A reappraisal of Theroteinus (Haramiyida, Mammaliaformes) from the Upper Triassic of Saint-Nicolas-de-Port (France)

The earliest mammaliaforms are difficult to assess because the fossil record is poor and because their distinctive morphologies cannot be directly compared with more recent mammaliaforms. This is especially true for the haramiyid genus Theroteinus, only known in the Saint-Nicolas-de-Port locality (Rhaetian, France). This study presents a new definition of the type-species Theroteinus nikolai. A new species Theroteinus rosieriensis, sp. nov., is named and distinguished by the lingual shift of distal cusps, a larger size, and a stockier occlusal outline. Comparisons with Eleutherodon, Megaconus and Millsodon suggest that Theroteinus has potential close relatives among the Jurassic haramiyids.


MATERIAL
This study describes 20 isolated teeth of haramiyids from Saint-Nicolas-de-Port. Denise Sigogneau-Russell and her co-workers have excavated only one stratigraphic level in the sand pit. Specimens collected during this fieldwork are kept both in the MNHN, with the acronym 'SNP,' and in the RBINS, with the acronym 'RAS.' Several amateur palaeontologists gathered their own collections alongside Sigogneau-Russell's team and donated them to MNHN and RBINS. The collection of Georges Wouters is identified by the suffix 'W' or 'FW,' and the collection of M. Marignac is identified by the suffix 'Ma.' However, there are no data on the exact, original stratigraphic level within the quarry of these collections. All the specimens described by Sigogneau-Russell, Frank & Hemmerlé (1986) and Hahn, Sigogneau-Russell & Wouters (1989)

Observations, drawings and measurements
All specimens were observed with a binocular microscope (CETI, Medline Scientific, Chalgrove, United Kingdom) at a magnification power of 36. A camera lucida mounted on the microscope was used for drawings. Measurements were taken with a digital readout for metrology (Heidenhain ND 1200, Traunreut, Germany). These measurements were used to make diagrams with Excel (Microsoft, Redmond, Washington, 2013) and statistical tests with the R statistical environment (R Development Core Team, 2016). The 3D images of studied teeth were obtained by X-ray computed tomographic (CT) scans at the AST-RX platform of the MNHN using phoenixjx-rayjvjtomejx L 240-180 CT scanner (GE Measurement & Control Solutions, Billerica, Massachusetts) (Table S1). The 3D data were processed with Materialise Mimics Innovation Suite 17.0 Research Edition (Materialise NV, Leuven, Belgium, 2014). The SEM photos were obtained by scanning electron microscope at the RBINS using a FEI QUANTA 200 ESEM (FEI, Hillsboro, Oregon) with a voltage of 15 kV and a dwell of 10 ms.

Dental nomenclature
The nomenclature used here to describe the haramiyid teeth is derived from Parrington (1947: Fig. 3), Hahn (1973: p. 5), Butler & Macintyre (1994: p. 435) and Butler (2000: p. 319). The row of cusps named a/A is characterized by less numerous, and well-individualized cusps. This row is lingual on lower teeth, but labial on upper teeth. The second row of cusps is named b/B. This second row is labial on lower teeth, but lingual on upper teeth. Both rows define a central basin. Additional cusps are named aa/AA when they are on the lateral flank of row a/A, and bb/BB when they are on the lateral flank of row b/B. In each row, cusps are numbered starting from number 1. On lower teeth, the numbering starts from the mesial extremity, while on upper teeth, it starts from the distal extremity. The term 'u-ridge' refers to the junction of crests which close the basin at its distal extremity on lower teeth and its mesial extremity on upper teeth. The term 'saddle' refers to the junction of two crests which delimits the basin at its open extremity, respectively mesial on lower teeth and distal on upper teeth. This nomenclature is used only in a descriptive purpose. The homonymy does not necessarily imply homology. Capital letters are used for upper teeth and lower case letters for lower teeth.
The descriptions of the wear facets are based on the nomenclature of von Koenigswald et al. (2013: p. 146) for jaw movements. This nomenclature is used to define the direction and the angle of the slope of the wear facets. The process and the pattern of the occlusion are beyond the scope of this article and will be dealt with in detail later on.

Methodology of characterization of Saint-Nicolas-de-Port Material
The haramiyid teeth are distinguished from other contemporary mammaliaforms by the presence of longitudinal rows of cusps separated by basins. For now, only two haramiyids are known in the Saint-Nicolas-de-Port material, Theroteinus and Thomasia Poche, 1908. The material referred to Theroteinus is distinguished from material referred to Thomasia by two main characteristics: low and obtuse cusps and a basin smaller in length and width.
The orientation of Theroteinus teeth follows the orientation described for Haramiyavia (Jenkins et al., 1997;Luo et al., 2015), Arboroharamiya (Zheng et al., 2013;Meng et al., 2014), Shenshou , Xianshou , and Thomasia (e.g., Butler, 2000). The lower molariforms are distinguished from the upper molariforms by the presence of, respectively, two or three rows of cusps and by the form of row a/A. In lower molariforms, the lingual row a includes the largest cusps. Cusp a1 is especially much larger than the other cusps, and it is located on the mesiolingual side of the crown. In upper molariforms, the labial row A shows three subequal cusps, when the central row B displays a cusp B2 larger than the other cusps of the row and located on the distal side of the crown.

