AFFINITIES OF FERREQUITHERIUM AND HOROLODECTES

To examine the evolutionary relationships of Ferrequitherium and Horolodectes, a phylogenetic analysis was performed using a newly constructed character–taxon matrix. The matrix consists of 81 characters, compiled through a combination of direct observation of specimens and descriptions from the literature. The specimens and literature used in the analyses, character list and data matrix are presented in Supporting Information, Appendices S2–S4. In addition to Ferrequitherium and Horolodectes, the analysis includes several of the taxa discussed in Scott et al. (2006) as potential near-relatives of Horolodectes, as well as additional taxa with dental morphologies that resemble that of Ferrequitherium.

Taxon sampling

The ingroup includes the following taxa: Horolodectes and Ferrequitherium: Ferrequitherium and its hypothesized close relative Horolodectes, are each monotypic and are represented by F. sweeti and H. sunae Scott, Fox & Webb 2006, respectively.

Pentacodontidae: The Pentacodontidae Simpson, 1937 are a family of unusual, Palaeocene–Eocene-age eutherians characterized by moderately to greatly enlarged upper and lower fourth premolars and relatively unspecialized molars (Simpson, 1937; Gingerich et al., 1983; Gunnell et al., 2008). Simpson (1937) originally restricted Pentacodontinae (then considered a subfamily) to Aphronorus Simpson, 1935 and Pentacodon Scott, 1892, but the family has since been expanded to include several other North American and, potentially, European taxa (McKenna & Bell, 1997; De Bast & Smith, 2017). The broader relationships of Pentacodontidae remain unclear, although the opinions of Matthew (1909, 1918), Simpson (1937) and McKenna (1975) that link the group with pantolestids, as either a subfamily of Pantolestidae or a family within a more inclusive Pantolesta, have been accepted in recent studies (e.g. McKenna & Bell, 1997; Boyer & Georgi, 2007; Rankin, 2014). Scott et al. (2006) noted several similarities in the dentitions of Horolodectes and Aphronorus, and the genus is therefore included in this study. Aphronorus was scored using A. fraudator Simpson, 1935, the type species, and A. orieli Gazin, 1969. A second putative pentacodontid, Bisonalveus Gazin, 1956, was also included in this analysis. In the parts that can be compared, the dentition of this genus is similar to that of Ferrequitherium, particularly the upper molars and lower fourth premolar. Bisonalveus has long been considered a pentacodontid (e.g. Van Valen, 1967; Krause & Gingerich, 1983; Youzwyshyn, 1988; Fox & Scott, 2005), although its evolutionary relationships have not been examined in detail, and at least one recent, largescale analysis has hypothesized a bizarre sister-group relationship between the genus and a clade, including pantodonts and tillodonts (Halliday et al., 2017; and see Zack, 2009 for discussion). Bisonalveus was scored using Bisonalveus browni Gazin, 1956, the type species, and an as-yet unnamed new species from the late Palaeocene of Alberta (Scott, 2008).

Bessoecetor Simpson, 1936: Bessoecetor is the oldest discovered pantolestid, known principally from Palaeocene deposits in the northern part of the Western Interior of North America (Russell, 1929; Simpson, 1936; Krause & Gingerich, 1983; Cifelli et al., 1995; Scott et al., 2002; Scott, 2008; Rankin, 2014). The genus was included in this analysis to test for potential broader pantolestan (i.e. exclusive of Pentacodontidae) affinities of Ferrequitherium and Horolodectes. Bessoecetor was scored using Bessoecetor septentrionalis (Russell, 1929): the species is one of the oldest and arguably most basal of the genus, and is particularly well documented, being known from dozens of specimens from Alberta and Montana (see, e.g. Simpson, 1936; Krause & Gingerich, 1983; Scott et al., 2002; Scott, 2008).

Cimolestes Marsh, 1889: The genus Cimolestes was recently split into several genera, in recognition of the morphological disparity among its included species (Fox, 2015). Fox (2015) limited Cimolestes to the type species, C. incisus Marsh, 1889, and C. stirtoni Clemens, 1973, both from Late Cretaceous deposits of the Western Interior of North America, with the former being included in this analysis. Cimolestes is considered a basal cimolestan (McKenna & Bell, 1997) and was included in the analysis to test for broader cimolestan (i.e. exclusive of Pantolesta) affinities.

