Mirusodens caii, Mao & Li & Hooker & Meng, 2023

Mao, Fangyuan, Li, Zhiheng, Hooker, Jerry J. & Meng, Jin, 2023, A new euharamiyidan, Mirusodens caii (Mammalia: Euharamiyida), from the Jurassic Yanliao Biota and evolution of allotherian mammals, Zoological Journal of the Linnean Society 199 (3), pp. 832-859 : 834-848

Species Mirusodens caii gen. et sp. nov.

( Figs 1–6 )

Holotype: A skeleton preserved in the part and counterpart of a split slab: part A, the less slab with most cranial elements and part B, the right slab ( Fig. 1 ; HT-b-Pm-0001, Hongtao Fossil Museum, Lingyuan , Liaoning).

Etymology: Species name is asser Mr Hongtao Cai, who collected and curates the holotype specimen.

Locality and age: Daohugou site, Nincheng County, Inner Mongolia, China; Bathonian–Callovian (168–164 Mya) ( Mao et al. 2021; see also: Ren et al. 2019, Gao et al. 2021, Yang et al. 2021).

Diagnosis: As for the genus, by monotypy.

Description

Skull: The crushed holotype skeleton is preserved in the main part and counterpart of a split slab ( Figs 1 , 2 ). The breakage runs though the right side of the skull so that the main part (part A) contains most (primarily the less side) of the skull. The less side of the skull is embedded in the matrix and thus well-preserved; its morphology is revealed by the CT-scan. In lateral view, the rostrum is deep to accommodate the enlarged upper incisor that has a strong and long root, similar to that of Dactylopsila trivirgata Gray, 1858 . The nasals project anteriorly, overhanging the external nostril. The anterior root of the zygomatic arch extends laterally at the position lateral to P4 and then continues posteriorly; the arch is deeper anteriorly and gently arching dorsally. The zygoma to the rostrum transition is not gradual but step-like, with the anterior root of the arch extending laterally and then posteriorly. A vague suture indicates that the jugal is probably sizable, which differs from the small jugal on the medial surface of the arch in multituberculates ( Hopson et al. 1989). There is one distinct and short infraorbital foramen. The orbit appears to be large. The glenoid fossa is orientated anteroposteriorly and does not have a postglenoid process, similar to that of multituberculates. The nuchal crest is prominent, projecting dorsoposteriorly. The mandible is typical of euharamiyidans, deep and short. The coronoid process inclines posteriorly. As in other euharamiyidans and multituberculates, the process extends on the labial side of m2 and blocks the tooth in lateral view. The mandible has a small angular process that bends medially. The mandibular condyle is lower than the occlusal surface of the dentition and the articular surface faces posterodorsally. The masseteric fossa extends anteriorly and reaches to the level below p4. As in other euharamiyidans the mental foramen is at the diastema between the lower incisor and p4.

Dentition ( Figs 2–4 ): The tooth morphology of Mirusodens caii is remarkable, particularly its upper incisors and ultimate premolars. Some of the teeth are exposed in the broken surface of the slab ( Fig. 2 ) but most are embedded in the matrix so that they are preserved in good condition, as revealed by CT scan ( Figs 2–4 ). The main slab (part A) contains the complete less upper dentition, the right upper incisor, P2–P4, and the less p4. The counterpart slab (part B) contains the right lower dentition, a segment of the incisor, the right M1–2, and the less m1–2. The lower jaws are partly preserved, which shows the general morphology and allows measurements of the mandible. The less upper dentition preserved in the main slab is interpreted as in its anatomic position. The teeth preserved in part B can be digitally re-associated to those in part A so that the upper and lower dentitions on both sides can be reconstructed. Measurements of teeth are in Table 1 .

Upper incisors: Mirusodens has one pair of enlarged upper incisors, which are regarded as I2. The less incisor is smaller than the right one, showing a degree of asymmetry. The tooth crown is larger (mesiodistally longer) than all cheek teeth except for P4 and supported by a single robust root. The crown is multi-cusped and shaped almost like a molar. All cusps bear fine enamel ridges (flutings) so that the morphology of the upper incisor is in sharp contrast to the single-pointed and smoothly surfaced lower incisor. The mesial half of I2 has two main cusps, which, for convenient description, we denote as cusp 1 and 2. Cusp 1 is mesial to cusp 2. A minor cusp is immediately medial to cusp 1; similarly, another minor cusp is mesial to cusp 2; we consider this as a spliưing of cusps, a unique feature that increases the cusps of the incisors. On the right I2, there is one more minor cusp on the labial side distal to cusp 2. The main and minor cusps are proportionally stronger on the right incisor. In lingual or buccal view, cusps 1 and 2 are high. Following cusp 2 are a few small cusps that decrease in height distally. Again, the number of the distal cusps differs on the two teeth. Except for the distal cusps, the cusps are on the buccal side of the tooth crown. The lingual side of the less I2 bears weak ridge-like cuspules, while the right I2 lacks lingual cusp. All cusps are on the buccal margin of the crown.

The lingual (medial) side of each I2 crown bears a large flat surface, which we interpret as the contact facet for the opposite I2, similar to other euharamiyidans, such as Qishou ( Mao and Meng 2019a) . The buccal side of the tooth is convex laterally. In life when the two incisors pair together, they formed a basinshaped structure that was surrounded by rugged cusps, a good device for holding food items. There is a ‘neck’ that delimits the transition of the crown and the root. The single root is robust and long, about twice the length of the crown length; it is implanted in the premaxilla at an angle of about 50° to the occlusal plane of the teeth.

The upper incisors of Mirusodens represent the extreme condition in known ‘haramiyidans’. They are proportionally larger and more complex in structure than those of Arboroharamiya Zheng et al. 2013 , Shenshou Bi et al. 2014 , Xianshou Bi et al. 2014 , Qishou, Wang et al. 2014 Maiopatagium Wang et al. 2014 , and Vilevolodon Luo et al. 2017 from the Linglongta phase of Yanliao Biota; they are also more complex than the upper incisors of any known Triassic and Jurassic ‘haramiyidans’ and multituberculates ( Hahn and Hahn 2006, Mao et al. 2022). In size and morphology; they are most similar to those of Butlerodon from the Middle Jurassic of United Kingdom ( Mao et al. 2022). Such a large and complex tooth may have functioned as a set of multiple upper incisors in the marsupial Dactylopsila trivirgata and Petaurus breviceps Waterhouse, 1839 ( Beck 2009, Burrows et al. 2020). As the right and less incisors fit together (in contact on their mesial surfaces), the pair forms a complex platform for sophisticated food picking and manipulation again the lower incisors (see comparison of lower incisors). The complex morphology of the upper incisors of Mirusodens is interpreted as a derived condition within euharamiyidans.

