Miguelsoria parayirunhor (Paula-Couto, 1952)

Billet, Guillaume, Muizon, Christian De, Schellhorn, Rico, Ruf, Irina, Ladevèze, Sandrine & Bergqvist, Lilian, 2015, Petrosal and inner ear anatomy and allometry amongst specimens referred to Litopterna (Placentalia), Zoological Journal of the Linnean Society 173 (4), pp. 956-987 : 961-974

CF. MIGUELSORIA PARAYIRUNHOR (PAULA- COUTO, 1952)

Referred specimens

UFRJ-DG 119-M, left isolated petrosal; UFRJ-DG 275-M, left isolated petrosal; UFRJ-DG 347-M, right isolated petrosal; UFRJ-DG 1035-M, left isolated petrosal; UFRJ-DG 1036-M, right isolated petrosal; UFRJ-DG 1037-M and UFRJ-DG 1038-M, incomplete isolated left petrosals (pars cochlearis broken) ( Fig. 2 ). Collected in 1948 or 1949. See the details for the referral to Miguelsoria in the section ‘Arguments for the referral to Miguelsoria ’ below after the anatomical description.

Locality and age

Late Palaeocene or early Eocene Itaboraí Basin, state of Rio de Janeiro, Brazil ( Gelfo et al., 2009; Bergqvist et al., 2011), northern and south-eastern parts of the basin.

Comments

Seven specimens were placed in this group, two of which tend to slightly depart morphologically from the others in various features. Given their overall resemblances however, we consider that these observed differences do not warrant referral to different species with the current state of our knowledge. We refer them all to cf. Miguelsoria parayirunhor ( Paula-Couto, 1952) in agreement with the phylogenetic, size, and abundance arguments detailed below. We suggest that the slight heterogeneity observed amongst the seven specimens may lie on their possible provenience from various white-fossils fissures (WFF; see below and Bergqvist, 2008), as the latter may be slightly asynchronous.

External anatomy

In this section, the illustrations refer mainly to UFRJ- DG 1036-M ( Figs 3–5 ), which exhibits well the general morphology present in the sample; other specimens are illustrated to depict important anatomical differences (e.g. vascular grooves on the squamosal face, Fig. 6 ).

Tympanic surface

The promontorium surface is weakly inflated and rounded in the vicinity of the vestibular fenestra and external aperture of the cochlear fossula ( Figs 3 , 4 ). This low relief corresponds internally to the position of the proximal part of the cochlear canal. Medially the promontorium is first bordered by a broad longitudinal crest. This crest is consistently present posteromedially but variably distinct anteriorly amongst the UFRJ-DG specimens under study. Its location on the ventral surface of the pars cochlearis led us to identify it as a rostral tympanic process ( MacPhee, 1981). Directly medial to this process is a thickened flange of the promontorium that exhibits anteriorly another crest, parallel and adjacent to the rostral tympanic process. It is distinctly more salient than the latter and forms the ventrolateral edge of a large longitudinal sulcus for the inferior petrosal sinus ( MacPhee, 1981). In addition to the salient crest laterally, this sulcus is also delimited by another sharp and long crest, which is more mediodorsally located and separates it from the cerebellar face of the petrosal ( Fig. 3 ). The medial surface of the petrosal is therefore mostly composed of these salient parallel crests: the rostral tympanic process, and the lateroventral and mediodorsal crest-like borders of the sulcus of the inferior petrosal sinus; there is no identifiable epitympanic wing and medial flange of the petrosal (sensu Wible et al., 2009).

The promontorium surface forms a rectangle, whose major axis is orientated anteroposteriorly ( Fig. 4A ). This rectangle is moderately elongated, its length being just 1.2–1.4 times its width (see measurements of promontorium in Table S1). The anterior apex of the promontorium appears truncated, when not broken, on the examined specimens. On all specimens, the promontorium and the tympanic surface of the petrosal as a whole largely parallel the ventral half of the squamosal (lateral) face and especially the extensive lateral face of the tegmen tympani (because the latter is not inflated; see below) ( Figs 3 , 4 ).

Two specimens (UFRJ-DG 1035-M and 1036-M) show distinct vascular sulci on the promontorium; others just exhibit a distinct single notch at the ventromedial edge of the fenestra vestibuli, which is presumably for the stapedial artery (e.g. see Wible, 2012: fig. 14). On UFRJ- DG 1035-M and 1036-M, this notch is continuous with a short transverse sulcus for the stapedial artery, which connects medially with a long sulcus ( Fig. 4 ). This sulcus has a curved course; from its posterior-most extremity at the posteromedial corner of the promontorium, close to the external aperture of the cochlear fossula, it first runs anterolaterally but gently turns more anteriorly at its junction with the sulcus of the stapedial artery in the vicinity of the fenestra vestibuli. Its subsequent anterior course is long and directed toward the anterior apex of the promontorium. This sulcus is viewed here as the impression left on the bone by a transpromontorial course of the internal carotid artery ( Wible, 1987).

The hiatus Fallopii opens at the anterior edge of the tympanic face. Posteriorly, the floor of the cavum supracochleare (which encloses the geniculate ganglion of the facial nerve, Voit, 1909) is partially broken on most specimens. Medioventral to the location of the geniculate ganglion, and just anterior to the fenestra vestibuli is an elongated fossa on the tympanic surface of the promontorium, which most probably received a part of the tensor tympani muscle. The tensor tympani fossa is large, and rectangular in outline with an anteroposterior elongation. It is delimited by sharp edges posteriorly and medially. The salient posterior edge of this fossa corresponds to the anterior edge of the fenestra vestibuli.

