Perissodactyl diversities and responses to climate changes as reflected by dental homogeneity during the Cenozoic in Asia

Abstract Cenozoic mammal evolution and faunal turnover are considered to have been influenced and triggered by global climate change. Teeth of large terrestrial ungulates are reliable proxies to trace long‐term climatic changes due to their morphological and physicochemical properties; however, the role of premolar molarization in ungulate evolution and related climatic change has rarely been investigated. Recently, three patterns of premolar molarization among perissodactyls have been recognized: endoprotocrista‐derived hypocone (type I); paraconule–protocone separation (type II); and metaconule‐derived pseudohypocone (type III). These three patterns of premolar molarization play an important role in perissodactyl diversity coupled with global climate change during the Cenozoic in Asia. Those groups with a relatively higher degree of premolar molarization, initiated by the formation of the hypocone, survived into Neogene, whereas those with a lesser degree of molarization, initiated by the deformation of existing ridges and cusps, went extinct by the end of the Oligocene. In addition, the hypothesis of the “Ulan Gochu Decline” is proposed here to designate the most conspicuous decrease of perissodactyl diversity that occurred in the latest middle Eocene rather than at the Eocene–Oligocene transition in Asia, as conventionally thought; this event was likely comparable to the contemporaneous post‐Uintan decline of the North American land fauna.


| INTRODUC TI ON
Living perissodactyls (odd-toed ungulates) represent the remnants of a major evolutionary sequence and comprise only six genera and 17 species with many in danger of extinction (Nowak & Walker, 1999). However, perissodactyls had a rich, diverse fossil record spanning 56 Myr, and have not only long been used as a strong evidence for evolution since Huxley (e.g., horses), but also for inferring climatic and environmental changes, and their coevolution with such changes (Franzen, 2010;MacFadden, 1992;Mihlbachler, Rivals, Solounias, & Semprebon, 2011;Secord et al., 2012;Simpson, 1951).
The most conspicuous fauna turnover during the Paleogene in Asia is considered to have occurred during the Eocene-Oligocene transition, and the Eocene perissodactyl-dominant faunas were abruptly replaced by the Oligocene rodent/lagomorph-dominant faunas (Meng & McKenna, 1998). This major faunal turnover, known as "Mongolian Remodelling," is attributed to climatic changes from the warm, humid Eocene to the cooler and more arid Oligocene (Meng & McKenna, 1998;Zachos, Pagani, Sloan, Thomas, & Billups, 2001).
Perissodactyls have been considered to have originated from within the radiation of phenacodont condylarths (Radinsky, 1966;Thewissen & Domning, 1992) or to be a sister group to Radinskya from the middle Paleocene of China (Holbrook, 2014;McKenna, Chow, Ting, & Luo, 1989), while more recent work has shown cambaytheres from the early Eocene of Indian subcontinent to be more closely related to perissodactyls than either of these previously considered taxa (Rose et al., 2014). Recent research on ancient proteins suggests that crown perissodactyls are the sister group to some extinct South American ungulates among more recent mammals (Welker et al., 2015). Teeth, composed of the hardest vertebrate tissues, are the best preserved material in perissodactyls as in other mammal fossils, and the most sensitive proxy to environmental changes (Mihlbachler et al., 2011;Secord et al., 2012). Previous studies on the dentition of extinct perissodactyls (and other ungulates) have focused on molar crown height, enamel stable isotopes, micro-mesowear, and overall morphology through the Cenozoic (Ackermans, 2020;Evans & Pineda-Munoz, 2018;Jernvall, Hunter, & Fortelius, 1996;Mihlbachler et al., 2011;Secord et al., 2012), but little attention has been paid to the evolutionary patterns of premolar molarization in perissodactyls (Butler, 1952;Holbrook, 2015).
Molarization of the premolars generally results in dental homogeneity and increases the grinding area of the dentition; this is especially important for hindgut fermenting ungulates such as perissodactyls as they are highly reliant on oral processing of the food before its initial ingestion (Clauss, Nunn, Fritz, & Hummel, 2009;Fletcher, Janis, & Rayfield, 2010). In contrast, the foregut fermenting artiodactyls, ruminants and camelids, are less reliant on oral processing: they only have partially molarized premolars, and their premolar complexity decreases from P4 to P2 (P1 is usually lost).
