Hadrocodium Luo, Crompton, and Sun, 2001

Luo, Zhe-Xi, Bhullar, Bhart-Anjan S., Crompton, Alfred W., Neander, April I. & Rowe, Timothy B., 2022, Reexamination of the mandibular and dental morphology of the Early Jurassic mammaliaform Hadrocodium wui, Acta Palaeontologica Polonica 67 (1), pp. 95-113 : 96-108

Genus Hadrocodium Luo, Crompton, and Sun, 2001

Type species: Hadrocodium wui Luo, Crompton, and Sun, 2001 ; see below.

Hadrocodium wui Luo, Crompton, and Sun, 2001 Fig. 1 .

Holotype: IVPP 8275 , a nearly complete skull.

Type locality: Dadi Locality, Lufeng Basin, Yunnan Province, China. Type horizon: Upper Red Beds of the Lower Lufeng Formation, Lower Jurassic .

Material.— IVPP 8275 holotype .

Emended diagnosis.—Dental formula I5/i4 C1/c1 P3/p3 M2/m2. It differs from species of Morganucodon , Erythrotherium , Dinnetherium , Haldanodon , and triconodontids in that primary lower cusp a occludes in the embrasure between the opposite upper molars, instead of the primary cusp a of the lower molar m2 occluding between cusps A and B of the upper M1 and M2. It differs from Megazostrodon rudnerae Crompton and Jenkins, 1968 , in lacking prominent labial cingulid cuspules of the upper and from kuehneotheriids in lacking the triangulation of molar cusps. It differs from morganucodontids, eutriconodonts, and kuehneotheriids in the presence of a much larger postcanine diastema. It differs from Sinoconodon and most cynodonts in the one-to-one precise occlusion of the upper and lower molars. It lacks the multicuspate rows on the teeth of Haramiyavia clemmenseni Jenkins, Gatesy, Shubin, and Amaral, 1997 , Kalaallitkigun jenkinsi Sulej, Krzesiński, Tałanda, Wolniewicz, Błażejowski, Bonde, Gutowski, Sienkiewicz, and Niedźwiedzki, 2020 , multituberculates, and the multiple-ridged teeth of docodonts.

Description.— Mandibles: The mandibular body (alveolar ramus) is gracile and dorso-ventrally shallow. Its ventral margin is slightly convex, and the tooth-bearing alveolar margin is straight along the postcanine row but noticeably curved upward in the postcanine diastema anterior to the first preserved premolar, and in the symphyseal region of the mandible. The symphyseal region of the mandible is even thinner than the diastema region and the mandibular body along the tooth row ( Figs. 1–3 ). The gracile nature of the mandibular part that bears incisors and canine suggests that there might have been bone resorption after the loss of the anterior premolars. The ventral margin of this mandibular part was slightly damaged by abrasion before the specimen was found. Due to the abrasion, the roots of incisors and canine are partly exposed ( Figs. 2–4 ).

The postcanine diastema is a pronounced feature between the canine and the first preserved premolar, herein designated as p2. Anterior to the p2 there are two dimples on the dorsal margin of the mandible, which clearly represent the plugged alveoli of a two-rooted premolar that was shed at this locus. These plugged (although clearly recognizable) alveoli suggest that there used to be an anterior premolar (p1) in this position. It cannot be ruled out that the postcanine diastema may have accommodated more than one premolar. But here we offer a conservative estimate that H. wui has three premolar positions, and we designate the lost premolar represented by the two plugged alveoli to be p1 (see Discussion).

The coronoid process rises from the alveolar line of the mandibular body, forming a 120–125° angle to the anterior border of the coronoid ( Figs. 1–3 ). The fully developed dentary condyle is preserved on the right mandible, but broken on the left mandible. The condyle is differentiated from the posterior part of the coronoid process by a constricted peduncle, a structure typical of stem mammaliaforms, including haramiyidans ( Luo et al. 2002; Gill et al. 2014; Luo et al. 2015a; Schultz et al. 2019; Panciroli et al. 2019; Sulej et al. 2020). The condyle is slightly compressed dorso-ventrally, less robust than the more bulbous and rounded dentary condyles of Morganucodon species ( Kermack and Mussett 1958; Kermack et al. 1973; Gill et al. 2014) and Sinoconodon rigneyi Patterson and Olson, 1961 ( Crompton and Sun 1985; Crompton and Luo 1993; Luo 1994; Luo and Wu 1994; Luo et al. 1995).

The anterior margin of the coronoid process is aligned mesio-distally with the postcanine row (including m2). Between the ultimate molar (m2) and the rising base of the coronoid process there is a gap, herein called the retro-molar space ( Figs. 3 and 4 : arrow). We interpret that the direct alignment of the toothrow in front of the coronoid process is an adult feature, by comparison to the adult pattern of other mammaliaforms with more complete series of ontogenetic stages, and this feature is also apomorphic in a phylogenetic context (details in Discussion).