Upper mesial and penultimate premolars (P2–3): A large diastema separates the upper incisor and the mesial premolars, similar to other euharamiyidans. This diastema may be interpreted as being created by loss of I3, canine, and perhaps mesial premolar (P1), a condition present in all known euharamiyidans. Differing from all known euharamiyidans where the dentitions are known, Mirusodens is unique in having three upper premolars. The mesial and penultimate premolars are here interpreted as P2 and P3, which are similar in general morphology in having a rounded or oval profile in occlusal view. These teeth are single-rooted with a ‘neck’ delimiting the transition of the root and the crown. P3 has the distal end of the root bent distally. As in the upper incisors, the less P2–3 are smaller and have fewer cusps than the right ones. They are similar in that the three buccal cusps are the largest on the crown and cusps in the centre of the crown are the smallest. All cusps are conical and bear fine enamel ridges. The occlusal surface of the tooth crown is oval-shaped and shallowly basined.

The premolar loci of euharamiyidans are not fully resolved. In earlier studies, the ultimate premolar was denoted as P4 and the penultimate as P3 ( Zheng et al. 2013, Bi et al. 2014, Luo et al. 2017, Meng et al. 2017). This is largely based on the assumption that the ancestral condition of the haramiyidan dentition has a full pack of premolars, as in Haramiyavia ( Jenkins et al. 1997, Luo et al. 2015). Reduction of the premolars’ number is a general trend in evolution of allotherians, which is best known in multituberculates ( Kielan-Jaworowska et al. 2004). In ‘haramiyidans’, if Haramiyavia is considered as an ancestral condition, reduction of teeth is also probably the evolutionary trend in ‘haramiyidans’. Thus, presence of P 2 in Mirusodens is probably a primitive condition. In other Yanliao euharamiyidans where the upper dentition is known, there are only two upper premolars, interpreted as P3 and P4 ( Zheng et al. 2013, Bi et al. 2014, Han et al. 2017, Lou et al. 2017, Meng et al. 2017, Mao and Meng 2019a, Wang et al. 2021). P3 of Mirusodens is larger and more complex than that of Qishou sp. ( Mao and Meng 2019a) and Maiopatagium (Meng et al. 2017) , possibly Shenshou as well (the upper premolars were broken) ( Bi et al. 2014), but smaller and simpler than that of Arboroharamiya ( Zheng et al. 2013, Han et al. 2017), Xianshou ( Bi et al. 2014) , or Vilevolodon ( Luo et al. 2017) ; nonetheless, P2 and P 3 in combination in Mirusodens would form a more complex structure than P3 alone in other taxa.

The cheek teeth of an arboroharamiyid (PIN, nos. 5087/16 and 5087/10), originally identified as upper molariforms ( Averianov et al. 2011: fig. 1), but believed to be upper premolars or P3s ( Meng et al. 2014, Averianov et al. 2019, Mao and Meng 2019a), show some similarity to the less P3 of Mirusodens in general shape and cusp number of the tooth crown; this endorses the identification of those isolated teeth as upper premolars. Similarly, the tooth (BDUC J 562) identified as a less lower molar of an undetermined haramiyid ( Butler and Hooker 2005: fig. 4B) was considered to be a P3 ( Mao and Meng 2019b), which also shows similar general shape to the less P3 of Mirusodens but has fewer cusps. Presence of P2 and P 3 in Mirusodens demonstrates the possibility that a similar condition could exist in other species, particularly the Triassic ones, such as Thomasia , which are represented only by isolated teeth that show diverse morphologies ( Sigogneau-Russell 1989, Debuysschere 2015).

Ultimate upper premolar (P4): P4 of Mirusodens is a remarkable tooth compared to those of other ‘haramiyidans’ and perhaps even any of other mammaliaformes. It is the largest cheek tooth and characterized by crown and root morphologies. The CT-image of the right P4 is not well separated, but it can be seen that the cusp orientations and size are somewhat different from the less P4, but the general morphologies of the two P4 are comparable. The occlusal outline of P4 (based on the less one) is heart-shaped with the apex pointing distally; the lingual side is curved, whereas the buccal side is straight. The largest cusp is at the distobuccal corner that we denote as A1. Another two main cusps on the mesiobuccal side were denoted as A2 and A3, although we do not assume any homology of these cusps in other species of euharamiyidans. Between A1 and A2, there are two minor cusps. On the crown surface, numerous cusps are arranged regularly as four or five parallel rows in a curved course from the lingual margin to the basin centre; the cusps decrease in size toward the centre. The buccal row bears larger cusps that form the buccal margin of the crown, whereas the row lingual to it consists of smaller cusps. At the very centre, the lowest area of the tooth basin, cusps are more randomly distributed and those in the deepest area are worn so that they became confluent. Such a remarkable arrangement of cusps on the broadly basined occlusal surface is unique; there seems no analogue to it, to our knowledge, in any known euharamiyidans or mammaliaformes. Such a cusp-floored basin is functionally similar to a coarse rasp so that food items can be firmly hold and then crushed as well as scraped against the main cusp of p4.

The complex crown is supported by four roots, denoted as root 1–4, that indicate considerable occlusal pressure sustained by the tooth. Root 1 supports the anterior one-third of the tooth crown; it is strong and anteroposteriorly compressed. The posterior two-thirds of the crown is supported by three roots. On the lingual side, root 2 as the strongest root is transversely compressed and obliquely extended, parallel to the oblique lingual edge of the tooth crown; there is no sign of division of this robust root, suggesting it is originally a single root. Labial to root 2 are two small, column-like roots of which the anterior one is longer and thicker. The two small roots are closely packed with a fused base. They support the mesiobuccal corner of the crown, primarily cusp A1. Such a complex root condition indicates considerable occlusal pressure sustained by P4.

P4 shows by far the most specialized structure in known euharamiyidans from the Yanliao Biota, including Arboroharamiya jenkinsi ( Zheng et al. 2013, Meng et al. 2014), Xianshou linglong , Shenshou lui (paratype 1, WGMV-001), Maiopatagium (Meng et al. 2017) , Vilevolodon ( Luo et al. 2017) , Qishou jizantang , and Qishou sp. ( Mao and Meng 2019a, b) in the Yanliao Biota. Among those taxa, P4 of Qishou is proportionally the smallest, in relation to the molars, and simplest in crown structures in having several main cusps but lacking a broad basin filled with multiple cusps. The wear paưern indicates a longitudinal trench through the tooth crown, suggesting a relatively horizontal jaw movement during mastication. P4 of Maiopatagium is similar to that of Qishou , but its occlusal paưern is fundamentally different from the laưer, as pointed out by Mao and Meng (2019b). Other aforementioned Yanliao taxa developed a proportionally larger and more complex P4, of which that of Arboroharamiya jenkinsi appears to be the most derived. The general paưern in these forms is that P4 has multiple main cusps and many cuspules; the tooth crown is transversely expanded (wider than long) and the distal end is concave for receiving the mesial end of M1. P4 of Mirusodens differs considerably from these forms. It is relatively larger, has more cusps that are arranged in a regular paưern on the crown. Its crown is triangular in shape in occlusal view and longer than wide with a pointed distal end. These differences in the Yanliao euharamiyidans can be viewed as derived features, compared to that of Qishou , perhaps adapted for different diets or food sources.