Apheliscidae Matthew, 1918: The Apheliscidae are a family of small-bodied eutherians that have been traditionally allied with ‘Condylarthra’, a widely recognized wastebasket taxon consisting of primarily Palaeocene to Eocene age, bunodont mammals (McKenna, 1960; Van Valen, 1967; Archibald, 1998; Zack et al., 2005b). ‘Condylarthra’ are thought to represent the initial radiation of ungulates, although compelling evidence linking any of the included lower-level ‘condylarth’ taxa with crown group ungulates remains elusive. A subset of Apheliscidae, the Apheliscinae Zack, Penkrot, Krause & Maas, 2005b, is characterized by enlarged upper and lower fourth premolars, a weak p4 paraconid and metaconid, poorly developed molar cingulids, and inflated cusps (Zack et al., 2005b; Penkrot et al., 2008). Apheliscines broadly resemble Ferrequitherium in these regards, and these are best seen in the dentition of Phenacodaptes Jepsen, 1930, from the Late Palaeocene of Wyoming, the upper molars and lower fourth premolar of which resemble those of Ferrequitherium. Phenacodaptes sabulosus Jepsen, 1930, the only known species of the genus, was scored in this analysis. In addition to Phenacodaptes, the Late Palaeocene Gingerichia Zack, Penkrot, Krause & Maas, 2005b was also included; Gingerichia is recognized as the basalmost apheliscine (Zack et al., 2005b). The genus is known from two species (G. geoteretes Zack, Penkrot, Krause & Maas, 2005b and G. hystrix Zack, Penkrot, Krause & Maas, 2005b), and scoring included both of these species. To test for broader apheliscid (i.e. exclusive of Apheliscinae) affinities, the North American Palaeocene genus Litomylus Simpson, 1935 was also included. Scoring decisions were based on Litomylus dissentaneus Simpson, 1935: Litomylus dissentaneus is reasonably well known and is represented by several well-preserved specimens documenting significant parts of the dentition.

Arctocyonidae: Webb (1996) and Scott et al. (2006) noted similarities in the dentition of Horolodectes and those of ‘Oxyclaenidae’ (sensu Matthew, 1915 and McKenna & Bell, 1997), a family of basal ‘condylarths’ with generally plesiomorphic dentitions (Archibald 1998). There is little consensus on either the monophyly of the various ‘oxyclaenids’ or how best to classify them (e.g. Cifelli, 1983; McKenna & Bell, 1997; Archibald, 1998; Muizon & Cifelli, 2000; Williamson & Carr, 2007; De Bast & Smith, 2013). Given these continuing uncertainties, the nomen Arctocyonidae is used here to refer to the possibly paraphyletic or even polyphyletic assemblage of basal ‘condylarths’ exclusive of Oxyclaenus (which may be more closely related to mesonychids; Williamson & Carr, 2007), while recognizing that some of the constituent arctocyonid taxa may be monophyletic (e.g. Arctocyoninae sensu Archibald, 1998). Two arctocyonids were included in this analysis: Protungulatum Sloan & Van Valen, 1965, and Loxolophus Cope, 1885. Protungulatum is widely regarded as the most basal ‘condylarth’, although its systematic position as a placental (i.e. within the eutherian crown group) or as a stem eutherian continues to be debated (e.g. Wible et al., 2007, 2009; Spaulding et al., 2009; Archibald et al., 2011; O’Leary et al., 2013; Halliday et al., 2016, 2017). The genus is first known from Late Cretaceous deposits of North America (Johnston, 1980; Fox, 1989; Archibald et al., 2011; Redman et al., 2015) but is best known from the Palaeocene of Montana (Sloan & Van Valen, 1965; Archibald, 1982; Lofgren, 1995). Dental characters for Protungulatum were scored using the type species, P. donnae Sloan & Van Valen, 1965, and an as-yet unnamed new species from the Puercan of Alberta. Loxolophus, a potentially non-monophyletic ‘condylarth’ (see Hunter et al., 1997) known from Puercan and Torrejonian deposits in several areas of the North American Western Interior, was included as a derived arctocyonid; dental characters were scored for Loxolophus hyattianus Cope, 1885.

Zhelestidae: The Zhelestidae are a family of Late Cretaceous, primarily Central Asian eutherians that have previously been linked to crown group ungulates (e.g. Archibald, 1998; Nessov et al., 1998; Archibald et al., 2001), but more recently have been considered stem eutherians (e.g. Ekdale et al., 2004; Wible et al., 2007; Chester et al., 2010; Archibald & Averianov, 2012). Scott et al. (2006) noted several similarities between zhelestids and Horolodectes, including teeth with low crowns and swollen cusps, molars with narrow stylar shelves, large conules, and lower molars with wide talonids and a closely approximated entoconid and hypoconulid. The dentition of Ferrequitherium resembles that of zhelestids in these features as well, but with additional similarities in the premolars, particularly the development of a metaconid on p4 (= p5 of zhelestids). Zhelestid characters were primarily scored for Aspanlestes atap Nessov, 1985; Aspanlestes is among the best-known zhelestids, with significant parts of the dentition, dermatocranium and basicranium now known (Archibald & Averianov, 2012). Because M3 is undiscovered for Aspanlestes, character 29 was scored for Parazhelestes mynbulakensis (Nessov, 1985).