Lingual aspect: The mandibular symphysis is marked by a distinct area on the lingual aspect of the anterior part of the mandible. Although the symphyseal surface is somewhat damaged on both mandibles, the preserved parts of symphysis show that it is slightly rugose. As preserved, the two hemimandibles are not in direct contact, indicating that the mandibular symphysis was connected by soft tissues that are not fossilized ( Figs. 1–3 ). The symphysis extends posteriorly to below the canine, and this posterior extent of the symphysis is much shorter than the symphyses that extend to the level of the p 3 in species of Morganucodon , or to the ultimate premolar (p5) or m 1 in docodontans. We interpret that the symphyseal contact likely was mobile, as in all known mammaliaforms, and a vast majority of crown mammals ( Kielan-Jaworowska et al. 2004; Scott et al. 2012).

The Meckel’s sulcus is present in the molar region of the mandibular body ( Figs. 1–3 ). The sulcus is short, and extends from below the foramen of the mandibular canal anteriorly toward the ventral mandibular margin; its anterior end fades out below the ultimate molar (m2). However, there is no Meckel’s sulcus on the lingual side of the mandible between the symphysis and the posterior premolars, which is confirmed on the scan slices. The Meckel’s sulcus is present in the premolar and symphyseal region in juvenile or young adult individuals of other mammaliaforms. For example, the Meckel’s sulcus is present in this region in juvenile growth stages of species of Morganucodon , Docodon , and Haldanodon but the sulcus disappears in this region in full adult specimens of the same taxa ( Kermack et al. 1973; Krusat 1980; Schultz et al. 2019; see also adult mandible of Borealestes serendipitus Waldman and Savage, 1972 ; Panciroli et al. 2019).

The postdentary trough is present on the medial side of the mandible, and a part of this trough is dorsally bordered by the medial ridge ( Figs. 1–3 ). The trough is shallow along its entire length and its anterior part contains the mandibular foramen. Posterior to the foramen the trough is open dorsally and lacks a border, likely for the passage of the inferior alveolar nerve (from the mandibular branch of the trigeminal nerve) into the mandibular canal. The posterior end of the trough becomes slightly wider and opens in a shallow embayment posteriorly below the posterior part of the medial ridge ( Figs. 1–3 ).

Some incomplete parts of middle ear elements are detected by CT scans, and can be visualized ( Fig. 4 ). The preserved parts certainly include a part of the ectotympanic (or “angular”) bone. Another preserved element may be a fragment of the Meckel’s element (or the “prearticular”). However, the surangular as interpreted for Morganucodon oelheri Rigney, 1963 , by Kermack et al. (1973), cannot be identified with certainty. The identities of these postdentary elements are difficult to establish and remain uncertain due to poor preservation. These elements cannot be fully resolved until better specimens of H. wui are recovered in the future.

On the medial side and near the base of the coronoid process, there is an oval and elongate depression, which we interpret to be the coronoid facet. Near the coronoid fossa of the right mandible, a small piece could belong to the coronoid bone but this identification remains tentative because this element is difficult to trace in coronal CT slices, and must be corroborated by additional specimens of H. wui recovered in the future. Although smaller, the fossa for the coronoid is identical in shape and position to the coronoid facet of Morganucodon watsoni Kühne, 1949 ( Kermack et al. 1968). The coronoid fossa is less similar to that in Kuehneotherium praecursoris D.M. Kermack, K.A. Kermack, and Mussett, 1968 ( Gill 2004; Gill et al. 2014) than to M. watsoni (Pamela Gill, personal communication 2021).

Hadrocodium shares plesiomorphic features related to the postdentary trough with basal mammaliaforms. Notwithstanding the minor differences of a shorter and thinner Meckel’s sulcus, a shallower trough, and much weaker medial ridge, Hadrocodium shares these plesiomorphic features with Sinoconodon , morganucodontans, and docodontans. This new revelation by CT scanning necessitates a major revision of the interpretation of these features by Luo et al. (2001).

The alveolar margin of the mandible is nearly straight and horizontal. The posterior part of this margin lacks the dental lamina groove (Crompton’s groove sensu Parrington 1971); this groove is a plesiomorphic feature present in species of Morganucodon and Sinoconodon rigneyi , and a wide range of non-mammaliaform cynodonts on the lingual side near the last molars ( Kermack et al. 1973; Crompton and Luo 1993). The absence of this feature is an apomorphy of Hadrocodium and is otherwise known in docodontans ( Schultz et al. 2019; Panciroli et al. 2019) and in the Late Triassic haramiyidans ( Jenkins et al. 1997; Luo et al. 2015a; Sulej et al. 2020).

Labial aspect: On the labial side of the mandible, the coronoid process has a shallow concave area, identified as the masseteric fossa. The anterior border of the fossa lacks a distinct border and grades into the mandibular body as in species of Morganucodon ( Kermack et al. 1973; Crompton and Luo 1993; Lautenschlager et al. 2017). The ventral extent of the masseter muscle attachment is not a clearly demarcated border in H. wui ; this is similar to Dinnetherium nezorum Jenkins, Crompton, and Downs, 1983 ( Jenkins et al. 1983; Crompton and Luo 1993) and to docodontans ( Krusat 1980; Ji et al. 2006; Schultz et al. 2019; Panciroli et al. 2019). There is no foramen in the masseteric fossa, as in some derived zatherian mammals ( Davis 2012). The lateral ridge extending from the dentary condyle and peduncle is distinctive. We interpret that the lateral ridge is a landmark boundary, approximately, between the concave surface for M. masseter profundus dorsally and a narrow, flat surface for the M. masseter superficialis on the moderately inflected angular process of the mandible, as reconstructed for other mammaliaforms, by comparison to extant mammals ( Turnbull 1970; Lautenschlager et al. 2017; Schultz et al. 2019).