Outside of the Yanliao Biota, the less P4 (PIN 5087/101) of Sharypovoia arimasporum from the Middle Jurassic (Bathonian), Western Siberia, Russia ( Averianov et al. 2019: Fig. 2 ) appears to be simple, more similar to that of Qishou or Maiopatagium . Moreover, the holotype tooth of ‘ Kermackodon multicuspis ’ ( Butler and Hooker 2005: Fig. 6 ; BMNH M46822), originally identified as a right M2 of a multituberculate, has been re-interpreted as P4 of an euharamiyidan ( Mao et al. 2022). The heart-shaped outline and wear paưern of the tooth is unlike any M2 of multituberculates, but similar to P4 of euharamiyidans. The tooth identified as the right upper premolar of Sineleutherus uyguricus ( Martin et al. 2010: Fig. 2H–J ; SGP 2005/3) is conspicuous. As interpreted by the authors, the lingual side of the tooth crown is remarkably flat and almost completely covered by an extensive wear facet; in addition, the base of the large distal cusp occupies the distal half of the crown and its tip bends mesially. The wear facet suggests strong and vertical shear from the lower tooth that has a similarly vertical shearing facet on the labial surface of the tooth, a condition that is unknown in any known ‘haramiyidans’. This tooth, if identified correctly, differs from all upper premolars of euharamiyidans preserved in associated dentitions and may represent a distantly related member of ‘haramiyidans’.

Upper molars: As in other euharamiyidans, Mirusodens has two molars in each jaw quadrant. Each molar has one massive root that is transversely narrow. From the groove on the lateral surface of the root shass, it can be inferred that the massive root was fused from two, mesiodistally arranged, roots. Unlike the upper incisor and premolars, the crown–root transition is not so distinct. M1 is longer and worn more deeply than M2, judging from cusp relief; minor cusps in the tooth basin of M1 are nearly erased by wear. As in other euharamiyidans, the distobuccal cusp (A1) is the largest of the tooth cusps and extends more distally than row B cusps. A concavity exists mesial to A1 and probably bears small cuspules, as in M2, but the cuspules are indistinct because of wear. There are two cusps on the mesiobuccal end of row A, denoted as cusp AA and Ax, which correspond to cusp BB and Bx in Kermackodon (‘ Eleutherodon ’) (see: Mao et al. 2022). Cusp AA is interpreted as homologous to the mesiobuccal cusp of M 1 in other euharamiyidans, such as A 5 in Arboroharamiya jenkinsi ( Meng et al. 2014) , A 2 in Sharypovoia arimasporum , or A 3 in Maiopatagium sibiricum [see Mao et al. (2022) for interpreting the cusp homologies]. Cusp Ax is weak, less developed than that in Kermackodon oxfordensis ( Butler and Hooker 2005, Mao et al. 2022). Ax is followed by a weak ridge. The central valley of the tooth crown is relatively straight, which is bounded by a borad and gentle buccal wall, and a narrow and steep lingual wall. The valley is not even in depth; its deepest area is buccal to cusp B2. There are six (or seven) B cusps of which the middle ones are higher than those on the two ends. As in other euharamiyidans, except for Maiopatagium ( Mao and Meng 2019b) , the buccal side of cusp row B, the lingual and buccal sides of A1 bear wear. Whether cusp row A bears wear on its buccal side is unclear, but in our interpretation, it most likely does. M2 is shorter than M1 and displays more cuspules because of minor wear. Cusp row Ax is more discernible in M2. There are small enamel ridges that extend from the buccal side of the crown to the basin floor. The two molars of Mirusodens are aligned in a typical ‘haramiyidan’ pattern in that they are aligned mesiodistally, differing from multituberculates that have the buccal cusp row of M2 aligns distally to the lingual row of M1, resulting in a different occlusal relationship for M1 and M2.

In lateral or medial view, the cheek teeth are not orientated with their occlusal surfaces horizontal; instead, they incline in different directions. For instance, the occlusal surfaces of P2 and P3 face ventroposteriorly, whereas that of P4 faces ventroanteriorly. Thus, the occlusal surfaces of the upper cheek teeth have the ‘en echelon’ (step-like) pattern ( Fig. 3A, C ), which was first recognized in the three upper molars of Haramiyavia ( Jenkins et al. 1997, Luo et al. 2015). In Mirusodens the most pronounced step-like region is between P2–3 mesially and P4 distally, whereas in other euharamiyidans with the full upper dentition preserved, it is between P3 and P4. The inclined occlusal surface of P4 is the longest with the mesial end higher than the distal end in position. The steps are proportionally small on the molars of Mirusodens , as in other euharamiyidans. The en echelon pattern of the upper dentition is associated with the tall cusp a1 of the lower molars, and the pronounced region between the penultimate premolar(s) and P4 corresponds to the enlarged cusp a1 of p4.

Upper molars of Mirusodens are singled rooted; the root is thick at the cervical region and gradually tapers toward the distal end. This paưern is present in most known euharamiyidans. The upper molars of Mirusodens are similar to Arboroharamiya ( Zheng et al. 2013, Meng et al. 2014), Shenshou lui (paratype 1, WGMV-001), Maiopatagium furculiferum (Meng et al. 2017) , and Qishou , in having cusp A1 not so distally extended, contrasting to those in Xianshou , Vilevolodon ( Luo et al. 2017) , and Sharypovoia ( Averianov et al. 2019) . However, A1 of Maiopatagium sibiricum appears extended distally, but this is partly due to a reduced row B that bears only one or two cusps. In addition, the upper molar of Mirusodens has a relatively longer row B that bears more cusps than other euharamiyidans except for Arboroharamiya jenkinsi , which is probably one of the reasons that cusp A1 does not appear so distally extended in these taxa. The important feature of the upper molar of Mirusodens , however, is the initial development of cusp Ax and row Ax, which is absent in other Yanliao euharamiyidans. These extra cusps are well developed in Kermackodon oxfordensis (‘ Eleutherodon ’ oxfordensis ) ( Kermack et al. 1998, Butler and Hooker 2005, Mao et al. 2022) and weakly so in the holotype specimen (M2) of Butlerodon from the Woodeaton locality. In general morphology, M2 of Mirusodens is most similar to that of Butlerodon from the Middle Jurassic Woodeaton locality of England ( Mao et al. 2022).

Lower incisor: As in other euharamiyidans, as well as in multituberculates and gondwanatherians, there is only one pair of enlarged lower incisors. CT-scan shows no tooth germ in the dentary. In buccal view, the incisor is in the shape of a curved dagger that tapers to a sharp tip that points dorsally. The buccal side of the tooth is convex and the lingual side is more flat. A sharp ridge extends along the dorsolabial rim of the crown, whereas the dorsolabial surface of the crown is rounded. From the exposed part, it is clear that the crown is covered entirely by enamel, and there seems to be no wear facet on the tip of the incisor. The root of the incisor is strong and extends backward to the level below m2; it is lingual to the roots of p4.

