An early Oligocene fossil demonstrates treeshrews are slowly evolving “living fossils”

Treeshrews are widely considered a “living model” of an ancestral primate, and have long been called “living fossils”. Actual fossils of treeshrews, however, are extremely rare. We report a new fossil species of Ptilocercus treeshrew recovered from the early Oligocene (~34 Ma) of China that represents the oldest definitive fossil record of the crown group of treeshrews and nearly doubles the temporal length of their fossil record. The fossil species is strikingly similar to the living Ptilocercus lowii, a species generally recognized as the most plesiomorphic extant treeshrew. It demonstrates that Ptilocercus treeshrews have undergone little evolutionary change in their morphology since the early Oligocene. Morphological comparisons and phylogenetic analysis support the long-standing idea that Ptilocercus treeshrews are morphologically conservative and have probably retained many characters present in the common stock that gave rise to archontans, which include primates, flying lemurs, plesiadapiforms and treeshrews. This discovery provides an exceptional example of slow morphological evolution in a mammalian group over a period of 34 million years. The persistent and stable tropical environment in Southeast Asia through the Cenozoic likely played a critical role in the survival of such a morphologically conservative lineage.

Oligocene of China. With an age of approximately 34 million years, the fossil nearly doubles the temporal length of the known fossil record of treeshrews, and represents the oldest definitive fossil of a crown-group treeshrew.
These new treeshrew fossils were discovered at the Lijiawa mammalian fossil locality near Qujing City in Yunnan Province, China. Among the numerous fossil mammal specimens recovered from this fossil site are those belonging to a large form of Gigantamynodon giganteus, an unnamed species of Cricetops, and a primitive Eucricetodon comparable with Eucricetodon caducus from the earliest Oligocene of Xinjiang, China. Those species all indicate an early Oligocene age for this fauna 31,32 .
Etymology. Specific epithet is derived from the name of Qilin District, in Qujing City. Qilin is the pinyin for kylin, a hoofed dragon-like beast of Chinese myth.
Hypodigm. IVPP (IVPP V20693), a right jaw fragment preserving p3-4 and the alveoli for i2-3, c, and p2 (IVPP V20694), an isolated right m1 (IVPP V20695), and a right jaw fragment preserving m2-3 (IVPP V20696, holotype). The alignment of the fossils is based on a left lower jaw fragment retaining a small portion of the i2-3 alveoli, alveoli and roots of c and p2, and p3-m3 (IVPP V20699, Fig. 2 and Supplementary Information). The i1-3, c, p2-4, and m1-3 are shown for P. lowii. (C), Lingual view of the lower dentition of P. kylin. (D), Buccal view of the dentition of P. kylin. Scale bar equals 5 mm. Diagnosis. Small treeshrew almost identical in size with Ptilocercus lowii. Differs from all tupaiids, but resembles P. lowii in: lacking dilambdodont upper molars; having a large hypocone on M1-2; strong cingulids on the lower molars; a p3 that is smaller than p2; and a premolariform lower canine. Differs from P. lowii in having: a relatively stronger paraconule and metaconule coupled with stronger pre-and post-conule cristae; a stronger lingual cingulum on M1-2; a relatively broader and taller p4 with less-developed paraconid, metaconid and talonid; and lacking a groove separating the hypoconulid and entoconid on the lower molars.
Comparative morphology. The upper molars of ptilocercids are very different from those of tupaiid treeshrews 33 . All tupaiids have a V-shaped, buccally inclined paracone and metacone on each upper molar. The pre-and post-paracrista and the pre-and post-metacrista are buccally directed, forming the so-called dilambdodont tooth pattern. A mesostyle is always present, and is in contact with the postparacrista and premetacrista. The protocone of tupaiid treeshrews also is V-shaped and buccally inclined. The trigon basin enclosed by the preprotocrista and postprotocrista is very deep and narrow. The mesial, lingual and distal borders of the protocone are usually smooth, and no cingulum is present. By contrast, the molar cusps of Ptilocercus kylin and Ptilocercus lowii are rounded and basally expanded (Fig. 1). The preparacrista, postparacrista and premetacrista extend mesiodistally. The preprotocrista and postprotocrista of the protocone form a U-shape, surrounding a broad, shallow trigon basin. The combination of these features produces a typical bunodont tooth, a feature also traditionally considered to be characteristic for basal primates [33][34][35] .
The upper molars of Ptilocercus kylin are slightly different from those of Ptilocercus lowii. The lingual cingula on M1-2 of P. kylin are stronger, and the paraconules and metaconules have better developed pre-and post-cristae. The hypocones of M1-2 of P. kylin are proportionally smaller than in P. lowii, and as a result, the posterior borders of M1-2 are less indented in P. kylin. The buccal shelves are as broad in P. kylin as in P. lowii, but the buccal cingula are less developed in P. kylin. In P. lowii, the buccal cingula are very thick. They enclose two small folds between the preparacrista and postmetacrista. In P. kylin, no such folds are present. The metacone of M3 in P. lowii is conical with a round and expanded base, whereas in P. kylin, the cusp has a sharper appearance.