The primary facial foramen is large and opens within the anterior half of the cavum supracochleare, close to the hiatus Fallopii. The secondary facial foramen opens lateral and slightly anterior to the fenestra vestibuli anterior edge (only visible in UFRJ-DG 1038-M and 275-M). It gives rise posteriorly to a wide facial sulcus for the facial nerve ( MacPhee, 1981). The crista parotica, which borders the facial sulcus laterally, has a straight course that defines most of the ventrolateral edge of the petrosal.

The tegmen tympani (as conceived in Billet & de Muizon, 2013) exclusively consists of an anterolateral massive process that extends from the anterior edge of the lateral face of the petrosal. This process is most inflated at its dorsolateral root and is slightly tapered ventrally. Ventromedially it contacts the anterolateralmost part of the floor of the cavum supracochleare ( Fig. 4A ). Together the tegmen tympani process and the anterolateral part of the floor of the cavum build up a bony arch through which a large canal exits the middle ear cavity ( Fig. 3A ). The tympanic opening of this canal is close anterolaterally to the secondary facial foramen and cavum supracochleare ( Fig. 3B ). Its course runs slightly oblique relative to the orientation of the cavum supracochleare on all specimens, forming an angle of approximately 45° with it. This canal through the tegmen tympani is likely to have provided a passageway for the ramus superior of the stapedial artery (e.g. Wible et al., 2001; Ladevèze et al., 2010) and is therefore named here the canal for the ramus superior. This term is regarded here as syn- onymous with the ‘tegmen tympani canal’ of Orliac & O’Leary (2014) because distinct evidence of the presence of a stapedial artery can be observed on the petrosals described here.

The petrosal portion of the epitympanic recess is a very shallow elongated fossa or groove that borders the crista parotica laterally from the tegmen tympani process anteriorly to the fossa incudis posteriorly. This structure represents the medial portion of the epitympanic recess; the lateral portion is likely to have been formed by the squamosal as it occurs in most early eutherians ( Wible, Novacek & Rougier, 2004; Wible et al., 2009). At the posterior end of the epitympanic recess, the fossa incudis forms a distinct ovoid fossa in the petrosal. It was also probably covered laterally by the squamosal. It lies in line laterally with the fenestra vestibuli, on the lateral wall of the crista parotica and faces in the same direction as the promontorium.

The fenestra vestibuli is oval and has a rather low stapedial ratio ( Segall, 1970) in all of the specimens studied, ranging from 1.36 to 1.66. This opening is rather small compared with the size of the promontorium.

The adjacent facial sulcus gently curves around the fenestra vestibuli laterally. The crista parotica, its lateral border at the level of the fenestra vestibuli, is not strictly in line with the channel of the cavum supracochleare but with the canal for the ramus superior of the stapedial artery. The facial sulcus is close posteriorly to the more medial stapedial fossa (for the stapedius muscle; Wible et al., 2001). This fossa is strictly posterior to the fenestra vestibuli, and dorsolateral to the postpromontorial tympanic sinus (see below). The depth and distinctness of the stapedial fossa from the facial sulcus vary within our sample, i.e. very distinct in UFRJ-DG 119-M, much less so in UFRJ-DG 347-M.

The external aperture of the cochlear fossula (sensu Wible et al., 2009; Wible, 2010) has a semicircular shape with a dorsal linear basis. It is distinctly larger than the fenestra vestibuli. Both are separated by a wide crista interfenestralis ( Wible et al., 1995). The postpromontorial tympanic sinus (sensu Wible et al., 2009) that posteriorly borders the external aperture of the cochlear fossula occupies an area of moderate size. The major axis of this oval surface is transversal. It is bordered laterally by a crest that represents a posterior continuation of the crista interfenestralis (see Ekdale et al., 2004; Wible et al., 2004). At the posteromedial extremity of this crest there is a faint medial section of the caudal tympanic process ( MacPhee, 1981), which generally delimits the posterior boundary of the postpromontorial tympanic sinus. This process is absent on UFRJ-DG 1036M ( Figs 3 , 4 ) and 119M. In the other specimens studied, the medial section of the caudal tympanic process is low and blunt and has a variable transverse extent. This process is connect- ed to the medial border of the external aperture of the cochlear fossula by a low medial crest running anteriorly. This crest is faintly notched in the vicinity of the latter aperture, possibly for the passage of the tympanic nerve (e.g. see Evans, 1993; Wible, 2010: fig. 3). Three specimens (UFRJ-DG 1035-M, 1037-M, and 347- M) conspicuously document the anatomy of the region immediately posterior to the medial caudal tympanic process ( Figs 3 , 4 ). A deep notch or a foramen is present on the posterior flank of this process in UFRJ-DG 1035-M and 1037-M, but it is shallower in UFRJ-DG 347-M. This might have provided a passage for the auricular branch of the vagus nerve (X) from the jugular ganglion (housed in the jugular foramen) to the facial nerve across the mastoid region ( MacPhee, 1981; Evans, 1993). A shallow notch is also present in UFRJ- DG 1036-M, 1038-M, and 119-M, as well as in UFRJ- DG 275-M.

The area of the tympanohyal and stylomastoid notch is only preserved on UFRJ-DG 347-M, one of the two specimens that anatomically depart the most from the others (see above). The preserved portion of the tympanohyal on this specimen barely projects ventrally over the ventral elevation of the crista parotica to which it is dorsally fused. It is short, massive, and slightly bent medially ( Fig. 3A ). The posteromedial flank of the tympanohyal process is notched. This is the stylomastoid notch for the exit of the facial nerve from the middle ear ( van der Klaauw, 1931; Wible, 2008).