Nomenclatural acts
The electronic version of this article in Portable Document Format (PDF) will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: urn:lsid:zoobank.org:pub:57401966-D5B5-468C-94FD-115C0C32FE00. The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central and CLOCKSS. Mammaliaformes Rowe, 1988 Order Haramiyida Hahn, Sigogneau-Russell & Wouters, 1989 Sub-order Theroteinida Hahn, Sigogneau-Russell & Wouters, 1989 Synonymy: Theroteinina Hahn & Hahn, 2006: p. 189.
Emended diagnosis: As for the type-family.
Distribution: As for the type-family.
Emended diagnosis: As for the type-genus.
Distribution: As for the type-genus.
Emended diagnosis: Haramiyids with lower and upper molariforms showing low cusps with more extended base and more massive aspect, short and narrow basins in relation to the size of the crown, presence of only two cusps in row a on lower molariforms (shared with some specimens of Thomasia), presence of a row BB on upper molariforms (potentially shared with Eleutherodon, Megaconus and Millsodon-see below), and an essentially vertical, masticatory movement.     Table 1.

Lower molariforms
The crown is dominated by two longitudinal rows of cusps which delimit a central basin, the lingual row a and the labial row b. The central basin is confined mesially by the saddle which joins cusps a1 and b2, and distally by the u-ridge which joins rows a and b. The central basin gets deeper and narrower from the mesial extremity to the distal extremity.
Cusp row a includes two cusps. Cusp a1 is the largest one of the teeth and rises vertically in lateral view. This cusp extends over the mesial half of the tooth. Cusp a1 shows a weak mesial carina which splits into two segments. One segment runs mesially and the other bends labially to join cusp b1. At the level of the base of cusp b1, the mesial segment turns  into a short, horizontal cingulum to join cusp b1. A distal crest starts from the distolabial side of the apex of cusp a1 to join cusp a2. This crest is straight in lateral view, but it is curved labially in occlusal view. A sulcus underlines the lingual side of this crest and descends to the base of cusp a1. A second crest, straight in occlusal and lateral views, starts from the labial side of the apex of cusp a1 to the base, where it takes part in the saddle. The distal and labial crests delimit a concave, narrow surface on the distolabial flank of cusp a1, which extends from the apex to the central basin. Cusp a2 is twice lower and labiolingually narrower, and much mesiodistally shorter than cusp a1. Cusp a2 is more lingual than cusp a1. The lingual flanks of cusps a1 and a2 are aligned and parallel to the mesiodistal axis of the tooth on MNHN.F.SNP 61 W, but deviate distolabially on MNHN.F. SNP 366 W, and MNHN.F.SNP 787. The occlusal outline of cusp a2 is semicircular with a convex, lingual side and nearly flat, labial side. The labial side shows a vertical, weak ridge in the middle. In distal view, the slope of the labial flank is more vertical than the slope of the lingual flank. The latter is slightly convex. In labial view, the mesial base of cusp a2 is higher than the distal base of the cusp. In lingual view, the bases of cusps a1 and a2 are at the same level. Cusp a2 shows two crests, respectively mesial and distal, straight in lateral and occlusal views, and aligned mesiodistally. The first crest starts from Table 3 Statistical comparisons of the means of lower molariforms of Theroteinus nikolai and Theroteinus rosieriensis from Saint-Nicolas-de-Port (Upper Triassic, France) by t-test. Normality of the data has been tested by Shapiro-Wilk Test (Table S2), p-values have been corrected by Holm-Bonferroni method, the alternative hypothesis is "true difference in means is not equal to 0."  the mesiolabial side of the apex to join the distolingual crest of cusp a1. The second crest starts from the distolabial side of the apex to the extremity of row a. The distal crest is much longer than the mesial crest. The slope of the mesial crest of cusp a2 is weaker than the slope of the distal crest of cusp a1 and the slope of the distal crest of cusp a2.