Two mental foramina are present on the labial side of the anterior (symphyseal) part of the mandible ( Figs. 1–3 ). The anterior of these foramina is small and located between i4 and the canine on the better preserved left mandible. The posterior mental foramen is oval shaped and much larger than the anterior-most foramen. It is located in the postcanine diastema region on both mandibles, just anterior to the level of the p1 position indicated by plugged alveoli of this tooth. These foramina are connected to the mandibular canal inside the mandibular body ( Fig. 5 ).

The mandibular canal on the left mandible is exposed by erosion of the thin cortical bone. From the exposed width of the canal, and from the cross section of the right mandible, the canal diameter is measured to be 0.12–0.13 mm, in a mandibular body that is 0.8–0.9 mm in height ( Figs. 2 , 5 ). The canal is oval in cross section ( Fig. 5 ) and is similar in H. wui and M. oehleri . The canal is positioned lateral to the root alveoli in cross section and is in the laterally bulging part of the mandibular body ( Fig. 5A ). The lateral placement of this canal relative to the postcanine alveoli is similar in H. wui and M. oehleri , but this placement is slightly different from the canal’s position in docodontans. In all docodontans examined so far ( Schultz et al. 2019; Panciroli et al. 2019), the mandibular canal is ventrolateral to the tooth root apices ( Fig. 5C, D ). The latter pattern of docodontans is more closely comparable to those of extant therians than in morganucodontans and Hadrocodium , and therefore more derived. In crown mammals of the Mesozoic, such as the eutriconodontans Juchilestes and Repenomamus , the mandibular canal is located directly ventral to the root apices Fig. 5E ).

The mandibular canal diameter is 14.5 to 15% the size of the mandibular body depth. If scaled to the width or the lingual side depth (= height) of the mandibular body, the mandibular canal shows about the same proportion to the width and height of the mandibular body ( Fig. 5 ) as the canals of docodontans and morganucodontans. The mandibular body of H. wui shows a distinct pattern in that the height of the mandibular body is lower on the labial side of the postcanine alveoli ( Fig. 5A 2 : asterisk), than on the lingual side of the alveoli ( Fig. 5 : arrow). The mandibular canal diameter appears to be slightly larger if it is scaled to the labial height of the mandible. Overall, the mandibular canal diameter is about the same relative to the mandibular body in H. wui , as in species of Morganucodon and docodontans. The mandibular canal of mammaliaforms (including H. wui ) is much smaller than those of Cretaceous Teinolophos trusleri Rich, Vickers-Rich, Constantine, Flannery, Kool, and Van Klaveren, 1999 , and the extant monotremes, than previously interpreted by Rowe et al. (2008).

Dentition: Here we are revising the dental formula of H. wui to I5/i4–C1/c1–P3/p3–M2/m2 ( Figs. 6–9 ). This change includes the alveoli of the first upper and lower premolars that were shed in the P1/p1 position. The new formula differs from the previous I5/i4–C1/c1–P2/p2–M2/m2 ( Luo et al 2001: fig. 1), but the previous study only counted the intact permanent teeth and excluded the upper P1 and lower p1 positions indicated by the visible (but plugged) alveoli of the shed premolars. In the following description we use a dental formula that includes the tooth positions represented by empty and plugged alveoli. We also follow Crompton and Jenkins (1968) and Crompton (1974) for their scheme of cusp designations.

Upper dentition: There are five upper incisors in the better-preserved right premaxilla ( Luo et al. 2001; Rowe et al. 2011). Crowns of I2–I4 are intact but I1 and I5 are broken, although their alveoli can be clearly identified ( Figs. 6 , 7 , 9 ). The left premaxilla has preserved only I4 (incomplete) and I5; the left I1–I3 and their corresponding left premaxillary part are damaged and lost. The I5 is separated from the upper canine by a gap that accommodates the tall and trenchant lower canine locking into a deep canine pit (the paracanine fossa) in the premaxilla when in full occlusion ( Fig. 6 ). This is visible from cross sections of the CT scans.

The upper canine shows a triangular outline in lateral profile and is slightly compressed medio-laterally ( Figs. 6 , 7 , 9 ). The canine has two strong roots divergent from each other at an angle of 20–25°. The conical shape of the crown as previously illustrated by Luo et al. (2001: fig. 1) was not correct. The posterior root of the two divided roots was not detected by manual fossil preparation in previous study ( Luo et al. 2001), as the root was covered by a mixture of frail bone and matrix in the specimen. The upper postcanine diastema is a prominent gap between the two-rooted canine and the preserved premolars (P2 and P3 on the right side, but only P3 on the left side). The right maxillary alveolar margin in the postcanine diastema shows a surface dimple corresponding to a single plugged alveolus ( Fig. 6 ). This is identified by CT visualization, and also recognizable on the specimen under the microscope. This plugged alveolus is interpreted to have held a single-rooted premolar that was shed. Its position is herein designated as “P1-position” ( Figs. 5–7 ).