Cerebellar and cerebral surfaces

The outline of the cerebellar surface is preserved on UFRJ-DG 1037-M, 119-M, 347-M, and 275-M. The anterior edge is formed medioventrally by a blunt prefacial commissure and more laterodorsally by a more salient crista petrosa. The latter varies in extent amongst the specimens. It is extremely short in UFRJ-DG 1037-M and 119-M, but clearly longer (as long as the prefacial commissure) in UFRJ-DG 347-M and 275-M. The varying length of this crest coincides with the expansion of the fossa subarcuata, one of the two major features visible on the cerebellar surface of the petrosal, the other being the internal auditory meatus. The fossa subarcuata is compressed in the medioventral−laterodorsal direction in the specimens that present a short crista petrosa. The dorsolateral rim of this fossa in these latter specimens is mainly oblique relative to the course of the inferior petrosal sinus groove that runs anteroposteriorly. In UFRJ- DG 347-M, the crista petrosa (a ridge in many specimens, but a bony lamina in UFRJ-DG 347-M) is orientated anteroposteriorly on its posterior portion. In its anterior portion it bends medially at almost 90°, bordering the anterior edge of the fossa subarcuata, and extends medially to contact the prefacial commissure. In most specimens a short sulcus is present on the dorsolateral border of the fossa subarcuata, on the medial edge of the crista petrosal ( Fig. 5A , sps). This sulcus is here considered a passageway for the superior petrosal sinus ( Wible, 1993).

The fossa subarcuata is grossly conical and relatively shallow in all specimens. The bottom of this fossa, which is located in the posterior half of the cerebellar face, faces anteromedially. It is pierced by a distinct foramen for the petromastoid canal, which provides a passageway for the subarcuate artery ( Gannon, Eden & Laitman, 1988; Krombach et al., 2002; O’Leary, 2010; Billet & de Muizon, 2013).

A low and blunt crest separates the fossa subarcuata from the more medioventral IAM. The two rounded foramina within the IAM, the FAS and FAI ( Wible et al., 2004), are separated at a shallow depth by a transverse crest. This crest is slightly oblique to subparallel to the blunt adjacent crest extending on the ventromedial edge of the fossa subarcuata. The largest foramen within the IAM is the posteroventromedial FAI. It contains anteriorly the spinal cribriform tract ( Meng & Fox, 1995) and posteriorly the foramen singulare ( Wible, 2010; Billet & de Muizon, 2013). The FAS contains a large canal for the passage of the facial nerve towards the cavum supracochleare in the middle ear region. Deep within the canal, a posteriorly directed and cribriform tract communicates with the superior vestibule area (see Wible, 2008; 2010).

A remarkable feature in cerebellar view, but also visible from the tympanic (ventromedial) aspect, is the notch within which the cochlear canaliculus (perilymphatic duct) opens at the posteromedioventral corner of the cerebellar face ( Fig. 5A ). Most particularly, the lateral wall of this external aperture of the cochlear canaliculus is notched in a conspicuous right angle in the UFRJ-DG 1035−1038-M and 119-M specimens, a notch that is however weak to almost indistinct in the UFRJ-DG 275-M and 347-M specimens. In at least half of the specimens this canaliculus has two exits; the main opening is medioventrally accompanied by a tiny accessory aperture.

More dorsolaterally, on the posterior rim of the cerebellar face, a large slit represents the aperture of the vestibular aqueduct (endolymphatic duct). It opens approximately at mid-height of the cerebellar face.

None of the specimens preserves a vestigial anterior lamina (see Billet & de Muizon, 2013) and there is apparently no substantial bony contribution to the middle cranial fossa.

Mastoid and posteromedial surfaces

Five out of the seven investigated specimens preserve the mastoid portion of the petrosal. It has a large, ventrodorsal, wing-like extension and a large exposure on the external surface of the skull. This is indicated by the position of its adjacent sutural areas with squamosal and occipital bones ( Figs 3B , 4A ). The suture with the medially adjacent exoccipital bone is particularly prominent and extends along the entire length of the medial wall of the mastoid part ( Figs 3B , 5A ).

The mastoid exposure is widely notched or grooved all along its ventral and posterior surfaces ( Fig. 4A ). Two crests border this large groove ventrally: the posteromedial one is salient whereas the anterolateral one is blunter and lower. A broad, concave surface or groove, grossly elongated anteroposteriorly, arises posterior to the stapedial fossa at the posterior edge of the middle ear cavity. From there it curves along the mastoid exposure posterolaterally as a posterior occipital groove ( Figs 3 , 4A ) and continues dorsally to a notch located at the same height as the dorsal root of the tegmen tympani process. This dorsal notch is very large, dorsoventrally elongated, and represents the medial edge of the post-temporal foramen, which opens between the petrosal and squamosal on the occiput ( Wible, 2008; Billet & de Muizon, 2013; Fig. 5B ).

On the dorsal-most part of the mastoid portion, a dorsal extension of the anterolateral crest (interrupt- ed more ventrally by the post-temporal notch) is present as a short vertical crest (UFRJ-DG 1035-M and 347- M). This crest, which receives the squamosal laterally, was probably exposed at the posterolateral corner of the skull and possibly contributed to a short section of the nuchal (lambdoid) crest.

Medially, just anterior to the conspicuous suture with the exoccipital, the petrosal exhibits a wide vertical sulcus for the sigmoid sinus ( Fig. 3A ) that would have connected more ventrally to the internal jugular vein ( Wible, 1990).

