly identified as P5/p5. If correct, then Sinodelphys would have had either five or six premolars, depending on whether the diastema between the first two preserved premolars was filled by another tooth. In either case, Sinodelphys would not have had a classic metatherian dental formula. Sinodelphys also lacks other character states often considered synapomorphic for Metatheria, such as a medially inflected angular process of the dentary. However, this character was not found to be a synapomorphy for Metatheria by O’Leary et al. (2013, see Tables 1 and 2).

We provisionally accept Sinodelphys as a basal metatherian, noting that it does share some features with other metatherians, and most published phylogenetic analyses recover it as a metatherian. If Sinodelphys is one of the earliest diverging metatherian lineages, then it may not be surprising that it lacks certain derived character states that are seen in all later metatherians (such as the inflected angular process and the characteristic metatherian tooth formula). In this scenario, some characters long considered metatherian hallmarks are actually diagnostic of a slightly less inclusive clade that does not include Sinodelphys, and possibly other early-diverging metatherians. Although we accept Sinodelphys as most likely belonging to Metatheria, this must be corroborated by additional phylogenetic analyses that include a large swath of Mesozoic mammals and a broad sampling of dental and non-dental characters.

Deltatheroidans were long regarded as eutherians (Gregory and Simpson 1926; Van Valen 1966) or stem boreosphenidan species (Fox 1974; Kielan-Jaworowska et al. 1979), but are now generally accepted as basal metatherians (Butler and Kielan-Jaworowska 1973; Kielan-Jaworowska and Nessov 1990; Rougier et al. 1998). Deltatheroidans are best known from Asia (Averianov et al. 2010; Rougier et al. 1998) where they were the most common metatherians in the Late Cretaceous. Asian genera include Sulestes from the Turonian of Uzbekistan and Deltatheridium and Deltatheroides from the Campanian of Mongolia and surrounding areas. However, the oldest occurrences are the genera Atokatheridium and Oklatheridium from the Early Cretaceous of North America (Davis et al. 2008; Kielan-Jaworowska and Cifelli 2001). The group persisted, but remained rare, through the Late Cretaceous of North America, surviving into the latest Cretaceous and most likely up to the Cretaceous-Paleogene boundary (Fox 1974; Wilson 2013; Wilson and Riedel 2010).

The sister taxon to Deltatheroida is a large clade that includes most remaining metatherians, including crown Marsupialia. The nomenclature of this clade and its major subclades has a confusing history in the literature, with different authors using various names and various taxonomic groupings, many of which are paraphyletic or polyphyletic based on more recent phylogenetic analyses. The non-deltatheroidan Cretaceous metatherians were placed in Ameridelphia in the influential review of Kielan-Jaworowska et al. (2004), although this taxon was originally established by Szalay (1982) to contain only North and South American taxa, including those from the Cretaceous (Szalay 1994b). However, because most workers consider australidelphians (the South American microbiotheres, including the living Monito del monte [Dromiciops gliroides], the living Australian marsupials, and their closest fossil relatives) to have arisen from Ameridelphia (Muizon and Cifelli 2001), Ameridelphia is paraphyletic as originally conceived (Kielan-Jaworowska et al. 2004). Another group name, Alphadelphia, is also paraphyletic; it is a cohort erected by Marshall et al. (1990) and revised by Case et al. (2005) as a group of stem metatherians on the line to marsupials (crown group metatherians). Furthermore, Didelphimorphia has been used informally as an unnatural group to include only Cretaceous North American metatherians (Kielan-Jaworowska et al. 2004). It must be paraphyletic if North American metatherians gave rise to Marsupialia and extinct South American clades.

More recently, Vullo et al. (2009) erected the taxon Marsupialiformes to refer to the clade that comprises the Marsupialia (crown group metatherians) and all taxa that are more closely related to them than to the Deltatheroida and the most basal metatherians (such as Sinodelphys). Under this taxonomic arrangement, taxa previously grouped together as “Ameridelphia” are a paraphyletic grade of stem marsupialiforms on the line to marsupials. Given the well-established sister-group relationship between the Deltatheroida and the Marsupialiformes, and the paraphyletic nature of once-used names like the Ameridelphia, our preferred taxonomic arrangement uses the Marsupialiformes and disposes of the Ameridelphia, Alphadelphia, Didelphimorphia, and other names that are no longer tenable in phylogenetic taxonomy.

There is a stem grade of early-diverging marsupialiforms that do not fit easily into the major Cretaceous marsupialiform clades. Taxa that appear to populate this grade, based on recent phylogenetic studies, including Adelodelphys, Sinbadelphys, Kokopellia, Arcantiodelphys, Anchistodelphys, Aenigmadelphys, Eoalphadon, Apistodon, Iugomortiferum, and Bistius. Bistius was considered to possibly represent a stem therian by Clemens and Lillegraven (1986), but Williamson et al. (2012) found it to be well nested within metatherians. Similarly, Cifelli and Eaton (2013) recently suggested that Dakotadens and the European Arcantiodelphys are not metatherians, but stem boreosphenidans. Williamson et al. (2012) did not include Arcantiodelphys in their analysis because it was too fragmentary, but they recovered Dakotadens as well nested within Metatheria. Cifelli (1990b) suggested that Iugomortiferum might also represent a stem boreosphenidan.

Recent phylogenetic analyses of Cretaceous metatherian taxa find little resolution of the relationships of the most basal marsupialiform taxa and only weak support for many proposed clades within Marsupialiformes. The Stagodontidae may be a monophyletic group, but inclusion of Pariadens is only weakly supported (Williamson et al. 2012). Although “Alphadontidae” has long been used to group taxa closely related to Alphadon marshi, it is now considered to be either paraphyletic (Kielan-Jaworowska et al. 2004) or polyphyletic (Williamson et al. 2012). Moreover, many taxa once placed within the genus Alphadon have now been split off into other genera (e.g., Protalphadon, Nortedelphys, Varalphadon, Turgidodon Case et al. 2005; Cifelli 1990a; Johanson 1996b). Little phylogenetic resolution was found among these taxa in a recent comprehensive phylogenetic analysis, although some small clusters of monophyletic groups (roughly corresponding to the major “alphadontid” genera) were recovered (Williamson et al. 2012).