Measurements tested Value of the test (t) 95% confidence interval p-value Adjusted p-value
The slope of the latter is more vertical than the slope of distal crest of cusp a1. Cusp row b includes four cusps, less distinguished from each other than the cusps of row a. Cusp b1 is the most mesial of the tooth. This cusp is subequal in size with cusp b4, or larger in MNHN.F.SNP 61 W. Cusp b1 is located in front of the saddle, but tends to rise lingually to join the mesiolabial carina of cusp a1. Cusp b2 is the largest cusps of row b. This cusp is slightly smaller than cusp a1, except in MNHN.F.SNP 366 W where cusp b2 is slightly larger and higher than cusp a1. Cusp b2 is labial to cusp a1, its base extends as mesially but much less distally, and its apex is slightly more distal, or much more distal in MNHN.F.SNP 61 W. Cusp b2 shows two crests, straight in occlusal and lateral views. The first crest runs labially, but mesiolabially in MNHN.F.SNP 61 W, to take part in the saddle. The second crest runs distally and joins cusp b3. Both crests define on the one side a slightly convex, distolingual occlusal outline, and on the other side a large arc of a circle. Cusp b3 is much smaller than cusps a1, a2, and b2 and slightly smaller than cusps b1 and b4. Cusp b3 is directly distal to cusp b2, except in MNHN.F.SNP 787 where it is slightly more labial. This cusp is more mesial than a1-a2 notch in MNHN.F.SNP 787, aligned with a1-a2 notch in MNHN.F.SNP 366 W, and more distal than a1-a2 notch in MNHN.F.SNP 61 W. The apex of cusp b3 is slightly higher than the apex of cusp a2, or at the same level in MNHN.F.SNP 61 W. The long axis of cusp b3 slightly deviates distolingually from the mesiodistal axis of the tooth, except in MNHN.F.SNP 787 where both axes are parallel. Cusp b4 is distal to cusp b3 and slightly more lingual. Consequently, the long axes of both cusps are aligned, except in MNHN.F.SNP 787. The apex of cusp b4 takes place slightly lower than the apex of cusp a2 and faces the distal crest of cusp a2. Cusp b4 shows a lingual carina, which is well developed in MNHN.F.SNP 787. The u-ridge is a low crest which extends row b and bends lingually to join the extremity of row a.  1C) and MNHN.F.SNP 497 W is heavily worn. For these reasons, they have not been incorporated in the description above. About MNHN.F.SNP 497 W, it may be noticed that row b is less developed than in other specimens, with a strong reduction of cusp b4 (Figs. 2C and 3D). In the absence of clear morphological characters, both specimens are referred to Theroteinus nikolai following morphometry (see Comparisons. Identification of Theroteinus species below).

Upper molariforms
The crown is dominated by three longitudinal rows of cusps: labial row A, central row B and lingual row BB. Rows A and B define a labial basin delimited distally by the saddle, constituted only by the lingual crest of cusp A2, and mesially by the u-ridge which joins rows A and B. Rows B and BB define a lingual basin, smaller than the labial basin, delimited distally by the meeting of cusps B2 and BB1, and mesially by the crest which joins rows B and BB. Both basins get deeper and larger mesially.
Row A includes three mesiodistally aligned cusps. The three cusps take place at the same level on the crown. Cusps A1 and A3 are subequal in length and width, cusp A1 is slightly higher than cusp A3. Cusp A2 is twice mesiodistally longer and higher, and labiolingually much wider than cusps A1 and A3. The occlusal outlines of cusps A1 and A3 show a semicircular, labial flank and a relatively flat, lingual flank, sometimes concave because of wear. Cusp A1 shows two crests, straight in occlusal and lateral views. The longest crest runs distolingually from the apex to cusp B1. The other crest runs mesially to cusp A2. The slopes of these crests are subequal. Cusp A3 shows two crests, straight in occlusal and lateral views. The longest crest runs mesiolingually from the apex to take part in the u-ridge. The other crest runs distally to cusp A2. The slopes of these crests are subequal. Cusp A2 shows three crests, straight in occlusal and lateral views. The first crest runs distally to cusp A1. The second crest runs mesially to cusp A3. The third crest runs lingually but does not join another structure. The distal crest is the shortest and shows the strongest slope. The lingual crest is much wider than both other crests. The lingual and mesial crests define a flat surface on the mesiolingual flank of cusp A2. A1-A2 and A2-A3 notches are equal in depth, but A1-A2 notch takes place higher than A2-A3 notch.
A small, supplementary cusp takes place under the labial flank of cusp A3. Row B includes five cusps. Cusp B1 looks like a distally curved semicircle. This cusp is slightly smaller in length and width than cusp A1, but much smaller in height. Cusp B1 is more distal and more lingual than cusp A1 and is mesiodistally aligned with the saddle. Cusp B2 is slightly smaller than cusp A2. Cusp B2 is much more lingual than cusp B1 and is labiolingually aligned with the A1-A2 notch. This cusp is cone-shaped and does not show any crest. Three small cuspules take place at the base of the mesiolingual flank of cusp B2. Cusp B3 is directly mesial to cusp B2. The apex of cusp B3 is slightly more labial the apex of cusp B2 and slightly more mesial than apex of cusp A2. Cusp B3 is subequal in size with cusp B1 and take place slightly lower than cusp B2. Cusp B4 is directly mesial to cusp B3. This cusp is smaller in all dimensions and takes place lower than cusp B3. Cusp B4 is labiolingually aligned with A2-A3 notch. Cusp B5 is the most mesial cusp of the tooth. This cusp is smaller in all dimensions, takes place lower, and is slightly more labial than cusp B4. The mesial extremity of cusp B5 shows two crests. One crest runs labially to take part in the u-ridge of the labial basin. The other crest runs lingually to mesially close the lingual basin.
Row BB includes two cusps. Cusp BB1 is sub-equal in size to cusp B4 and takes place at the same level. Cusp BB1 is placed right next to cusps B2 and B3, directly lingual to the B2-B3 notch. Cusp BB2 is mesial to cusp BB1 but slightly more lingual. This cusp is subequal in length and width with cusp B5, but slightly higher. A cusp BB3 was possibly present, but this part of the crown is broken. Row BB is extended by a crest which runs mesially and then bends labially to close the lingual basin.