Lipotyphla Haeckel, 1866: Whereas the dentition of Horolodectes provides little indication of a potential relationship with lipotyphlans, the less specialized dentition of Ferrequitherium shows broad similarities with those of various putative lipotyphlans from the Palaeocene, particularly those that have been classified in the Erinaceomorpha (hedgehog-like lipotyphlans; sensu Novacek et al., 1985; McKenna & Bell, 1997; but see the contrasting opinion of: Penkrot & Zack, 2016). The teeth of Litocherus Gingerich, 1983 and Litolestes Jepsen, 1930, particularly the lower fourth premolar, resemble those of Ferrequitherium, and the two former genera were included on that account, with Litocherus being potentially among the basalmost erinaceomorphs, and Litolestes being potentially the earliest erinaceid (see Novacek et al., 1985; O’Leary et al., 2013). Characters were scored for Litocherus notissimus (Simpson, 1936) and Litolestes ignotus Jepsen, 1930. In addition to Litocherus and Litolestes, the genus Diacocherus Gingerich, 1983 was also included in the analysis: Diacocherus has been considered a basal erinaceomorph (e.g. Krishtalka, 1976a; Novacek, 1982, 1985; Novacek et al., 1985; Butler, 1988), although its status, both as a genus distinct from Adunator Russell, 1964, and as a lipotyphlan, continues to be debated (Krishtalka, 1976a; Gingerich, 1983; Secord, 2008; Hooker & Russell, 2012; Rose, Storch & Krohmann, 2015). Dental characters were scored for D. meizon Gingerich, 1983. Finally, the genus Leptacodon Matthew & Granger, 1921 was included in order to test for potential broader lipotyphlan affinities, exclusive of Erinaceomorpha. Scoring decisions for Leptacodon were based on L. munusculum Simpson, 1935, a relatively well-known species that is considered to be among the basalmost members of Nyctitheriidae Simpson, 1928 (see Manz & Bloch, 2015).

Analytical protocol

Four heuristic parsimony analyses were performed in PAUP4.0a (Swofford, 2002) using the random addition sequence option; default conditions were used for all other heuristic search options. Multistate characters were unordered and multistate terminals were considered polymorphic. PAUP4.0a was also used to perform bootstrap analyses (default conditions) and to calculate Bremer support. Uncertainty with respect to the nearest relative of the ingroup and conflicting character states among potential outgroups prompted the decision to run separate analyses, with the resulting trees examined for any congruence in topology. The inclusion of Zhelestidae as part of the ingroup poses several difficulties for outgroup choice: zhelestids were once considered near-relatives of ungulates (e.g. Archibald, 1998; Nessov et al., 1998; Archibald et al., 2001), but subsequent analyses have positioned the family outside of Placentalia and, in some cases, basal to several stem eutherians, including Kennalestes Kielan-Jaworowska, 1969 and Maelestes Wible, Novacek, Rougier & Asher, 2007 (e.g. Wible et al., 2009; Archibald & Averianov, 2012). On the contrary, more recent large-scale analyses of many of the relevant taxa included in this analysis recover Zhelestidae crownward of Kennalestes and Maelestes (Hooker, 2014; Penkrot & Zack, 2016). Given these uncertainties, three analyses were run, each with a different outgroup. In the first analysis, the Late Cretaceous therian Prokennalestes trofimovi Kielan-Jaworowska & Dashzeveg, 1989 served as the outgroup; Prokennalestes has in previous analyses been consistently recovered stemward of the ingroup taxa included here, including Zhelestidae (e.g. Wible et al., 2009). The second and third analyses included the Late Cretaceous therian Kennalestes gobiensis and the Late Cretaceous therian Maelestes gobiensis, respectively, as outgroups. Given the uncertain phylogenetic position of Zhelestidae relative to Kennalestes and Maelestes, two subanalyses were performed with each outgroup in order to examine the effects of including or excluding Zhelestidae from the ingroup. Four characters (78–81) were excluded from the subanalyses where Zhelestidae were excluded, as states for these characters were rendered either invariable for all taxa, or invariable in the ingroup. A fourth analysis was performed using Prokennalestes, Kennalestes and Maelestes together in order to examine the effects of multiple outgroups on the resulting topology. Character states were optimized using the DELTRAN option, a criterion that favours convergence over reversal.