In general, the lower incisor is similar in all euharamiyidans, the only feature worth noting is that there is no indication of the lower incisor germ within the preserved dentary of Mirusodens , as in some euharamiyidans (Mao et al. 2019). Lack of the germ occurs in those that have relatively enlarged p4. The enlarged lower incisor and the deep mandible accommodating the enlarged tooth are highly similar to those of the marsupial Dactylopsila trivirgata and Petaurus breviceps ( Beck 2009, Burrows et al. 2020). These arboreal animals use the lower incisors to extract wood-boring larvae ( Cartmill 1974, Kay and Hylander 1978, Rawlins and Handasyde 2002), as well as to consume gums ( Rawlins and Handasyde 2002). The highly similar lower jaw and incisor morphologies between these arboreal marsupials and Mirusodens (other euharamiyidans as well) suggest that the laưer may have lived a similar life as the aforementioned extant arboreal animals, in the Jurassic forests, although these morphologies and lifestyles certainly evolved independently.

Lower premolar: As in other euharamiyidans that have the lower dentition preserved, there is only one lower premolar in each lower jaw of Mirusodens , and the tooth is denoted as p 4 in correspondence with P 4 in position and function. Both p4s are preserved in the holotype with the less p4 crown slightly crushed. The p4 is double-rooted and the roots are strong, with the distal root being thicker than the mesial one and the root of the molar. The roots are also long, ventrally passing the middle line of the incisor root on the lateral side of the laưer. The tooth is implanted slightly inclined in the dentary, with the crown leaning mesially. The tooth crown is transversely compressed and in buccal and lingual views, the crown outline is triangular, nearly symmetrical. The crown is formed almost completely by the hypertrophied cusp a1, and there is no distal cusp or heel, except that there is a small concave area on the distolingual base of the crown. The p4 is much larger than the molars and its length is longer than the total length of m1–2, suggesting that p4, working along with P4, is the primary functional tooth in processing food. There are four or five serrations along the mesial half of the blade-like crown; the distal one ends at the summit in the middle point of the crown. On either the lingual and buccal sides, a fine ridge that extends from the summit mesioventrally in parallel to the mesial edge of the crown and delimits a narrow band that is flat or slightly concave on the mesiobuccal surface.

There are roughly two types of p 4 in previously known euharamiyidans, except for Kermackodon oxfordensis : those (type I) that have a relatively small cusp a1 but large and multiple cusped distal portion (two rows of cusps) and a basin; the others (type II) have a large a1 but a simple heel with only a few small cuspules. Type I p4 is present in Shenshou , Qishou , Sineleutherus ( Martin et al. 2010, Averianov et al. 2011), and Sharypovoia ( Averianov et al. 2019) , whereas type II p4 is present in Arboroharamiya , Xianshou , and Vilevolodon . Type I p4 is similar to that of Thomasia ( Sigogneau-Russell 1989, Debuysschere 2015), although the identification of the isolated teeth in Thomasia is not certain ( Mao et al. 2022). Type I p4 probably represents the plesiomorphic condition, whereas type II p4 is derived in euharamiyidans. The p4 of Mirusodens represents perhaps the most specialized condition among known euharamiyidans in which the tooth has an enlarged crown that is longer than the total length of m1 and m2 and cusp a becomes predominant at the expenses of the distal heel and cusps. In addition, the blade-like crown has some serrations; in other euharamiyidans, p4 is transversely compressed but cusp a remains more or less conical, without any serration.

The p4 of Mirusodens is most similar to that of Kermackodon oxfordensis (BMNH M46684) ( Butler and Hooker 2005) in their general shapes. In particular, p4 of K. oxfordensis has three or four serrations. The serrations in these two euharamiyidan p4s, however, are different from those in multituberculates in being few and uneven. In multituberculates, such as the Middle Jurassic Tashtykia primaeva (PIN 5087/52; Averianov et al. 2020: fig. 6), the p4 crown is relatively low and has a rounded or squared outline in buccal view; the serrations are many and evenly distributed along the entire crown. Fine ridges extend from the serrations on both lingual and buccal sides in parallel arrangement. In function, p4 of multituberculates is different from that of euharamiyidans in that it shears against the upper teeth (except for some more advanced cimonodontans), while in euharamiyidans p4 bites against the basined P4, primarily for crushing.

As recognized by Averianov et al. (2020), BMNH M46684 is unique in having mixed characters of euharamiyidans and multituberculates. Its high sectorial crown and posterior basin surrounded by small cusps are typical of euharamiyidans, whereas p4 of multituberculates lacks a posterior basin. In addition, the mesial end of BMNH M46684 lacks a vertical groove for holding the preceding premolar, another euharamiyidan feature. In BMNH M46684, there is still a small heel with multiple cuspules and the enamel ridges or serrations are on the distal half of the tooth crown. However, in p4 of Mirusodens the serrations are on the mesial half of the tooth and the distal basin is absent. The p4 of Mirusodens also lacks the groove on the mesial end of p4; this is because there is no additional lower premolar mesial to p4, which should be regarded as an euharamiyidan feature.

As discussed above, the upper premolar of Sineleutherus uyguricus is unique ( Martin et al. 2010). Several teeth identified as the lower premolars in the same species are also interesting. Among those teeth, SGP 2004/6, identified as a less ultimate lower premolar, is similar to type In p4 of euharamiyidans. We concur with this identification except that the tooth is most likely a right p4. Of the other three teeth, SGP 2004/15 has only one row of three cusps, whereas SGP 2004/16 and SGP 2004/17 bear some small cuspules on the distolingual base. A common feature shared by the three teeth is a large flat wear facet on the buccal side of the crown ( Martin et al. 2010: fig. 3), which matches or is similar to that on the lingual side of the tooth identified as the upper premolar (SGP 2005/3). Several implications can be made based on these teeth: First, because SGP 2004/6 was considered as an ultimate lower premolar, then the other three premolars with simpler cusp morphologies may be mesial premolars, as inferred by Martin et al. (2010): ‘it [SGP 2004/15] may derive from a more anterior position in the tooth row’. Also, the authors noted ‘a prominent projection at the base of the crown for the interlock with the following tooth’ on SGP 2004/16-17. This means that S. uyguricus has more than one lower premolar. Second, because the type I SGP 2004/6, identified as the ultimate lower premolar, does not have a large flat wear facet to match that on SGP 2005/3, then the laưer is unlikely to be the ultimate upper premolar; this suggests that there is more than one upper premolar. Third, given the identification of the large wear facet on the lingual side of the upper premolar (SGP 2005/3) and buccal side of the lower premolars (SGP 2004/15-17), it could be inferred that the wear facets were created by shearing contact between these presumably mesial lower and upper premolars. However, because of the cusp orientation, these shearing facets would be created in an awkward way in which the tallest cusp of the upper premolar is distal, whereas the tallest cusp on the lower premolar is mesial. How such extensive wear facets could be formed remains a challenging issue to be explored. New evidence may prove S. uyguricus to be a unique species that has a lower molar similar to those of euharamiyidans in the almost coeval Yanliao Biota but possesses additional mesial premolars that are yet unknown in any other ‘haramiyidans’.

