Lactodens sheni, Han & Meng, 2016

Han, Gang & Meng, Jin, 2016, A new spalacolestine mammal from the Early Cretaceous Jehol Biota and implications for the morphology, phylogeny, and palaeobiology of Laurasian ‘ symmetrodontans’, Zoological Journal of the Linnean Society 178 (2), pp. 343-380 : 350-368

LACTODENS SHENI GEN. ET SP. NOV.

Etymology

Named after Mr Wentao Shen, the collector of the holotype specimen.

Diagnosis of species

Same as for the genus. Holotype

A partial skeleton with partial cranium and associated dentaries with some teeth preserved in situ ( Fig. 1 ). The specimen ( HG-M016 ) is housed in the Bohai University Paleontology Center , Jinzhou City, Liaoning Province.

Type locality and age

Shangheshou area, Caoyang City, Liaoning Province, China; Jiufotang Formation, Early Cretaceous (Aptian) ( He et al., 2004; Pan et al., 2013). The other known mammal from the formation is Liaoconodon hui ( Meng, Wang & Li, 2011) .

Description

Upper dentition: The upper right dentition is completely preserved ( Figs 3 – 7 ), showing the dental formula I3-C1-P3-M6. The left cranium and teeth must be in the counterpart of the slab that was not collected. However, two left upper molars, a partial M3, and a complete M4 are preserved ( Fig. 8 ). Tooth measurements are provided in Table 1.

The first upper incisor (I1) is displaced ( Figs 3 , 4 ). I2 and I3 are bent and are broken between the crown and roots but the tooth crowns are intact and lay by the broken roots, which are preserved in situ. The upper incisors are single-rooted as revealed by the breaks, but the root of each incisor shows a tendency of bifurcation distally, which is rare amongst Mesozoic mammals where upper incisors are known. The incisors appear vertically implanted, in contrast to the mesially inclined lower incisors. There is a narrow gap between adjacent incisors and between I3 and the upper canine. The upper incisors are similar in crown morphology; they are conical, with a convex mesial surface and a flat distal surface that is delimited by sharp edges at the lingual and labial sides. The crown length is slightly greater than its width; the latter is difficult to measure for I2 – 3 because of the condition of preservation. I1 is the smallest of the three upper incisors, with its distal side exposed so that only its width can be measured. I2 is slightly larger than I3.

The upper canine is double-rooted, with the roots diverging dorsally ( Figs 3 , 4 ). The crown is noncaniniform (or premolariform), being extremely low and transversely narrow; it is considerably longer than the incisor. In lingual view, a cusp (tentatively denoted as A) occupies the mesial half of the crown and has a triangular profile. There is no cusp mesial to cusp A. Cusp A slopes distally and is followed by a small cusp (tentatively denoted as C) at the distal end of the tooth, separated from cusp A by a notch in lingual view. There is no cingulum. On the canine and incisors, including the lower incisors, the enamel of the anterior surface is darker in colour and thicker than the distal surface.

There are three upper premolars, which are denoted as P1 – 3, respectively. A small diastema (0.4 mm) separates the canine from the mesial premolar (P1), whereas the diastema between P1 and P2 and that between P2 and P3 is 0.6 and 1.1 mm, respectively. Compared to the tooth length ( Table 1), these diastemata are substantial. Each of the upper premolars is double-rooted. The roots are transversely narrow, with the distal one being stronger than the mesial one. All three upper premolars are extremely low crowned, transversely thin, and mesiodistally elongate ( Figs 3 – 6 , 9 ; Table 1). P1 and P2 are similar in shape except that P1 is slightly smaller. Each of the teeth has a main cusp A (damaged on P1) that is slightly mesially positioned in lateral view, a small cusp B mesially and cusp C distally. These cusps are aligned mesiodistally and the premolars essentially have the ‘triconodont’ tooth pattern. The cusps, more distinct on cusp A, have a convex labial surface and relatively flat lingual surface. There is no cingulum on P1 – 2. The ultimate premolar, P3, is significantly larger and more complex than P1 and P2. Cusp A is much higher and proportionally more mesially positioned on P3. A distinct difference between the upper and lower premolars is that cusp A is proportionally small and low in the upper premolars so that their crowns appear even lower than those of the lower premolars. P3 has a small distal cusp C, a weak lingual cingulum running nearly along the full length of the tooth, and a weak labial cingulum on the distolabial base of the tooth. All the upper and lower premolars have a morphology that is typical of mammalian deciduous teeth, as discussed by Cifelli (1999) (see Discussion).

The diastema separating P3 from M1 is 0.3 mm. Unlike the premolar series, there is no diastema between any adjacent pair of molars ( Figs 3 , 5 , 7 , 9 ). Corresponding to the lower molars in the dentary, the labial alveolar margin of the maxilla is ventrally much lower than the lingual one. This morphology is related to the tooth crown height and orientation in which the upper molar is relatively high crowned lingually and low crowned labially, whereas the lower molars show opposite crown heights. This relationship makes sense functionally because the strong labial main cusps of the lower molars would occlude with the strong lingual main cusps of the upper molars. The crown height, also reflecting the height of the paracone (cusp A), increases distally from M1 to M4 (the cusp tip of M3 is broken) and then decreases from M4 to M6. The length decreases distally from M1 to M6 and the angle defined by the three main cusps, with the paracone being the apex, decreases distally from M1 to M5 and then increases from M5 to M6 ( Table 1). As in other spalacotheriids, crests are developed between the paracone and labial cusps, forming continuous prevallum and postvallum shearing surfaces ( Crompton & Jenkins, 1967; Crompton, 1971). The specimen represents an adult individual because even the postparacrista of M6 bears a wear facet. In occlusal view, the outline of the upper molar crown, particularly for M2 – M6, is triangular, with the distal edge longer than the mesial one. By contrast, the mesial edge of each of the triangular lower molars, especially m2 – m5, is longer than the distal one. The triangular space, or embrasure, between two upper molars generally reflects the crown outline of the corresponding lower molar. Along with the stronger and longer protocrista of the upper molars and the longer mesial crest of the lower molars, embrasure shear (Patter- son, 1956) is largely restricted to the opposing surfaces of corresponding upper and lower molars.

Owing to breakage of the maxilla on the lingual side of P3 – M3, it can be seen that the roots of these teeth are short relative to the crown height, contained in the thin palatal process of the maxilla. M1 has two roots with the distal one being stronger. Differing from P3, the roots of M1 and the other upper molars are not transversely compressed. Moreover, there is a trend that the distal root of the molars is more labially positioned in the distal molars. M1 is subequal to P 3 in length but is much higher and transversely wider than the latter ( Figs 6 , 7 , 9 ). The width ratio between M1 and P3 is considerably greater than that between m1 and the ultimate lower premolar. Compared with M2, M1 does not show a typical angulation; instead, the lingual side of the tooth has a broad curved outline in occlusal view. The paracone of M1 is strong and mesiodistally long; it is similar to other molars in having the cusp tip tilting distally in lateral view. The stylocone (cusp B) and the metastyle are slightly labial relative to the paracone; the three cusps define an open angle (see Table 1 for tooth angle). The stylocone and paracone are connected by a low ridge, the preparacrista. Mesiolabially, the stylocone merges with the base of a subequally sized parastyle at the mesiolabial tip of the tooth. By contrast, the metastyle is connected with the paracone by a strong ridge, the postparacrista, that descends from the distolingual side of the apex of the paracone and bears distinct wear. The lingual cingulum is weak and irregular, but more or less complete. The labial side of the paracone is shelf-like and has an uneven surface; it is not in a trigon shape. There is a cuspule at the distal end of the tooth, distolabial to the metastyle. Further discussion of the identification of this tooth as the first molar is provided in the Discussion.