Comments on MNHN.F.SNP 722
Only the distal part of MNHN.F.SNP 722 is preserved, with cusps A1, B1, B2, BB1, and a part of cusps A2 and B3 (Fig. 1B). Since morphometry is not applicable, this specimen is referred to Theroteinus nikolai following the position of cusp B2 in relation to cusps B3 and BB1. However, some doubts remain because cusp B3 is fragmentary. Etymology: rosieri-: a Latinized form of 'Rosières' from 'Rosières-aux-Salines' another name used for the study site; -ensis; suffix added to a toponym to form an adjective.

Lower molariforms
The crown is dominated by two longitudinal rows of cusps which delimit a central basin, the lingual row a and the labial row b. The central basin is confined mesially by the saddle which joins cusps a1 and b2 and distally by the u-ridge which joins rows a and b. The saddle is very high compared with the u-ridge, except in MNHN.F.SNP 309 W where the difference is weaker. The central basin gets deeper and narrower from the mesial extremity to the distal extremity. Row a includes two cusps. Cusp a1 is the largest cusp of the tooth and rises vertically in lateral view. This cusp extends over the mesial half of the tooth, even more in MNHN.F.    Table 1). and occlusal views, and aligned mesiodistally. The first crest starts from the mesiolabial side of the apex to join the distolingual crest of cusp a1. The second crest starts from the distolabial side of the apex to the extremity of row a. The distal crest is much longer than the mesial crest. The slope of the mesial crest of cusp a2 is weaker than the slope of the distal crest of cusp a1 and the slope of the distal crest of cusp a2. The slope of the latter is more vertical than the slope of the distal crest of cusp a1. These crests are not preserved in MNHN.F.SNP 487 W and RBINS.RAS 74 FW. Row b includes four cusps, less distinguished from each other than the cusps of row a. Cusp b1 is the most mesial of the tooth. This cusp is subequal in high and mesiodistal length with cusp a2 but wider and more voluminous. Cusp b1 is located in front of the saddle, but tends to rise lingually to join the mesiolabial carina of cusp a1. Cusp b2 is the largest cusps of row b. This cusp is slightly smaller than cusp a1, except in MNHN.F. SNP 309 W and RBINS.RAS 62 FW where it is much smaller but still larger than other cusps. Cusp b2 is labial to cusp a1, its base extends less mesially and distally, and its apex is slightly more distal, or much more distal in MNHN.F.SNP 309 W. Cusp b2 shows two crests, straight in occlusal and lateral views. The first crest runs labially to take part in the saddle. The second crest runs distally and joins cusp b3. Both crests define on the one side a slightly convex, distolingual occlusal outline, and on the other side a large arc of a circle. This part of the crown has been removed by wear in MNHN.F.SNP 487 W. Cusp b3 is much smaller than cusps a1, a2, b1, and b2 and slightly smaller than cusp b4. Cusp b3 is distal and slightly lingual to cusp b2. This cusp is labiolingually aligned with the a1-a2 notch. The base of cusp b3 is slightly lower than the base of cusp a2. The long axis of cusp b3 deviates slightly distolingually from the mesiodistal axis of the tooth. This part of the crown has been removed by wear in RBINS.RAS 74 FW and RBINS.RAS 77 FW. Cusp b4 is distal to cusp b3 and slightly more lingual. Cusp b4 is mesiodistally aligned with the saddle and labiolingually aligned with the distal crest of cusp a2. The base of cusp b4 is slightly lower than the base of cusp b3. In MNHN.F.SNP 487 W, RBINS.RAS 62 FW, and RBINS.RAS 77 FW, a low crest extends row b and bends lingually to join the extremity of row a. In MNHN.F.SNP 309 W and RBINS.RAS 74 FW, this crest splits into two segments. The first segment bends lingually to join the extremity of row a. The second segment bends labially and runs down the side of the crown and turns into a thin bulge which extends into the base of cusp b2.

Comments on RBINS.RAS 800
The occlusal surface of RBINS.RAS 800 is not well preserved. As a consequence, the cusps are difficult to describe. For these reasons, this specimen has not been included in the description above. In the absence of clear morphological characters, this specimen is referred to Theroteinus rosieriensis following morphometry (see Comparisons. Identification of Theroteinus species below).