The lower dentition of ptilocercids also shows many characteristics that are not seen in tupaiid treeshrews. In tupaiids, the lower canine is a large peg-like tooth with a sharp apex. Their premolars are not closely spaced. The size of the premolars gradually increases from p2 to p4, and p3 has two roots. Their molars have trenchant cusps connected to one another by sharp crests. The molars usually have smooth buccal faces, and their trigonids are much higher than their talonids. By contrast, the lower canine of Ptilocercus lowii is a premolar-like tooth. It has a procumbent crown with a well-developed lingual cingulid and distal cusp, and has an obvious neck between the crown and root. The premolar region is shortened, and p2 and p3 are all single-rooted and oblique. The p3 is smaller than p2 and is much smaller than p4, a peculiar condition rarely seen in mammals. The lower molars of P. lowii are widely considered primate-like [33][34][35] . They have conical cusps connected by relatively weak crests. The buccal cingula of the lower molars, however, are very strong, and the trigonids of the molars are only slightly higher than the talonids. These characteristics that distinguish P. lowii from tupaiid treeshrews are all also present in Ptilocercus kylin (Fig. 1).
One lower jaw fragment (IVPP V20694) preserves the alveoli of i2-3, the lower canine and p2 (Fig. 2). The alveolus of i2 is much larger than the i3 alveolus situated immediately distal to it. The orientation of both alveoli suggests that i2 and i3 were very procumbent. The canine alveolus is larger than those of i3 and p2, and the alveolus of p2 is in turn larger than that of p3 (Fig. 2). An isolated lower canine shows that this tooth is premolar-like, with a procumbent crown, an obvious neck between the crown and root, a strong lingual cingulid and a distal accessary cusp. Based on the relative sizes of the alveoli, p3 is much smaller than p4, and smaller than p2, a peculiar feature shared with Ptilocercus lowii. The p4 of Ptilocercus kylin is quite large, retaining a generalized morphology (with a very low paraconid and metaconid and a heel-like talonid) that is shared with tupaiids, insectivores and many stem primates. The paraconid and metaconid of this tooth are quite small, and are much lower than the sharp, trenchant protoconid. The talonid is not fully developed, forming only a short heel. However in P. lowii, p4 is relatively small, and is molariform, with a larger paraconid, a larger metaconid and a longer talonid. The combination of these features suggests that the p4 of P. lowii is specialized and represents an apomorphic condition. The lower molars of P. kylin are almost identical in shape and size with those of P. lowii, except for slight differences present at the distal ends of the teeth. Extant tupaiid and ptilocercid treeshrews all have a lingually positioned hypoconulid, which is separated from the entoconid by a sulcus. In P. lowii, the hypoconulid is more like a rounded terminal swelling of the lingually extended hypocristid, and the distal end of the entoconid is also rounded. Meanwhile in P. kylin, the hypoconulid is a cuspid, and is connected to the entoconid via short cristids. The m3 hypoconulid of P. kylin is relatively long and forms a short heel, and that feature is seen in many stem primates but not in P. lowii or other crown treeshrews.

Discussion
Although a few diagnostic traits distinguish Ptilocercus kylin from Ptilocercus lowii, the two species are exceptionally similar to each other. Therefore, it is unsurprising that a phylogenetic analysis based on large data matrix placed P. kylin as the sister species to P. lowii (Fig. 3). The monophyly of the Ptilocercus branch is strongly supported (Supplementary Information). At the same time, the monophyly of the tupaiid clade also has significant character Scientific RepoRts | 6:18627 | DOI: 10.1038/srep18627   support. Within Tupaiidae, the interspecific relationships are similar to those based on phylogenetic analysis of molecular data 5,36-38 . More specifically, our analysis also supports referring "Urogale" everetti to the genus Tupaia 5 . Various methods of phylogenetic inference and molecular divergence date estimation have suggested that the split between treeshrews and their close relatives (primates and colugos), and that between Ptilocercidae and Tupaiidae (within treeshrews) occurred in the Palaeogene 3-5 . The inter-generic diversification within tupaiid treeshrews began as early as 35 Ma 5 . Detailed morphological comparisons between the early Oligocene Ptilocercus kylin and extant Ptilocercus lowii reveal very few diagnostic differences that could support placing these species in separate genera. Ptilocercus treeshrews appear to have evolved at a very slow pace (bradytely) and accumulated very few morphological changes over the last 34 million years. In many ways they are even more conservative than their sister group Tupaiidae. Given its very ancient divergence time from its sister group, the known species diversity within Ptilocercidae is roughly ten times lower than that in Tupaiidae. Only three ptilocercid taxa (P. lowii, P. kylin, and a late Miocene unnamed fossil from Yunnan China) are known at present, in comparison to approximately thirty known living and fossil tupaiid taxa. However, the species diversity of Tupaiidae itself is quite low when compared to primates.