M2 has a much more acute crown morphology than M1. The angle is 80°, 30° smaller than that of M1 (110°; Table 1). The angle is defined by nearly straight preparacrista and postparacrista, the ridges that form the mesiolingual and distolingual edges of the triangular tooth crown. The paracone has a convex lingual surface and is much higher than the labial side of the tooth. The tip of the paracone slightly recurves distolabially. The labial surface of the paracone is also convex but the convexity decreases apicobasally; the labial surface of the cusp base is flat or slightly concave, which gradually turns into the floor of the labial (trigon) basin of the tooth. The preparacrista starts as a weak ridge at the mesial side of the paracone; it extends mesiodorsally for a short distance and curves labially as a straight ridge that forms the nearly transverse anterior rim of the tooth crown. In mesiolingual or distolabial view, the preparacrista between the base of the paracone and the stylocone is broadly curved. There is a wear facet but no cusp on the preparacrista. The stylocone is a blunt ridge that merges with the preparacrista. The distolabial corner of the stylocone comprises a narrow crest that gradually ascends distally along the labial margin of the crown. There seems to be no parastyle on the mesial or mesiolingual side of the stylocone, which differs from the condition on M3 and M4. The postparacrista initiates at the distal side of the paracone and extends distolabially; it is much higher and thicker (mesiodistally) than the preparacrista. Similar to the preparacrista, there is no cuspule on the crest. The distolabial tip is broken, but the presence of the metastyle is discernible from the mesial base left by the metastyle. The distal side of the metastyle is embraced in a concave area mesial to the stylocone of M3. A ridge-like distal stylar cusp (partly broken) is present mesial to the metastyle; it extends along the labial border of the tooth and gradually reduces its thickness toward the deepest point of the extoflexus. In occlusal view, the ectoflexus between the stylocone and distal stylar cusp on M2 and other distal upper molars is shallow and gently concave. The mesial, lingual, and distal sides of the tooth bear a more or less continuous cingulum, collectively termed as the medial cingulum, but there is no cingulum on the labial side of the tooth. The wear facet on the postparacrista is clear and even. Labial to the paracone is the broad and fairly deep basin, which has been described as the ‘trigon basin’ in some studies ( Fox, 1985; Cifelli, 1999; Lopatin et al., 2010) but not in others ( Sweetman, 2008). A deep ‘trigon basin’ on the upper molariforms has been recognized as a derived feature in Spalacolestes and Symmetrodontoides ( Tsubamoto et al., 2004) .

The third and fourth upper molars (M3 – 4) are more similar to each other than to any other cheek tooth. The differences are that M3 is stronger and has a greater trigon angle and parastyle in occlusal view, whereas the paracone of M4 is higher in lingual view. In both of these molars, the paracone is mesiodistally more compressed than that of M2 and becomes distinctively acute-angled. The preparacrista is transverse and straight. The parastyle projects mesially to form a lobe. On the mesial side of the tooth at the junction of the preparacrista and parastyle a concave bay is formed to accommodate the metastyle of the preceding tooth. This relationship may be interpreted as an interlocking structure between the teeth. The distal stylar cusp is present, but again there is no cusp on either the preparacrista or postparacrista. The mesial and distal sur- faces are flat and nearly vertical (slightly tilting distally), implying that the prevallum and postvallum functioned as shearing surfaces. Again, the postparacrista is much higher and thicker than the preparacrista. The trigon basin is fairly deep. The continuous cingulum on the mesial, lingual, and distal sides is present but weaker than that of M2. The extoflexus is shallow.

Two left upper molars are preserved ( Fig. 8 ). By the tooth angle and shape, they are identified as M3 – 4. The mesial half of the right M3 was broken but M4 is complete. These two teeth were heavily treated with glue and so it was difficult to view the crown surface. In addition to the morphology described for the right M3 – 4, the preparacrista of the left M4 shows an swelling in the middle region, suggesting a possible small cusp C, which is not discernible in the right upper molars.

The penultimate upper molar (M5) is reduced in size relative to the earlier ones (length, height, and width). In occlusal view, the tooth outline is nearly symmetrical as the preparacrista and postparacrista become subequal in length, although the paracone leans more distally. An unusual condition is that the M5 crown is positioned dorsal to M4 and M6 as evidenced by the fact that the preparacrista and postparacrista of M5 are ventrally overlapped by the metastyle of M4 and parastyle of M6, respectively. This relationship, if representing a normal anatomical position, suggests that M5 probably erupted later than both M4 and M6, a tooth eruption sequence previously known in other ‘symmetrodontans’.

The ultimate upper molar (M6) is the smallest upper molar and is significantly smaller than M5. This tooth is more lingually positioned than other molars. Its distolabial corner is broken, but it is clear that the postparacrista is shorter than the preparacrista. On the parastyle at the mesiolabial corner of the tooth, a wear facet is present. The wear on the postparacrista is also distinct, indicating that the holotype of Lactodens represents an adult individual.

Lower dentition: All teeth except for i1 – 2 on the left jaw are broken, whereas the right i1 – 3, canine, p2, and p4 – m1 are intact ( Figs 3 , 5 , 7 ). For the right m2 – 5, however, only the labial base is variably preserved. Based on the preserved teeth and alveoli in both dentaries it can be unequivocally determined that the lower dental formula is i3-c1-p5-m6. A major difference between the lower and upper dentitions is that there are five lower premolars, in sharp contrast to the three upper premolars. In the upper dentition, diastemata between premolars are significant, whereas there is only a small gap between adjacent lower premolars; it is clear, however, that the lower premolars are not in contact with each other. Thus, none of the premolars bears tooth – tooth contact facets on its mesial or distal end.

The lower incisors are simple, single-rooted, and mesially inclined ( Fig. 4 ). Similar to the upper incisors, the first lower incisor is the smallest of the three, but differing from the uppers, i3 is the largest lower incisor. This size difference, however, is minor. The crown of each lower incisor is slightly wider than the root and convex labially and slightly concave lingually; thus the crown is slightly wider than long. The incisors are closely packed and there seems to be no space between neighbouring incisors.

Similar to the upper canine, the lower canine is also noncaniniform; it is shorter and simpler than the upper canine ( Fig. 4 ). It has two roots that are compacted at the junction with the crown but diverge cervically, as shown by the roots of the right canine revealed at the breakage of the dentary. The diameter of each canine root is subequal to, or smaller than, that of the incisor. In the left dentary, the canine crown is broken, revealing the two packed roots, with the mesial one being thicker and positioned more labially. The jawbone containing the canine and i3 is thicker than the part containing the i2 and postcanine teeth. The canine crown is low and has a main cusp with a height similar to that of the incisor, from which a ridge runs distally but does not terminate as a cusp. The crown is mesiodistally longer than transversely wide.

The first (p1) and third (p3) premolars are not preserved, but the impressions of the right p1 and p3 on the matrix provide an outline of the tooth crown ( Fig. 4 ). With the alveoli of the left p1 and p3, the sizes of the teeth can be estimated. The lower premolars increase in size distally. Similar to the upper premolars, the lower premolars are double-rooted and the roots are widely separated, particularly so for p3 – 5. The cross-sections of the roots are oval, with the long axis being mesiodistal, which differs from the molar roots, which have a circular or transversely wider than long cross-section. Because of the mesiodistally wide separation of roots and the low crown, the premolars are quite similar to deciduous teeth present in Mesozoic mammals (see Cifelli, 1999 and Discussion below). Lower premolars are of the ‘triconodont’ tooth pattern in having a main central cusp a as well as at least cusps b mesially and c distally, all aligned mesiodistally. Detailed comparison with a typical triconodont tooth shows that cusp a in Lactodens , and in other related symmetrodonts as well, is proportionally large and forms a broader triangle in lateral view; it is more mesially positioned so that the premolar is asymmetrical.

Judging from the crown impression and alveoli, p1 is slightly narrower and shorter than p2. The lingual side of the right p2 is fully exposed. Its distal root is thicker than the mesial one. The crown of p2 has three cusps, denoted as a, b, and c, of which the central cusp a is the largest, but the size difference of these cusps is proportionally not so distinctive compared with those of p4 and p5. Cusp b is rudimentary at the mesial base of cusp a. Cusp c is well developed and distantly separated from cusp a in lingual view. A faint cingulid runs for the full length of the lingual base of the p2 crown. The impression of the right p3 shows a weak buccal cingulid and a crown shape intermediate between p2 and p4. The shape of the alveoli indicates a transversely compressed tooth crown.

The p4 and p5 are similar in having widely separated roots that are transversely compressed; the gap between the roots is longer than the longitudinal diameter of the distal (larger) root. Revealed by breakage of the shallow dentary bone at p4, the distal root of p4 is short, ending immediately dorsal to the roof of the mandibular canal. Similar root lengths can be assumed for all postcanine lower teeth. The crown of p4 is well preserved except for a crack between cusps a and b. It is mesiodistally elongate, transversely thin, and distinctively low crowned. Cusp a is triangular in lingual view and has a convex mesial slope and a gently concave distal one, with the former being slightly shorter and steeper. Cusp b is well developed at the mesial end of the crown and slightly smaller than cusp c; both cusps are conical and slightly compressed transversely. Cusps b and c are slightly lingual to cusp a so that in occlusal view the outline of the tooth crown is slightly convex labially and concave lingually. Distal to cusp c is a small cusp, presumably cusp d, at the distal extremity of the tooth and emerging from the cingulid. On both the lingual and labial sides of p4, cingulids are complete and extend along the crown base.