Upper molariforms
The crown is dominated by three longitudinal rows of cusps: labial row A, central row B and lingual row BB. Rows A and B define a labial basin delimited distally by the saddle, constituted only by the lingual crest of cusp A2, and mesially by the u-ridge which joins rows A and B. Rows B and BB define a lingual basin, smaller than the labial basin, delimited distally by the meeting of cusps B2 and BB1, and mesially by the crest which joins rows B and BB. Both basins get deeper and larger mesially. The lingual basin is very shallow in MNHN.F.SNP 335 W.
Row A includes three cusps. In MNHN.F.SNP 335 W, cusp A3 is slightly more labial than cusps A1 and A2. In MNHN.F.SNP 2 Ma, cusp A1 is more lingual than cusp A2 and cusp A3 is more labial than cusp A2. The three cusps are located at the same level on the crown. Cusps A1 and A3 are subequal in height and width, cusps A1 is slightly longer than cusp A3. Cusp A2 is twice mesiodistally longer and higher, and much labiolingually wider than cusps A1 and A3. In MNHN.F.SNP 2 Ma, cusps A1 and A3 are less wide than cusp A2. In occlusal view, cusps A1 and A3 show a semicircular labial flank and a relatively flat, lingual flank. Cusp A1 shows two crests, straight in occlusal and lateral views. The longest crest runs distolingually from the apex to cusp B1. The other crest runs mesially to cusp A2. In MNHN.F.SNP 2 Ma, the mesial crest is present but cusp A1 shows a flat side in front of cusp B1. Cusp A3 shows two crests, straight in occlusal and lateral views. The longest crest runs mesiolingually from the apex to take part in the u-ridge. The other crest runs distally to cusp A2. The slope of the distal crest is more vertical than the slope of the mesial crest. Cusp A2 shows three crests, straight in occlusal and lateral views. The first crest runs distally to cusp A1. The second crest runs mesially to cusp A3. The third crest runs lingually but does not join another structure. The distal crest is the shortest, and the slopes of the three crests are subequal. The lingual crest is much wider than both other crests. The lingual and mesial crests define a concave surface on the mesiolingual flank of cusp A2. A1-A2 notch is less deep and takes place higher than A2-A3 notch.
Row B includes four cusps. Cusp B1 is subequal in size with cusp A1 in MNHN.F.SNP 2 Ma, but smaller in MNHN.F.SNP 335 W. Cusp B1 is more distal and more lingual than cusp A1 and is mesiodistally aligned with the saddle. Cusp B2 is slightly smaller than cusp A2. Cusp B2 is much more lingual than cusp B1 and is labiolingually aligned with the A1-A2 notch. This cusp is cone-shaped and does not show any crest. One small cuspule takes place at the base of the mesiolingual flank of cusp B2. B2-B3 notch is labiolingually aligned with cusp A2. Cusp B3 is more labial than cusp B2 and slightly more lingual than cusp B1 (MNHN.F.SNP 2 Ma) or mesiodistally aligned to cusp B1 (MNHN.F. SNP 335 W). Cusp B3 is much smaller than cusps A2 and B2 and slightly larger than cusps A1, A3, and B1 (MNHN.F.SNP 2 Ma) or subequal with cusps A1 and A3 (MNHN.F.SNP 335 W). In MNHN.F.SNP 2 Ma, cusp B3 is wider than long. Cusp B3 takes place slightly lower than cusp B2. B3-B4 notch is labiolingually aligned with the A2-A3 notch. Cusp B4 is directly mesial to cusp B3. This cusp is smaller in all dimensions and located lower than cusp B3. In MNHN.F.SNP 335 W, a cusp B5 was potentially present but is now removed by wear. The mesial extremity of row B shows two crests. One crest runs labially to take part in the u-ridge of the labial basin. The other crest runs lingually and closes the lingual basin at the mesial side.
Row BB includes three cusps in MNHN.F.SNP 2 Ma. In MNHN.F.SNP 335 W, the cusps cannot be described because of the wear. Cusp BB1 is sub-equal in size to cusp B4 and takes place at the same level. Cusp BB1 is placed right next to cusps B2 and B3, slightly more mesial than the B2-B3 notch. Cusp BB2 is mesial to cusp BB1 but slightly more lingual. This cusp is smaller and takes place lower than cusp BB1. Cusp BB3 is the mesialmost of the tooth. This cusp is mesiodistally aligned with cusp BB1. A crest extends row BB and runs labially to mesially close the lingual basin.

Comments on MNHN.F.SNP 14 FW
Only the distal part of MNHN.F.SNP 14 FW is preserved, with cusps A1, B1, B2, BB1, and a part of cusps A2 and B3. Since morphometry is not applicable, this specimen is referred to Theroteinus rosieriensis following the position of cusp B2 in relation to cusps B3 and BB1. MNHN.F.SNP 14 FW differs from other teeth described above by a less developed cusp A1 and more developed cusp B3.