Lower molars: As in the upper dentition, lower molars are relatively small compared to the enlarged p4. As in other euharamiyidans, the mesiolingual cusp (a1) is the largest and tallest, and positioned near the longitudinal axis of m1. The m2 is considerably smaller than m1 and its row b is reduced. There is a central valley on both m1 and m2. A distinct feature, however, is the presence of a distinct cusp b1 mesiobuccal to the base of a1. On the less m1, this cusp was broken but its base is still discernible, while b1 is distinct on the right m1. Presence of b1 is a common feature in Triassic ‘haramiyidans’ and European Middle Jurassic species ( Mao et al. 2022). Both lower molars are single-rooted. The strong root gradually tapers distally. The root of m1 is shorter than those of p4, whereas the root of m2 is the shortest and weakest. As suggested by the crown size and shape, the root condition also indicates the functional role played by the teeth decreases from p4 to m2.

Presence of b1on m 1 in Mirusodens is an interesting and critical feature. On the lower molars of Thomasia , b1 is a distinct cusp, mesiobuccal to the base of a1, although it is lower than b2. A similar cusp was denoted in Haramiyavia ( Hahn and Hahn 2006, Luo et al. 2015). This cusp condition is highly similar to that of Mirusodens . In other euharamiyidans, b1 was denoted in some euharamiyidans, such as Arboroharamiya ( Meng et al. 2014) , Sineleutherus uyguricus , and ‘ Sineleutherus ’ issedonicus ( Averianov et al. 2019) , but it is a small cusp distobuccal to a1 and is within the row of b cusps (not lower than b2). Whether b 1 in Arboroharamiya and Sineleutherus is homologous with that of Mirusodens and Thomasia is uncertain ( Mao et al. 2022). Presence of this cusp on m1 of Mirusodens and European taxa is another feature that suggests possible relationship between euharamiyidans from the two areas.

Postcranium ( Figs 1 , 5 ): The dorsal axial skeletal elements and ribs were gone or preserved as carbonized film, suggesting that these elements are less ossified or more gracile than the limb bones. Although the dorsal vertebrae are not preserved, the impressions of ribs indicate that the thoracolumbar transition is distinct, as in other euharamiyidans ( Bi et al. 2014, Mao and Meng 2019a), which is a feature in extant mammals. The long tail consists of at least 16 caudal vertebrae. Some remains of caudal vertebrae are preserved and each caudal vertebra has a long and thin centrum. The number and length of the caudal vertebrae suggest a tail with possible prehensile ability, similar to those in other euharamiyidans.

Most of the limb elements are preserved but in split condition ( Fig. 1 ). The hindlimbs are splayed out and forelimbs are overlapped in preservation, but it is clear that the former are longer than the laưer. For the hindlimb, the femur is shorter than the tibia and fibula. Where the bones are present, they largely remain in original articulation, except for the right pes and both manus that are displaced or partly missing. As in other euharamiyidans reported from the Yanliao Biota, the limb skeleton is gracile with elongated elements, displaying features characteristic of arboreal and even gliding locomotion. For instance, the ulna is proportionally long but the olecranon process is extremely short. The digits of both the pes and manus are slender and long; in lateral view they are curved and dorsoventrally thickened (the depth is greater than the width of each phalanx). Digit III appears to be the longest, whereas digit I is the shortest of the five, as best shown in the well-preserved less pes. Pedal digit I (dI) is the shortest; dII, dIII, and dIV are long and subequal in length, with dIII slightly longer than the other two. The dV is the second shortest digit. The ankle is compact (proximodistally short), as in mammals, such as Jueconodon , but different from the mammaliamorphs, such as Fossiomanus ( Mao et al. 2021) ; both Jueconodon and Fossiomanus are from the Early Cretaceous Jehol Biota ( Mao et al. 2021).

While the general skeletal morphology is similar to other euharamiyidans, the less pes displays some additional features. Viewing the pes as a whole with the fur impressions on it the toes are proportionally long, whereas the sole is relatively short; this reflects the osteological structures ( Fig. 5 ). The five long and well-developed pedal digits are all separate and more or less evenly spaced; it does not show the opposite arrangement of digit I, as in some arboreal marsupials ( Fig. 5G, H ). However, for arboreal or gliding small mammals, the digit arrangements may not be so different compared to those of terrestrial species ( Fig. 5I–L ).

It is also clear that, as in other euharamiyidans, the proximal or intermediate phalanges are subequal to, or longer than, the metatarsals and metacarpals. Thus, the length of each finger or toe is much longer than the corresponding metapodial ( Fig. 5E, F ). This feature differs from those of small mammals ( Fig. 5G–M ), which is best illustrated in the bone elements of the arboreal Glaucomy ( Fig. 5M ). The elongated digits, along with the sharp and curved claw sheath, suggest capability of manual and pedal prehension, consistent with the interpretation that euharamiyidans are primarily arboreal animals ( Zheng et al. 2013, Bi et al. 2014, Luo et al. 2015, Meng et al. 2015, Han et al. 2017, Mao and Meng 2019a, Wang et al. 2021).

The long fingers and toes could provide the capability for holding small tree branches. However, as tree branch or trunk gets thicker, the manus and pes of these small animals cannot reach around to grip by prehension of digits. Also, the toes (fingers may well be the same) of Mirusodens do not have expanded plantar pads to provide sufficient pad friction to keep the animal from falling. It is the claws that contribute to the ability for Mirusodens to cling to structures that have a sizable diameter, as in extant squirrels ( Cartmill 1974). The long and sharp manual and pedal claws extend well beyond the apical pads of each toes, allowing the animal to cling on to tree trunks in different orientations, even vertical, by digging the claws into the substrate as anchor points; this would increase vertical agility of the animal on tree trunks and allow the animal to move in all directions or at an angle across the climbing surface. Because of the divergent toes and fingers, as well as the long limbs, the claws can be spread out and positioned at relevant places so that the body (centre of gravity) is kept close to the tree and secure the body mass being evenly distributed across tree trunk and branches against the gravity, preventing the animal from falling.

Sofl tissues ( Figs 1 , 5 , 6 ): Soss tissues refer to impressions or potential remains of the integumentary system (fur impressions, keratin sheath, pads of the digits, and carbonized skin and hair) ( Figs 1 , 5 , 6 ). It should be pointed out that the specimen is split into part and counterpart, and the breakage goes through the body of the animal such that the fur on the body surfaces would be preserved in the matrix of each split part (blocked by the skeleton, organic residues of the body, and potential stomach remains); thus, only those on the periphery of the body are beưer exposed. Although fur impressions of mammaliaforms from the Yanliao Biota have been reported from previous studies (Ji et al. 2006, Meng et al. 2006), the exquisitely preserved fur impressions of Mirusodens display some new details of fossilized integumentary system. The entire body of the small animal was insulated by dense fur, a good indicator of endothermy. Hair has differentiated into the guard hairs and under hairs with a diversity of length and density. At different areas of the body, hair varies in length and thickness, and density. The guard hairs are long, sparse, and coarse; they extend outward in the preserved impressions. The under hairs are short, fine, and dense; they are concentrated close to the body, such as the back of the skeleton, as expected. In the body area, the carbonized layer of organic remains, potentially the hair and perhaps skin, become thicker and darker. Hairs around the limbs are straight and long relative to the length of limb bones; some hairs measure up to 20 mm. By the preserved condition, these long hairs are on the ventral side of the body. This is similar to the condition displayed in the extant gliding mammals, such as the marsupial Petaurus and the placental Glaucomys , where hairs on the ventral side of the body, especially around the limbs, are long. In contrast, those on the toes are fine and short. The thickest hairs are along the caudal vertebra of the tail. The thickness for the short and fine hair ranges from 6 μm to 20 μm and can be as thick as up to 90 μm in those of the tail.