However, it is unclear what the role of premolar molarization is in perissodactyl evolution and diversity, how perissodactyl diversity and premolar molariform changed through the Cenozoic in Asia, and whether perissodactyl diversity tracked the Cenozoic climatic changes. Here, we show that three recently proposed patterns of premolar molarization in perissodactyls may have played an important role in their response to climatic and environmental changes during the Cenozoic. Further, based on analysis of perissodactyl diversity with an updated Cenozoic timescale of China in Asia, we note that the most distinct change occurred during the latest middle Eocene, rather than at Eocene-Oligocene transition as conventionally considered, and is likely comparable to the contemporaneous post-Uintan decline of North American land fauna (Prothero, 1994).

| Premolar molarization
Five categories of premolar molarization, initially used for rhinoceroses, were assigned to the P2-4 of perissodactyls: nonmolariform, premolariform, submolariform, semimolariform, and molariform (Qiu & Wang, 2007) (Figure 1). In the premolariform morphology, the hypocone is united with the protoloph, but the metaloph is not completely formed (i.e., is separate from the hypocone) ( Figure 1a). In the submolariform morphology, the metaloph is completely formed (connected to the hypocone) and united with the protoloph on the lingual side (Figure 1b). In the semimolariform morphology, the protocone and hypocone are distinctly separate, but still connected by F I G U R E 1 The degree of premolar molarization in perissodactyls as shown by rhinoceros premolars (modified from Qiu and Wang (2007)). (a) Premolariform (assigned value 2); (b) submolariform (assigned value 3); (c) semimolariform (assigned value 4); (d) molariform (assigned value 5). hy, hypocone; mel, metaloph; pr, protocone; prl: protoloph an enamel ridge (Figure 1c). In the molariform morphology, the protoloph and metaloph are completely separate (Figure 1d). Numbers from one to five were assigned to the five categories, respectively, similar to those proposed by Prothero (2005). These numbers were assigned to each premolar (P2-4) according to their degree of premolar molarization (Appendix Tables A1-A3). The mean values for P2-4 were then calculated, which represent the degree of premolar molarization of each genus. We then calculated the mean value of the degree of premolar molarization in each perissodactyl family during each Asian Land Mammal Age (ALMA) (Appendix Table A2).
Among perissodactyls, brontotheres are characterized by bunodont-lophodont teeth with relatively weak transverse lophs on the upper premolars, unlike the condition in other perissodactyls that are more strictly lophodont, and so we characterized their teeth in a somewhat different fashion. We assigned premolariform (value 2) to the brontotheres with a "lingual crest" (Mihlbachler, 2008) on the upper premolars, semimolariform (value 4) to those with distinct hypocones on the "lingual crest," and molariform (value 5) to those with distinct hypocones completely separated from the protocone.
Three patterns of premolar molarization among perissodactyls have been recognized, including an endoprotocrista-derived hypocone (type I), paraconule-protocone separation (type II), and a metaconule-derived pseudohypocone (type III) (Bai, Meng, Mao, Zhang, & Wang, 2019;Holbrook, 2015). The premolar molarization in most perissodactyls was initiated by the type I pattern, which means that the hypocone developed from a crista posterior to the protocone (endoprotocrista) (Holbrook, 2015). However, the P2-4 of deperetellid tapiroids, the P3 of the Eocene equids, and the P2 of the hyrachyid Metahyrachyus adopted the type II pattern, which entails the separation of the paraconule from the protocone, and the paraconule is enlarged and lingually extended .
Only amynodontid rhinocerotoids had the type III pattern, where the metaconule is separated from the protocone and is lingually extended . In contrast, the molar hypocones evolved either from the postprotocingulum, as in most eutherians, or from the metaconule, as in artiodactyls (Hunter & Jernvall, 1995).