The upper P2 is preserved only on the right maxilla, but this tooth on the left maxilla was lost without being replaced, and its two alveoli are plugged though still recognizable. The right P2 is two-rooted but its crown is broken. The tooth is separated anteriorly from the plugged P1 alveolus by a small interdental gap. It is also separated posteriorly from P3 by a very small interdental gap. P3 is the largest of all teeth in height and in length ( Table 1). Its large main cusp (cusp A) is slightly asymmetrical with a straight anterior crest and a slightly concave posterior crest. At the cingular level, the tooth shows no distinctive cingular line on either the lingual or the labial aspect although in the cingular positions there are slight bulges. There are two miniscule cingular cuspules on the mesial and the distal ends of the crown. However, the large mesial cuspule illustrated in previous study ( Luo et al. 2011: fig. 1) is an artefact, as shown by CT visualization. The left and right P3s show an asymmetrical variation on two sides: the left P3 has a weakly developed cusp on the posterior crest, which is entirely absent on the right P3 ( Fig. 7 ). We could not rule out that this difference is caused by preservation differences of the left P3 versus the right P3.

The upper M1 is better preserved on the right side than the left. This tooth bears three main cusps: the primary cusp A in the middle, a lower but pointed cusp B on the mesial end of the crown, and a low posterior cusp C (worn off on the left M1). Additionally, a miniscule cingular cusp D is on the distal end of the crown ( Figs. 6 , 7 ). The base of the crown has a slightly bulging labial cingulum although lacking any cingular line or cuspules. The lingual aspect of the tooth exhibits a weak cingulum, which is otherwise featureless. M1 has two fully divided roots and the posterior root is curved or slanted mesially, in both the left and the right M1. The distal part of left M1 appears to be either heavily worn or damaged in fossilization, such that cusp C and cuspule D are lost.

M2 is characterized by the taller cusp A and shorter cusps B and C located respectively at the mesial and distal ends of the crown. But it lacks distal cuspule D. The tooth crown is antero-posteriorly symmetrical with respect to the middle cusp A. M2 has two fully confluent roots connected by dentine over the root length. The M2 roots are much shorter than those of M1. The gradient of variation of the successively more confluent and shorter roots in the more posterior molars, as seen here in H. wui ( Fig. 7 ), is also well documented in specimens of Morganucodon watsoni , in which the last upper molar (M4) has the smallest (also the shortest) crown of all postcanines, and often with a fusion of the two roots ( Parrington 1971; Kermack et al. 1981; Jäger et al. 2019). The upper molars of Kuehneotherium praecursoris also show a gradient of variation in root fusion of molars ( Mills 1984; Gill 2004). In the latest reconstruction of the toothrow in Kuehneotherium praecursoris, Gill (2004) placed the smaller to smallest upper molars with more fused roots in the ultimate toothrow position(s), although the degree of fusion of the two roots can be variable in the posterior molars ( Gill 2004, personal communication 2021).

The upper M2 is very small, its crown is much shorter than the M1 crown, and lower in height than P3 and M1 (50% of P3 length and 65% of the M1 length). Taken together, P3, M1, and M2 form a posteriorly decreasing size (length) and height gradient. The posteriorly decreasing size of molars is also present between m2 and the ultimate molar (m4 or m5) in morganucodontans and docodontans, but the steep gradient of size reduction in P3–M1–M2 is an unusual and distinct feature of H. wui .

Thus the size reduction of the ultimate molar crowns M2/m2), and the confluence and fusion of the roots of M2/ m2, as seen in H. wui are similar to the ultimate molars of Morganucodon watsoni ( Parrington 1971; Kermack et al. 1981; Gill 2004; Jäger et al. 2019). Kuehneotherium praecursoris is also interpreted to have very small ultimate upper and lower molars ( Gill 2004). The comparative morphology across mammaliaforms is a strong indication that the small upper M2 with fused roots in H. wui type specimen is the terminal tooth in the entire dentition.

Lower incisors and canine: The left incisors are well preserved and are the basis for this description. Of the four lower incisors, i1 is twice the thickness of other incisors. The elongate i1 root extends posteriorly and medial to the roots of i2 through i4. Although the apical end of the long i1 root is not preserved (as far as can be assessed by CT), the alveolar canal of this root reaches posteriorly to a point medial to the 4 root. The i2, i3, and i4 are peg-like, and they form a size gradient with i2 being the shortest and i4 being the tallest.