Unlike some gliding euharamiyidans ( Han et al. 2017, Meng et al. 2017) and Volaticotherium ( Meng et al. 2006) , the impressions of the integumentary system on the holotype of Mirusodens do not show a clear sign of a patagium (the gliding membrane). Usually, the patagium can be recognized by its well-defined edge and the dark area (in contrast to the matrix) that bears fine impressions of hairs between body parts. However, in extant gliding mammals, either marsupials or placentals, the edge of the patagium is not sharply defined by skin but by hair, although the hair along the edge of the patagium is relatively short and even in length ( Fig. 2H, I ). If the gliding membrane is not fully stretched, the edge of the membrane is invisible. Nonetheless, the fur outline in gliding mammals is broader than those of non-gliding ones. In the holotype of Mirusodens , the distributions of the fur impressions and the dark area that is apparently derived from the animal body are broad, extending between the head and forearms, between limbs, between hindlimbs and the tail, and along the tail. Such a distributional paưern does not seem to be less only by hairs from a non-gliding animal. Compared to the extant gliding mammals and other gliding species of euharamiyidans from the Yanliao Biota ( Han et al. 2017, Luo et al. 2017, Meng et al. 2017), we interpret that the dark areas are less by skin membrane and hair. Thus, Mirusodens represents another gliding species in the Yanliao Biota.

The less pes has been preserved in such an exceptional condition that the fur and pedal skin impressions of the toes, in association with the bones, are visible, representing an unprecedented example of pes morphology in Mesozoic mammals. Hairs are short, fine, and dense on the ankle and instep, and gradually reduce in density toward the toe tips. Some hairs extend to pass the apical pedal tip and reach to one-third or halfway along the claw. The hairs on the pes of Mirusodens are similar in length and distribution to those of extant small mammals ( Fig. 5G–L ). Impressions of the pedal skin or digital pads are visible, which have well-defined edges in contrast to the hair impressions overlapping it. They show that each toe ends distally at the proximal base of the distal phalanx. The skin impression outlines the shape of the long and separated toes. A slight curvature is present at the joint of the proximal and intermediate phalanges of digit II. The plantar pads do not show any expansion so that the toe gradually tapers toward the claw. The relationship of the hair and pedal skin suggests that the dorsal side, instep, of the pes is exposed because there is no sign of the foot pad, which should be on the plantar side of the pes; this orientation is consistent with the curvature of the phalanges. Lack of the digital pads differs from the pes of extant mammals ( Voss and Jansa 2009, 2021).

As shown by the impressions or moulds, the manual and pedal digits bear well-developed claw sheaths that are sharp, transversely compressed, and dorsoventrally recurved. Possible remains of the keratin sheath are present in some digits. The horny claw sheath extends from the proximal base of the distal phalanx beyond the apical pads of each toe. The length of the sheath is about twice that of the distal phalanx length; it increases the length, sharpness, and curvature of the claw, which allows the capability of climbing on thick tree trunks (see below). Previous studies have recognized the arboreal limb structures of euharamiyidans, particularly the long digits relative to the metacarpals and metatarsals ( Zheng et al. 2013, Bi et al. 2014, Luo et al. 2015, Meng et al. 2015, Han et al. 2017, Mao and Meng 2019a, Wang et al. 2021). The new morphologies of the pes reinforce the arboreality of these euharamiyidans and furnish additional arboreal features. These pedal morphologies reported here are unknown in any non-mammalian tetrapod but highly similar to those of extant mammals, such as squirrels ( Cartmill 1974) or arboreal marsupials ( Voss and Jansa 2009, 2021) ( Fig. 5G–L ).

Phylogenetic analyses

With the new species and morphological features, we conducted a phylogenetic analysis based on a data matrix with 131 taxa and 573 characters (see Material and Methods and Supporting Information, File S1, S7). The euharamiyidan data have been refined to reflect revision of some allotherian species and their dental morphologies. In particular, teeth previously assigned to four Jurassic genera of ‘haramiyidans’ and multituberculates have been reinterpreted as from different tooth loci of the same euharamiyidan species, Kermackodon oxfordensis ( Mao et al. 2022) . The resulted consensus tree of the analyses is illustrated in Figures 7 , 8 . In the phylogeny, Mirusodens is grouped closely with arboroharamiyids from the early Late Jurassic Linglongta phase of the Yanliao Biota and the European Jurassic species ( Kermack et al. 1998, Butler and Hooker 2005, Mao et al. 2022), to which shenshouids from the Yanliao Biota and Siberia form the outgroup. This is consistent with the view that ‘haramiyidans’ diversified and had cosmopolitan distributions in the Middle Jurassic. Our analysis places Euharamiyida (see below for definition) as the sister-group of gondwanatherians with Cifelliodon and Thomasia being the successive outgroups; this clade further pairs with multituberculates to form a sister-group that has Haramiyavia and Theroteinus as the outgroups. The Allotheria as a clade is deeply nested within Mammalia and forms the sister-group of the clade leading to therians, which is largely similar to the results of other studies ( Krause et al. 2014, 2020a, Hoffman et al. 2020, Wang et al. 2021). ‘Symmetrodontans’ and eutriconodontans are the successive outgroup taxa of this allotherian–therian clade.

Our analysis reinforces the view that multituberculates, ‘haramiyidans’, and gondwanatherians constitute the clade Allotheria within Mammalia, which has long been recognized ( Butler 2000, Kielan-Jaworowska et al. 2004, Hahn and Hahn 2006) and has been supported by recent phylogenetic analyses ( Luo et al. 2007, Zheng et al. 2013, Bi et al. 2014, Krause et al. 2014, 2020a, Han et al. 2017, Hoffmann et al. 2020, Mao et al. 2020, 2021, Wang et al. 2021), although competing hypotheses exist ( Luo et al. 2015, 2017, Huưenlocker et al. 2018). Given the preferred phylogeny and the morphological evidence, it is most probable that the Late Triassic Haramiyavia and Theroteinus represent the primitive morphotypes of allotherians that gave rise to the Jurassic euharamiyidans and multituberculates, a view previously put forward by others ( Van Valen 1976, Hahn et al. 1989, Butler 2000, Butler and Hooker 2005, Hahn and Hahn 2006) and supported by several phylogenetic analyses ( Bi et al. 2014, Krause et al. 2014, Han et al. 2017, Mao et al. 2021, Wang et al. 2021), and that mammals originated in the Late Triassic ( Bi et al. 2014, Mao et al. 2021).