Stagodontids are a small clade that includes the Campanian Eodelphis and the Maastrichtian Didelphodon, both from North America. This clade was defined by Williamson et al. (2012) as the most inclusive clade (stem-based) containing Didelphodon vorax, but not Glasbius intricatus, Dakotadens morrowi, Turgidodon praesagus, Herpetotherium fugax, Peradectes elegans, Pediomys elegans, or Didelphis virginiana.

Stagodontids are the largest metatherians of the Mesozoic and in all faunas where they are present they are the among the largest therian species. They are also commonly recognized in the fossil record by their distinctive teeth. The premolars are massive and inflated (Clemens 1966; Clemens 1968a; Clemens 1979; Fox and Naylor 2006; Lofgren 1992). The lower molars have an enlarged paraconid with a well-developed carnassial notch in the paracristid. The upper molar postmetacrista is also well developed, indicating the importance of postvallum/prevallid shear (Fox and Naylor 2006; Muizon and Lange-Badré 1997). The lower molar metaconid and upper molar preparacrista, on the other hand, are relatively small, indicating that prevallum/postvallid shear was relatively less important (Fox and Naylor 2006). However, these shearing features were obliterated through wear in older animals, which may have relied more on a “crushing and grinding” function (Fox and Naylor 1995; 2006).

Unequivocal stagodontids are known only from the middle Campanian through Maastrichtian of North America (Fox and Naylor 2006). However, Cifelli (2004) and Cifelli and Eaton 1987) tentatively referred the Albian-Cenomanian North American Pariadens to Stagodontidae based on its large molar size, relatively large paraconids, and a distinctive lower molar paraconid keel. The phylogenetic analysis of Williamson et al. (2012) supports this assignment, placing Pariadens as the earliest-diverging stagodontid, the immediate outgroup to a clade consisting of Eodelphis and Didelphodon.

The Pediomyidae is one of the few marsupialiform clades that is strongly supported by explicit synapomorphies and high tree-support values in phylogenetic analyses (Davis 2007; Williamson et al. 2012). Williamson et al. (2012) defined this clade as the most inclusive clade (stem-based) containing Pediomys elegans, but not Peradectes elegans, Herpetotherium fugax, Didelphis virginiana, or Didelphodon vorax. Robust phylogenetic evidence that such a clade exists was provided by Davis (2007), who revised the taxonomy and systematics of the Pediomyidae and demonstrated that a suite of taxa historically referred to Pediomys (but now assigned to a variety of species of Pediomys, Protolambda, and Leptalestes) can be grouped together exclusive of other Cretaceous metatherians. The phylogenetic analysis of Davis (2007) found Aquiladelphis and Glasbius as more closely related to the various species of Pediomys than to other Cretaceous metatherians, and therefore as pediomyids following the definition of Williamson et al. (2012). The analysis of Williamson et al. (2012) found Aquiladelphis deeply nested within Pediomyidae but Glasbius as a more distantly related, non-pediomyid taxon.

Pediomyids first appear in the Santonian Milk River Formation of Alberta (Fox 1971) and reach a relatively high taxonomic and morphological diversity in Late Cretaceous faunas in northern North America. Larger-bodied taxa such as Aquiladelphis and Protolambda resemble stagodontids in emphasizing crushing and postvallum/prevallid shear.

Glasbius is one of the most dentally distinctive Cretaceous metatherians (Archibald 1982; Clemens 1966, 1979; Davis 2007; Williamson and Weil 2008; Wilson 2013). Its molars are relatively small, but have low and bulbous cusps that include a large and elongate protocone with a lingually positioned metaconule. The stylar cusps are large, nearly as large as the paracone and metacone, and positioned lingual to the buccal margin of the tooth. The metacone is larger than the paracone. In addition, some of the upper molars also typically possess a mesial basal cingulum, an unusual feature among Cretaceous metatherians. The lower molars possess a broadly expanded talonid, a buccal cingulid that sometimes sports accessory cuspules, a cristid obliqua that meets the trigonid buccal to the protocristid notch, and an ultimate molar that is reduced in size relative to the other molars.

Glasbius has been found as a close relative or member of Pediomyidae by several phylogenetic analyses, including the analysis of Rougier et al. (1998) and its many offshoots (e.g., Averianov et al. 2010; Rougier et al. 2004; Wilson and Riedel 2010) and that of Davis (2007). Case and others allied Glasbius with Hatcheritherium, a poorly known latest Cretaceous taxon represented only by upper molars (Case et al. 2005). However, Williamson et al. (2012) did not find support for either of these relationships. Instead, their phylogenetic analysis placed Glasbius as the sister taxon to the South American Paleogene taxon Roberthoffstetteria, corroborating the close relationship between these two genera noted previously by several workers (Case et al. 2005; Goin et al. 2003; Marshall et al. 1990). If this topology is correct, it would therefore require that the Glasbius + Roberthoffstetteria clade crossed the K-Pg boundary; one of the few metatherian mammal lineages to do so (Williamson et al. 2012).