The lower canine is mediolaterally compressed and the lateral profile of its crown has a lanceolate outline; the crown is slightly asymmetrical with the anterior crest being more convex and the posterior crest more concave. The canine has mediolaterally compressed roots, which are incipiently divided, with a shallow groove on the medial aspect along the root length, and a figure 8 shape in horizontal cross-section. Internally the canine roots have two divided root canals, as visualized in CT slices. The canine roots are long and penetrate the entire depth of the mandible. The lower canine of H. wui shows similarity to lower canines of docodontans in having incipiently divided roots with side grooves ( Meng et al. 2015; Schultz et al. 2019; Panciroli et al. 2019). It is almost identical to the isolated lower canines of Kuehneotherium praecursoris identified from the Lower Jurassic fissure-fill deposits of the UK ( Gill 2004: fig. 3.16), and bears some resemblance to the lower canines of Morganucodon watsoni ( Mills 1971; Parrington 1971). But the root structure and crown shape differ from the more conical crown and single-rooted canine (with singular root canal) of Morganucodon oehleri Rigney, 1963 ( BMNH 2858) ( Luo et al. 1995) and of Sinoconodon rigneyi ( Fig. 9 ).

Lower premolars: There is a long postcanine diastema posterior to the canine and anterior to p 2 in both mandibles. On the right mandible, there are two partially plugged alveoli, indicating that a two-rooted premolar was shed ( Figs. 4 , 6 , 8 ). Although this anterior premolar is no longer present, its position is marked by a vestige of its alveoli. We designated this as p1 position, and we propose that the first premolar position, despite that the tooth is already lost, should be included in the tooth count and the dental formula. The comparison of the postcanine diastema will be further detailed below (see Discussion section).

The lower p2 crown, as preserved on right p2, shows a triangular outline in lateral view with a central cusp a and a posterior cusp c. The p2 is separated anteriorly by an interdental gap from the plugged p1 alveoli, also posteriorly separated by an interdental gap from p3. The tooth has two roots. The lower p3 is the tallest and most trenchant tooth of all lower teeth. Its crown is triangular in lateral outline. Its main cusp is asymmetrical in tilting anteriorly with a nearly vertical anterior crest that is shorter than the slanting posterior crest. The posterior crest bears a small cuspule, separated from the distal cingulid cuspule. The distal end of the p3 crown shows no interlocking with m1.

Overall, the ultimate premolar of H. wui more closely resembles the posterior premolar type of Kuehneotherium praecursoris ( Gill 2004; Gill et al. 2014) rather than Morganucodon watsoni ( Kermack et al. 1973) . The largest premolar (p3) of IVPP 8275 is nearly identical to premolars that were interpreted to be the posterior premolars of Kuehneotherium praecursoris by Gill ( Gill 2004: figs. 3–18; Gill et al. 2014) and by Kermack and co-workers ( Kermack et al. 1968: text-fig. 3).

Lower molars: Lower m1 has a tall cusp a, and a much shorter cusp b on the mesial end of the tooth but no anterior cingulid cusp e. The posterior main cusp c is also a short cusp, separated from a distinctive distal cusp d. Cusps a–d are aligned in a straight line. The base of the crown shows no cingulid line and lacks cusps on either the labial or lingual side of the tooth.

The lower m2 has a tall cusp a and two well-developed cusps b and c. Both cusps b and c are distinctive in rising higher from the base, and different from their counterparts on m1. The main cusps a–c of m2 are aligned in a straight line similar to the straight-line cusp pattern of m1. Cusp b is closely twinned with a small (although discernible) anterolingual cuspule e. The posterior main cusp c is lower than the anterior cusp b, and c is more divergent posteriorly from the primary cusp a, such that cusp a and cusp c are farther apart from each other than cusp a and cusp b. The distal cusp d is recognizable but not well developed. The crown shows no lingual or labial cingulid structure, unlike the ultimate molars of species of Morganucodon that show a distinct cingulid line in m4 of M. oehleri and distinctive lingual cingulid plus strong cingulid cusps (such as cusp g) (sensu Mills 1971; Kermack et al. 1973; Crompton 1974; Butler 1988; Fig. 10 ). The lower m2 has much more confluence of its two roots than m1. Hadrocodium wui has a molar interlocking mechanism between the distal cusp d of m1 and the flat surface between cusp b and the very small cingulid cuspule e of m2. But there is not tooth-to-tooth interlock between p3 and m1.

The lower m2 is the smallest tooth of the p3–m1–m2 series. The left m2 is smaller than the left m1, and about 84% of the latter in length. The left m2 is also smaller than the left p3, and about 93% of the latter in length. The lower p3– m2 constitute a gradient in decreasing crown height in the successively more posterior teeth. There is a greater degree of fusion of roots in the most posterior molar as compared to the preceding postcanines.

The posteriorly decreasing size gradient of the posterior lower molars, as seen here in H. wui , has long been recognized for M. watsoni and M. oehleri ( Parrington 1971; Kermack et al 1973, 1981; Crompton and Luo 1993; Jäger et al. 2019). This plesiomorphic pattern is also typical of docodontans, many of which are now known by complete lower molar series ( Krusat 1980; Ji et al. 2006; Luo and Martin 2007; Averianov et al. 2010; Meng et al. 2015; Rougier et al. 2015; Schultz et al. 2019; Panciroli et al 2019; Zhou et al. 2019). A similar gradient in posterior decrease of molar size has been interpreted for Kuehneotherium praecursoris Gill 2004 ), and has also been documented for Haramiyavia Jenkins et al 1997 ; Luo et al. 2015a). Therefore, we interpret that the posteriorly decreasing gradient of lower molars is a plesiomorphic feature of perhaps all mammaliaforms of Late Triassic to Early Jurassic age, including Hadrocodium . The eruption of the posterior-most molar, the m 2 in the case of H. wui , which is also the smallest of the postcanine size gradient, indicates the end of tooth development.