Squamosal surface

The lateral or squamosal surface of the petrosals under study has a triangular outline. It is dominated anteriorly by the massive dorsal root of the process of the tegmen tympani ( Fig. 5B ). Posteroventrally, the most prominent feature is the oval sutural surface with the squamosal, with the greatest diameter orientated grossly anteroposteriorly. This suture is located on the concavity exhibited by the lateral face of the wingshaped pars mastoidea just anterolateral to the large mastoid groove described above. On the dorsolateral edge of the mastoid exposure, almost at the dorsal apex of the petrosal, is another articular surface for the squamosal on the lateral side of the short crest mentioned in the preceding section. The two articular surfaces are separated by a large notch in the lateral crest of the mastoid groove, which forms the medial edge of the post-temporal notch ( Fig. 6 ). The latter is greatly elongated. On most of the specimens it is distinctly notched at its dorsal and ventral ends and was apparently constricted in the middle as evidenced by the slight bony elevation between the two notches.

Between the ventral suture for the squamosal and the dorsal root of the tegmen tympani process, and immediately dorsal to the fossa incudis, is an oblique, sinuous, and elongated groove whose depth and extent varies amongst the specimens ( Figs 4A , 5B ). This groove is always the most conspicuous at the point where it is wedged between the posterior face of the root of the tegmen tympani and the fossa incudis. In the specimens where it is the best-marked (UFRJ-DG 1035- M, 1037-M, 275-M, and 347-M), this groove extends from the anterodorsal corner of the fossa incudis anteroventrally to the posterodorsal-most extent of the sutural surface with the squamosal posterodorsally. It is unknown whether or not this groove may have housed a vessel. However, we suspect that it may alternatively represent the mark of a nested articulation between the petrosal and squamosal.

A variably excavated and grossly circular fossa is present just anterior, or slightly anterodorsal, to the ventral notch of the post-temporal foramen ( Fig. 5B ). This fossa is probably the posterior extremity of a vascular sulcus (the post-temporal sulcus) that may have connected the post-temporal foramen with the foramen for the ramus superior at the anterolateral corner of the petrosal. The condition of this fossa is variable. In UFRJ-DG 347-M it is elongated and is a wide sulcus rather than a fossa and is distinctly part of the posttemporal canal. In other specimens (e.g. UFRJ-DG 1035-M and 1036-M; Fig. 5B ) a connection of the fossa with the post-temporal sulcus is less conspicuous. The post-temporal sulcus is sometimes markedly sigmoid but has a variable trajectory. It is also variably developed, being more distinct on UFRJ-DG 275-M, 347- M, 119-M, 1035-M, and 1036-M ( Figs 4B , 5B ) than on UFRJ-DG 1037-M and 1038-M. This post-temporal sulcus is likely to have transmitted the arteria diploëtica magna, emitted by the ramus superior of the stapedial artery, running from the superior ramus foramen to the post-temporal foramen. This sulcus generally also transmits the vena diploëtica magna, which is emitted by the capsuloparietal emissary vein ( Wible, 1987). The latter is one of the three veins emitted by the transverse sinus, the other two being the superior petrosal sinus, which runs on the dorsolateral surface of the petrosal to the cavernous sinus, and the sigmoid sinus, which descends from the posteromedial corner of the petrosal to the jugular foramen. The capsuloparietal emissary vein of placentals is a branch emitted by the transverse sinus, which generally runs between the petrosal and squamosal and exits the skull via the postglenoid foramen ( Wible, 1993; 2011). It is therefore conceivable that some marks of the passageway of this vessel are to be found on the lateral aspect of the petrosal. On three of the specimens described here (UFRJ-DG 1035-M, 1038-M, and 347-M) a distinct sulcus dorsal to the post-temporal sulcus runs anteroposteriorly, the two sulci being roughly parallel ( Fig. 6 ).

This sulcus dorsal to the post-temporal sulcus arises posteriorly from the dorsal-most part of the petrosal, a region where the trifurcation of the transverse sinus generally occurs ( Fig. 6 ; Wible, 1993). The dorsal tip of the petrosal is only preserved in the specimens UFRJ- DG 275-M and 347-M, on which it is bifid and deeply notched. This condition is especially obvious on UFRJ- DG 347-M (which is the best preserved specimen), on which the sulci for the superior petrosal sinus and sigmoid sinus are easily discriminated dorsally and posteromedially, respectively (the sulcus for the sigmoid sinus is also visible at a similar location on UFRJ- DG 1036-M, Fig. 3A ). The sulcus dorsal to the posttemporal sulcus may therefore be for the capsuloparietal emissary vein ( Fig. 6B ). On the other specimens showing this sulcus which is tentatively referred to the capsuloparietal emissary vein (UFRJ-DG 1035-M and 1038-M), the dorsal-most part of the petrosal is not so well preserved but the posterior end of the sulcus also originates dorsally ( Fig. 6A ). However, some variation exists, and on UFRJ-DG 1036-M there is no sulcus dorsal to the anterior portion of the post-temporal sulcus ( Figs 4B , 5B ). Furthermore, the latter is very deep and wide and has a diameter clearly larger than that of the foramen for the ramus superior. It is therefore hypothesized that, in this specimen, the anterior portion of the post-temporal sulcus conveyed both the arteria diploëtica magna and the capsuloparietal emissary vein. Besides, a large posteromedial sulcus in continuity with the anterior portion of the post-temporal sulcus may have conveyed the posterior portion of the capsuloparietal emissary vein in this specimen; this posteromedial sulcus is directed transversally and located dorsomedial to the circular fossa and posttemporal notch ( Fig. 5B ). It is also distinct in UFRJ- DG 1035-M ( Fig. 6A ). Thus, it was possibly not providing a passage way to the arteria diploëtica magna, which would have diverged more ventrally towards the circular fossa and post-temporal notch. The pattern exhibited by several other specimens, such as UFRJ- DG 119-M, is more uncertain because their sulci are more difficult to distinguish.