The taxon Glasbiidae was not defined by Williamson et al. (2012) and has not to our knowledge previously been defined. To provide nomenclatural stability and make it more explicit to assign taxa to Glasbiidae based on phylogenetic analyses, we here restrict Glasbiidae to the genus Glasbius. According to the topology of Williamson et al. (2012), Roberthoffstetteria, a South American taxon that is commonly considered a basal polydolopimorph, is the sister taxon to Glasbiidae. Clemens (1966) in his original description of the genus considered a wide array of possible taxonomic affinities for Glasbius, with particular attention to what he concluded was convergent evolution of bunodont molars within the South American Microbiotheriidae and Caroloameghiniidae. Some authors have since rejected this notion (but see Kirsch et al. 1997) and variously argued for close phylogenetic relationship of Glasbius to either the caroloameghiniids (e.g., Marshall et al. 1990), more broadly within (Case et al. 2005) or sister to the Polydolopimorphia (Goin et al. 2003), or within the Paucituberculata (Kielan-Jaworowska et al. 2004). Kirsch et al. (1997) joined the Polydolopimorphia and Paucituberculata (including the living caenolestids) into a new cohort Pseudiprotodontia, although they excluded Glasbius from this arrangement. In a recent cladistic analysis of bunodont metatherians, Goin et al. (2009) rejected the concept of Pseudiprotodontia and instead recovered a clade comprising Microbiotherium, Glasbius, and Polydolopimorphia (including Roberthoffstetteria) that was supported by three synapomorphies. Similarly, a consensus tree from Oliveira and Goin (2011) also showed a possible sister-taxon relationship between the Microbiotheria and the Polydolopimorphia (including Roberthoffstetteria). Because these cladistic analyses do not have comprehensive taxon sampling or well-supported trees, the possible phylogenetic relationships between Glasbius and South American taxa (polydolopimorphs, paucituberculatans, and microbiotherians) should be viewed cautiously.

Herpetotheriidae was defined by Williamson et al. (2012) as the most inclusive clade (stem-based) containing Herpetotherium fugax, but not Peradectes elegans, Pediomys elegans, Didelphodon vorax, or Didelphis virginiana. This is one of the most species-rich North American metatherian clades of the Paleogene (Korth 2008), and they are represented by nearly complete skulls and skeletons (Horovitz et al. 2008; Horovitz et al. 2009; Sánchez-Villagra et al. 2007). Recent phylogenetic analyses that include relatively complete cranial material of Paleogene peradectid (Mimoperadectes) and herpeotheriid taxa indicate that herpetotheriids are the sister taxon to crown clade Marsupialia (Horovitz et al. 2009).

Unequivocal herpetotheriids are relatively common in the fossil record of North America after the Cretaceous-Paleogene boundary. There is some evidence, however, for more basal herpetotheriids in the Cretaceous. Case et al. (2005) argued that the poorly known Campanian taxon Ectocentrocristus represents a herpetotheriid, along with another genus, Nortedelphys, that includes three species from the Maastrichtian of North America. Ectocentrocristus remains problematic because it is based on what may be a deciduous P4, possibly of a Turgidodon-like taxon (Beck et al. 2008; Cifelli in Kielan-Jaworowska et al. 2004; see Williamson et al. 2012). The phylogenetic analysis by Williamson et al. (2012) and here (below) supports the placement as a basal herpetotheriid of Ectocentrocristus, but not Nortedelphys, which is found to be a more basal marsupialiform. This provides additional evidence for a Late Cretaceous origin of Herpetotheriidae, suggesting that this clade crossed the K-Pg boundary.

This clade was defined by Williamson et al. (2012) as the most inclusive clade (stem-based) containing Peradectes elegans, but not Herpetotherium fugax, Pediomys elegans, or Didelphis virginiana. This clade comprises almost entirely post-Cretaceous taxa, and was recently hypothesized to be the sister taxon of living opossums (Didelphidae) and other members of crown clade Marsupialia based on a comprehensive higher-level phylogenetic analysis of fossil and living marsupials (Horovitz et al. 2009). Historically, workers have assigned various Cretaceous species to Peradectidae, including some species of Alphadon (Crochet 1979; 1980). These species are either not clearly referable to Peradectidae based on phylogenetic analyses (e.g., Williamson et al. 2012) or are now considered as Paleocene-Eocene in age (Sigé et al. 2004).

The phylogenetic analysis of Williamson et al. (2012) recovered a massive polytomy comprising many species traditionally referred to Peradectidae as well as many non-peradectid metatherians. Because of this, when the phylogenetic definition of the clade is strictly applied according to the Williamson et al. (2012) topology, only two taxa fall into a restricted Peradectidae (the name-bearing definitional taxon Peradectes elegans and its sister taxon Peradectes californicus). Williamson et al. (2012) discussed how this result was likely due to missing data in several taxa, and anticipated that future phylogenetic analyses would successfully group together all or most traditionally regarded peradectids into a clade. This clade may or may not include the Cretaceous taxon Maastrichtidelphys.

As for the vast majority of Cretaceous mammals, Cretaceous metatherians are usually represented by isolated teeth. Therefore, most phylogenetic analyses and taxonomic arrangements rely heavily on dental characters, especially size and relative size, shape, and position of various tooth features (Fig. 5). Identification is complicated by the change in tooth shape according to position within the tooth row, but teeth typically change in a predictable fashion. Further, few upper dentitions are found in direct association with lower dentitions. Fortunately, because molars of Mesozoic therian mammals have precise occlusion, the shape of upper molars can be used to predict the shape of the occluding lower molars and vice versa. Moreover, during mastication, upper and lower molars contact each other, resulting in distinctive wear facets. These facets can provide clues to the shape of the opposing teeth and factors regarding the mandibular movement and occlusion (Crompton and Hiiemae 1970; Crompton and Kielan-Jaworowska 1978). Teeth of metatherians form an intricate functional complex that is intimately related to diet (Gordon 2003b) and tooth size is highly correlated with mass of the animal (Gordon 2003a).

Although teeth can usually be used to confidently distinguish metatherians from other mammals and give some insight into body size and diet, there are difficulties associated with using teeth for higher-level metatherian taxonomy and phylogeny. Some of the differences used to distinguish teeth of various taxa are subtle. In addition, some taxa are represented only by small samples, in some cases only a few fragmentary teeth. Therefore, the range of morphological variation within some taxa is poorly understood. This, no doubt, partly explains why there is considerable disagreement and skepticism surrounding some identifications of some taxa and the validity and taxonomic decisions regarding synonymies of other taxa (see Eaton 2013; Eaton and Cifelli 2013; Williamson et al. 2012 for recent discussions regarding the taxonomy of metatherian taxa). This also helps explain rampant homoplasy and polytomies in some phylogenetic analyses, as well as why competing analyses with different taxon- and character-sampling regimes produce different results.

Figure 5. 