The two roots of the lower m 2 in H. wui are fused along most of their length ( Figs. 8 , 10 ); the roots are fused by more than two-thirds of the root length in right m2 ( Figs. 8 , 10 ), or by half of the root length in left m2. This is structurally comparable to the terminal lower molar (m4) in morganucodontans with roots fused either partially (as in M. oehleri ), or entirely (as in some m4s of M. watsoni ) ( Fig. 10 ). These fused roots of the m4 of Morganucodon correspond to a singular alveolus, instead of two alveoli as in preceding molars ( Fig. 11 ; see Parrington 1971; Kermack et al. 1973).

The pattern of the smallest ultimate molar (m4) with partially or entirely fused roots is now documented in many specimens of M. watsoni ( Parrington 1971; Kermack et al. 1973; Jäger et al. 2019). The same pattern is also present in the full-grown adult specimens of the docodontan Borealestes serendipitus , in which the last two lower molars show tightly compressed and partially fused roots ( Panciroli et al. 2019). Here we also corroborate this pattern in an unerupted m4 of M. oehleri ( Fig. 10 ). Although complete lower jaws with intact teeth are not available for kuehneotheriids, the best available postcanine reconstruction for this group ( Gill 2004) suggests similarly that the last molars have tightly compressed or fused roots in Kuehneotherium praecursoris ( Kermack et al. 1968; Gill 2004). A similar pattern also occurs in Kuehneotherium stanislavi Debuysschere, 2017 ( Debuysschere 2017). Based on this comparison, we interpret that the lower m2 of the H. wui type specimen, with its fused roots, is the last and terminal molar to erupt, and equivalent to the terminal molar m4) of species of Morganucodon ( Fig. 8 ) and last molars (m5 and m6) of B. serendipitus ( Panciroli et al. 2019) .

Molar occlusion: The upper and lower molars, now visualized from CT scans ( Figs. 6–8 ), show several features previously hidden by occlusal association of upper and lower teeth. This is helpful to rectify several incorrect interpretations of molar occlusion in our earlier study ( Luo et al. 2001: fig. 1). The new pattern of correspondence between cusps of upper and lower molars is different by half of a cusp from the previous incorrect reconstruction ( Luo et al. 2001: fig. 1).

The upper postcanines show more of a lingual inclination than the more vertically oriented lower molars. The upper molar lingual inclination is common for the triconodont-like dentition of other mammaliaforms, such as Morganucodon , and of eutriconodontans ( Jäger et al. 2019, 2020, 2021). This pattern is obvious from the better preserved right side of the skull, which shows more lingually inclined upper teeth in CT sections ( Figs. 4 , 5 ).

The lower m1 cusp a is aligned with the embrasure between the upper P3 and M1 ( Fig. 6 ). The anterolabial surface of the large cusp a of m1 occludes with the postero-lingual aspect of the cusp A of the upper P3. Because the cusp b on m1 is relatively low, the entire anterolabial aspect of m1 likely forms a continuous contact facet to P3. The posterolabial aspect of cusp a of m1 occludes with the anterolingual surface of upper M1 across both cusp A and cusp B. As cusp c of the m1 is relatively small (better preserved on the left m1), the m1 crown forms a continuous wear surface between cusps a and c. It also appears that m1 is more strongly worn than m2, and this is consistent with the greater degree of wear of m1 as the first erupted molar of the molar row, as in other mammaliaforms with diphyodont tooth replacement. The lower m2 primary cusp a occludes into the embrasure between the posterior cusp C of M1 and the anterior cusp B of M2, while m1 cusp b can contact the lingual surface of cusp C of the upper M1 ( Fig. 6 ).

The upper M1 primary cusp A occludes into the embrasure between the cusp c of lower m1 and cusp b of m2. The lingual surface of cusp C of the M1 has a discernible facet that matches the labial facet across cusps a and b of the lower m2.

A striking pattern of the molar occlusion is the unique occlusal match of M1–M2 vs. m2 as a consequence of size disparity of the very small upper M2 vs. larger lower m2. Because of the much reduced size of the upper M2 relative to m2, cusp A of the M2 directly corresponds to the shallow notch between cusp a and cusp c of m2. On the labial side of the lower m2, there are wear facets on the shallow valley. Overall the lower p3–m1–m2 occlude in the embrasure of the upper P3–M1–M2, but due to the size difference of M2, the M2 cusp A occludes between cusps a and c of the lower m2.

