A new Miocene baleen whale from Peru deciphers the dawn of cetotheriids

Cetotheriidae are an iconic, nearly extinct family of baleen whales (Mysticeti) with a highly distinct cranial morphology. Their origins remain a mystery, with even the most archaic species showing a variety of characteristic features. Here, we describe a new species of archaic cetotheriid, Tiucetus rosae, from the Miocene of Peru. The new material represents the first mysticete from the poorly explored lowest portion of the highly fossiliferous Pisco Formation (allomember P0), and appears to form part of a more archaic assemblage than observed at the well-known localities of Cerro Colorado, Cerro los Quesos, Sud-Sacaco and Aguada de Lomas. Tiucetus resembles basal plicogulans (crown Mysticeti excluding right whales), such as Diorocetus and Parietobalaena, but shares with cetotheriids a distinct morphology of the auditory region, including the presence of an enlarged paroccipital concavity. The distinctive morphology of Tiucetus firmly places Cetotheriidae in the context of the poorly understood ‘cetotheres’ sensu lato, and helps to resolve basal relationships within crown Mysticeti.


Introduction
Cetotheriidae are a group of relatively small and mostly extinct baleen whales sharing a highly distinct cranial morphology [1][2][3]. The existence and, for the most part, scope of this family are widely agreed on [1,2,[4][5][6][7], but this consensus is a relatively recent phenomenon. The term 'cetothere' has a long history as a wastebasket taxon covering any extinct chaeomysticete (toothless baleen whale) that does not unambiguously fall within one of the living families. A detailed re-examination of this group led to a revision dividing 'cetotheres' into  [23]). The drawing of a cetotheriid marks the provenance of the holotype of Tiucetus rosae from allomember P0. To the right of the section is a partial overview of the Pisco Formation as exposed in the Sacaco Basin (modified from [18,24]), showing the proposed order of the vertebrate horizons of Cerro la Bruja (CLB), El Jahuay (ELJ) and Aguada de Lomas (AGL). Note the mismatch between the proposed ages for CLB and ELJ.
rostrum; from balaenopterids and eschrichtiids in having a supraorbital process that gradually descends from the skull vertex, a broadly triangular ascending process of the maxilla, and a pars cochlearis that is not cranially elongated; from Isanacetus, Parietobalaena and 'Diorocetus' chichibuensis in having a narrow body of the periotic with no sign of lateral inflation, and a more prominent inner posterior prominence of the tympanic bulla that is not medially and anteriorly retracted; from Diorocetus in having the inner posterior prominence of the tympanic bulla developed as a bulbous dorsal projection on the involucrum, and in lacking a ridge posteriorly bordering the facial sulcus on the posterior process; from all cetotheriids and Titanocetus in having a broadly triangular ascending process of the maxilla, a welldeveloped ascending process of the premaxilla extending posteriorly as far as the maxilla, and a nasal that only moderately tapers posteriorly; and from all cetotheriids except Cephalotropis and Joumocetus in having parietals that are well exposed on the skull vertex.