Representative upper and lower molars of Cretaceous and early Paleogene metatherians. Scale bars equal 1 mm.

Metatherians underwent a substantial radiation during the Late Cretaceous, and while several Cretaceous metatherians are known from the Late Cretaceous of Europe and Asia (e.g., Averianov et al. 2010; Martin et al. 2005; Rougier et al. 2004; Szalay and Trofimov 1996; Vullo et al. 2009), the majority are known from North America where they are represented almost exclusively by jaws and teeth.

The interrelationships of most major metatherian subclades are unresolved. Several recent cladistic analyses have looked at the higher-level phylogenetic relationships of mammals and have included some Cretaceous and Paleogene metatherian taxa (Ji et al. 2002; Kielan-Jaworowska et al. 2004; Luo et al. 2003; Luo et al. 2002). Other analyses have focused on the relationships within Metatheria and the position of Metatheria within Mammalia (Horovitz et al. 2008; Horovitz and Sánchez-Villagra 2003; Rougier et al. 2004; Sánchez-Villagra et al. 2007), and others have explored the relationships of a small number of Cretaceous taxa within Metatheria (Davis 2007; Johanson 1996a; Martin et al. 2005; Vullo et al. 2009).

Williamson et al. (2012) provided the first analysis to include nearly complete, species-level coverage of Cretaceous and early Paleogene metatherian taxa. The resulting phylogeny was used to examine the pattern of metatherian survivorship across the K-Pg boundary and the origin and evolution of early Paleogene North American metatherian taxa. The dataset consists only of dental characters, because the majority of Cretaceous-Paleocene metatherians are represented solely by teeth. We here present a revised analysis of Mesozoic and early Paleogene metatherian taxa built upon that of Williamson et al. (2012).

The revised analysis presented here includes 95 taxa and 83 characters. It adds three non-metatherian taxa to those included in the Williamson et al. (2012) analysis: Juramaia sinensis, Asioryctes nemegetensis, and Ukhaatherium nessovi. The analysis thus includes four basal eutherians, six deltatheroidans, 59 Cretaceous marsupialiaforms, and 26 Paleogene marsupialiforms. Juramaia is used as the outgroup taxon. We have modified the character descriptions to reflect the postcanine dental homologies as proposed by O’Leary et al. (2013). Based on this, the first tooth in the molar series in Metatheria is homologous to the deciduous p5 (dp5) of eutherians (O’Leary et al. 2013) (Fig. 4).

We added five new characters (characters 7, 15, 51, 52, and 68; see Suppl. material 1), slightly changed character 85 (now character 83), and deleted one character (Williamson et al. 2012, character 15). As with the previous Williamson et al. (2012) analysis, we excluded taxa that are represented only by lower dentitions that are very incomplete (e.g., Esteslestes, Arcantiodelphys), or that we believe are invalid (see Williamson et al. 2012). We have also excluded Pappotherium and Holoclemensia, poorly known taxa that may be outside of Metatheria (see Williamson et al. 2012) and that acted like wildcard taxa when included in a preliminary version of our analysis.

Minor changes or corrections were made to the scoring of some taxa from the Williamson et al. (2012) dataset; a total of 10 changes are listed in Suppl. material 1.

We conducted a parsimony analysis using TNT v. 1.1, September 2013 (Goloboff et al. 2008). We first analyzed the matrix (Suppl. material 2) under the “New Technology Search” with Sectorial Search, Rachet, Drift, and Tree Fusing, finding the minimum length tree 10 times. The 21 recovered trees were then analyzed under traditional tree bisection reconnection (TBR) swapping, to more extensively explore each tree island. This resulted in 2,008 most parsimonius trees of length 500 (consistency index = 0.210; retention index = 0.703).

The skulls of Asiatherium and Sinodelphys are relatively crushed, making reconstruction problematic. Extant didelphid marsupials and some early Paleocene taxa retain many plesiomorphic therian features and their skulls closely resemble the limited fossil cranial material of Cretaceous metatherians in many respects, meaning that they can be used as a guide for reconstructing Cretaceous metatherian skull morphology (Wible 1990, 2003; Argot 2001) (Fig. 8). In particular, Pucadelphys and Mayulestes from the early Paleocene of Bolivia are useful models for approximating the cranial anatomy of Cretaceous taxa, although they are probably more closely related to extant Marsupialia than to any Cretaceous taxon (Horovitz et al. 2009; Luo et al. 2003; O’Leary et al. 2013). Their skulls have both been described in detail (Pucadelphys: [Macrini et al. 2007; Marshall and Muizon]; Mayulestes [Muizon 1998]), as has the skull of the extant didelphid Monodelphis brevicaudata (the red-legged short-tailed opossum) (Wible 2003). These are all similar in size to many Cretaceous taxa, including Asiatherium, which had a skull of about 3 cm in length, about the same as that of the extant didelphid Monodelphis domestica (Szalay and Trofimov 1996).

The skulls of Cretaceous metatherians have a pronounced postorbital constriction of the cranium when seen in dorsal view, a braincase with well-developed sagittal and nuchal crests and high and robust zygomatic arches. However, these sagittal and nuchal crests were probably not as highly developed as they are in the living didelphid Didelphis, which is similar in size to Didelphodon, the largest Cretaceous metatherian (Gordon 2003a). Asian deltatheroidans as well as Didelphodon had a relatively short rostrum compared to extant didelphids.

The mammalian palate has incisive foramina situated within the premaxilla, and these are particularly large in metatherians compared to basal eutherians, extending posteriorly to the premaxillary-maxillary suture. In addition, many basal metatherians have two pairs of large palatine vacuities. The larger of these is anterior to the anterolateral part of the maxillary-palatine suture, and there is a smaller pair on the maxillary-palatine suture, at the posterolateral margins of the palatal exposure of the palatine. However, these are variably present in basal metatherians, living marsupials, and basal eutherians, so it is not clear if these are a synapomorphy of Metatheria or just a feature seen in many (but not all) species. The pterygoids of metatherians are not fused to the alisphenoids as they are in eutherians. Metatherians have a relatively larger alisphenoid than do eutherians. In some taxa, the alisphenoid contributes to the glenoid fossa and in others it extends posteriorly to form a tympanic process that partially conceals the petrosal as an alisphenoid bulla. An alisphenoid bulla is present in the undescribed Guriliin Tsav skull, Asiatherium, Didelphodon, and many extant marsupials, suggesting that it may be a feature that evolved at or near the base of Metatheria, potentially as a synapomorphy uniting the clade. However, it is lacking in Pucadelphys and Mayulestes, possibly indicating that it has evolved multiple times within Marsupialia (Muizon 1994).