On some specimens (UFRJ-DG 1035-M and, possibly, 119-M), an additional short sulcus runs posteriorly from the posterodorsal end of the sulcus, which may have housed the capsuloparietal emissary vein to the dorsal notch of the large post-temporal foramen. This passageway may have conveyed the vena diploëtica magna. This implies that its trajectory would have been, therefore, independent from that of the arteria diploëtica magna in these specimens ( Fig. 6A ). This vascular interpretation should however be regarded as tentative until additional and/or better preserved material is investigated.

Bony labyrinth of the inner ear

The shape of the inner ear of the five investigated specimens (UFRJ-DG 1035-M, 1036-M, 119-M, 275-M, and 347-M) is very homogenous ( Table 1), which contrasts with the larger morphological variation observed on the external surfaces of the corresponding petrosals.

Cochlear canal

The cochlea has 2.25−2.4 coils ( Fig. 7 ). The height of the cochlear canal decreases abruptly just past the level of the cochlear fossula. Distal to this point, the height of the cochlear canal remains almost constant and barely decreases towards the apex. The tightness of the coiling of the cochlea is moderate. Successive turns touch each other, except for the first (most proximal) 90° of coiling. In all specimens, there is an especially conspicuous gap between the first and second turns at the level of the bottleneck that affects the height of the cochlear canal distal to the cochlear fossula. On the external wall of the bony cochlear canal, at the level of this bottleneck, and just dorsal to the groove for the secondary bony lamina (see below), the imprint of a vessel is visible on the reconstructed bony labyrinth ( Figs 7 , 8 ). In its most proximal part, this vessel first runs along the groove for the secondary lamina but turns abruptly dorsally and progressively lateral in its most distal course, making a U-turn distal to the cochlear bottleneck. At this point, it joins the medial base of the cochlear canaliculus, close to the dorsomedial extremity of the cochlear fossula. The course of this rarely observed vessel imprint on reconstructed fossil bony labyrinths evokes the vena aquaeductus cochleae ( Frick, Leonhardt & Starck, 1992; Feneis, 1993), which drains the blood from the basal cochlear turn and the vestibule by several smaller veins and leaves the bony labyrinth through the cochlear canaliculus.

The cochlea of the investigated specimens shows the typical therian pattern in terms of a cribriform modiolus and a well-developed primary bony lamina with the canal for the spiral nerve ganglion (Rosenthal’s canal, CN VIII; Fig. 9 ). The secondary bony lamina is also present in our specimens but shows some variation: it extends as a distinct lamina from the fenestra cochleae to less than one-quarter (specimen UFRJ- DG 347-M) to almost three-quarters (specimen UFRJ- DG 275-M) of the basal turn. The lamina continues as a very smooth ridge at least to the end of the first turn in all of the specimens under study ( Fig. 9 ).

In profile, the cochlea has a height that is approximately two-thirds that of its width ( Table 1). The plane of its basal turn creates an angle of 23–30° with that of the plane of the LSC.

The fenestra cochleae is barely larger than the external aperture of the cochlear fossula on all of the reconstructed bony labyrinths of the group. It presents a similar outline to the latter and extends just slightly over its external margin. The cochlear canaliculus is short and directed posteromedially. It is often composed of double or even triple parallel ducts, of which the main one is always more than twice the width of the other(s). The smaller canals might have housed blood vessels such as the vena aquaeductus cochleae.

Vestibule and semicircular canals

The part of the vestibule directly adjacent to the cochlear canal is not separated from the latter by a sharp break in slope. The vestibule is moderately elongated in the (sub)horizontal plane and anteroposterior direction. The aqueductus vestibuli (endolymphatic duct) has its origin just anterior to the proximal base of the common crus. From this point, it runs mostly parallel to the common crus, but is slightly more inclined posteromedially towards its dorsal apex than the latter. Its course is slightly curved on most specimens. The aqueductus vestibuli reaches its external aperture approximately at the level of the dorsal apex of the common crus. For most of its course, the aqueductus vestibuli is much thinner than the ducts of the semicircular canals (SCs; Fig. 7 ).

The ASC is the largest of the SCs. The PSC and LSC are very close in size, the PSC being just slightly larger. The ASC and PSC both have a very rounded shape (curvature), as measured by the aspect ratio (SC L/W; Table 1). The ASC is however relatively more elongat- ed in the direction of its length diameter (see an illustration of these diameters in Billet & de Muizon, 2013: fig. 1) than is the PSC. Conversely, the LSC is rather elongated in the direction of its width diameter. All SCs show a weak to null planar deviation. The common crus is cylindrical in most specimens but faintly conical in a few, with a larger duct at its ventral base than at its dorsal apex. A secondary common crus, the bony confluence of the posterior limb of the LSC and the ventral limb of the PSC, is present in all investigated specimens ( Figs 7–9 ). The dorsal-most extent of the ASC is more than the apex of the common crus, and that of the PSC is just slightly more dorsal or level with this latter structure in all specimens. The three ampullae are well inflated. The angles between the SCs are all at or slightly below 90°, the most acute for each specimen being the angle between the ASC and LSC, at 79–83°.