Lower molariforms
In MNHN.F.SNP 309 W, RBINS.RAS 62 FW, RBINS.RAS 74 FW, and RBINS.RAS 77 FW, all cusps are abraded by wear. The labial side of row b shows a large, concave surface of wear which extends from the distal extremity of cusp b2 to cusp b4. It is difficult to say if this concavity was present before the wear occurred or not, but it shows traces of wear, like the sides of the basin. MNHN.F.SNP 487 W also shows wear on the entire surface of the tooth but several facets are present. Cusp a1 shows a steep, distolabial facet. Cusp a2 shows a steep, distal facet on its apex connected with a steep, distolingual facet on its lingual side. Cusp b1 shows a horizontal facet. Cusp b2 is partially truncated by a concave, shallow, labio-distolabial facet, which extends on cusp b4. The apex of cusp b4 shows a horizontal, distal facet.

Upper molariforms
In MNHN.F.SNP 2 Ma, only the apices of the cusps are abraded by wear. In MNHN.F.SNP 335 W, the cusps are more strongly abraded and show several facets. Cusp A1 shows a steep, distal facet. Cusp A3 shows possibly a steep, mesial facet. Cusp B2 shows a shallow, mesiolabial facet. Other cusps of row B show one steep, mesial facet. Row BB shows one steep, mesial facet. In MNHN.F.SNP 14 FW, the wear seems more considerable. Cusp A1 shows a shallow, distal facet. Cusp A2 shows a large, horizontal, labial facet. Cusp B1 shows a horizontal distal facet. Cusp B2 shows a horizontal facet. Cusp B3 possibly shows a shallow, mesial facet but is partially broken. Cusp BB1 shows a shallow, mesiolabial facet.

Reconstruction of the dental row of Theroteinus
In such a poorly known group such as Haramiyida, the reconstruction of the dental rows from isolated teeth is notoriously difficult. Although five genera with complete or partial dentitions have been discovered in the last twenty years (Jenkins et al., 1997;Zheng et al., 2013;Zhou et al., 2013;Bi et al., 2014), there is no comparative study to provide elements on interspecific and ontogenetic variations. The reconstruction of the dental row of Theroteinus is complicated by two additional problems: (i) the small number of specimens (n = 20) which prevents to evaluate the intraspecific variations, and (ii) the absence of premolariform specimens.
In upper molariforms, the variations of the development of cusps A1 and B1 and of the number of elements on the distolingual side of cusp B2 can be related to the tooth position but also to individual or ontogenetic variations.
In lower molariforms, three specimens show characters possibly related to tooth position. MNHN.F.SNP 61 W has cusp b2 more distal in comparison with cusp a1 than other specimens. The first molar of Haramiyavia shows a similar character which may be a clue for a more mesial position in the dental row (Jenkins et al., 1997;Luo et al., 2015). MNHN.F.SNP 487 W has row b less high than in other specimens, cusp b2 is especially much smaller in comparison with cusp a1. This difference of height is present in premolariforms of some haramiyids such as Thomasia as well, and it may consequently be a clue for a more mesial position in the dental row. However, MNHN.F.SNP 487 W does not show the distal shift of cusp b2 seen in MNHN.F.SNP 61 W and the difference of height may also be related to ontogenetic variations. MNHN.F.SNP 497 W shows a distally reduced row b, especially cusp b4. Since this specimen does not show characters of the two other specimens, this reduction of row b may be a clue for the last locus in the dental row. Indeed, this locus displays often a partial reduction of the crown in other groups of mammaliaforms (see e.g., Debuysschere, 2016). Since the reduction of row b could modify the occlusal function of the tooth, this interpretation may imply either that the last upper locus displays an equivalent reduction, or that this part of the tooth does not occlude with opposite teeth (i.e., a more mesial position of the last upper locus).

COMPARISONS Identification of Theroteinus species
The reappraisal of Theroteinus nikolai and the erection of Theroteinus rosieriensis sp. nov. are based on morphometric and morphologic characters.

Morphometry
Measurements of the Theroteinus material are presented in Table 1 and descriptive statistics in Table 2. Because of the small number of upper molariforms (n = 4), no statistical test can be made to support this discussion. Statistical tests are possible on lower molariforms, but they need to be interpreted with caution because of the limited number of specimens available (n = 12). Means have been compared by the Welch's t-test, which is a variant of the Student's t-test (command`t.test()' in R software). This test assumes that data are normally distributed. This hypothesis has been tested by the Shapiro-Wilk test (command`shapiro.test()' in R software), without rejection of the null hypothesis (Table S2). Since there are multiple comparisons, p-values have been corrected by the Holm-Bonferroni method (command`p.adjust(method = "holm")' in R software) in order to control the family-wise error rate. The results of the t-test are presented in Table 3. Graphically, the Fig. 7A shows two sets of upper teeth which do not overlap either by length or by width, and two sets of lower teeth which slightly overlap. Figure 7B shows that the same sets are present in length/width ratio, but with a more considerable overlapping between sets of lower teeth. Since differences of means in length, width, and length/width ratio are statistically significant (Table 3), specimens are divided between elongated small teeth and stocky large teeth.