The ultimate lower premolar (p5) is mesiodistally slightly longer but transversely notably wider than p4. In occlusal view, the outline of the tooth crown is spindle-shaped, although the lingual side is not as convex as the labial side. Cusp a is higher and more inflated, but cusp b is slightly smaller than that of p4. Mesial to cusp b is a small projection that may be considered as cusp e. Cusp c is also slightly smaller than that of p4, but cusp d is distinct and more distantly separated from cusp c than in p4. In occlusal view, cusp c is slightly lingual to cusp a. Both lingual and labial cingulids are present. At the level of cusp a, the base line of the tooth crown curves dorsally.

In the molar series only m1 is nearly completely preserved; the m2 – 5 crowns are largely broken, with the labial base of each tooth preserved in different conditions ( Figs 3 , 5 , 7 ). The ultimate lower molar (m6) is not preserved and is represented only by its alveoli. From the preserved m1 and basal portions of m2 – m5, some trends of the lower molars are recognizable. The length and trigonid angle of the lower molars reduce but the width increases distally (unknown for m6). The angle can be measured from the preserved bases of m2 – m5; the length and width of the crown can also be measured or estimated with reasonable confidence. The paracristid is orientated obliquely and the protocristid is generally transverse. The lingual cingulid is present on all lower molars (unknown in m6). The alveoli on the left dentary also show that the roots of the molars become more mesiodistally compressed distally and transversely wider than mesiodistally long. The tooth crown is more ventrally extended on the labial side than the lingual side, indicating that the labial side of the crown is higher.

The m1 is well preserved except for cusp d being broken. The critical transition of dental morphology between the ultimate premolar and molars is evident. The m1 differs from p 5 in having a much higher crown ( Figs 3 , 5 – 7 ; Table 1). It is doublerooted but with the mesial root mesiodistally longer than the distal one, opposite to the condition of p5. The protoconid (cusp a) of m1 is triangular in lingual and labial views and much higher than that of p5. Its lingual surface is flat whereas the buccal surface is convex. Cusp a (presumably the protoconid) of p5 is positioned on the mesial half of the tooth, with the mesial surface steeper than the distal one. By contrast, the protoconid (cusp a) is located on the distal half of m1, leaning distally and with its distal slope steeper than the mesial one. The paraconid (cusp b) and metaconid (cusp c) are much larger and more distinctive than those of p5. More importantly, these two cusps are distinctly lingual to the protoconid so that in occlusal view the crown outline is broadly triangular and considerably wider transversely than in p5. The paraconid is conical with a rounded tip and is lower than a third of the height of the protoconid; it is also separated by a notch from the mesial ridge descending from the protoconid. The metaconid is about twice the height of the paraconid, and similar to the protoconid, it leans slightly distally. In lingual view, the metaconid is a column-like cusp, but in labial view, it merges with the protoconid. A concave and shallow trigonid basin is present on the lingual sides of the protoconid and bounded mesially and distally by the paraconid and metaconid, respectively. The cingulid is complete on both lingual and labial sides, with the lingual one more distinctive and better defined. The distal cingular cusp (d), if present, is distolingual to the metaconid. The mesial cingular cusp (e) is small and mesiolingual to cusp b on the cingulid at the mesial tip of the crown. In lingual view, the base line of the crown is also convex dorsally, but differing from p4, the convexity is more pronounced and at the midpoint of the tooth.

The penultimate molar (m5) has more of the crown base preserved than the other broken lower molars. The preserved part shows that the crown was high and mesiodistally compressed. Although the actual crown height is unknown, it is inferred to be similar or slightly lower than that of m1, based on the preserved portion of the crown, the crown height of the corresponding M5, and the crown height gradient of the upper molars. In lingual view, it is clear that the mesial cingular cusp is distinct and that it continues to the mesial cingulid that extends ventrolabially and around the labial side of the crown. The lingual cingulid is also present on the preserved part of the tooth and the lingual base line of the crown is dorsally convex. In occlusal view, the cross-section of the broken paraconid is circular and separated from the metaconid (not preserved) by a notch on the lingual side of the crown. The paraconid is connected to the protoconid by a paracristid. The mesial (prevallid) surface is flat and nearly vertical. The crosssection of the protoconid is triangular. From the reduced alveoli of m6, particularly the circular distal one, it can be inferred that m6 is transversely narrower and mesiodistally shorter than m5 and that the distal end of m6 is narrower than the mesial half of the tooth. As m2 – 6 are broken, it is unclear whether the interlocking pattern is present [the hypoconulid (cusp d) placed labial to cusp e of the succeeding tooth ( Kielan-Jaworowska et al., 2004)]. However, in occlusal view a concavity at the distal ends of m1 and m2 suggests contact with cusp e of the succeeding tooth.

Dentary: The skull is smashed and therefore the cranial morphology is indeterminable, but the length can be estimated as 23 mm from the impression of the rostrum tip to the distal end of the skull left in the matrix ( Fig. 1 ). The two dentaries are largely preserved and morphologically informative ( Figs 1 , 3 , 10 ). The lengths of the skull and dentary are proportionally long compared with the lengths of the limb bones (see Table 1), indicating that Lactodens sheni had a long rostrum and relatively short limbs.

The deepest region of the horizontal ramus is at the position of m3, measuring 1.5 mm on the lingual side from the alveolus rim of m3 to the ventral bor- der of the dentary and 1.0 mm on the labial side. The transverse width of the horizontal ramus is 0.9 mm at the m3 alveoli. The dentary becomes shallower distally toward m6 and then deepens distally into the pterygoid region. Thus, the horizontal ramus bows ventrally below m4 but dorsally below m6. The horizontal ramus is proportionally very slim and long. It has a small symphysis, which apparently was not fused in life. The dentary thickens at the region housing i3 and the lower canine. The thinnest region is where the alveoli of p2 are contained, and it gradually thickens distally until the region containing the alveoli of m4, then it continues to thin again distally to the end of m6, where the coronoid process is developed and the dentary morphology becomes complex. The lingual surface from p5 to the mandibular condyle is preserved fairly well. It is smooth and in cross-section is gently convex lingually. Because of the breakage, the mandibular canal is revealed at the position of p 4 in the right dentary. The ventrodorsal height of the canal is close to one third of the dentary height at this position; it is thus quite sizable relative to the depth of the dentary. The canal probably ends somewhere anteriorly at p1. As mentioned above, the roots of the cheek teeth are short and terminate immediately dorsal to the roof of the canal.

The mandibular canal opens distally at a position distoventral to m6; its lateral wall continues distally as a groove that extends into the pterygoid fossa. The pterygoid crest starts as a weak ridge anterior to the mandibular foramen. This condition differs from that of Spalacolestes ( Cifelli & Madsen, 1999) in which the crest initiates below the alveolar margin at the junction of the horizontal and ascending rami. In this region and in the more anterior portion of the dentary, there is no trace of the Meckelian groove ( Fig. 10 ). Posterior to the mandibular foramen, the crest becomes much stronger. The pterygoid crest is broken on the right dentary but is largely preserved on the left. It is a distinctive crest that extends mediodorsally so that a broad groove or pocket is formed as the floor of the fossa. The mandibular foramen levels with the floor of this pocket. The pterygoid fossa has a shallow depression that extends anterior to the mandibular foramen and reaches a point level with the anterior edge of the ascending ramus; this is in contrast to the broad and more distinct concave area on the dentary in Spalacolestes ( Cifelli & Madsen, 1999) . The bone forming the ascending ramus and the pterygoid fossa is thin, but the structure in the region is quite complex. In lingual view, the pterygoid fossa is large and deep, with the deepest area around the mid region between the mandibular foramen and the condyle, dorsal to the pterygoid crest. This area forms an oval concavity and is bordered posteriorly by a strong and blunt ridge that we denoted as the medial ridge of the pterygoid fossa. Distal to the ridge, the fossa (the posterior portion of the pterygoid fossa) becomes shallower but may still be considered as part of the pterygoid fossa. This shallow concave area is ventrally and posteriorly delimited by a thickened edge of the dentary. The distal extremity extends slightly distal to the condyle. There is no angular process. The thickened edge continues dorsally and is separated from the stem or neck of the condyle by a narrow notch. The right condyle is completely preserved and positioned at a level considerably higher than the alveolar line of the teeth. The condyle is small and transversely wider than the thick stem that extends dorsally and slightly posteriorly. The convex articular surface is oval, with the long axis orientated transversely, and faces posterodorsolabially. The articular surface is small and, along with the short stem of the condyle, suggests a limited rotational angle of the lower jaw. Anterior to the condyle, the bone thickness gradually reduces anteriorly.