The bony labyrinth within the petrosal

In all investigated specimens, the bony labyrinth of the inner ear appears to present similar proportions relative to the surrounding petrosal dimensions. More precisely, our calculation of the ratio IEH/PET size (see Material and Methods) is close to 0.70–0.80 in each of these specimens, which suggests that the inner ear is not much reduced relative to the petrosal and occupies most of the internal volume of the latter. This is fully visible when bone transparency for the petrosal is reconstructed ( Fig. 9 ). As should be expected in such a configuration, the vestibular aqueduct and cochlear canaliculus are rather short when compared with the inner ear dimensions.

Arguments for the referral to Miguelsoria

Phylogenetic arguments

Our analysis resulted in 48 most parsimonious trees [MPTs; 507 steps, consistency index (CI) = 0.37, retention index (RI) = 0.72], the strict consensus of which (518 steps, CI = 0.36, RI = 0.71) is shown in Figure 10 . No exclusive phylogenetic affinities for UFRJ-DG 119- M, 1035−1038-M, 275-M, and 347-M petrosals with any group of SA placental can be proposed, as they appear as sister to a clade that contains Neogene Litopterna , Astrapotheria, Xenarthra, Pyrotheria , and Notoungulata (i.e. all SA ungulates of Neogene age considered in this paper). Moreover, this analysis failed to recover a monophyletic Litopterna as the genera Miguelsoria and Protolipterna are located in a polytomy much closer to the cladogram root than to later litopterns. In order to avoid potential conflicts amongst different anatomical partitions within the data set and to try to get more information for the attribution of the UFRJ-DG 119- M, 1035−1038-M, 275-M, and 347-M petrosals, we ran a reduced analysis based on features of the petrosal and inner ear anatomy only using well-documented taxa (see Material and Methods). This resulted in three MPTs (52 steps, CI = 0.54, RI = 0.72) and in a more informative strict consensus tree (62 steps, CI = 0.45, RI = 0.61; Fig. 11 ). The latter supports the UFRJ-DG 119-M, 1035−1038-M, 275-M, and 347-M petrosals as the sister group of a clade that contains the Litopterna and Astrapotherium , whereas the Itaboraí MNHN-F-BRD 23 isolated notoungulate petrosal ( Billet & de Muizon, 2013) clusters with the Notoungulata . Even though the node containing the Litopterna− Astrapotherium −UFRJ- DG 119-M, 1035−1038-M, 275-M, and 347-M petrosals is not very well supported (Bremer support of the node = 1; mean of Bremer supports = 1.1), this analy- sis clearly shows that these UFRJ-DG petrosals have preferential affinities with the Litopterna and Astrapotheria and might belong to one of these groups (note that the xenungulate Carodnia was not includ- ed in any of these phylogenetic analyses because data for it are too fragmentary and its petrosal is unknown, but see size arguments below). Furthermore, as didolodontid condylarths are possibly related to Litopterna ( de Muizon & Cifelli, 2000) and as they are present at Itaboraí ( Bergqvist et al., 2006), the UFRJ- DG 119-M, 1035−1038-M, 275-M, and 347-M petrosals might also belong to this group (whose petrosal anatomy is unknown as yet from direct evidence).

Size arguments

The linear regression of M2 area on the promontorium area resulted in the following predictive equation: M2 = 1.6905 × Prom − 9.4692; R 2 = 0.9541; Wilks’ lambda test = 0.0459; P (regr) = 2.152E- 09 ( Fig. 12 , Table S1). The high R 2 values show that these models based on promontorium area explain a very large part of the variation of the M2 area.

The equation above predicts a value of ∼ 29–37 mm 2 for the M2 area associated with the UFRJ-DG petrosals under study here: UFRJ-DG 1035-M: M2 = 29.42 mm 2; UFRJ-DG 1036-M = 30.59 mm 2; UFRJ-DG 119-M: 30.32 mm 2; UFRJ-DG 347-M: 37.49 mm 2; UFRJ-DG 275- M: 35.97 mm 2.

We compared these predictions with the areas of M2 of all Itaboraí litopterns, astrapotheres, and ‘condylarths’ (see phylogenetic arguments; Table S1). Amongst them, the only taxa close to this range are Protolipterna ellipsodontoides Cifelli, 1983b (∼ 19–38 mm 2; mean = 24.54 mm 2; LB measurements on 58 individuals) and Miguelsoria parayirunhor (∼ 26–46 mm 2; mean = 33.52 mm 2; LB measurements on 24 individuals), with the mean of the latter fitting the best. Other taxa have values much farther from the predicted M2 area ( Table S1).

Therefore, besides the litopterns Miguelsoria and Protolipterna , the other SA placentals at Itaboraí were either found to be phylogenetically unrelated to these petrosals (e.g. Notoungulata, Xenarthra ) or to have a clear incompatibility of size with them [e.g. Carodnia (Xenungulata) , Tetragonostylops (Astrapotheria) , see Paula-Couto, 1952].

Taphonomic and abundance arguments

All of the seven UFRJ-DG petrosal specimens belonging to the group under study present a white and cream coloration. This indicates that they possibly come from various fissure fill deposits but not from the fissure with brownish fossils recovered at Itaboraí in 1968 ( Bergqvist, 2008). The ‘brownish fossil fissure’ (BFF) or ‘1968 Fissure’ presents a different taxonomic composition and abundances from the finds made in the various ‘white fossil fissures’ (WFF) ( Bergqvist, 2008: table 6.3). The UFRJ-DG 119-M, 275-M, 347-M, and 1035−1038-M petrosals are, to our knowledge, the smallest placental petrosals recovered in the WFF. Small SA ungulate taxa represented by dental material in the WFF notably include Miguelsoria ( Bergqvist, 2008: tables 6.1, 6.3). Protolipterna , the other litoptern that is compatible in size with the UFRJ-DG petrosals under study, has only been found in the BFF ( Bergqvist, 2008: table 6.3). Thus Miguelsoria seems to be the best referral for the UFRJ-DG 119-M, 275-M, 347-M, and 1035−1038-M petrosals amongst the size-compatible litoptern taxa.