Ear region. The incorporation of postdentary bones into the basicranial region of the braincase is one of the most extraordinary examples of morphological evolution in the fossil record (see Luo 2007). This transition independently occurred multiple times, including in a recent common ancestor of metatherians and eutherians. This means that metatherians have a single bone in the lower jaw (the dentary) and that the numerous postdentary bones of more basal mammaliaforms have been modified and many incorporated into the ear region. The bones of this region of the mammalian skull are morphologically complex and contain evidence of the cranial neurovasculature. Consequently, the ear region has been an important area of focus for assessing both the interrelationships of metatherians and the position of Metatheria within higher-level mammalian phylogeny.

Extant placental and marsupial mammals possess very different basicranial vasculature (Wible 1990). Stem therians possessed two major vessels running through the middle ear: a lateral head vein and a stapedial artery, which is a branch of the medial carotid artery. Eutherians possess a basicranial vasculature in which the prootic sinus and the lateral head vein completely disappear by adulthood. A new vein, the capsuloparietal emissary vein, leaves the cranial cavity through the capsuloparietal foramen and exits the skull through the postglenoid foramen (Wible 1990). In extant marsupials, and evidently also in some Cretaceous metatherians (Wible 1990), the posterior or tympanic portion of the lateral head vein becomes reduced during ontogeny and in most extant taxa, is totally lacking in the adult (Wible 1990, 2003; Wible et al. 2009). A new vessel, the sphenoparietal emissary vein, forms in marsupials as a new outlet for the prootic sinus.

Fossils may help us to better understand the evolution of this basicranial vasculature in metatherians and other early mammals. The mammalian basicranial region, especially the petrosal bone, can preserve osteological correlates associated with each vessel of the basicranial vasculature, meaning that their presence or absence can be assessed in well-preserved fossils.

Wible described several isolated petrosals from the Bug Creek Anthills of Montana, some of which he assigned to Cretaceous metatherians (Wible 1990). He found that the most complete Cretaceous petrosal, which he referred to as ‘Type A’, lacks the vascular sulci for the stapedial artery and contains a short prootic canal. In fact, the vascular grooves and canals of the Type A petrosal (Fig. 9) are essentially the same as in some recent didelphids including Didelphis, and he therefore concluded that it possessed the same fundamental basicranial vasculature pattern. This same pattern has also been found in the deltatheroidan metatherians Deltatheridium (Rougier et al. 1998) and Sulestes (Averianov et al. 2010). It appears, therefore, that this pattern (the carotid artery takes a medial course, the stapedial artery is absent) may characterize Metatheria. Indeed, Rougier et al. (1998), Kielan-Jaworowski et al. (2004), Wible et al. (2009) and other authors have considered this pattern to be a diagnostic character of Metatheria. Based on phylogenetic character optimization, however, O’Leary et al. (2013), concluded that this pattern is actually an ambiguous synapomorphy for Theria that is later reversed in Eutheria. Therefore, it would be plesiomorphic for Metatheria, not apomorphic. Because the O’Leary et al. (2013) analysis had a limited sampling of metatherian taxa and did not include the stem cladotherian Vincelestes or any Cretaceous metatherians, this optimization is subject to change with additional fossil evidence.

Mandible. The dentary of most Cretaceous metatherians has a shallow, horizontally oriented, and unfused mandibular symphysis. The canines diverge laterally, which is a synapomorphy of metatherians that was identified in the analysis by O’Leary et al. (2013; character 1411). An inflected angular process on the medial surface of the posterior mandible, which is so large that it forms a shelf, is often regarded as a synapomorphy for Metatheria, but it is absent from Sinodelphys which instead has a rod-like angular process that projects posteriorly (Luo et al. 2003) as in eutherian taxa. Therefore, the inflected angular process is a synapomorphy of a more exclusive group within Metatheria (e.g., Marsupialiformes + Deltatheroida), and as such its presence in a fossil is still a clear indication that the fossil is a metatherian.

Metatherians typically have several mental foramina on the mandible, and the posterior-most one is located below a position between p4 and dp5 (following the tooth terminology of O’Leary et al. 2013) or more posteriorly (this feature could not be corroborated as a synapomorphy of Metatheria, but was found to be an ambiguous synapomorphy of Theria by O’Leary et al. [2013]). Lack of a mandibular foramen was listed as a synapomorphy for Metatheria in some phylogenies (Rougier et al. 1998; Wible et al. 2009). However, Cifelli and de Muizon (1997) identified a small lateral mandibular foramen near the base of the masseteric fossa in the middle Cretaceous metatherian Kokopellia, and suggested that one might also be present in a specimen of Alphadon. However, no lateral mandibular foramen has been found in other metatherian specimens, including deltatheroidans (Averianov et al. 2010; Wilson and Riedel 2010). O’Leary et al. (2013) were also unable to corroborate this as a synapomorphy, but optimizations of this character might change with the incorporation of additional taxa in their matrix. Kokopellia may also preserve a remnant of a Meckelian groove on its lingual face (Cifelli and Muizon 1997), but this feature has not been detected in other metatherian specimens.