In extant mammals, wear facets developed in the ultimate molars would indicate the terminal stage of dental ontogeny ( Anders et al. 2011), as the ultimate molars are the last to erupt in the full permanent dentition in extant marsupials ( van Nievelt and Smith 2005a; Astua and Leiner 2008) and placentals ( Smith 2000; Anders et al. 2011). Thus, the presence of a wear pattern on M2/m2 suggests that the H. wui holotype specimen is an adult, or approaching the adult stage (see Discussion for further details on growth stage assessment).

Stratigraphic and geographic range.—Upper Red Beds of Lower Lufeng Fromation, Yunnan, China; Dadi Locality in Lufeng Basin.

Figures

Fig. 1. Lower jaws of comparative mammaliaforms. A. Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275), from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. Right mandible in lingual (A1) and labial (A2) views. B, C. Morganucodontan Dinnetherium nezorum Jenkins, Crompton, and Downs, 1983, from the Kayenta Formation, Lower Jurassic of Arizona, USA. B. An older adult (MCZ 20910).C.A young adult (MCZ 20870); modified from Crompton and Luo 1993). Younger individuals of D. nezorum have a full set of five premolars, while older individuals have lost up to two anterior premolars without replacement. The postcanine diastema developed by shedding premolars without replacement is an aging-related character of many mammaliaforms including Hadrocodium.

Fig. 2. CT rendering of the mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. A. Right mandible in lateral view (stereopairs); solid triangle indicates the retro-molar space.B. Left mandible in lateral B1, stereopairs) and ventral (B2, stereopairs) views.

Fig. 3. CT rendering of the mammaliaformHadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. A. Left mandible in medial view (stereopairs). B. Right mandible in tilted dorsal view (B1, stereopairs), p1 is lost but the p1 position is indicated by plugged alveoli; arrow indicates the retro-molar space; and medial view (B2, stereopairs).

Fig. 4. CT rendering of the mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. A. Mandibles and upper teeth (as preserved) in ventral view. B. Mandibles in association with upper teeth in dorsal view. The upper teeth are more inclined (as preserved) and are oblique to the lower teeth. Post-mortem distortion caused the upper postcanines to shift relative to lower teeth, by half a cusp length.

Fig. 5. Comparative morphology of mandibular canals and tooth alveoli in cross sections. A. The mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China; transverse section of mandible through anterior root of p3 (A1): the main cusp of a tall p3 occludes into its maxillary pit on the palate, and the upper P3 is more lingually inclined than the lower p3; transverse section through anterior root of m1 (A2): the mandibular canal is lateral to the root alveoli of the postcanines; the upper M1 is more inclined lingually than the lower m1; the main cusp of lower m1 occludes into its pit on the palate; the asterisk indicates that the labial alveolar margin is lower on the lingual alveolar margin at the point of the arrow. B. The morganucodontan Morganucodon oehleri Rigney, 1963, (BMNH 2858) from the Lower Lufeng Formation, Lower Jurassic of China; cross section through the anterior root of m2: oval outline with long diameter at 0.25 mm, and short diameter at 0.15 mm; mandibular canal is lateral to the side of roots. C, D. The docodontan Docodon victor Schultz, Bhullar, and Luo, 2019 (as revised by Schultz et al. 2019) from the Morrison Formation, Upper Jurassic of Wyoming, USA. C. YPM 11826, cross section through a root alveolus of m5; mandibular canal cross section in oval outline with the long diameter at 0.8 mm and short diameter at 0.6 mm. D.YPM 11823, mandibular canal cross section through anterior root of m5; circular outline diameter 0.6 mm.The mandibular canal is ventro-lateral to the apices of roots in docodontans. E. The eutriconodontan Juchilestes liaoningensis Gao, Wilson, Luo, Maga, Meng, and Wang, 2009 (DMNH 2607) from the Yixian Formation, Lower Cretaceous of China, as a representative for root canal pattern for crown mammals, in which the mandibular canal is ventral to the root apices.

Fig. 6. Composite reconstruction of full dentition of Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China.A. Left upper teeth (as preserved) in labial view (CT visualization). B. Lower teeth (as preserved) in lingual view. C. Composite reconstruction of upper and lower teeth (left labial view). D. Occlusal relationship of cusps between lower m1–m2 and upper P3–M1– M2 (revised from Luo et al 2001: fig. 1; cusp designation following Crompton and Jenkins 1968 and Crompton 1974).

Fig. 7. CT rendering of mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. A. Left upper teeth (as preserved and virtually extracted from the upper jaws): incomplete incisors, canine, incomplete premolars and two molars in lingual (A1, stereopairs) and labial (A2, stereopairs) views. Left dentition has five incisor positions:two broken incisors, plus three additional incisors represented by alveoli visible on the fossil skull (but not visualized in teeth). Of the three premolars, P1 is not preserved, but P1 position is represented by a plugged alveolus on right side (not visualized here). P2 is preserved on right side, but is lost and represented only by empty alveoli on the left. P3 is intact on both sides.B. Right upper teeth:five upper incisors (I1 and I5 broken), canine, two premolars, two molars in lingual (B1, stereopairs) and labial (B2, stereopairs) views.