Furthermore, relative taxon abundance in the WFF supports this statement. When considering nonpetrosal material only, Miguelsoria constitutes the most abundant small- to medium-sized SA ungulate taxon in the WFF when notoungulates are omitted ( Bergqvist, 2008: table 6.3). In the UFRJ-DG collection there is a total of 23 isolated placental petrosals from the WFF at Itaboraí. Within this sample, the UFRJ-DG 119- M, 275-M, 347-M, and 1035−1038-M petrosals under study here constitute the most abundant group, except for the notoungulate petrosal specimens (identifiable on the basis of characters provided in Billet & de Muizon, 2013). Therefore, abundance data match well the referral of these isolated petrosals to Miguelsoria .

Additional important information comes from the BFF petrosal material (‘1968 Fissure’), which G. Billet had the opportunity to study briefly. Within the BFF, the most abundant morphology-based group of petrosals (MCT 1755M) represents almost two-thirds of the total number of these bones (59 amongst 94). These petrosals are very similar in anatomy and size (similar or slightly smaller) to the seven WFF UFRJ-DG petrosals described in this study ( Fig. 13 ). Their great abundance within the BFF strongly suggests that they belong to the taxon Protolipterna ellipsodontoides , the dominant taxon in all of the nonpetrosal placental material of the BFF ( Bergqvist, 2008: table 6.3). Their relatively small size also indicates that referral to Protolipterna is correct. However, we cannot eliminate the possibility that a small number of these 59 petrosals (MCT 1755M) may represent Miguelsoria instead of Protolipterna . Indeed, the two genera are very close in size and morphology, and Miguelsoria is also represented, but much less abundant, in the BFF ( Cifelli, 1983b; Bergqvist, 2008). In any case, these abundant BFF petrosals, which are certainly attributable in the main to Protolipterna , provide indirect support for the tentative attribution of the WFF UFRJ-DG 119- M, 275-M, 347-M, and 1035−1038-M petrosals to the related taxon Miguelsoria . As in the case of phylogenetic arguments, these data also disfavour their referral to notoungulates, which are less abundant than Protolipterna in the BFF material ( Bergqvist, 2008: table 6.3).

In conclusion, given these several lines of congruent evidence (phylogenetic affinities with Litopterna , size and abundance matching best with Miguelsoria , resemblance to Protolipterna BFF petrosals), we therefore refer the UFRJ-DG 119-M, 275-M, 347-M, and 1035−1038-M petrosals to cf. Miguelsoria parayirunhor .

Figures

Figure 2. Ventrolateral views of the seven UFRJ-DG petrosal specimens studied in this paper and attributed to cf. Miguelsoria parayirunhor. UFRJ-DG specimen number is indicated below each specimen. Note that specimens UFRJ-DG 347-M and 1036-M have been mirrored in this figure.

Figure 3. Virtual 3D reconstruction and outline drawing of petrosal UFRJ-DG 1036-M (mirrored), details of anatomy; A, posterior view; B, posteroventral view. Abbreviations: a.b.X n., notch/sulcus for the auricular branch of the vagus nerve (X); acan n, notch housing the external aperture of the cochlear canaliculus; acf, external aperture of the cochlear fossula; a.e. fv, anterior edge (salient) of fenestra vestibuli; br.thyl, broken tympanohyal; ci, crista interfenestralis; cpa, crista parotica; epr, epitympanic recess; exs, exoccipital articulation surface of petrosal; fs, facial sulcus; fv, fenestra vestibuli; hf, hiatus Fallopii; ips, sulcus of inferior petrosal sinus; lv.ips, crest-like lateroventral border of sulcus for inferior petrosal sinus; md.ips, crest-like mediodorsal border of sulcus for inferior petrosal sinus; pg, posterior occipital groove; pps, postpromontorial tympanic sinus; pr, promontorium; ptn, post-temporal notch; rsup f, foramen for the ramus superior of stapedial artery; rtp, rostral tympanic process; stf, stapedial fossa; tnn, tympanic nerve notch; tt, tegmen tympani; v.e.m., ventral extent of mastoid surface; vss, vertical sulcus for sigmoid sinus.

Figure 4. Virtual 3D reconstruction and outline drawing of petrosal UFRJ-DG 1036-M (mirrored), details of anatomy; A, ventral view; B, anterior view. Abbreviations: a.b.X n., notch/sulcus for the auricular branch of the vagus nerve (X); acf, external aperture of the cochlear fossula; ci, crista interfenestralis; cpa, crista parotica; epr, epitympanic recess; fi, fossa incudis; fs, facial sulcus; fsa, fossa subarcuata; fv, fenestra vestibuli; hf, hiatus Fallopii; iam, internal auditory meatus; ias, internal carotid artery sulcus; lv.ips, crest-like lateroventral border of sulcus for inferior petrosal sinus; obl.g, oblique groove; pff, primary facial foramen; pg, posterior occipital groove; pmc, petromastoid canal; pps, postpromontorial tympanic sinus; pr, promontorium; pts, post-temporal sulcus; rsup f, foramen for the ramus superior of stapedial artery; rtp, rostral tympanic process; stf, stapedial fossa; st. s., stapedial artery sulcus; tnn, tympanic nerve notch; tt, tegmen tympani; ttf, tensor tympani fossa; v.e.m., ventral extent of mastoid surface.