The extant Ptilocercus lowii is arboreal and nocturnal, and occurs largely in tropical rainforest areas with abundant vines and undergrowth in southern Thailand, Malaysia, Brunei, and Indonesia (Fig. 4) 39 . Individuals of that species consume large amounts of alcoholic nectar, and have a symbiotic pollination relationship with the tropical bertam palm (thought to be at least 55 million years old) 40 . The conserved morphology of Ptilocercus clade, as revealed by the specimens of Ptilocercus kylin, could be the result of a relatively stable ecological niche in the forests of Southeastern Asia through the Cenozoic [40][41][42] , and conversely, the morphological similarity between the fossil and extant Ptilocercus also may suggest that P. kylin occupied a dense tropical rainforest habitat, like its living sister species.
The Eocene-Oligocene transition was marked by dramatic changes in global climate and sea level [43][44][45] , and those changes are associated with the expansion of open habitats, and a major retraction of tropical rainforest to low latitudes. However in south Asia, mid-latitude areas were connected to the equatorial region by uninterrupted terrestrial habitats 46 , and closed canopy tropical rainforests and monsoonal forests still extended over a vast area during this transitional period 47 . The presence of early Oligocene Ptilocercus far to the north of its current range is consistent with the hypothesis that rainforest environments were much more widespread in Asia in the early Oligocene than they are today (Fig. 4). We posit that the geographic distribution of Ptilocercus treeshrews has tracked the expansion and retraction of tropical forests in southern Asia through the Cenozoic.

Methods
Measurements. Specimens were measured under the ZEN Pro 2012 system stored with a Zeiss stero-microscope (Discovery V20), and were calibrated from the caliper. The results are listed in Supplementary Information S- Table 1.
Phylogenetic methods. The phylogenetic analysis was based on a data matrix derived from published data in ref. 12 48 and saved in the NEXUS format. Specimens checked and scored and the arguments for the characters are listed as notes in the NEXUS file. Parsimony analysis of the total evidence dataset (dental, cranial, postcranial, soft tissue, and molecular data) was undertaken using TNT, Tree analysis using New Technology, a parsimony analysis program subsidized by the Willi Hennig Society 49 . We ran multiple replications, using sectorial searches, drifting, ratchet and fusing combined (Supplementary Information: S- Table 2). Random sectorial search, constraint sectorial search and exclusive sectorial search were used. Ten cycles of tree drifting, 10 cycles of ratchet and 10 cycles of tree fusing were performed in the search. Default parameter settings for random sectorial search, constraint sectorial search, exclusive sectorial search, tree drifting, ratchet and fusing were used. The search level was set as 10 for 177 taxa. Optimal scores were searched with 10000 replications. Some characters are set as ordered (listed in the Supplementary Information), but the outgroups were not used as reference for ordering the character states. All characters have equal weight. The relationships of some extant taxa (based on Springer et al.'s gene super-matrix) 50 were used as a backbone constraint or molecular "scaffold", but the relationships among the treeshrews were not constrained.
Approximately 20 hours were required to finish the unconstrained search. More than 769 billion rearrangements were examined. Forty-seven trees with a best score of 13067 were retained. Nearly 21 hours were required to finish the backbone-constrained search. More than 728 billion rearrangements were examined. Twenty trees with a best score of 13141 were retained. The best trees and the strict consensus of these trees were described in PAUP* 51 . The results of backbone-constrained and unconstrained searches were compared in S- Table 3. The backbone-constrained trees (Supplementary Information: S- Fig. 3) are preferred because they incorporated more phylogenetically-relevant information (morphological and molecular information).
We used absolute Bremer support and Relative Bremer Support 52,53 , calculated in TNT (Supplementary Information: S- Table 4), to describe the stability of the phylogenetic result (Supplementary Information: S- Fig. 3, S- Fig. 4).
We define the Order Scandentia as a crown group, including the clade stemming from the most common ancestor of extant ptilocercids and tupaiids. We described the strict consensus tree of the 20 most parsimonious trees from the backbone-constrained searches in PAUP* 51 . The ACCTRAN (accelerated transformation 51,54 ) method of character-state optimization is used. Gaps are treated as "missing". Multistate taxa interpretation depends on "{}" versus "()" designation ("min" values for CI, RI, and RC are minimum-possible character lengths). Synapomorphies for Ptilocercidae, Tupaiidae and Scandentia are listed in the supplementary tables (Supplementary Information: S- Table 5 to S- Table 7). The character numbers listed in S- Table 5 to S- Table 7 correspond to those in the NEXUS file.