Fig. 8. CT rendering of mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China.A. Left lower teeth (as preserved and virtually extracted from mandible): incisors, canine, incomplete premolars molars in labial (A1, stereopairs) and lingual (A2, stereopairs) views. The first premolar (p1) was shed and the vestige of its former alveoli recognizable on right mandible; penultimate premolar (p2) crown broken; ultimate p3 intact. B. Right lower teeth extracted by CT visualization, in labial (B1, stereopairs) and lingual (B2, stereopairs) views.

Fig. 9. Comparative morphology of incisors and canine among mammaliaforms. A, B. The morganucodontan Morganucodon oehleri Rigney, 1963, from the Lower Lufeng Formation, Lower Jurassic of China. A. CUP-FMNH 2320, type specimen. B. BMNH 2858, whole tooth and oblique-horizontal slice to visualize the single root canal.C. The mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China; C1, upper incisors and canine; C2, lower incisors and canine. D. The docodontan Haldanodon exspectatus Kühne and Krusat, 1972 (Gui-Mam 41/75) from the Guimarota Coal Mine, Upper Jurassic of Portugal (Ruf et al. 2013; Huttenlocker et al. 2018). E. The docodontan Docofossor brachydactylus Luo, Meng, Ji, Liu, Zhang, and Neander, 2015b (BMNH 131735) from the Tiaojishan Formation, Upper Jurassic of China (Luo et al. 2015b). F. The docodontan Agilodocodon scansorius Meng, Ji, Zhang, Liu, Grossnickle, and Luo, 2015 (BMNH 00138) from the Tiaojishan Formation, Upper Jurassic of China (Meng et al. 2015b); F1, left upper incisors and canine in lingual view; F2, left lower incisors and canine (flipped). Hadrocodium wui is similar to Haldanodon exspectatus and Docofossor brachydactylus in the upper canine with fully divided roots, and to all docodonts in bilaterally compressed lower canine with partially divided (grooved) root(s) and the separated root canals inside the root(s). Hadrocodium wui is also similar to Kuehneotherium praecursoris is also similar to Kuehneotherium in this feature (Gill 2004). Hadrocodium wui differs from Morganucodon oehleri and Sinoconodon rigneyi both of which are characterized by single-rooted and tubular upper and lower canines.

Fig. 10. Comparative morphology of lower molars of Hadrocodium and Morganucodon. A. The morganucodontan Morganucodon oehleri Rigney, 1963, from the Lower Lufeng Formation, Lower Jurassic of China. BMNH 2858, ultimate lower molar (right m4) lingual (A1) and oblique occlusal (A2, stereopairs) views. B. The morganucodontan Morganucodon watsoni Kühne, 1949, from the Lower Jurassic fissure fills of Wales. UMZC Eo.CR1, ultimate molar (m4); lingual view (B1) (Jäger et al. 2019, image courtesy of Kai Jäger), lingual (B2) and occlusal (B3) views (Gill et al. 2014, images courtesy of Pamela Gill). C. The mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001, from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China. Holotype, IVPP 8275, right m1 and m2 in occlusal (C1, stereopairs) and lingual (C2) views.

Fig. 11. Comparison of the mandible of Hadrocodium with the mandibular growth stages of Morganucodon and Docodon. A. The mammaliaform Hadrocodium wui Luo, Crompton, and Sun, 2001 (holotype, IVPP 8275) from the Lower Lufeng Formation, Lower Jurassic of Yunnan, China (arrow indicates the retro-molar space and feature of adult). B, C. Mandibular growth stages of the morganucodontan Morganucodon watsoni Kühne, 1949 (= “Eozostrodon parvus”) from the Lower Jurassic fissure fills of Wales (Parrington 1971); adult (B) and the oldest-known adult (C) (respectively specimens D60 and D120 in Parrington 1971: fig. 3). D–F. Mandibular growth stages of the docodontan Docodon victor Schultz, Bhullar, and Luo, 2019, from the Morrison Formation, Upper Jurassic of Wyoming, USA. D. YPM 23748, juvenile. E. YPM 11823, adult. F. YPM 11826, the oldest-known adult. In H. wui, the ultimate molar is positioned in front of the coronoid process base. This is an adult feature well documented in the successively older (or the oldest-known) adult individuals in the growth series of other mammaliaforms. Morganucodon watsoni shows such a growth pattern: a positional shift of the ultimate molar (m4) anteriorly to the base of coronoid process in successively older individuals; the ultimate molar has a single alveolus for its fused or confluent root(s) in most cases (Parrington 1971; Pamela Gill, personal communication 2021). Docodon victor shows a similar growth pattern of a shift of the coronoid process relative to the last molar(s) of the toothrow. The youngest-available individual D. victor (YMP23748) shows the last molar (m5) is medial to the coronoid process. In the adult (YPM11823) the last molar (m7) is shifted more anteriorly. In the oldest-available individual (YPM11826), the ultimate molar (m8) is shifted to the anterior, as the coronoid process is shifted posteriorly relative to the toothrow (Schultz et al. 2019).The placement of the ultimate molar with a retro-molar space anterior to the base of the coronoid in the oldest-available adult specimen of D. victor is similar to that of H. wui. D–F, stylistic illustrations based on CT visualizations by Schultz et al. (2019).