The ascending ramus is partly broken, but its general shape is discernible based on its well-preserved base, fragments, and impression left on the matrix. At the junction of the horizontal and ascending rami, there is a small projection, which we term as the coronoid tubercle; it possibly represents the vestigial coronoid that fused to the dentary. A weak line partly surrounding the process may suggest the suture, but this is not so convincing. The anterior edge of the ascending ramus inclines posteriorly and forms an angle of 116 ° with the horizontal ramus. Unlike Zhangheotherium in which the coronoid process is slim and bends posteriorly, the coronoid process of Lactodens sheni is quite broad (anteroposteriorly long). The left dentary shows that the massetetic fossa is broad and there is a low crest that defines the anterodorsal margin of the fossa. In dorsal view of the left dentary, the ascending process and probably the condyle are more labially positioned.

Cranium: The cranium is crushed and very little can be identified. At the posterior side of the cranial region, there is a bone plate that bears a curved edge that is definitely original. The shape and position indicate that this is part of the rim of the magnum foramen. This structure indicates that, although crushed, the posterior border and general shape of the cranium are largely preserved and it provides a reasonable reference point for the posterior end of the skull so that the skull length can be reasonably estimated and is somewhat consistent with the measured length of the dentary.

Clavicle: Most of the anterior skeleton is on the slab but in a poor condition of preservation. There is no clear shape discernible of the vertebrates, scapula, interclavicle, or manubrium in the crushed bone chips. A disarticulated clavicle is preserved in fairly good condition, although it is fractured near the distal (lateral) end and the distal end is missing ( Fig. 1 ). The bone is 5.5 mm long (based on the preserved part and the impression the broken distal end left in the matrix) and 0.4 mm thick at the midpoint of the shaft. The thickest area is around the middle portion of the bone, from where the shaft gradually thins towards both ends and then flares at the articulations. The anatomical orientation of the bone can be determined from the articular facet on the proximal (medial) end. The articular facet for the interclavicle has a smooth, teardrop-shaped surface, being broad medially. The facet is at a low angle with or nearly parallel to the long axis of the shaft, indicating its overlapping articulation with the interclavicle (presumably present). The articular facet and the anterior arching of the bone indicates that this is the right clavicle with its dorsal (roughly) side exposed. The impression left by the distal end suggests a facet at nearly a right angle to the shaft and thus an end-on articulation with the acromion of the scapula. The clavicle is long relative to the forelimbs (humerus, ulna, and radius), which indicates that the animal had a low body posture when it stood. Judging by the size of the articular facet, it is certain that the joint between the clavicle and interclavicle was mobile, perhaps in a fashion of pivotal motion as interpreted in some zhangheotheriids ( Hu et al., 1997, 1998; Ji et al., 2009), ‘symmetrodontans’ ( Chen & Luo, 2013), multituberculates ( Sereno & McKenna, 1995), and eutriconodonts ( Ji, Luo & Ji, 1999; Hu, 2006).

Humerus: The general morphology of the forelimbs is discernible and provides a reference to estimate the proportional skull and dentary lengths. Both humeri are preserved in crushed condition ( Fig. 1 ). The outlines of both humeri measure 7.5 mm long, suggesting that this is the actual humeral length. The midpoint of the diaphysis is 1.0 mm wide, but the width is more likely to have been affected by crushing than the length of the bone. Owing to the damage, structures on the humerus, such as the head, greater and lesser tubercles, and the diaphysis torsion are not discernible. The proximal portion of the humerus is wider than the rest of the bone, indicating the presence of the deltopectoral crest and the tubercle or crest for the teres major muscle. A ridge-like process on the right humerus is probably the deltoid tuberosity. The distal end expands slightly transversely. The right humerus shows a projection, probably the medial epicondyle.

Ulna: The right ulna is preserved in a crushed condition; some bone chips of the shaft are gone, but the preserved part and the impression provide some general morphological information for the bone. The ulna is the longest forelimb bone, with a length of 9.2 mm and a width of 0.5 mm at the midpoint of the diaphysis. The olecranon process at the proximal end of the ulna is blunt and bends medially. The outline of the semilunar notch is a gentle curve that distally ends at the level of the proximal end of the radius. The longitudinal length of the notch is subequal to the length of the olecranon process. On the distal end of the ulna and radius, there is no indication of a styloid process.

Radius: The radius is preserved in a better condition than the ulna. It is more robust than the shaft of the ulna (distal to the semilunar notch). The radius head expands at the distal edge of the semilunar notch. The diaphysis is slightly curved and extends slightly more distally than the ulna; its distal part flares and has a flat end. The diaphysis bears a low crest that extends with a curvature for a distance about a fifth of the radial length; this crest is probably for insertion of the pronator teres.

Carpals: Because of the preservation, only the concave ventral sides of the metacarpals and phalanges are visible ( Fig. 11 ). The relationships of the preserved carpals with the metacarpals are largely retained but the bones are not in their anatomical positions. Because of the disarticulation and dislocation of the phalangeal elements, the accurate morphology of the manus cannot be determined. There are six elements observed in the wrist region, five of which are identified based on their relative positions. One small element at the mesial side of the preserved wrist cannot be identified, and is marked with a question mark. From the medial to lateral side, the four distal carpals are the trapezium, trapezoid, capitate, and hamate ( Fig. 11 ), which are in contact with various proximal phalanges, respectively. Proximal and dorsal to the hamate is the triquetrum that is partly exposed in ventral view. Between the distal end of the radius and the trapezoid, there is a concave area, which is probably the space for the scaphoid. Judging from the size of this concave area, the scaphoid, if present, would be sizable, probably larger than any of the preserved elements. The lunate could be at the position between the radius and capitate. Based on the preserved carpals and proximal phalanges and with the assumption they largely remain in their original positions, we interpret that the trapezium articulated distally with carpals I (not preserved, see below) and II. The trapezoid has a proximal projection that would have been dorsal to the scaphoid in life. Proximally, the trapezoid articulates with the lateral portion of the proximal end of metacarpal (Mc) II and the medial portion of the proximal end of Mc III. The capitate is more transverse and primarily in articulation with Mc IV and with a smaller articulation with Mc III. The hamate is sizable and appears to articulate primarily with Mc V and may have a small contact with Mc IV. There is no trace of a pisiform.

Metacarpals and phalanges: We only identified four metacarpals and three distal phalanges from the holotype specimen ( Fig. 11 ). It is unlikely that Lactodens sheni had lost one digit. It is most probable that either digit I or V was not preserved. In Akidolestes , digit I is reduced in the left manus ( Chen & Luo, 2013: fig. 8; note that digits I and II were labelled as IV and V in fig. 7C) but the same digit in the right manus appears to be normal (ibid: fig. 9). Comparing to the manus of Akidolestes in which Mc II appears to be more robust, we interpret the four preserved metacarpals as representing Mc II – V. This is also supported by the fact that digit III so interpreted is the longest in the holotype of L. sheni , similar to the common condition in early mammals. Of the metacarpals, Mc II is the thickest one. In ventral view, the middle point of the shaft is thinnest; it flares toward the two ends of the element. Compared with Mc III, Mc II is not so straight in ventral view. The bone bows dorsally. The proximal end is flat, whereas the distal end is convex, with the projection being slightly greater laterally than medially. On the lateral side of the distal end there is a contact concavity for the medial tubercle on the distal end of Mc III. Mc III is longer and slimmer than Mc II. It also bows notably dorsally. Its proximal end is rounded and proximally convex. The curvature of the shaft is greatest at the point about a third from the proximal end. Its distal end is transversely expanded, much wider than the proximal end, and is the widest of the four metacarpals preserved. The distal surface is gently convex and roughly even transversely. Two small sesamoid ossicles are present on the ventral side of the distal of Mc III; they were displaced and in life probably positioned at the junction between the metacarpal and proximal phalanx. Mc IV is broken in its proximal portion. The distal end is wide and similar to that of Mc III except that the articular surface is not transversely even, but has its medial side slightly more distal than the lateral side. Mc V is significantly shorter than Mc IV. In ventral view, the distal end of all metacarpals bears three weak tubercles that confine two very shallow articular grooves for the proximal phalanx.

The proximal phalanx of digit II is missing, but proximal phalanges III – V are preserved. The proximal end of phalange V is complete, showing two convex projections that confine a central groove on the proximal end. The two projections articulate with the two shallow grooves on the distal end of the metacarpal. The ventral surface of the shaft bears two flexor sheath ridges or crests that are slightly flared (projecting ventrally and slightly outwards). The proximal ends of the flexor sheath ridges reach to the midpoint of the bone. The flexor sheath ridges are relatively proximally positioned. The distal end of each is transversely not so expanded and bears three weak tubercles that confine two faint articular grooves for the intermediate phalanx.