Maxilla, premaxilla and nasal
In dorsal view, the ascending process of the premaxilla is robust and parallel-sided, and extends posteriorly as far as the maxilla (figures 2 and 3). The suture between the premaxilla and the maxilla is open along its entire preserved length, suggesting rostral kinesis. Centrally on the rostrum, the premaxillae define a broad and elongate narial fossa. Anterior to the narial fossa, the premaxillae are lost but, judging from the shape of the adjacent maxilla, appear to have widened anteriorly. The nasal is relatively elongate and extends anteriorly far beyond the inferred level of the antorbital notch. Posteriorly, the nasal tapers, but still makes a robust contact with the frontal and extends to about the same level as the premaxilla and maxilla. The ascending process of the maxilla is broadly triangular and extends roughly halfway across the length of the frontal (figures 2 and 3). Adjacent to the anterior portion of the nasal, each maxilla bears a cluster of three large dorsal infraorbital foramina with associated, posteromedially oriented sulci. The posteriormost of these foramina opens posteriorly in the direction of the ascending process of the maxilla, and may hence be homologous with the primary dorsal infraorbital foramen of other cetotheriids (sensu [7]). Additional foramina, some of comparable size, open further anteriorly on the maxilla and are associated with (mostly anteriorly) oriented sulci.
In lateral view, the dorsal margin of the rostrum is straight, with the premaxilla and nasal being flush with the maxilla (figure 4a,b). The lateral edge of the maxilla is dorsoventrally flattened. Posteriorly, the medial portion of the maxilla gradually expands in height, with its ventral border gradually descending towards the inferred level of the orbit. Breakage has exposed the infraorbital canal, including what appears to be its bifurcation into the dorsal alveolar canal and the infraorbital canal proper.
In ventral view, the maxilla is transversely concave, with its medial portion steeply descending along the vomer to form a pronounced palatal keel (figures 4b and 6). Posteriorly, each maxilla preserves the posteriorly concave, anteromedially directed outline of the maxillo-palatine suture, similar to that of Parietobalaena palmeri (USNM 16119). Somewhat anterior to this suture, approximately 20-25 mm from the medial border of the maxilla, a poorly preserved groove probably represents the palatine sulcus. The anterior portion of each maxilla bears several elongate palatal nutrient foramina and sulci. The more lateral of these sulci are oriented anteriorly, whereas the more medial ones point anteroventrally or even anteroventromedially and extend far on to the palatal keel.

Frontal
In dorsal view, the frontal is exposed on the cranial vertex for a length of about 18 mm (figure 3). The interfrontal suture is evident, and there is no narial process. A sharp orbitotemporal crest runs from the vertex along the centre of the supraorbital process, before gradually becoming less distinct and finally disappearing completely roughly halfway between the sagittal plane and the inferred position of the orbit. The dorsal surface of the supraorbital process is smooth, with no obvious foramina. In anterior view, the supraorbital process is straight to slightly concave, and gently slopes ventrally as it descends from the vertex (figure 7a). In lateral view, the portion of the frontal exposed at the vertex is oriented at an angle relative to the rostrum, and rises dorsally towards its suture with the parietal. In ventral view, there is a well-developed, sharp preorbital ridge, bordered anteriorly by a wide, slightly concave area that presumably once accommodated the infraorbital plate of the maxilla (figure 6).

Parietal
In dorsal view (figure 3), the parietals are broadly exposed on the vertex and form a broad (18 mm) sagittal ridge that contrasts with the narrow crest of Parietobalaena (P. palmeri: USNM 10677, 16119; P. yamaokai: HMN F00042). Anteriorly, the parietal overrides the posterior portion of the frontal, extending all the way to the orbitotemporal crest and on to the posteromedial corner of the supraorbital process. Where it forms the intertemporal constriction, the parietal is convex transversely, but then becomes partially concave (both anteroposteriorly and dorsoventrally) inside the temporal fossa. In lateral view, the parieto-squamosal suture is sigmoidal and descends almost vertically from the nuchal crest, before turning 90°and terminating at the alisphenoid (figure 5