Phylogenetic definition of Allotheria

A phylogenetic definition for Allotheria was given by Sereno (2006: 319) as: ‘the most inclusive clade including Taeniolabis taoensis Cope 1882 , but not Mus musculus Linnaeus 1758 ’. This definition has been adopted by Wang et al. (2021). However, the phylogenetic relationships of allotherians and other extinct groups have been unstable, and an ideal phylogenetic definition remains challenging. For instance, the above definition fits well to the phylogeny of Krause et al. (2020a) but not necessarily so for others, such as Krause et al. (2014), Luo et al. (2015), Wang et al. (2021: extended data figs 8, 9), and this study because the definition does not sufficiently accommodate taxa that do not belong to the inclusive clade including Taeniolabis taoensis , nor are they closely related to the clade containing Mus musculus or Ornithorhynchus anatinus . In our phylogeny ( Figs 7 , 8 ), for instance, such taxa include ‘symmetrodontans’, eutriconodontans, Tinodon , and Fruitafossor . Employing species such as Taeniolabis taoensis as the anchor point to formulate the definition could promote stability of the definition (Sereno 2006), compared to that using higher-level taxa to formulate the definition ( Rowe 1988), but this may also be problematic for practical reasons. For instance, in some studies of early mammaliamorphs, the terminal taxa used for the phylogenetic analysis are primarily fossils, such as Krause et al. (2014), that do not include Mus musculus , and in others the terminal taxa are at the generic or higher taxonomical rank, such as cimolodontans in Luo et al. (2015), which makes the definition per se semantically unclear if contrasting to those phylogenies. Considering these factors, we propose an alternative phylogenetic definition: Allotheria is the most inclusive clade containing taxa of Multituberculata (sensu McKenna and Bell 1997) but not those belonging to the clades of therians, monotremes, or any falling between the laưer two clades. In this definition, ‘taxa’ implies any number of taxa at either the species, generic, family, or even higher level. Given that the allotherian phylogenies are still unstable, the definition hopefully offers the flexibility that could fit to phylogenetic analyses regardless how many terminal taxa and at which ranks are used.

Phylogenetic definition of ‘Haramiyida’

Wang et al. (2021: their supporting information) proposed a phylogenetic definition for ‘Haramiyida’: ‘The most inclusive clade including Thomasia antiqua Plieninger 1847 , but not Cifelliodon wahkermoosuch Huưenlocker et al. 2018 , Taeniolabis taoensis Cope 1882 , or Adalatherium hui Krause et al., 2020 ’. The intention of this definition was perhaps to include Thomasia as the anchor taxon in the group because it is the typical ‘haramiyidan’ in the conventional view. However, this definition excludes Haramiyavia , another key member in the conventional taxonomy of ‘haramiyidans’. The most problematic aspect of this definition is its being narrowly phylogeny-specific. This definition is based on, and thus reflects, the phylogeny reconstructed by Wang et al. (2021); it is hardly applied to many other phylogenies proposed in recent studies or, if applied, it would result in a distortion of what ‘haramiyidans’ mean in a broadly accepted sense. For instance, in the phylogeny obtained by Huưenlocker et al. (2018: fig. 4), ‘Haramiyida’ as defined by Wang et al. (2021) contains only the clade of Haramiyavia and Thomasia that forms the sister-group of the clade of Eleutherodontida. However, all eleutherodontidans are not ‘haramiyidans’ by definition because the Eleutherodontida contains Cifelliodon , a taxon that is excluded from ‘Haramiyida’ by the definition. In the phylogeny of Krause et al. (2020a), for instance again, the pair of Haramiyavia – Thomasia was clustered with tritylodontids and placed outside of mammaliaforms; thus, by the definition of Wang et al. (2021) tritylodontids are ‘haramiyidans’ but all euharamiyidans are not. Similarly, the definition of Wang et al. (2021) cannot apply to the phylogeny we have here, because Thomasia is the outgroup of the clade that contains Cifelliodon , gondwanatherians, and euharamiyidans; again, the clade that contains Thomasia antiqua also contains taxa that are excluded by the definition of Wang et al. (2021). All these cases result from the fact that Thomasia is generally primitive in dental morphology and represented by poor specimens; thus its phylogenetic position is unstable. Therefore, the definition of Wang et al. (2021) for ‘Haramiyida’ creates instability in the phylogeny and taxonomy of ‘haramiyidans’. There seems no meaningful phylogenetic definition of the potentially paraphyletic ‘haramiyidans’ (in the conventional sense) based on available evidence. For the sake of histological and practical reason, however, we think it is still useful to keep the widely used term ‘Haramiyida’ (‘haramiyidans’), placed in quotation marks. This usage is more or less similar to ‘Triconodonta’ (‘triconodontans’) vs. Eutriconodonta (eutriconodotans).

Phylogenetic definition of Euharamiyida

It is possible to reach a practical and stable definition of Euharamiyida for the reason that this group is supported by mounting derived features and thus becomes increasingly stable phylogenetically. Based on our phylogeny ( Figs 7 , 8 ) and other recent phylogenies ( Huưenlocker et al. 2018, Krause et al. 2020a, Wang et al. 2021), we propose the following phylogenetic definition: Euharamiyida is the most inclusive clade including taxa of arboroharamiyids (Arboroharamiyida, Xianshou , and Vilevolodon ) but not those belonging to the clades of multituberculates, gondwanatherians, monotremes, therians, or any clade falling between the named clades. Again, in this definition, ‘taxa’ implies any number of taxa at the species, generic, family, or even higher level, from a clade that are used for reconstructing the phylogeny.

Figures

Figure 1. Holotype of Mirusodens caii gen. et sp. nov. (HT-B-PM-0001). A, main part (part A or ‘less part’ when referring to anatomic orientation of the split skeleton) in which most cranial structures and upper teeth were preserved; B, counterpart (part B or ‘right part’) in which most lower teeth were preserved. Red boxes 1–7 correspond to figure panels A–G in Figures 6 and 8–10 and to A–D in Figure 5. Note that the skeleton is preserved in association with numerous valves of conchostracans that are typical invertebrate fossils in the Daohugou strata. The dark area probably represents residues of the pelage, possibly the patagium (see Figs 5 and 6 for comparison with extant gliding mammals). Abbreviations:l-fe, less femur; l-fi, less fibula; l-hu, less humerus; l-I2, less and presumably the second upper incisor; l-ra, less radius; l-ti, less tibia; l-ul, less ulna; r-I2, right and presumably the second upper incisor.