The intermediate phalanx of digit III is displaced, but those of digits II and IV are largely in anatomical place; they are very similar in shape. The ventral shaft surface of the intermediate phalange is gently concave. The element is transversely wider than dorsoventrally deep. The proximal end is considerably wider than the shaft; the lateral and medial corners project more ventrally to become tubercle-like and a concave area is formed between them. The proximal articular surface faces proximodorsally. The distal articular surface is transversely not so expanded, but is characterized by a single and deep groove or trochlea. By contrast, the lateral and medial crests or ridges confining the trochlea are distinctively high. These ridges are not symmetrical. The outer side is more vertical, whereas the side toward the groove is convex. For digit II the lateral ridge on the distal end of the intermediate phalanx is slightly higher than the medial one, and in digit IV, the medial one is slightly higher. In lateral view, this trochlea-like articulation is in a semicircular shape, indicating perhaps a 180 ° freedom of rotation of the ungual phalanx. The intermediate phalanx of digit V is reduced compared with its proximal phalanx and with other intermediate phalanges.

The three ungual phalanges (II – IV) are similar in size and shape. The ungual phalanx is quite long com- pared with the length of the phalanges, and is transversely compressed. The proximal articular surface is divided into two lateral and medial facets by a central ridge that extends ventrodorsally. The articular facets fit into the articular groove (trochlea) on the distal end of the intermediate phalanx. The ungual phalanx has a posterodorsal process that extends proximally and tapers proximally in dorsoventral and transverse dimensions. The dorsal surface of the ungual phalanx is angular in cross-section and it runs like a ridge longitudinally; its ventral side bears a digital flexor tubercle and distal to it is a flat or concave ventral surface. The tip of the ungual phalanx is sharp and not cleaved. On the lateral and medial sides of the distal ungual phalanx there is a narrow but distinct groove. The groove extends in parallel to the curved claw. Ventral to the groove the lateral ridge projects outward. The ventral surface of the ungual phalanx is concave dorsally in lateral view but convex ventrally in cross-section.

COMPARISON

Non-‘symmetrodontan’ mammals

Lactodens differs from nonmammaliaform cynodonts, basal (Late Triassic and Early Jurassic) mammaliaforms, Sinoconodon , morganucodontids, australosphenidans (including monotremes), eutriconodontans, allotherians (‘haramiyidans’, multituberculates, and perhaps gongwanatherians), kuehneotheriids, and cladotherians (including crown Theria) in having the following combination of characters: a single dentary-boned lower jaw that has the condyle fully developed as the jaw articulation (lacking the postdentary trough and bones); angular process of the mandible absent; triangulated molar cusp pattern with angle decreasing distally in upper and lower molars; and lack of a talonid on the lower molars ( Kielan-Jaworowska et al., 2004). These characters also characterize Lactodens as a ‘symmetrodontan’.

Nonspalacotheriid ‘symmetrodontans’

Amongst ‘symmetrodontans’, the acute angled upper and lower molars, particularly the distal ones, distinguish Lactodens from those with ‘obtuse-angled’ molar teeth, such as Tinodon and Gobiotheriodon ( Marsh, 1887; Trofimov, 1980; Averianov, 2002; Kielan-Jaworowska et al., 2004; Lopatin et al., 2005), the more primitive taxa that have been assigned to Kuehneotheriidae and Tinodontidae ( Kielan-Jaworowska et al., 2004) . The acute angled upper and lower molars of Lactodens also differentiate it from zhangheotheriids ( Zhangheotherium and Maotherium ) and Kiyatherium , which have molars with a less acute angle. As noted by Cuenca-Bescos et al. (2014), the trigonid angle cannot be determined for Maotherium ( Rougier et al., 2003a; Ji et al., 2009) and Zhangheotherium ( Hu et al., 1997, 1998), but the trigon angle of Maotherium can be measured, which largely corresponds to the trigonid angle. Moreover, the labial surfaces of the lower molars in Zhangheotherium and Maotherium are visible and are gently convex, contrasting with the acute and thus mesiodistally compressed labial surface of the lower molars in Lactodens .

Although zhangheotherids may be considered as being more primitive than spalacotheriids in having less acute-angled and more cuspute molars, specialization of dental morphology may have evolved independently in the former. One of the features distinguishing zhangheotheriids and Kiyatherium ( Lopatin et al., 2010) from Lactodens is the lower number of postcanines regardless of different interpretations of the ultimate premolar vs. mesial molar in zhangheotheriids ( Hu et al., 1997, 2005a; Rougier et al., 2003a). As Cifelli (1999: 267) pointed out, Zhangheotherium ( Spalacotherium as well) ‘appear to be derived with respect to Spalacolestinae (assuming four premolar loci, with replacement at each, in the North American forms) in the loss of one or more premolars’. New evidence shows that the primitive condition of the premolar number is probably five instead of four, as in the lower dentition of Lactodens . By contrast, an extreme example of zhangheotheriids exists in Mao. asiaticus ( Ji et al., 2009) , in which there is only one premolar in both the upper and lower jaws so that a long diastema is created between the premolar and canine, a unique condition amongst known ‘symmetrodontans’.

The upper molars of zhangheotheriids have a distinct cusp B ̕ on the preparacrista and C on the postparacrista. In addition, the tooth size varies differently in the dentition among these taxa. In Mao. asiaticus ( asiaticum ?) ( Ji et al., 2009), for instance, the central molar (M3) is the largest molar, away from which the upper molar size decreases. The lower molars decrease in size (primarily the length) distally from m1 to m6. A similar pattern exists in Mao. sinensis (sinense?), although the latter has only four upper molars. In Zhangheotherium , m3 is the largest and the rest of the teeth decrease in size mesially and distally; the size variation of the upper molars is unclear. In Lactodens , both upper and lower molars decrease in size (length) distally.

The dentary of zhangheotheriids and Kiyatherium has a long and narrow coronoid process that inclines posteriorly at an angle of approximately 40 °; the long condylar process extends dorsally and is separated from the coronoid process by a deep notch that extends ventrally to the level of the alveolar margin or below, and a transversely compressed condyle. Zhangheotheriids and those with ‘obtuse-angled’ molar teeth further differ from Lactodens in having the Meckelian groove (for those with the dentary preserved) on the medial surface of the dentary. Zhangheotherium ( Meng et al., 2003) and Maotherium ( Ji et al., 2009) also have the ossified Meckel’s cartilage preserved.

Akidolestes

Akidolestes cifellii Li & Luo, 2006 , is an Asian spalacotheriid from the Lower Cretaceous Yixian Formation at the Dawangzhangzi Locality, Lingyuan, Liaoning. The rock unit underlies, and thus is slightly older than, the Jiufotang Formation where the holotype of Lactodens came from. It is probably the species that is most similar to Lactodens . The original description of Ak. cifellii was brief ( Li & Luo, 2006), but follow-up works provided more detailed dental and postcranial morphologies of the species ( Chen & Luo, 2008, 2013). Lactodens is similar to Ak. cifellii in having a long rostrum, cusp triangulation less than 50° on the posterior molars, and lacking cusp B-1 between the stylocone and the paracone on the upper molar. However, Lactodens differs from Ak. cifellii in several aspects. The tooth formula of Ak. cifellii is I4.C1.P5(?).M5(?)/i4.c1.p5.m6. With the uncertain number of upper postcanine teeth, Ak. cifellii at least differs from Lactodens in having four upper and lower incisors. Chen & Luo (2008) described P3 and P5 of Ak. cifellii as being similar to the corresponding lower premolars. Thus, Ak. cifellii has two more upper premolars than Lactodens , and the upper molars are one fewer than in Lactodens . In Lactodens , because there are three upper and five lower premolars, there is no one-to-one corresponding relationship between the upper and lower premolars.

In Ak. cifellii the lower canine has a single and column-like root, similar to that of the incisor in shape and size ( Li & Luo, 2006; Chen & Luo, 2008), contrasting with the closely packed double-rooted lower canine in Lactodens . Akidolestes cifellii is similar to Lactodens in having relatively long premolars, but their heights are difficult to evaluate because their orientations appear different from those of the molars ( Li & Luo, 2006: fig. 2b; Chen & Luo, 2008: fig. 1C). The m1 of Ak. cifellii is significantly mesiodistally compressed and has a trigonid angle (79°) much smaller than that of Lactodens (110°). In Lactodens , p5 and m1 are subequal in length, whereas the m1 of Ak. cifellii is significantly shorter than p5. One may argue that the m1 of Lactodens is actually the ultimate premolar, which we think is unlikely (see below), but if this is true, then Ak. cifellii would differ from Lactodens in the number of lower premolars and molars. Cusps c and b of m 1 in Ak. cifellii are about a third of the height of cusp a ( Chen & Luo, 2008); in Lactodens cusp b of m1 is low but the height of cusp c is more than half of cusp a, higher than cusp c on m1 of Ak. cifellii . The other lower molars (m2 – 6) of Lactodens are broken, but the trigonid angle in the mesial molars is generally greater than that of Ak. cifellii ( Chen & Luo, 2008) .