Squamosal
The right squamosal is complete, except for some damage to the zygomatic process and some minor erosion of the postglenoid process. In dorsal view, the squamosal, parietal and frontal define a temporal fossa that is wider transversely than it is long (figure 2). The zygomatic process is robust, oriented slightly anterolaterally, and bears a low, rounded supramastoid crest along its posterior portion. The squamosal fossa is elongate, convex anteriorly and concave posteriorly. Inside the fossa, there is a short (approx. 20 mm) squamosal cleft. Anteroventrally, the border of the squamosal is smooth, with no sign of a squamosal crease. Posteriorly, the nuchal crest terminates anterior to the level of the occipital condyles. In lateral view, the postglenoid process is robust, triangular in outline, and oriented ventrally or slightly posteroventrally, with its posterior border being slightly concave (figure 4b). The lateral face of the postglenoid process is flattened and projects beyond the outer surface of the zygomatic process, thus creating a bony eminence where the two processes meet. The zygomatic process is also robust, with approximately parallel dorsal and ventral borders. Dorsal to the base of the postglenoid process, there is a well-developed, bluntly triangular squamosal prominence (figure 5). Immediately posteroventral to the prominence, the squamosal is excavated by a moderately sized, deep fossa for the sternocephalicus, which is separated from a second, smaller and shallower fossa just below by a low, rounded horizontal ridge.
In posterior view, the squamosal projects laterally well beyond the lateral border of the exoccipital (figure 7b). Unlike in certain cetotheriids, such as Metopocetus (M. hunteri, NMR9991-07729) and 'Cetotherium' megalophysum (USNM 205510), the posterior meatal crest does not clearly extend on to the base of the postglenoid process, except for a faint, short (approx. 20 mm) ridge present on the left side only. The postglenoid process is wide and, accounting for breakage, appears to have been broadly parabolic in outline, as seen in a somewhat more elongated form in Parietobalaena (e.g. P. campiniana, IRSNB M399-R4018; P. palmeri, USNM 16229). Ventrally, the postglenoid process extends well below the level of the paroccipital process.  In ventral view, the falciform process is robust. Its posterior border is slightly notched, closely apposed to the anterior border of the tympanic bulla, and underlaps the anterior process of the periotic (figures 6 and 8). The postglenoid process is oriented transversely, with no evidence of twisting as seen in Piscobalaena and herpetocetines (figure 6). The anterior surface of the postglenoid process is convex transversely, except for its medialmost portion, which is markedly concave and appears as if pinched. The transition from the convex to the concave portion of the postglenoid process is aligned with a bony protuberance, which is located close to the rim of the temporal fossa. Medially, the postglenoid process directly borders the shallow sigmoid fossa, with no intervening anterior meatal crest. The external acoustic meatus is parallel-sided, with the posterior meatal crest descending only slightly on to the posterior process ( figure 8).

Supraoccipital, exoccipital and basioccipital
In dorsal view, the supraoccipital is triangular with a narrow, pointed apex and sigmoidal lateral borders (i.e. nuchal crests). As in Isanacetus and Parietobalaena, but unlike in other cetotheriids, the tip of the supraoccipital extends anteriorly just beyond the level of the zygomatic process of the squamosal (figure 2). The nuchal crest is oriented dorsally and does not overhang the temporal fossa ( figure 5).
Posterior to its apex, the supraoccipital shield is initially flattened, but then abruptly turns concave transversely and develops a short, but tall, external occipital crest. Halfway towards the foramen magnum, the supraoccipital remains concave along the midline, but turns convex along the nuchal crest. Yet further posteriorly, the surface of the shield becomes flattened dorsal to the foramen magnum, but strongly concave transversely as it merges with the exoccipital. The exoccipital is slightly thickened, but does not extend posteriorly beyond the level of the occipital condyle. The condyles themselves are situated on a short, barely constricted neck. In posterior view, the foramen magnum occupies approximately two-thirds of the total height of the occipital condyles (figure 7b). The dorsal condyloid fossae are present, but shallow. Lateral to the condyle, the jugular notch is relatively wide and open. The paroccipital process is squared and rather narrow-considerably more so than in Isanacetus (MFM 28501), Parietobalaena (e.g. P. palmeri, USNM 16119; P. yamaokai, HMN F0042) and other cetotheriids (e.g. Piscobalaena nana, MNHN.F. SAS1616). The basioccipital crest is inflated and has a sharp, ridge-like medial border that partially floors the area between the choanae and the intercondyloid notch.
In ventral view, the lateral edges of the basioccipital crests are approximately parallel ( figure 6). The ventral condyloid fossae are virtually indistinguishable. Posterolateral to the basioccipital crest, there is a fused but still distinct medial exoccipital crest forming the medial border of the jugular notch. The jugular notch itself is rounded and, on the right side, houses a distinct hypoglossal foramen. The right hypoglossal canal is rather short, with its internal opening and the hypoglossal foramen being a mere 6 mm apart. On the left side, the hypoglossal canal appears to be anteriorly open, so that only a sulcus with no distinct dorsal or ventral foramina remains (

Vomer
In dorsal view, the trough-like lateral sheets of the vomer floor the mesorostral groove ( figure 2). In ventral view, the portion of the vomer between the maxillae is transversely rounded and smooth ( figure 6). As preserved, the vomer is clearly exposed along the midline of the rostrum, but obvious   crests. As far as can be told, the vomerine crest extends posteriorly to the end of the vomer, or very nearly so, remaining sharp along the entire way.