Associated partial postcranial skeletons (Fig. 7) are known for only two Cretaceous metatherian taxa: Asiatherium (Szalay and Trofimov 1996; Trofimov and Szalay 1994) and Sinodelphys (Luo et al. 2003). Szalay (1994a; b) referred isolated tarsal bones from the latest Cretaceous of western North America to “Pediomys” (probably Protolambda, following Davis [2007] and Didelphodon). These bones were collected from the Bug Creek Anthills, now regarded as a mixed assemblage of reworked latest Cretaceous and earliest Paleocene fossils (Lofgren 1995). We assume that Szalay (1994a, b) referred the isolated tarsals to “Pediomys” based on their large size and presumed Cretaceous age. They are too large to belong to Thylacodon montanensis, the only earliest Paleocene metatherian reported from eastern Montana (e.g., Wilson 2013; 2014). Longrich (2004) described isolated caudal vertebrae from latest Cretaceous North America as Didelphodon, but the referral of these and other isolated bones to Metatheria has been questioned by Fox and Naylor (2006). Horovitz illustrated a partial calcaneum of Deltatheridium from the Late Cretaceous of Mongolia (Horovitz 2000). More recently, Szalay and Sargis (2006) have described isolated tarsal elements including several calcanea and one partial astragalus, from the Bissekty local fauna of the middle Cretaceous of Uzbekistan. Other isolated elements described from the Bissekty Formation are an isolated partial humerus (Chester et al. 2010) and two partial femora (Chester et al. 2012). It is unclear how many taxa these various postcranial bones from the Bissekty Formation represent; they may belong to as many as four species, or alternatively may all belong to one taxon, Sulestes karakshi (Averianov et al. 2010; Chester et al. 2010; Chester et al. 2012; Szalay and Sargis 2006).

Axial skeleton. Little is known of the Cretaceous metatherian axial skeleton as it is only incompletely preserved in two specimens, the holotypes of Asiatherium and Sinodelphys. Asiatherium was found to lack an anticlinal vertebra and has two fused sacral vertebrae (Szalay and Trofimov 1996).

Scapula. Damaged scapulae are preserved with the partial skeletons of Asiatherium and Sinodelphys. The scapula of Asiatherium (Fig. 7) possesses a long and prominent scapular spine and a prominent acromion process (Szalay and Trofimov 1996). The scapula of Sinodelphys has a much wider supraspinous fossa than infraspinous fossa at midlength, with a strongly sigmoidal cranial profile (Luo et al. 2003); a feature found in Paleogene metatherians, but absent in Cretaceous eutherians (Luo et al. 2003). The scapula of Asiatherium differs in having a wider infraspinous fossa than supraspinous fossa at midlength and a nearly straight cranial margin.

Clavicle. A robust clavicle is preserved in both Asiatherium and Sinodelphys.

Humerus. Asiatherium and Sinodelphys both preserve humeri. In Asiatherium (Fig. 7) the humerus has a distinct and well-developed greater tuberosity and a large deltoid crest that extends about a third of the length of the bone. The distal end of the humerus has a relatively large and rounded capitulum and a large and well-developed supinator crest and capitular tail. The humerus of Sinodelphys has a similarly large supinator crest (Luo et al. 2003).

The three distal humerus fragments from the Bissekty Formation (Fig. 11) of Uzbekistan possess a spherical capitulum, a trochlea separated from the capitulum, and a well-developed lateral epicondylar crest. One fragment that might represent a third metatherian taxon possesses a wide capitular tail (Chester et al. 2010). The presence of a capitular tail is recognized as a synapomorphy for Metatheria by O’Leary et al. (2013; ch. 3072).

Radius and ulna. The head of the radius of Asiatherium (Fig. 7) is oval rather than round as in most didelphids (Szalay and Trofimov 1996; Trofimov and Szalay 1994). The anterior edge of the olecranon process of the ulna twists medially as in terrestrial didelphids and caenolestids (Trofimov and Szalay 1994).

Carpus. The carpals of both Sinodelphys and Asiatherium (Figs 7, 11) have a large cuneiform, a large unciform, and scaphoid and a small, bean-shaped trapezium, unlike non-metatherian therians and eutherians (Luo et al. 2003). In living didelphids, the enlarged hamate, triquetrum, and scaphoid in the carpus are thought to be related to increased capacity for gripping (Luo et al. 2003).

Pelvis. Asiatherium possesses rod-shaped ilia and epipubic bones (Fig. 7). Although often erroneously thought to be a feature of marsupials that evolved in order to support the pouch, epipubic bones are plesiomorphic for Theria and are found in basal eutherians, as well as non-therian mammals such as monotremes and multituberculates (Novacek et al. 1997; Reilly et al. 2009). Although they do support the pouch in living marsupials, epipubic bones may also relate to abdominal muscle function (Reilly et al. 2009).

Femur. The femur of Asiatherium (Fig. 7) lacks a distinct patellar groove, indicating that it lacked a patella, a common condition in extant marsupials (see Szalay and Sargis 2001). It lacks a third trochanter as in all extant marsupials except caenolestids and peramelids (Szalay and Sargis 2001). The lateral condyle of the femur is larger than the medial condyle. Chester et al. (2012) described two isolated metatherian femora from the Bissetky Formation (Fig. 12). These also lack a distinct patellar groove and a third trochanter. The lack of a patellar groove (and patella) may indicate arboreal ancestry (Argot 2004; Chester et al. 2012; Szalay and Sargis 2001).

Fibula. The fibula of Asiatherium (Fig. 7) contacts the distal femur as is typical of living marsupials, but it is not proximally expanded as in extant didelphids.

Tarsus. Associated tarsals of a Cretaceous metatherian are known only for Sinodelphys. (Fig. 13G, J–K). Isolated tarsals from the Late Cretaceous sediments and referred to “Pediomys” (likely Protolambda sp. following the revision of pediomyids of Davis [2007]) were described by Szalay (1994a)(Fig. 13L–M). Isolated tarsal bones tentatively referred to the deltatheroidan Sulestes (Fig. 13A–C, D–F) were described from the Bissetky Formation of Uzbekistan by Szalay and Sargis (2006).

The astragalus of Cretaceous metatherians is distinct from other mammals because the neck is asymmetrical, with the navicular facet of the astragalus extending medially along the length of the neck (Fig. 13).