An important difference between the lower molars of Ak. cifellii and Lactodens , and probably other spalacotheriids as well, is the relative lengths of the paracristid and protocristid. Akidolestes retains distinctive cusps on the ultimate lower molar that has a symmetrical crown, and as illustrated ( Li & Luo, 2006: fig. 2b; Chen & Luo, 2008: fig. 1C), the protocristid is longer than the paracristid on the mesial lower molars (m1 – 4) so that the metaconid is more lingually extended than the paraconid. By contrast, the opposite is true in Lactodens and some other spalacotheriids ( Fox, 1976; Cifelli & Gordon, 1999; Cifelli & Madsen, 1999;; Tsubamoto et al., 2004; Hu et al., 2005a; Sweetman, 2008). The longer paracristid matches the longer postparacrista of the upper molars in Lactodens so that the postvallum-prevallid shearing is emphasized, in which the mesial end of the lower molar shears against the distal end of the upper one. However, this condition that a long paracristid of a lower molar matches a long postparacrista of the upper molar does not always seem to be the case, as shown in Ak. cifellii . In describing Spalacolestes cretulablatta, Cifelli & Madsen (1999: 182) wrote: ‘On posterior molars, there is a tendency for the paracristid to be slightly longer than the protocristid, whereas the reverse is true for Symmetrodontoides ’. Moreover, in describing Spalacotheridium noblei, Cifelli & Madsen (1999) also noted that the protocristid is slightly longer than the paracristid on m4. The upper molar morphology of Ak. cifellii is little known, but the lower molar crown with the protocristid being longer than the paracristid of Ak. cifellii perhaps suggests an emphasis on prevallum-postvallid shearing, which differs from that of Lactodens .

Li & Luo (2006) considered that within spalacotheriids, the ultimate lower molars of Ak. cifellii are symmetrical, with three distinctive cusps, which was considered to be a primitive retention that differs from Symmetrolestes , Heishanlestes , Spalacotheridium , Spalacolestes , and Symmetrodontoides , in which the ultimate lower molar is asymmetrical. The ultimate lower molar is not preserved in Lactodens , but the preserved alveolus shows that the distal root of m6 is reduced in size and has a circular cross-section, in contrast to the transverse mesial root. This root condition is similar to that of Spalacolestes cretulablatta ( Cifelli & Madsen, 1999: fig. 9, OMNH 27421), which has an asymmetrical ultimate lower molar. It is reasonable to assume that m6 of Lactodens is also asymmetrical, with its distal half being reduced. Akidolestes cifellii further differs from Spalacotherium , Heishanlestes , Spalacotheridium , Spalacolestes , and Symmetrodontoides in that the labial cingulid is not complete in the penultimate lower molars. In Lactodens , m5 has a complete labial cingulid. In general, the labial cingulids are better developed in Lactodens than in Ak. cifellii , judging from the figure in Li & Luo (2006: fig. 2). The upper molar morphology is largely unknown in Ak. cifellii ( Li & Luo, 2006; Chen & Luo, 2008), hampering further comparison.

The mandible of Ak. cifellii was reported to be nearly identical to those of Zhangheotherium and Maotherium , and thus Kiyatherium , in having an elongate and gracile coronoid process and a mediolaterally compressed dentary condyle ( Li & Luo, 2006; Lopatin et al., 2010). In Lactodens , the mandibular condyle has a short neck and is transversely (mediolaterally) wider than anteroposteriorly long and the coronoid process is more similar to those of Spalacotherium , Spalacolestes , and Heishanlestes in being broader (anteroposteriorly long). The dentary of Lactodens also differs from that of Zhangheotherium and Kiyatherium in lacking the Meckelian groove on the medial surface of the dentary, but has a deep pterygoid fossa that is ventrally defined by a medially projected pterygoid crest, similar to the condition seen in Spalacolestes ( Cifelli & Madsen, 1999) . The medial side of the mandible of Ak. cifellii is not exposed; thus it is unknown whether or not the Meckelian groove is present.

Measurements of the teeth, dentary, and skull are unknown in Akidolestes ( Li & Luo, 2006; Chen & Luo, 2008), but those of the skeletal elements are available ( Chen & Luo, 2013). The lengths of the humerus, ulna, and radius are 8.8 – 9.1, 11.0 (listed as 110 in Chen & Luo, 2013: table 1), and 8.8 – 8.9 mm, respectively. The lengths of the corresponding elements in Lactodens are 7.5, 9.2, and 7.0 mm ( Table 1). In terms of the length ratios of these bones, Lactodens is about 15 – 20% smaller than Akidolestes . To sum up, it is justified that Lactodens is distinct from Akidolestes .

Spalacolestes

Spalacolestes is the type genus of the subfamily to which Lactodens is assigned. Species of Spalacolestes almost certainly possessed deciduous premolars in life (Cifelli, 1999) that are highly similar to those of Lactodens (see Discussion). Spalacolestes is similar to Lactodens in having acutely angled trigonids on the posterior molars and in having the m1 paraconid much lower than the metaconid. The upper molars of Spalacolestes are similar to Lactodens in reduction of the stylocone, in lacking cusps B1 and ‘C’, and in having the preparacrista lower than the postparacrista and deep trigon basins. Both genera are similar in having a distinct pterygoid crest and a deep anterior portion of the pterygoid fossa.

Spalacolestes differs from Lactodens in being smaller and having seven molars. However, Cifelli & Madsen (1999) noted that a complete dentition in North American spalacotheriids remains unknown; thus the dental formula cannot be established unambiguously and should be regarded as tentative. The upper molars of Spalacolestes have a more bulbousbased paracone with a gently curving (not tightly arced or folded) lingual face. By contrast, the paracone of the distal molars (M3 – M6) of Lactodens is more angular and mesiodistally compressed, judging from the preserved labial parts of the latter. The parastyle of M1 – 2 and the distal stylar cusp on the upper molars of Spalacolestes are more developed than those of Lactodens . Although both genera have the paraconid lower than the metaconid on m1, the condition is more pronounced in Lactodens . In lingual view the paraconid of m 1 in Lactodens is more mesially extended relative to the protoconid, and the latter is more distally inclined so that m1 is mesiodistally longer and asymmetrical. In addition, from what has been preserved in the holotype of Lactodens , the cingular cusps on the lower molars are less developed than in Spalacolestes . The pterygoid crest in Spalacolestes is more pronounced and starts at the junction of the horizontal and ascending rami.

Some isolated teeth of an unusual design were identified as deciduous premolars of spalacotheriids, including Spalacolestes (Cifelli, 1999) . Although it is still challenging to identify those teeth, the dentition of Lactodens echoes the conclusion made by Cifelli (1999). The tooth tentatively identified as a deciduous lower canine is single-rooted, but bears a faint labial groove on the root (Cifelli, 1999: fig. 1, OMNH 60623), which differs from the double-rooted upper and lower canines of Lactodens , although the root division of the lower canine is not so definite at the crown base. In addition, the two closely packed roots are positioned diagonally with the mesial root more labial relative to the distal one. Although singlerooted, the lower canine (Cifelli, 1999: fig. 1C, OMNH 60623) is proportionally longer than that of Lactodens . However, the lower canine of Lactodens is similar to the canine OMNH 60623 in being noncaniniform and having a mesiodistally aligned weak crest descending gently from the apex of the primary cusp, but in Lactodens there is no distobasal cuspule.

Heishanlestes

Heishanlestes is a spalacolestine discovered from Badaohao, Heishan County, Liaoning Province, and is represented by specimens of the dentary and lower teeth. Heishanlestes is subequal to or slightly larger than Lactodens , and similar to the latter in the structure of the pterygoid fossa. As Hu et al. (2005a) described, the pterygoid fossa extends deeply into the dentary and is delimited both ventrally and medially by an elevated pterygoid crest that encloses the fossa as an elongate pocket that opens dorsomedially. This morphology is very similar, if not identical, to that of Lactodens . However, the geologically younger taxon, Heishanlestes , is highly specialized in several aspects. It has four lower premolars that are closely spaced and have transversely inflated crowns that incline mesially with the mesially projected crown overlapping the distal crown of the proceeding tooth; lower molars have inflated cusps; m1 is proportionally short and m5 – 6 are more reduced in size; the dentary is more robust and the anterior edge of the coronoid process is vertical to the horizontal ramus.