| Hypsodonty
Different methods for calculating the Hypsodonty Index (HI) have been proposed (Fortelius et al., 2002;Janis, Damuth, & Theodor, 2002;MacFadden, 1992;Van Valen, 1960), but all entail the comparison of the unworn tooth crown height with some other dental linear measurement. Here, we follow Fortelius et al. (2002) in using HI as a ratio of height to length of the second molar (upper or low), although the deficiency of using length instead of width has been noticed (Damuth & Janis, 2011). Three classes of hypsodonty were proposed by Fortelius et al. (2002) based on the following criteria: brachydont teeth with a ratio of less than 0.8 (assigned a value of 1); mesodont teeth with a ratio of 0.8-1.2 (assigned a value of 2); and hypsodont teeth with a ratio of greater than 1.2 (assigned a value of 3). For the Neogene perissodactyls from Asia, the classes of hypsodonty for almost all genera can be found in the NOW (New and Old World) database (NOW: http://www.helsi nki.fi/scien ce/ now/) (Appendix Tables A1 and A4). However, data for the classes of hypsodonty of Paleogene perissodactyls in Asia are almost entirely lacking. We considered all Eocene perissodactyls to have brachydont teeth except for the amynodontids Huananodon (mesodont) (You, 1977) and Hypsamynodon (hypsodont) (Gromova, 1954).
Indeterminate taxon identifications and taxonomic modifications such as "cf." or "?" or some cases of "sp." were ignored in our analyses. If more than one species of a genus was known from an ALMA, the type species of the genus was selected; otherwise, the most common and well-preserved species was selected if the type species of the genus was absent during the period in Asia. If intraspecific variation was present, the characteristics of the dentition of the holotype were followed. In the middle Miocene, four genera of elasmotheres were considered to be synonyms of Hispanotherium (Deng & Chen, 2016); however, we treated them all as valid genera pending the discovery of more complete material. The subgenera of Hipparion from the late Neogene have been elevated to generic levels in recent analyses (Bernor, Wang, Liu, Chen, & Sun, 2018;Sun, Zhang, Liu, & Bernor, 2018a), and we followed this taxonomy.

| RE SULTS AND D ISCUSS I ON
Taking advantages of the updated Cenozoic timescale in China (Deng et al., 2019a;Wang et al., 2019), and revisions of perissodactyl fossils from China (Bai, Wang, Li, et al., 2018;Deng & Chen, 2016 Tables A1 and A3). In general, perissodactyl generic diversity fluctuated in relation to paleoclimatic changes (Bai, Wang, Li, et al., 2018;Zachos, Dickens, & Zeebe, 2008). At the beginning of the early Eocene, perissodactyls had a relatively high diversity, which is consistent with the notion that different lineages of perissodactyls diverged as early as the earliest Eocene during Paleocene-Eocene Thermal Maximum (PETM) (Bai, Wang, & Meng, 2018b). An abrupt increase of diversity from the Arshantan to the Irdinmanhan is likely related to the rising temperatures of the Mid-Eocene Climatic Optimum (MECO), but the high diversity in the Irdinmanhan may be biased by the overestimation of generic numbers (Bai, Wang, Li, et al., 2018). Deperetellids, helaletids, and paraceratheriids attained a relatively high or the maximum degree of premolar molarization (mean values ranging from 2.5 to 4.7) in the middle Eocene ALMA, the Irdinmanhan and Sharamurunian (Figure 3b) (Appendix Table A4); this may reflect an increasing ability to process larger amounts of low quality vegetation and the adoption of a relatively more open habitat (Bai, Wang, Li, et al., 2018;Bai, Wang, & Meng, 2018c;Gong et al., 2019Gong et al., , 2020, following the temperature decline after the MECO and the intense of seasonality (Figueirido, Janis, Perez-Claros, De Renzi, & Palmqvist, 2012;Janis, 1989). However, the degree of premolar molarization shows a degree of fluctuation rather than an average sustained increase at the family level during the Paleogene.