The calcaneus of Cretaceous metatherians (Fig. 13) possesses a calcaneocuboid facet that is oblique (approximately 60 degrees) to the long axis of the calcaneus, a calcaneoastragalar facet that is nearly vertical, and a sustentacular facet that is medially facing (Luo et al. 2003; Szalay 1994a; Szalay and Sargis 2006) and adjacent to a large distally positioned plantar tubercle. This may be related to the habitual inversion of the distal part of the pes (Luo et al. 2003). The sustentacular process forms a pointed triangle.

The navicular of Sinodelphys (Fig. 13G) is transversely wide and short and the cuboid facet faces mediodistally.

Figure 7. 

Cast of holotype of Asiatherium reshetovi. A postcranial skeleton B close-up of right forelimb C ventral view of skull. Scale bars equal 1 cm.

Figure 8. 

Skull of Monodelphis brevicaudata, modified from Wible (2003). Skull (A–D) and mandible (C) in dorsal (A), ventral (B), left lateral (C), and caudal (D) views. Abbreviations: an angular process; as alisphenoid; astp alisphenoid tympanic process; bo basioccipital; bs basisphenoid; coc coronoid crest; con mandibular condyle; cor coronoid process; ctpp caudal tympanic process of the petrosal; ec ectotympanic; eo exoccipital; fc fenestra cochleae; fm foramen magnum; fr frontal; ham pterygoid hamulus; inf incisive foramen; iof infraorbital foramen; ip interparietal; jf jugular foramen; ju jugal; lac lacrimal; lacf lacrimal foramen; m malleus; maf masseteric fossa; mapf major palatine foramen; mf mental foramen; mpf minor palatine foramen; mx maxilla; na nasal; oc occipital condyle; pa parietal; pal palatine; pcp paracondylar process of the exoccipital; pe petrosal; pmx premaxilla; pgp postglenoid process; ppt postpalatine torus; pptf foramen in the postpalatine torus; ps presphenoid; pt pterygoid; ptn posttemporal foramen; ptp posttympanic process; rtpp rostral tympanic process of the petrosal; so supraoccipital; sq squamosal; tl temporal line.

Figure 9. 

Right petrosal Type A (Protolambda) after Wible (1990) showing reconstruction of vasculature. Abbreviations: ac aqueductus cochleae; adm arteria diploetica magna; cev capsuloparietal emissary vein; cp crista parotica; ctpp caudal tympanic process of petrosal; er epitympanic recess; fc fenestra cochleae; fi fossa incudis; fs facial sulcus; fv fenestra vestibuli; hF, hiatus Fallopii; If, lateral flange of petrosal; lhv lateral head vein; It, lateral trough of petrosal; me mastoid exposure; mp mastoid process; pc prootic canal; pcv vein ofprootic canal; pff primary facial foramen; pp paroccipital process; pr promontorium; ri ramus inferior of stapedial artery; rs ramus superior of stapedial artery; rtpp rostral tympanic process of petrosal; sa stapedial artery; sev sphenoparietal emissary vein; sf stapedius fossa; sff secondary facial foramen; sips sulcus for inferior petrosal sinus; spd sulcus for perilymphatic duct; vdm vena diploetica magna.

Figure 10. 

Skeleton of Monodelphis domestica. Scale bar is 1 cm.

Figure 11. 

Humerus and carpi of Cretaceous metatherians. A Distal humerus of unidentified metatherian (cf. Sulestes?) from the Bissetky Locale, Uzbekistan after Chester et al. (2010) in anterior view. Carpus of the Cretaceous stem therian Eomaia (B) and the Cretaceous metatherians Sinodelphys (C) and Asiatherium (D) after Szalay and Trofimov (1996) and Luo et al. (2003). Scale bars equal 1 mm.

Figure 12. 

Partial femora of unidentified metatheria (cf. Sulestes?) from the Bissetky Locale, Uzbekistan after Chester et al. (2012). A–C right distal femur in anterior (A), posterior (B), and distal (C) views D–F right proximal femur of a possible unidentified metatherian in proximal (D), anterior (E), and posterior (F) views.

Figure 13. 

Tarsus of the Cretaceous stem therian Eomaia and Cretaceous metatherians. A–F isolated calcaneum (A–C) (cf. Sulestes?) from the Bissetky Locale, Uzbekistan in dorsal (A), ventral (B), and distal (C) views after Szalay and Sargis (2006). D–F isolated astragalus (cf. Sulestes?) from the Bissetky Locale, Uzbekistan in dorsal (D), ventral (E), and distal (F) views after Szalay and Sargis (2006). G articulated partial pes of Sinodelphys in ventral view H–I calcaneum and astragalus, respectively, of Eomaia in ventral view J–K calcaneum and astragalus of Sinodelphys, respectively, in ventral view L–M calcaneum and astragalus of Protolambda, respectively, in ventral view G–M are after Luo et al. (2003). Scale bars equal 1 mm.

The most recent molecular clock analyses estimate the marsupial-placental split between ~140 and 190 million years ago, depending on the tree topology, calibrations, and rate models used (Bininda-Emonds et al. 2007; dos Reis et al. 2014; dos Reis et al. 2012; Meredith et al. 2011; Phillips et al. 2009; van Rheede et al. 2006). The recent study of dos Reis et al. (2012), which is the most rigorous and comprehensive yet published, estimated this split at approximately 168–178 million years ago, which corresponds to the late Early Jurassic-early Middle Jurassic. Most of these studies were conducted prior to the discovery of what is currently considered the oldest unequivocal eutherian fossil, the ca. 160 million-year-old holotype of Juramaia (Luo et al. 2011). If Juramaia is correctly identified as a eutherian, then the marsupial-placental split must have occurred prior to 160 million years ago. The dates provided by dos Reis et al. (2012) therefore comply with the age of Juramaia.