In more detailed morphology, the lower tooth crown defined by the paracristid and protocristid is nearly U-shaped in Heishanlestes , in contrast to the V-shaped molars in Lactodens . More distinctively, the m1, m2 – 4, and m5 – 6 of Heishanlestes form three morphologically discontinuous subsets of the molar series. Judging from wear, the m 1 in Heishanlestes appears to have been transitional in function between the premolar and molar series; it may have joined the premolars in crushing prey as well as having a shearing function ( Hu et al., 2005a). The m2 – 4 of Heishanlestes are generally similar to the lower molars of other spalacotheriids, including Lactodens , in having continuous prevallid and postvallid shearing surfaces, but its m5 – 6 are unique in possessing a large central cusp. Although the distal molars of Lactodens are broken, judging from the corresponding upper molars, the lower molars would have had the normal spalacotheriid lower molar pattern, without the central cusp. In any case, the lower canine, premolars, and m1 of Lactodens are distinctively different from those of Heishanlestes . The anterior teeth in Lactodens are more suited for a piercing or ‘grasping’ function, rather than for crushing.

Aliaga

Aliaga from the upper Castellar Formation (lower Barremian), Spain, is represented by fragmentary material ( Cuenca-Bescos et al., 2014). Teeth assigned to the genus are smaller than those assigned to cf. Aliaga henkeli . The genus differs from other spalacotheriids in having distinct accessory cusps on the cingulids of the lower cheek teeth. Aliaga molinensis is further distinct from other spalacotheriids in having the lower molar cingulid incomplete lingually. The m1 cingulid of Lactodens is complete lingually and the lingual base of m5 also has a clear and complete cingulid. There is no accessory cusp on the cin- gulid of the lower molars in Lactodens . The trigonid angle of Aliaga is smaller than that in Lactodens and most species of spalacotheriids ( Cuenca-Bescos et al., 2014: table 2, fig. 4). The premolars assigned to cf. Al. henkeli Krebs, 1985 ( Cuenca-Bescos et al., 2014) have larger and relatively high crowns, distinct from the premolars of Lactodens , which resemble deciduous teeth.

Spalacotheridium

Spalacotheridium Cifelli, 1990 , is from the early Late Cretaceous Straight Cliffs Formation, Garfield County, and the upper Cedar Mountain Formation (upper Abian or lower Cenomanian) ( Cifelli & Madsen, 1999), Emery County, Utah. The type species of the genus, Spalacotheridium mckennai is represented by a lower molar, originally thought to be possibly a m4 ( Cifelli, 1990), but in light of more specimens of the second species, Spalacotheridium noblei Cifelli & Madsen, 1999 , the holotype was identified as m2 ( Cifelli & Madsen, 1999). One of the diagnostic features for the genus is its small size, which distinguishes Spalacotheridium from Lactodens . Additional specimens of Spalacotheridium noblei allow comparison of other teeth between Spalacotheridium and Lactodens . The m1 of Spalacotheridium noblei is much shorter and has a more acute trigonid angle than that of Lactodens . Similarly, M1 of Spalacotheridium noblei is also more acute and has a distinct parastyle, although in light of the dentition of Lactodens , the tooth identified as M1 (OMNH 26429; Cifelli & Madsen, 1999: fig. 16A, B) could be a M2 or M3. Even so, the parastyle in relation to the rest of the tooth is still more distinct in Spalacotheridium noblei . In M3, the paracristid is much lower than the protocristid. Perhaps partly because of wear, the upper molar cusp seems lower than that of Lactodens .

Spalacotherium

Spalacotherium contains five species from the Early Cretaceous strata of several localities in Europe [ Spalacotherium tricuspidens Owen, 1854 (= Peralestes ), Spalacotherium evansae Ensom & Sigogneau-Russell, 2000 , ‘ Spalacotherium ’ henkeli Krebs, 1985 , Spalacotherium taylori Clemens & Lees, 1971 , and Spalacotherium hookeri Gill, 2004 ], amongst which Spalacotherium taylori is represented by one tooth and Spalacotherium tricuspidens by dentaries and most of the teeth ( Simpson, 1928; Kielan-Jaworowska et al., 2004). ‘ Spalacotherium ’ henkeli Krebs, 1985 , was assigned to cf. Al. henkeli ( Cuenca-Bescos et al., 2014) . Spalacotherium represents a primitive member of the spalacotheriids ( Cifelli & Madsen, 1999; Tsubamoto et al., 2004; Sweetman, 2008) in having the Meckelian groove and a small pterygoid fossa. Where the dentition is known, it has seven molars ( Kielan-Jaworowska et al., 2004). Compared with Lactodens , the molars have a less acute angle; upper molar cusps are more robust and less conical in shape and have smaller cusp B1 but lack the parastyle ( Kielan-Jaworowska et al., 2004). These features in combination distinguish Spalacotherium from Lactodens . Similar to Spalacolestes (Cifelli, 1999; Cifelli & Gordon, 1999), some isolated teeth referred to Spalacotherium evansae ( Ensom & Sigogneau-Russell, 2000) and Spalacotherium hookeri ( Gill, 2004) were also identified as deciduous premolars. In light of the dentition of Lactodens , the lower tooth (DORCM GS 373) is probably a lower canine, whereas the one identified as an upper tooth (DORCM GS 312) (Ensom & Sigogneau- Russell, 2000: fig. 6) is possibly a distal lower premolar because it is more similar to p4 or p5 instead of upper premolars of Lactodens .

Spalacotheroides

Spalacotheroides Patterson, 1955 (see also Patterson, 1956: fig. 1) is from the Early Cretaceous Antlers Formation (Aptian – Albian), Texas, USA. This monotypic genus (type species Spalacotheroides bridwelli ) differs from Lactodens in being smaller, having an incomplete labial cingulid on the lower molars, and upper molars that possess a strong parastyle, cusps B1 and C, and a relatively strong stylocone. The only lower molar preserved on the dentary fragment (holotype) was interpreted as possibly being one of m2 – m5 ( Patterson, 1955; Fox, 1976; Cifelli & Madsen, 1986, 1999), and this lower molar differs from those of Lactodens in being relatively low crowned.

Symmetrodontoides

Symmetrodontoides Fox, 1976 , is a North American Late Cretaceous genus including three species [ Symmetrodontoides canadensis Fox, 1976 (type species); Symmetrodontoides foxi Cifelli & Madsen, 1986 ; Symmetrodontoides oligodontos Cifelli, 1990 : see also Fox, 1985; Cifelli & Gordon, 1999]. The bestpreserved specimen is the holotype of Symmetrodontoides canadensis , a right dentary fragment with three molars identified as m3 – 5 ( Fox, 1976). Under that identification, the corresponding teeth of Lactodens , either in labial or occlusal view, are smaller and have a more acute trigonid angle and slimmer cusps and cristids than Symmetrodontoides canadensis . The labial side of the lower molars in Symmetrodontoides canadensis is more rounded than that of Lactodens . Upper molars of Symmetrodontoides display a characteristic inflation of one of the stylar cusps ( Fox, 1985; Cifelli & Gordon, 1999). The enlarged stylar cusp arises from the posterior lobe of the stylar shelf and occupies the lobe nearly com- pletely, but because it is located posterior to the ectoflexus and does not join with the preparacrista, it cannot be a stylocone ( Fox, 1985). The new evidence from Lactodens supports Cifelli’s (1990) interpretation that the tooth (UALVP 12086) tentatively identified as ml of Symmetrodontoides canadensis ( Fox, 1976) probably occupied the ultimate premolar locus (see Discussion below).

Symmetrolestes

Symmetrolestes parvus Tsubamoto & Rougier, 2004 (in Tsubamoto et al., 2004) is from the Early Cretaceous (late Hauterivian to Aptian) Kitadani Formation, Kitadani-cho, Katsuyama City, Fukui Prefecture, central Japan. The species is represented by a fragmentary right lower jaw with the incisor and five postcanine teeth. Symmetrolestes is smaller than Lactodens and has only four molars. The identification of the incisors of Symmetrolestes is challenging because of the poor condition of preservation, but we concur with the identification of p 5 in the lower dentition of Symmetrolestes by Tsubamoto & Rougier (in Tsubamoto et al., 2004). Tsubamoto & Rougier presented a good argument in identifying the ultimate premolar, even though it has a triangular outline in occlusal view, but the tooth does not have a distinct paraconid and metaconid lingual to the protoconid and is much less molariform compared with m1 of Lactodens . The shape of the horizontal ramus in Symmetrolestes has a more arched ventral border; the deepest region is at m3 and the depth of the dentary gradually becomes shallower anteriorly.