Pterygoid
The right pterygoid is nearly complete, except for the hamular process and some minor damage to the rim of the Eustachian notch ( figure 6). In ventral view, the pterygoid is broadly exposed between the palatine and the falciform process of the squamosal, and forms approximately one-quarter of the rim of the foramen pseudovale ( figure 8a). Laterally, a small portion of the pterygoid extends on the wall

Alisphenoid
In lateral view, the area of the temporal fossa exposing the alisphenoid is damaged. Nevertheless, it appears that the external surface of this bone occupies a relatively small space between the parietal, squamosal and pterygoid, and-as far as can be told-anteriorly contributes to the rim of the orbital fissure. Details of this region, including the position and size of the orbitosphenoid, are too poorly preserved for further description.

Periotic
Both periotics are preserved in situ, but only the ventral surface of the right periotic is currently accessible. In ventral view, the anterior process is slightly longer than the pars cochlearis and blade-like, with its entire ventral border forming a sharp crest (figures 8 and 9). A laterally compressed anterior process also occurs in some cetotheriids, such as Herpetocetus (e.g. H. transatlanticus, USNM 182962) and Kurdalagonus mchedlidzei (NMRA 10476/1), but strongly contrasts with the markedly inflated posterior portion of the anterior process in Parietobalaena (e.g. P. palmeri, USNM 10668, 16119). There is no anterior bullar facet. The anterior pedicle is aligned with the anterior border of the pars cochlearis, and firmly fused to the body of the periotic. A rounded tubercle immediately above the pedicle (figure 9) is presumably homologous with the fused dorsal portion of the accessory ossicle, but there is no evidence of a suture or any part of the fovea epitubaria. The lateral tuberosity is robust, short, somewhat pointed and located posterolateral to the anterior pedicle. The border of the mallear fossa is indistinct. The distal opening of the facial canal is located slightly anterior to the fenestra vestibuli. Both the latter and the fossa for the stapedial muscle are obscured by the in situ stapes. The pars cochlearis is bulbous, with a smooth ventral surface and a rounded anteromedial corner ( figure 8). The fenestra cochleae is flush with the medial and posterior borders of the pars cochlearis, rather than being recessed as in most balaenopterids. The posterior process is elongate, somewhat squared in outline and slightly expanded distally. The facial sulcus occupies the posterior half of the posterior process, and is partially floored by a narrow posteroventral flange (sensu [7]) (figure 9). Along its anterior border, the posterior process furthermore bears a well-developed anteroventral flange (sensu [7]), which delimits the extension of the paroccipital concavity anterior to the facial sulcus, and may correlate with the presence of an enlarged cartilaginous tympanohyal [15]. A similar condition exists in virtually all known cetotheriids [15], but is absent or rudimentary in Parietobalaena (e.g. P. palmeri, USNM 10668, 16119; Parietobalaena sp. SMNH VeF62). Laterally, the anteroventral flange becomes lower and terminates approximately 13 mm before the end of the facial sulcus.
In medial view, the outline of the anterior process is obscured by matrix and the lateral lamina of the pterygoid, but it appears that at least the anteroventral angle is pointed and triangular (figure 8b). Anterior to the pars cochlearis, the anterior process bears an approximately 12 mm long ridge for the origin of the tensor tympani muscle. A deep, step-like promontorial groove runs across the entire length of the pars cochlearis and terminates just anterior to the fenestra cochleae in a large (approx. 5 mm), posteromedial embayment. There is no posterior cochlear crest, and the pars cochlearis consequently does not approach the crista parotica. In posterior or posterolateral view, the fenestra cochleae is circular and well separated from the aperture for the cochlear aqueduct ( figure 9).
On the posterior process, just posterior to the anteroventral flange, a small (1 mm) foramen opens into a posteroventral sulcus (new term) that runs across the ventral surface of the posteroventral flange ( figure 9). Lateral to the facial sulcus, the outer portion of the posterior process slightly widens, abruptly flattens and becomes oriented dorsolaterally, thus forming a distinct external surface that is exposed on the lateral skull wall. The same condition, often more pronounced, is characteristic of all cetotheriids [1,3]. A somewhat similar morphology also occurs in some specimens of Parietobalaena (e.g. P. palmeri, USNM 10668), but the external surface of the posterior process, if present at all, is generally smaller, less well defined, not laterally expanded and not clearly offset from the facial sulcus.