Many of these molecular clock studies also estimate the origin of Marsupialia. dos Reis et al. (2012) estimated marsupials to have originated 64–84 million years ago. Previous studies also reported similar age ranges. For example, Bininda-Emonds et al. (2007) estimated this divergence at 71.4–93.7 million years ago, and Meredith et al. (2011) at 67.9–97.2 million years ago. These findings generally agree that the origin of Marsupialia likely occurred in the Late Cretaceous, although there is a small possibility using the dos Reis et al. (2012) results that marsupials were a strictly post-Cretaceous radiation.

Juramaia sinensis was recently described as the oldest eutherian mammal, and therefore the oldest therian as well (Luo et al. 2011). Luo et al. (2011) estimated its age to be ca. 160 Ma based on radiometric dates of between about 159 Ma and 165 Ma for the Tiaojishan Formation, the rock unit in China in which the holotype was discovered (Liu and Liu 2005; Liu et al. 2006). More recent dates have further constrained the age of the horizon in which Juramaia was discovered to between 160.5–161.0 million years ago (Liu 2012). This would place Juramaia in the earliest part of the Late Jurassic (Gradstein et al. 2012). This is substantially older than the previously regarded oldest eutherian, Eomaia, from the Early Cretaceous (ca. 125 million years old) of China. It is also older than what were previously considered to be the oldest boreosphenidan taxa (Tribactonodon and Aegialodon) from the Lower Cretaceous of England (Berriasian Purbeck Group and Valangininan Wealden Group, respectively) and similar in age to the oldest australosphenidan (Ambondro) from the Middle Jurassic of Madagascar (Flynn et al. 1999).

The oldest proposed metatherian, Sinodelphys szalayi, is Early Cretaceous in age (ca. 124–131 million years old [Yang et al. 2007]). However, some workers (e.g., Averianov et al. 2010) have voiced doubts about whether this taxon has the classic metatherian dental formula, as was originally described by Luo et al. (2003). If its dental formula is different from the metatherian norm, it may mean that Sinodelphys is not a metatherian (see above). Disregarding Sinodelphys, the oldest certain metatherians are the deltatheroidans Oklatheridium szalayi and Atokatheridium boreni from the Lower Cretaceous (Albian, ca. 104–108 million years old) Antlers Formation of Oklahoma (Averianov et al. 2010; Cifelli and Davis in press; Davis et al. 2008; Kielan-Jaworowska and Cifelli 2001) and of an indeterminate genus and species from the Lower Cretaceous (Albian) Cloverly Formation of Montana (Cifelli and Davis in press). These are older than the oldest unequivocal deltatheroidans from Asia, which are Sulestes karakshi and Sulestes sp. from the Upper Cretaceous (Turonian, ca. 94–90 million years old) Bissekty Formation of Uzbekistan (Averianov et al. 2010; Kielan-Jaworowska and Nessov 1990).

The first appearance of marsupialiforms (i.e., non-deltatheroidan metatherians) in North America is near the Albian-Cenomanian boundary, at the base of the Late Cretaceous, approximately 100 million years ago (Cifelli 2004; Cifelli and Muizon 1997; Eaton 1993; Cifelli and Eaton 1987). Arcantiodelphys, a marsupialiform from the Cenomanian of Europe (Vullo et al. 2009), is nearly as old.

The oldest unequivocal crown marsupials appear in the early Paleocene with Peradectes and other taxa (Horovitz et al. 2009; O’Leary et al. 2013; Williamson et al. 2012). The South American Paleocene Pucadelphys was considered to be outside of Marsupialia by some workers (Horovitz et al. 2009), but within the crown clade by O’Leary et al. (2013). The latest Cretaceous North American Glasbius may represent a Cretaceous member of Marsupialia if it is found to be a member of, or sister taxon to, the extinct, mostly South American clade, Polydolopimorphia (Goin et al. 2003), which some workers have allied with the extant clade Caenolestidae, in Paucituberculata and other workers have allied with the extant Microbiotheria (Horovitz and Sánchez-Villagra 2003; Kirsch et al. 1997).

Both molecular clock estimates and fossil evidence (if correctly identified) currently agree that the metatherian-eutherian split, and therefore the origin of Metatheria, occurred by 160 million years ago. Crown-group metatherians (Marsupialia) probably originated within the 10 million years before the Cretaceous-Paleogene boundary, although their oldest unequivocal fossils are from the earliest Paleocene and molecular clocks cannot completely rule out a post-Cretaceous origin. Both deltatheroidans and marsupialiforms first appear in the fossil record in the middle part of the Cretaceous, but these are minimum divergence estimates.

Reconstruction of the phylogenetic relationships of Cretaceous and Paleogene metatherians is important for understanding the pattern of metatherian survivorship across the K-Pg boundary and the origin of the crown-clade members of Metatheria, the Marsupialia. North American metatherian faunas underwent a catastrophic decline at the boundary. The K-Pg boundary section of eastern Montana, which contains the most intensely studied latest Cretaceous terrestrial vertebrate fauna in the world, shows a drop from 11 species in the uppermost Cretaceous Hell Creek Formation to only one, Thylacodon montanensis, in the lowest Paleocene Tullock Formation (Williamson et al. 2012; Wilson 2014). Some previous workers have suggested that Thylacodon (often referred to as “Peradectes,” but see Williamson et al. 2012) is closely related to “Alphadon” or other North American metatherian taxa (e.g., Clemens 2010; Wilson 2014) and thus represents a “resident taxon.” However, our phylogenetic analysis does not support such a close relationship between Thylacodon and any Cretaceous metatherian taxa. Nevertheless, our analysis indicates that there were at least four boundary crossings, and perhaps as many as 13 depending on the resolution of polytomies. The Paleogene taxa Herpetotheriidae and “Peradectidae sensu lato” probably originated in the Late Cretaceous. Only two clades, the one containing Glasbius and Roberthoffstetteria and the other containing Ectocentrocristus and Herpetotheriidae, show well resolved relationships with sister taxa on either side of the K-Pg boundary. However, the taxa bracketing the boundary are either found on different continents (Glasbius, Roberthoffstetteria) or are separated by a significant temporal gap (Ectocentrocristus, Golerdelphys). This analysis supports the conclusion that metatherians underwent a profound extinction event at the end of the Cretaceous.