It is noteworthy that a few amynodontids possessed mesodont or hypsodont teeth in the middle-late Eocene, when artiodactyls with teeth that were more lophodont and higher-crowned (e.g., the F I G U R E 2 Asian perissodactyl composition (in percentage of genera) and diversity of Tapiroidea   Similarly, the entire mammalian fauna from China showed a similar abrupt decrease after the Sharamurunian in terms of number of both species and genera (Wang, Meng, Ni, & Li, 2007). We named this event the "Ulan Gochu Decline," which is likely comparable to the contemporaneous post-Uintan decline of the North American land fauna (Berggren & Prothero, 1992;Prothero, 1994;Stucky, 1990) and the beginning of the White River Chronofauna in the late Duchesnean (Woodburne, 2004). The mammalian fauna turnover in the late Duchesnean was considered to have had more influence on the North American fauna than did events at the EOT (Berggren & Prothero, 1992;Meng & McKenna, 1998). The "Ulan Gochu Decline" was probably related to the sustained cooling following the MECO. The diversity of perissodactyls somewhat increased in the late Eocene (Ergilian) when the temperature rose again slightly ( Figure 3a).
The Asian mammalian fauna turnover during the EOT is known as the "Mongolian Remodelling," with the perissodactyl-dominant fauna replaced by the rodent/lagomorph-dominant fauna, a turnover that was attributed to the dramatic drop of temperature at the end of the Eocene (Meng & McKenna, 1998). Recent integrated analyses, however, suggest that the faunal turnover predated the Eocene-Oligocene boundary (Wasiljeff et al., 2020). The climatic deterioration during the EOT also affected primates in southern Asia, F I G U R E 3 Cenozoic Perissodactyl diversity in Asia and the degree of premolar molarization in different perissodactyl lineages. (a) Perissodactyl diversity and dental hypsodonty in relation to global climatic change (modified from Zachos et al. (2008)) with the most conspicuous decrease of diversity occurring during the latest middle Eocene ("Ulan Gochu Decline"); (b) degree of premolar molarization as represented by the mean value and standard error in different perissodactyl lineages during the Cenozoic in Asia. Equoids are excluded because they were scarce in Asia during the Paleogene. The red, yellow, and blue bars show the time periods of the "Ulan Gochu Decline," "Mongolian Remodelling," and the beginning of the Neogene, respectively. The degree of premolar molarization is assigned into five categories (Qiu & Wang, 2007): nonmolariform (1), premolariform (2), submolariform (3), semimolarifom (4), and molariform (5)

Bu Ar
Paraceratheriidae favoring the survival of strepsirhines over haplorhines (Ni, Li, Li, & Beard, 2016). In North America, primates (all nonanthropoids) were in decline through the middle Eocene and were essentially extinct by the late Eocene, although a single taxon is known from the latest Oligocene/earliest Miocene (Gunnell, Rose, & Rasmussen, 2008).
Among perissodactyls, only brontotheres went extinct during the transition in Asia, as they did in North America (Figures 2a and 3a).
The mean value of premolar molarization in hyracodontids gradually increased from 1.4 in the middle Eocene to 3.4 in the late Oligocene, approaching a moderate extent before their extinction by the end of Oligocene (Figure 3b) (Appendix Table A2). Amynodontids reached a peak of premolar molarization in the late middle Eocene with a relatively low value (mean value 2.3), which then decreased gradually to the lowest value (value 1) by the end of the Oligocene before their extinction (Figure 3b). In contrast, paraceratheriids, rhinocerotids, and tapirids, which had higher values of premolar molarization (>3), all survived into the Neogene (Figure 3b), with the implication that this greater degree of premolar molarization contributed to their advantage over those with lower values during the Oligocene/Miocene transition. In addition, premolar molarization type I was likely more advantageous than types II and III, as inferred from the fact that the groups with latter two types of dentitions went extinct before the end of the Paleogene, while almost all perissodactyls with type I dentitions survived into Neogene. Furthermore, the cascade of premolar molarization in a species varied in different groups, such as the premolars of paraceratheres becoming molarized from anterior to posterior teeth (Qiu & Wang, 2007), whereas those of tapiroids took place from posterior to anterior teeth.