Morphology
In upper molariforms, two sets can be defined by the position of cusp B2 which is either mesiodistally aligned with cusps B3 and B4, or lingually shifted to face the lingual basin. In lower molariforms, two sets can be defined by the position of cusp b4 which is either aligned with row b, or lingually shifted to face the saddle. Both of these variations are related together, because of the occlusal pattern. Indeed, cusp B2 occludes lingually to cusp b4, consequently the latter cannot be shifted lingually if the former is not shifted either. Table 4 presents other morphological differences between sets defined above. However, in the current stage of knowledge, it is difficult to say if these differences are related to taxonomic, ontogenetic or individual variations.
Evaluating if observed differences are individual, populational, ontogenetic, sexual, or systematic variation is a hard question; especially with haramiyids for which these different variations, except systematic, are terra incognita for now. Such a question will be addressed when an adequate material is discovered. In this framework, it is more careful and parsimonious for the taxonomy to consider the most important observed differences as systematic variations, as do other studies (e.g., Bi et al., 2014).
The sets defined by morphological and morphometric characters perfectly match. The lower and upper molariforms are associated following characters presented above and the two sets are considered as two species of the genus Theroteinus. The set including MNHN. F.SNP 78 W (defined as holotype by Sigogneau-Russell, Frank & Hemmerlé, 1986) is identified as Theroteinus nikolai and the second set is identified as Theroteinus rosieriensis sp. nov. Since the new hypodigm of T. nikolai includes all specimens referred to Theroteinus sp. described by Hahn, Sigogneau-Russell & Wouters (1989), they are attributed to Theroteinus nikolai.

Comparisons with other haramiyids
Theroteinus differs from all other known haramiyids by low and massive cusps, separated by very shallow notches and by short and narrow basins in comparison with the size of the tooth. This genus differs also by a small number of cusps in each row, especially only two cusps in row a (character seen only in some specimens of Thomasia).
Theroteinus possibly shares the presence of a supplementary upper lingual row BB with Eleutherodon Kermack et al., 1998 (Middle Jurassic, England), Megaconus (Middle Jurassic, China), and Millsodon Butler & Hooker, 2005 (Middle Jurassic, England). However, recognizing this similarity depends on the different interpretations of the specimens concerned, especially on the orientation of the teeth.
Following the orientation of upper molariforms of Eleutherodon proposed by Kermack et al. (1998);Butler (2000) named the labial row A, the middle row B, and the lingual row BB, which corresponds to the pattern of Theroteinus. However, Meng et al. (2014: p. 29) proposed a second interpretation based on the comparison of the wear pattern of Eleutherodon with the wear pattern of Arboroharamiya. In this second interpretation, the labiolingual axis is inverted (Meng et al., 2014: Fig. 13). Although Meng et al. (2014) did not explicitly explain how they named the rows, it seems that they considered row A of Butler (2000) as row B, row B as row A and row BB as supplementary elements on the labial side of the tooth. According to the interpretation of Meng et al. (2014), Eleutherodon does not share the presence of row BB with Theroteinus. Zhou et al. (2013) did not name the rows of cusps of upper teeth of Megaconus. However, since the ultimate tooth shows only two rows, it is more parsimonious to consider these rows as rows A and B, which implies that the third lingual row present in the two previous teeth would be a row BB. This interpretation is consistent with the few published comments on the occlusion of Megaconus such as that of Zhou et al. (2013: Supplementary Information: p. 6): '[l]ower molars have two multicusp rows that alternately occlude in the two valleys between the three rows of cusps of M1 and M2.' However, Meng et al. (2014) questioned the orientation of the upper dentition of Megaconus. They proposed a reversal of the labiolingual axis and seemed to consider the labial row as a row AA (Meng et al., 2014: Fig. 14). It is difficult to decide between both interpretations upon the available data. It must be emphasized that both orientations are given with few details on the definition of rows and on the relationships between them. In this framework, several interpretations are possible, which prevents to conclude on the presence of row BB in Megaconus molariforms. Table 4 Summary of differences between lower and upper molariforms of Theroteinus nikolai and Theroteinus rosieriensis which are not included in diagnoses.