Tympanic bulla
The right tympanic bulla is complete and preserved in situ, whereas the left bulla is detached but missing its ventral half and the posterior portion of the outer lip. In dorsal view (figure 10a), the involucrum is narrow adjacent to the Eustachian outlet and then gradually widens posteriorly. There is no sign of an involucral incisure (sensu [33]), but the involucrum does bear fine transverse sulci all along its dorsal surface. The involucral and main ridges (sensu [14]) are convex medially.
The portion of the outer lip anterior to the lateral furrow is inflated and rounded (figure 10e). The dorsal border of the sigmoid process is oriented approximately transversely, not obviously twisted, and located at just under two-thirds of the total length of the bulla. Posteriorly, the sigmoid process does not overhang the conical process, but is connected to it via a narrow ridge. As far as can be told, the conical process itself is straight and narrow, with its apex being neither thickened nor deflected. The posterior pedicle is elongate and located close to the posterior border of the bulla, with no sign of a posterior extension of the involucrum as seen in, for example, Balaenoptera musculus [34]. Internally, the pedicle is excavated by a branch of the tympanic cavity that probably marks the ancestral position of the posterior sinus [15].
In lateral view, the lateral furrow is well developed, step-like and oriented approximately vertically (figure 10d). Dorsal to the furrow, the mallear ridge is low and narrow, and oriented posterodorsally. The sigmoid cleft is straight and also largely vertical, resulting in the absence of a distinct ventral border of the sigmoid process (figure 10e). The conical process is parabolic in outline, and separated from the posterior pedicle by a narrow, posterolaterally directed channel.
In medial view, the posterior portion of the bulla is divided into inner and outer posterior prominences (figure 10b). As in all described chaeomysticetes except eomysticetids, the bulla has rotated medially around its long axis (relative to the condition in archaeocetes), so that the inner posterior prominence now faces dorsally, and the outer posterior prominence ventrally. As a result, the dorsal surface of the involucrum is divided into (i) a slightly depressed anterior portion marking the position of the Eustachian outlet; (ii) a horizontal central portion; and (iii) the bulbous inner (dorsal) posterior prominence. The involucral and main ridges are well developed and separated from each other by a broad, but relatively shallow, median furrow. The main ridge is oriented obliquely and traverses the medial face of the bulla from its posteroventral to its anterodorsal corner. At the level of the Eustachian outlet, the main and involucral ridges converge.
In ventral view, the anteromedial corner of the bulla is somewhat angled and the ventral surface of the bulla transversely convex (figures 6 and 10c,e). There is no anterolateral shelf. Inside the tympanic