Figure 2. Skull of Mirusodens caii gen. et sp.nov.(holotype, HT-B-PM-0001). A, partial skull in the main part of the slab (part A) in which the less half of the skull and most teeth are preserved; B, partial skull in the counterpart of the slab (part B); C, D, two micro-CT slices through different positions of the partial skull in part A, showing the preserved condition of the specimen and the resolution of the scan; E, F, CT-rendered skull in preserved part A: E, the exposed or right side of the preserved skull in slab A, which is mostly broken; F, the less side of the skull that is embedded in the matrix and thus beưer preserved. Abbreviations:amf, anterior extremity of the masseteric fossa; cop, coronoid process; glf, glenoid fossa; l-i, less lower incisor; l-I2, less upper incisor (I2); l-M1, less upper first molar; l-m1, less lower first molar; l-m2, less second lower molar; l-M2, less second upper molar; l-P2, less second upper premolar; l-P3, less third upper premolar; l-p4, less ultimate premolar (p4); mac, mandibular condyle; mef, mental foramen; nuc, nuchal crest; r-I2, right upper incisor (I2); r-P2, right upper second premolar; r-P3, right upper third premolar; r-P4, right upper ultimate premolar (P4); zya, zygomatic arch.

Figure 3. Upper teeth of Mirusodens caii gen. et sp. nov. (holotype, HT-B-PM-0001). A, upper dentition in less side view (labial for the less dentition and lingual for the right one); B, occlusal view of the upper teeth; C, upper teeth in right side view (labial for the right dentition and lingual for the less one). Note the root condition in each tooth. In (C) the root of the right incisor is broken, whereas the tooth crown of the less one is blocked by the right one. The teeth are restored digitally from the main and counterpart slabs and represent the preserved condition in the matrix.The white line outlines of the occlusal surfaces of the upper cheek teeth in lingual and buccal views, showing the en echelon (steplike) paưern, as first noted in Haramiyavia (Jenkins et al. 1997). Abbreviations:l-, indicates less side; mf-r, medial contact facet of the right upper incisor; r-, indicates right side; row-B, cusp row B of the upper molar; rt1–4, root 1, 2, 3, and 4 of the ultimate upper premolar (l-P4).

Figure 4. Lower teeth of Mirusodens caii gen.et sp. nov.(holotype, HT-B-PM-0001). A, skull (semi-transparent) in less side view, showing relationships of teeth in the skull; B, less side view of upper and lower teeth that are digitally restored from the main and counterpart slabs. Note that p4 is considerably longer than P4 and that the root of the lower incisor is on the lingual side of p4 roots; C, less upper and lower dentitions in lingual view; D–F, right lower p4–m2 in labial, lingual, and occlusal views.Note the position of cusp b1 in m1 (broken in m2); G–I, less lower m1–m2 in occlusal, lingual, and labial views; cusp b1 is broken in both m1 and m2.

Figure 5. Pes and manus morphologies and soss tissues of Mirusodens caii gen.et sp. nov.(Holotype, HT-B-PM-0001) in comparison with those of extant mammals.A, C, D, correspond to red-boxed areas 8–10 in Figure 1. A, less hindlimb, showing the hair impressions in related to the limb bones. The hair distribution is narrowest at the ankle area and becomes much broad toward the knee; such an area does not seem to be all aưributable to only hair because the hair impressions appear fine and short; it is most likely that the hair is on the patagium that stretched from the hindlimb to the tail and to the trunk of the body; B, the less pes in dorsal view, showing the toes, claws, potential impressions of the pedal skin or digital pads that outline the shapes of the toes that are long and well separated.Hair are short and fine on the instep and toes; C, disarticulated right pes, showing the shapes and relative lengths of the footbones; D, a claw in the manus, showing the remain and impression of keratinous claw and relationship of the phalanx (partial impression) and claw sheath; E, bony elements of the pes of euharamiyidan (unpublished figure) in ventral view, showing the relative length of the bony elements; F, the bony elements of the manus of Shenshou; G–I, pes morphologies of extant marsupials in dorsal view (G, gliding Petaurus, AMNH 196914; H, arboreal Marmosa, AMNH 266428; I, terrestrial Monodelphis, AMNH 263547); J–L, pes morphologies of extant placentals in dorsal view (J, gliding Glaucomys, AMNH 188250; K, arboreal Microsciurus, AMNH 32497; L, terrestrial Geosciurus, AMNH 83652); M, bony elements of the pes of Glaucomy in dorsal view (AMNH 267293). Red arrows in (B) point to the edge of the toe skin edges.Red and blue lines in (E), (F), and (M) indicate the relative length of metapodial and proximal phalanx of digit I. Some images in (E–M) are photographically reversed for comparison.Abbreviations:ast, astragalus; csh, claw sheath (impression and remain); dph, distal phalanx (phalanges); iph, intermediate phalanx; I–V, digits from I to V; l-fi, less fibula; l-ti, less tibia; mt, metatarsal; pat-h, possible patagium around the hind limb; pph, proximal phalanx (phalanges); tcsh, tip of claw sheath; tdph, tip of distal phalanx (impression).

Figure 6. Hair impressions of Mirusodens caii gen. et sp. nov. (holotype, HT-B-PM-0001) in comparison with extant mammals. A–G, correspond to red-boxed areas 1–7 in Figure 1. A, hair impressions at the outer area of the body fur; these hairs are long, fine, and somewhat curly.B, imaged area in the middle of the body fur impression; hair impressions are still visible but not so distinct compared to (A). C, imaged area near the skeleton, where the hair impression is unclear; this area may represent organic remains less by skin.D, imaged area on top of the skull, showing short, fine, and dense hair.E, imaged area around the forearm, showing long hair in comparison with the limb bones. F, sampled area along part of the caudal vertebrae; F’, close-up view of the red-boxed area in (F). The hairs along the caudal vertebrae are long, thick, and straight. Numerous unidentified spherical particles are caught among the coarse hair, as pointed by the two white arrows and exemplified in the upper right corner.G, imaged area near the chest, showing carbonated films of the rib and dark areas that are possible organic remains less by skin.Note that in all these areas, there seems no evidence, such as a clear membrane edge, that suggests a patagium.However, it seems unlikely that all the dark areas represent fur. For instance, the dark area along the caudal vertebrae is much broader than the area bearing the coarse hair; such a wide area does not seem to be formed only by hair but possibly suggests presence of the patagium; H, from top to boưom: marsupial Petraurus (AMNH 196914), gliding; Marmosa (AMNH 266428), arboreal; and Monodelphis (AMNH 263547), terrestrial; I, placental Glaucomys (AMNH 188250), gliding; these show extension of the pelage in the body and tail morphology of gliding species in contrast to nongliding species.

Figure 7. Strict consensus tree of 990 trees retained in heuristic search using (PAUP*, v.4.0) (Swofford 2002). Tree length = 2968; consistency index (CI) = 0.3245; homoplasy index (HI) = 0.6755. See Material and Methods and Supporting Information, File S1.

Figure 8. Key phylogenetic nodes and clades within mammaliamorphs. The phylogeny is condensed from the strict consensus tree Supporting Information, File S1, showing the placement of Mirusodens caii within Euharamiyida, Allotheria and Mammalia. The phylogenetic frame form the basis for the definitions of Euharamiyida and Allotheria (see text).