In short, the Eocene mammal faunas from Asia showed two pulses of decline in diversity that may be related to global climatic changes. The first one (the "Ulan Gochu Decline"), comparable to the post-Uintan decline of North American land fauna, took place after the MECO when temperatures declined slowly, and was reflected most clearly in changes in the diversity of perissodactyls. The second one (the "Mongolian Remodelling"), comparable to the European "Grande Coupure," was at the EOT and may have been a response to the sudden global drop of temperature. It is noteworthy that the endemic Asian taxa sporadically dispersed to North America during the Paleogene, but there was apparently little dispersal in the opposite direction (Beard, 1998). A very few taxa of Asian perissodactyls, such as early equids and palaeotheres, are considered to have dispersed from North America or Europe to Asia during the Paleogene (Bai, 2017;Bai et al., 2018b;Woodburne, 2004), so the immigrants would have had limited impact on the Paleogene Asian perissodactyl diversity.
In the early Miocene, rhinocerotids replaced paraceratheriids and dominated the perissodactyl groups (Figure 2c). The Chinese rhinocerotid diversity and responses to the Neogene climatic change have been investigated by Deng and Chen (2016) and Deng and Downs (2002), and it is not necessary to replicate them here. However, the following statements need to be addressed: The appearance of the rhinocerotid Hispanotherium with hypsodont teeth in the middle Miocene indicates a more abrasive diet and a slightly more open and drier habitat, which is enhanced in the late Miocene with the spread of Old World savanna palaeobiome (Kaya et al., 2018). The decrease of rhinocerotid diversity and the rise of equid diversity during the mid-late Miocene transition have been mainly attributed to the cooling event and the dispersal of hipparionine equids from North America to Eurasia in the late Miocene (MacFadden, 1992). During the late Miocene, the hypsodont equids and rhinocerotids coexisted with brachydont or mesodont ones (Figure 3a; Appendix Table A1), but the proportion of hypsodont groups gradually increased. The decreased diversity of rhinocerotids in the early Pliocene was probably due to the expansion of C 4 grasses (Han, Wang, & Liu, 2002), although the climate was relatively warm and humid except in the high altitude, cold Tibetan Plateau (Deng et al., 2019b). The hypsodont equids became the dominant taxa among perissodactyl groups in the Pliocene (Figure 3a; Appendix Table A1). In addition, the distribution of Chinese mammals has also been influenced by the East Asian Monsoon (Qiu & Li, 2005), which was probably initiated in the Eocene and intensified in the late Miocene, driven by the uplift of the Tibetan Plateau (Qiu & Li, 2005;Quan et al., 2014). The degree of premolar molarization in tapirids, rhinocerotids, and equids remained high (mean value >4.2) and virtually stable through the Neogene, and the main modification of the teeth in the latter two clades lay in increasing the height of the molar/premolar crowns and the complexity of the occlusal enamel (Figure 3, Appendix Table A4) (Deng & Chen, 2016;Famoso, Davis, Feranec, Hopkins, & Price, 2016;Fortelius et al., 2002;Simpson, 1951). Paraceratheriids, with a relatively lesser degree of premolar molarization and persistently me- Deng, Ying-Qi Zhang, and Bo-Yang Sun for discussion. We are grateful to Yong Xu for the drawing of Figure 1. Thanks to two anonymous reviewers for constructive comments that greatly improve this manuscript.

CO N FLI C T O F I NTE R E S T
None declared.  (1997), Rose et al. (2014), and updated data. Each family name is followed by information (in parentheses) listing (separated by slashes): the mean value of premolar molarization, the pattern of premolar molarization, and the hypsodonty level. The ancestral mean values were reconstructed using the parsimonious criterion with the linear cost assumption in Mesquite 3.6 (Maddison & Maddison, 2018 Tong and Wang (2006), Wang and Tong (1996) Orientolophus ? Bai et al. (2018b), Ting (1993) Hela.  (2000) Rh-Dic. Protaceratherium 5 1 2 Antunes and Ginsburg (1983) (1868) Note: Almost all Eocene perissodactyls from Asia were considered to have brachydont teeth, which were left blank in the table.