T. nikolai T. rosieriensis
Lower teeth -a vertical, weak medial ridge in the middle of the labial side of cusp a2 -alignment of the long axes of cusps b3 and b4 -a lingual carina on cusp b4 -a very high saddle -cusp a2 twice times smaller than cusp a1 -sub-equality of the slopes of lingual and labial sides of cusp a2 -a mesiodistally less extended base of cusp b2 -cusp b3 much smaller than cusp b1 -cusp b3 slightly lingual to cusp b2 Upper teeth -mesiodistal alignement of three cusps A -a small cusp under the labial side of cusp A3 -a distally curved semi-circular cusp B1 -cusp A1 slightly longer than cusp A3 -a concave surface on the mesiolingual side of cusp A2 -four cusps in row B -cusp B1 mesiodisally aligned with the saddle -cusp B3 more labial than cusp B2 -three cusp in row BB -cusp BB1 slightly more mesial than B2-B3 notch The comparison with Millsodon is based on a specimen BDUC J 3, which is considered as a probable upper molar by Butler & Hooker (2005: p. 192). If this interpretation is accepted, this specimen shows a row BB as Theroteinus, but it differs strongly from the latter by relationships of size and position between other cusps. Indeed, the pattern of cusps is very peculiar to an upper molariform and the two published interpretations of the specimen are very different from each other (Butler & Hooker, 2005contra Hahn & Hahn, 2006 (Figs. 8A and 8B). The sole argument supporting the interpretation of BDUC J 3 as an upper molariform is the presence of a third cusps row. The rest of the crown looks more like a lower molariform, and can be described as follows: (i) a first row of cusps including a cusp much larger than others, (ii) a second row of cusps, which are similar in size with small cusps of the first row, (iii) in the second row the largest cusp is close to the large cusp of the first row but not labiolingually aligned with it, and (iv) a cusp located at one extremity of the tooth, aligned with the second row, but separated from it by the large cusp. No one other haramiyid upper tooth matches this pattern, unlike lower molariforms of Thomasia and Haramiyavia. This new interpretation of specimen BDUC J 3 as a lower molariform (Fig. 8C) is favoured here and implies two consequences. First, the referral of specimen BDUC J 3 to the genus Millsodon needs a reassessment. Second, the specimen can be compared with the lower molariforms of Theroteinus. Several characters are shared by these teeth: (i) the presence of few cusps by row, (ii) low and obtuse cusps, (iii) and a short and narrow basin. Moreover, the third row of BDUC J 3, which is labial in this interpretation, can be considered as development of the labial bulge present in some specimens of Theroteinus (Figs. 4B, 4C and 5). However, new examinations of BDUC J 3 would be necessary to discuss further these points, which is impossible as this specimen is lost according to Butler & Hooker (2005: p. 191).

DISCUSSION
Theroteinus is referred to Haramiyida because of the presence of parallel rows of cusps. Moreover, its molariforms show a pattern of cusp rows a/A and b/B, in size and relative position of cusps, which is strongly similar to patterns seen in Thomasia and Haramiyavia. In addition, the occlusal pattern of Theroteinus is similar to the pattern of Thomasia with row B occluding into the lower basin. However, Theroteinus is very peculiar among haramiyids. The genus is defined by characteristic morphological characters (see above) and by a different masticatory movement. Indeed, Theroteinus is the only haramiyid for which the wear pattern does not highlight a horizontal movement of the jaw during mastication (Sigogneau-Russell, Frank & Hemmerlé, 1986 and see above). Such a wear pattern and the small size of the basins support an essentially vertical masticatory movement. Because of these differences, Theroteinus has occupied since a long time a special place in the systematics of haramiyids, either as sister-group of the whole order Haramiyida (Hahn, Sigogneau-Russell & Wouters, 1989) or isolated in a sub-order (Butler, 2000;Hahn & Hahn, 2006). In the absence of a relevant cladistics analysis including Theroteinus, the suborder Theroteinida is conservatively used in order not to complicate the taxonomy of haramiyids, which has already seen many changes. In the same purpose, the name 'Theroteinida' is used unchanged although it would be best to modify it. As underlined by Hahn & Hahn (2006: p. 189), the suffix of the name of a sub-order should be different from the suffix of the name of the including order. However, the suffix '-ina' suggested by Hahn & Hahn (2006: p. 189) cannot be used since it is reserved for the name of a subtribe by article 29.2 of the ICZN (International Commission on Zoological Nomenclature, 2000). The only taxon closely related to Theroteinus is Millsodon, which is considered as a member of Theroteinidae by Hahn & Hahn (2006). Butler & Hooker (2005: p. 192) compared the upper tooth of Millsodon with the upper molariforms of Theroteinus and suggested that Millsodon could be 'a derivative of the Theroteinidae or a specialized relative of the Haramiyidae.' However, Butler & Hooker (2005) considered Millsodon as indeterminate at familial rank and did not compare its lower molariforms with lower molariforms of Theroteinus. Hahn & Hahn (2006) considered that lower molariforms of Millsodon can be derived from lower molariforms of Theroteinus. This interpretation is based on specimen MNHN.F.SNP 226 W. Hahn & Hahn (2006: p. 184) considered that differences between this specimen and other lower molariforms of Theroteinus cannot be explained only by wear and that this specimen represents 'a new taxonomical unit (perhaps a genus and a species)' and is intermediate between Theroteinus and Millsodon. This interpretation is questionable. First, the specimen MNHN.F.SNP 226 W is very poorly preserved, not only because of wear during life but also probably because of taphonomic processes. Creating a new genus and outlining an evolutionary scenario only on the base of such a poorly preserved specimen is problematic. Second, comparisons between Theroteinus and Millsodon are difficult. On the one hand, the description of the upper tooth of Millsodon is questionable (see above). On the other hand, all lower teeth of Millsodon are heavily worn (e.g., Butler & Hooker, 2005: Figs. 1D and 1E), and the cusps are difficult to characterize. However, all specimens of Millsodon clearly show a well-developed basin, which is distinct from Theroteinus. Consequently, the family Theroteinidae is considered here as monogeneric.