Malleus and stapes
In anterior view, the anterior process of the malleus is robust and excavated by the groove for the chorda tympani. Dorsally, the anterior process is closely apposed to the dorsomedial corner of the sigmoid process, without, however, being entirely fused to it as in recent balaenopterids. At its base, the anterior process is surrounded by a weakly developed rim, which posteriorly appears to be confluent with the tympanic sulcus. In posterior view, the head of the malleus is rounded and separated from the tubercle by a deep oblique groove (figure 10f ). The incudal facets are perpendicular to each other, with the vertical facet being roughly twice as large as the horizontal one. The vertical facet is flattened, whereas its horizontal counterpart is convex and somewhat dome-like. The tubercle is oriented medially and terminates in a ventrally curved, pronounced manubrium. The muscular process appears to be indistinct. Most of the stapes is obscured by its in situ position inside the fenestra vestibuli. Nevertheless, as far as can be told, the head of the stapes is separated from the crura by a short neck, and the stapedial foramen is patent.
Most studies agree that some or all crown mysticete lineages originate from within 'cetotheres' [6,8,24], but the structure of this early phase of baleen whale evolution has remained largely obscure. Up to a point, this situation is unavoidable: basal taxa share fewer of the defining features of the clade they belong to, and hence are also more difficult to identify. In this context, the particular morphology of Tiucetus rosae is both striking and informative. Our phylogenetic analysis recovers Tiucetus as the basalmost cetotheriid (figure 11), in line with its relatively large posterior process, enlarged paroccipital concavity, and possession of a posteroventral sulcus (see below). At the same time, the morphology of the cranial vertex of Tiucetus entirely differs from that of typical cetotheriids-e.g. in lacking posteriorly convergent ascending processes of the maxillae, and in having a large exposure of the parietal on the vertex-and instead closely resembles that of Diorocetus and Parietobalaena. Tiucetus therefore, more than any other described species, bridges the morphological gap between 'cetotheres' sensu lato and one of the major mysticete families, and sheds light on the cranial architecture of basal cetotheriids.
Sister to, and possibly included within, Cetotheriidae is a previously identified clade comprising Diorocetus hiatus and Thinocetus arthritus [2,4]. Both species resemble cetotheriids in having a relatively large posterior process, a deep facial sulcus, and a somewhat posteriorly elongated angular process of the mandible bearing a fossa for the medial pterygoid muscle (as inferred for Herpetocetus morrowi [6]). Moreover, the placement of D. hiatus as a basal cetotheriid is supported by the results of another recent analysis [8]. Nevertheless, the phylogenetic position of D. hiatus has generally been extremely variable, leading to its interpretation as a stem balaenopteroid [2,10,15], stem plicogulan [38], sister to a clade comprising balaenopteroids and cetotheriids [6,39], or member of a clade including both cetotheriids and other 'cetotheres' sensu lato [9,11]. Given this level of disagreement, we prefer not to formally reassign Diorocetus and Thinocetus to Cetotheriidae, pending a more detailed re-examination of these species.
in I. laticephalus; unclear in P. campiniana). Nevertheless, most of these species remain poorly known, owing to various issues with the known (type) material: the holotypes of 'Aglaocetus' patulus, 'Diorocetus' shobarensis, Parietobalaena yamaokai and Tiphyocetus temblorensis are badly damaged; those of Parietobalaena campiniana and P. palmeri are juveniles; and the periotic morphology remains largely or entirely unknown for 'Diorocetus' chichibuensis, 'D.' shobarensis, Pinocetus polonicus and T. temblorensis. Further preparation and redescription, e.g. of one of the paratypes of P. yamaokai (HMN F00042), is required to clarify taxonomy and relationships. fossil specimens where breakage or even a thin layer of matrix could easily obscure their presence. In cetotheriids, the sulcus appears particularly broad and well-defined, and in this form may characterize either the family, or a subclade within it. The function of the foramen and sulcus currently remain uncertain, although their location is suggestive: in Eubalaena, the sulcus opens near the unusually large, ossified tympanohyal (figure 12e), which is rivalled in size only by Metopocetus [7]; in balaenopterids and cetotheriids, it opens into the anterior portion of the paroccipital concavity, which likewise may house an enlarged cartilaginous tympanohyal [15]. Together, these observations suggest that the posteroventral foramen and sulcus may carry nerves or blood vessels associated with the articulation of the hyoid apparatus to the basicranium. MicroCT scans of isolated periotics and dissections of the auditory region of extant mysticetes may shed further light on this issue.