Figure 5. Virtual 3D reconstruction and outline drawing of petrosal UFRJ-DG 1036-M (mirrored), details of anatomy; A, cerebellar view; B, dorsal view. Abbreviations: aav, external aperture of the vestibular aqueduct; acan n, notch housing the external aperture of the cochlear canaliculus; cev s?, sulcus for capsuloparietal emissary vein?; circ. fs, (vascular?) circular fossa; cn, canal and cribriform tract for cochlear nerve; cp (br.), crista petrosa (broken); cpa, crista parotica; exs, exoccipital articulation surface of petrosal; fai, foramen acusticum inferius; fas, foramen acusticum superius; ff, facial foramen; fi, fossa incudis; fsa, fossa subarcuata; fsg, foramen singulare; f sup vest, foramen for the superior vestibule area; ips, sulcus of inferior petrosal sinus; lv.ips, crest-like lateroventral border of sulcus for inferior petrosal sinus; md.ips, crest-like mediodorsal border of sulcus for inferior petrosal sinus; obl.g, oblique groove; pg, posterior occipital groove; pmc, petromastoid canal; pr, promontorium; ptn, post-temporal notch; pts, post-temporal sulcus; sps, sulcus of superior petrosal sinus; sqs, squamosal articulation surface of petrosal; st.s., stapedial artery sulcus; tt, tegmen tympani.

Figure 6. Dorsolateral views of the petrosals UFRJ-DG 1035-M (A) and 347-M (B). Abbreviations: cev s?, sulcus for capsuloparietal emissary vein?; circ. fs, (vascular?) circular fossa; ptn, post-temporal notch; pts, post-temporal sulcus; rsup f, foramen for the ramus superior of stapedial artery; sqs, squamosal articulation surface of petrosal; vdms?, possible sulcus for the vena diploëtica magna.

Figure 7. Virtual 3D reconstruction and outline drawing of the bony labyrinths of UFRJ-DG 1035-M and 347-M (outline drawing only); A, anterior (a) view; B, dorsal (d) view; C, lateral (l) view; D, posteromedial (p) view. Abbreviations: aa, anterior ampulla; acf, external aperture of the cochlear fossula; asc, anterior semicircular canal; av, vestibular aqueduct; can, cochlear canaliculus; cc, common crus; cf, cochlear fossula; co, cochlear canal; fsg, foramen singulare; fv, fenestra vestibuli; la, lateral ampulla; lsc, lateral semicircular canal; pl. s., primary bony lamina sulcus; psc, posterior semicircular canal; scc, secondary common crus; sl. s., secondary bony lamina sulcus; vasc. sulc., impression of vascular sulci left on the bony cochlear canal.

Figure 8. Virtual 3D reconstruction and outline drawing of the bony labyrinth of UFRJ-DG 1036-M, details of anatomy in posteromedial view. Abbreviations: aa, anterior ampulla; asc, anterior semicircular canal; av, vestibular aqueduct; can, cochlear canaliculus; cc, common crus; cf, cochlear fossula; co, cochlear canal; fsg, foramen singulare; f sup vest, foramen for the superior vestibule area; lsc, lateral semicircular canal; psc, posterior semicircular canal; scc, secondary common crus; sl. s., secondary bony lamina sulcus; vasc. sulc., impression of vascular sulci left on the bony cochlear canal.

Figure 9. Computed-tomography slices through the isolated petrosal UFRJ-DG 1035-M. Numbers refer to individual slices that are re-located on the virtually reconstructed translucent petrosal and bony labyrinth at the lower right corner. Abbreviations: aa, anterior ampulla; aav, external aperture of the vestibular aqueduct; A.b.X n., notch/sulcus for the auricular branch of the vagus nerve (X); acan n, notch housing the external aperture of the cochlear canaliculus; acf, external aperture of the cochlear fossula; asc, anterior semicircular canal; cc, common crus; cgsn, canal for the ganglion of spiral nerve; cn, canal and cribriform tract for cochlear nerve; co, cochlear canal; fai, foramen acusticum inferius; fn, canal for facial nerve; fs, facial sulcus; fsa, fossa subarcuata; fsg, foramen singulare; f sup vest, foramen for the superior vestibule area; fv, fenestra vestibuli; la, lateral ampulla; lsc, lateral semicircular canal; lv.ips, crest-like lateroventral border of sulcus for inferior petrosal sinus; pa, posterior ampulla; pl, primary bony lamina; psc, posterior semicircular canal; rtp, rostral tympanic process; scc, secondary common crus; sl, secondary bony lamina; ves, vestibule.

Figure 10. Strict consensus cladogram (518 steps, consistency index = 0.36, retention index = 0.71) of the analysis with 50 taxa searching for the relationships of the UFRJ-DG petrosals under study.

Figure 11. Strict consensus cladogram (62 steps, consistency index = 0.45, retention index = 0.61) of the reduced analysis (19 taxa and petrosal and inner ear characters only) searching for the relationships of the UFRJ-DG petrosals under study; Bremer support values are indicated at the nodes.

Figure 12. Second upper molars (M2) area vs. promontorium area in a sample of 14 Palaeogene early diverging placental specimens with associated petrosal and teeth remains (Table S1).

Figure 13. View of the large group of resembling petrosals held at the Departamento Nacional de Produçao Mineral from the ‘brownish fossil fissure’ (see text), with a close-up of some specimens; lateral views.