Yaverlestes

Yaverlestes gassoni Sweetman, 2008 , is based on specimens collected from several sites from the Early Cretaceous (Barremian) Wessex Formation, the Isle of Wight, UK. It differs from Lactodens in being smaller, having three lower premolars and five molars, and lacking a cingulum on upper molars. The upper molars of Yaverlestes have a distinct cusp B1 (B ̕), cusp C, and well-developed stylocone. The m1 has a more acute trigonid angle than that of Lactodens , and the same is probably true for the other lower molars. These features readily distinguish Yaverlestes from Lactodens . In addition, the total length of the lower premolar series and of each individual premolar are shorter that those of Lactodens .

Infernolestes

Infernolestes rougieri Cifelli, Davis & Sames, 2014 , from the Lakota Formation (Lower Cretaceous, upper Berriasian – Valanginian) of South Dakota, is based on a well-preserved right lower molar, identified as m1. This tooth is generally similar to m1 of Lactodens . Minor differences exist, however; for instance, the m1 of Infernolestes is shorter and proportionally higher; the paraconid is proportionally higher, positioned closer to the protoconid and slightly more distally inclined; and the lingual cingulid is flat. In labial view, the mesial cingular cusp and the paraconid are more distantly separated, the distal cingular cusp is more distinct, and the labial cingulid is incomplete. Cifelli et al. (2014) considered the m1 alone not to be informative enough for subfamilial relationships so that they conservatively referred Infernolestes to the Spalacotheriidae only. In light of the m1 morphology of Lactodens , Infernolestes is probably referable to Spalacolestinae , perhaps representing the earliest and most basal species amongst known spalacolestines.

Shalbaatar

Shalbaatar Nessov, 1997 , is based on a right edentulous dentary fragment with alveoli of m4 – 7? from the Turonian Bissekty and?Santonian Aitym Formations of Uzbekistan. The genus was originally referred to Multituberculata (?Plagiaulacoidea) ( Nessov, 1997). In reporting another specimen, a fragmentary left edentulous dentary with alveoli of m5 – 7?, assigned to cf. Shalbaatar sp. , Averianov (2002) relocated Shalbaatar into Spalacotheriidae . However, Averianov (2002: 713) predicted that ‘If possible attribution to Spalacolestinae is corroborated by further specimens, Shalbaatar would be the only non- American member of this subfamily’. In a later study ( Averianov & Archibald, 2003, 2013), Shalbaatar was still placed within Spalacotheriidae , not in Spalacolestinae . The holotype of Lactodens lends support to the identification of Shalbaatar as a spalacotheriid instead of a multituberculate, even though the specimens are fragmentary and edentulous. Judging from the fragmentary dentary, Shalbaatar is notably larger than Lactodens .

Figures

Figure 1. The holotype specimen of Lactodens sheni gen. et sp. nov. (HG-M016). A, the partial skeleton with skull and dentitions preserved in a slab of siltstone (HG-M016, housed in the Bohai University Paleontology Center, Jinzhou City, Liaoning Province); B, interpretive line drawing to illustrate the main skeletal structures preserved; C, area exposed by preparation to reveal labial sides of the right upper and lower dentitions from the backside of the slab; the approximate area of the excavation is shown by the rectangular box in B (image in C is slightly enlarged); D, the preserved skull. Abbreviations: c, lower canine; C, upper canine; cl, clavicle; cp, coronoid process; lhr, left humerus; lma, left manus (mostly impression); lra-u, left radius and ulna (mostly impression left by both bones); md, mandibular condyle; mf, mandibular foramen; mgf, magnum foramen; pf, pterygoid fossa (pocket part in dorsal view); rca, right carpals; rhr, right humerus; rmc, right metacarpals; rra, right radius; rul, right ulna; rup, right ungual phalanges; vt, vertebra (probably thoracic). Scale bar in A = 1 mm.

Figure 3. The upper and lower dentitions of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). The image displays the lingual sides of the upper and lower right dentitions and the alveoli of the lower left dentition. Abbreviations: C, upper canine; c, lower canine; I, upper incisor; i, lower incisor; M, upper molar; m, lower molar; P, upper premolar; p, lower premolar. The number indicates tooth locus (e.g. I1 denoting the first upper incisor and i1 the first lower incisor). The suffixes –l and –r indicate left and right tooth, respectively.

Figure 4. The anterior teeth and alveoli of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). A, optic image; B, scanning electron microscope image. The suffixes -l, -r, -i, and -a in A represent left, right, impression, and alveolus, respectively. Note the orientation of the left canine alveoli (i3-l-a) in which the mesial root is larger and more labially positioned. Abbreviations: C, upper canine; c, lower canine; I, upper incisor; i, lower incisor; P, upper premolar; p, lower premolar. The number indicates tooth locus. The suffixes -a, –l and –r indicate alveolus, left and right tooth.

Figure 5. Optical images of postcanine teeth of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). A, lateral view of the right upper and lower postcanine teeth; B, occlusal view of the right upper postcanine teeth; C, occlusal view of the right lower postcanine teeth and preserved parts of broken molars. Abbreviations: M, upper molar; m, lower molar; P, upper premolar; p, lower premolar. The number indicates tooth locus.

Figure 6. Scanning electron microscope images of mesial postcanine teeth of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). A, lingual view; B, labial view. To assist comparison, image in A has been horizontally flipped. Abbreviations: M, upper molar; m, lower molar; P, upper premolar; p, lower premolar; wf, wear facet.

Figure 7. Scanning electron microscope images of molars and ultimate premolars of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). A, lingual view; B, labial view; C, occlusal view of the upper teeth. To assist comparison, image in A is horizontally flipped. Abbreviations: M, upper molar; m, lower molar; P, upper premolar; p, lower premolar. The number indicates tooth locus.

Figure 8. Occlusal view of left M3 (partial) and M4 of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). Abbreviations:?B1, possibly cusp B1 (B̕); dsc, distal stylar cusp; M3, third upper molar; M4, fourth upper molar; mts, metastyle; pa, paracone; pas, parastyle; poc, postparacrista; prc, preparacrista; stc, stylocone.

Figure 9. Labial and lingual views of the upper and lower dentitions of Lactodens sheni gen. et sp. nov. The reconstructed teeth or parts of teeth are shown by grey lines (the outline of premolars are largely based on the impression left in the slab). Abbreviations: C, upper canine; c, lower canine; I, upper incisor; i, lower incisor; M, upper molar; m, lower molar; P, upper premolar; p, lower premolar. The number indicates tooth locus.

Figure 10. Dentary morphology of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). A, medial view of the posterior portion of the dentary; B, pterygoid fossa (pocket) in dorsal view; C, medial surface of the dentary showing no trace of the Meckelian groove; D, reconstruction of the lower jaw in medial view (the reconstructed teeth are shown in grey lines). Abbreviations: apc, anterior portion of the pterygoid fossa; arc, anterior ridge of the coronoid process; cmf, crest that defines the anterodorsal margin of the masseteric fossa; cp, coronoid process; ct, coronoid tubercle; md, mandibular condyle; mf, mandibular foramen; mrp, medial ridge of the pterygoid fossa; pc-b, pterygoid crest (broken); pf-d, pterygoid fossa in dorsal view; ppc, posterior portion of the pterygoid fossa.

Figure 11. Optical and scanning electron microscope images of the right manus of Lactodens sheni gen. et sp. nov. (holotype, HG-M016). The images show the ventral view of the manus. Abbreviations: ca, capitate; fsr, flexor sheath ridges on the proximal phalanges III and IV; ha, hamate; ip-II–V, intermediate phalanx II–V; lc-ip-II, lateral crest of trochlea on the distal end of intermediate phalanx II; lf-up-III, lateral facet of articulation on the proximal end of ungual phalanx III; lr-up-II, lateral ridge of ungual phalanx; Mc II, III and V, metacarpal II, III and V; mc-ip-II, medial crest of trochlea on the distal end of intermediate phalanx II; mf-up-III, medial facet of articulation on the proximal end of ungual phalanx III; pp-V, proximal phalanx V; sb, sesamoid bone; td, trapezoid; tg, tubercles (three) confining two articular grooves on the distal end of proximal phalanx III; tm, trapezium; tr, triquetrum; up-III, ungual phalanx III; up-IV, ungual phalanx IV. The question mark ‘?’ indicates part of a possible carpal element in the wrist.