A new member of fossil balaenid (Mysticeti, Cetacea) from the early Pliocene of Hokkaido, Japan

The family Balaenidae includes two genus and four extant species. Extinct balaenids are known for at least four genus and 10 species. The oldest known record of members of the Balaenidae is known from the early Miocene, but still need more early members of the family to provide better phylogenetic hypotheses. FCCP 1049 from the lower part of the Chippubetsu Formation, Fukagawa Group (3.5–5.2 Ma, Zanclean, early Pliocene) was preliminary described and identified as Balaenula sp. by Furusawa and Kimura in 1982. Later works discussed that FCCP 1049 is different from the genus, and is placed in different clade from Balaenula astensis. The result of our phylogenetic analysis places FCCP 1049 basal to Balaenella brachyrhynus, and is again separated from B. astensis. In this study, FCCP 1049 is re-described and named as Archaeobalaena dosanko gen. et sp. nov. Archaeobalaena dosanko is distinguishable from other balaenids by having a deep promontorial groove of the pars cochlearis of the periotic. Archaeobalaena dosanko can be differentiated from other balaenids, except Morenocetus parvus by having a slender zygomatic process, and posteriorly oriented postorbital process in dorsal view. Archaeobalaena dosanko adds detailed skull, periotic and bulla morphologies for the earlier balaenids.


Introduction
The family Balaenidae includes two genus and four extant species. Extinct balaenids are known for at least four genus and 10 species [1][2][3]. The modern balaenid body length reaches 17-20 m [4]. The oldest known nominal species of the members of the Balaenidae is known from the early Miocene of Patagonia, Argentina [1], and its body length was estimated as 4.8-6.2 m [5]. Our knowledge of balaenid past diversity is growing [1,3,[5][6][7][8], but still needs earlier members of the family to hypothesize their phylogeny.
A Pliocene balaenid, the genus Balaenula had been treated as 'a taxonomical basket, where all the small-sized balaenids were put' in history [2]. The type species of the genus Balaenula balaenopsis from Antwerp was established by Van Beneden [9], but the holotype is doubtfully recognized as an individual [1]. Supposed similar cases of establishing new species with mixed individuals was centered on a cetotheriid, Herpetocetus scaldiensis of Van Beneden [9], which was summarized by Deméré et al. [10]. The second species, Balaenula astensis from the late early Pliocene of Villafranca d'Asti, was established by Trevisan [11] and re-described by Bisconti [8].
Furusawa & Kimura [12] preliminary identified FCCP 1149 from an early Pliocene sediment in Hokkaido, Japan, as Balaenula sp. based on having a low triangle-shaped occipital with a depression at the centre, an acute angle of the nuchal crest against the plain in posterior view and a rounded exoccipital placing below to the ventral margin of the occipital condyle. The study mentioned that these features can be seen only on B. balaenopsis. Later, Bisconti [8] mentioned that FCCP 1049 is different from Balaenula based on an extended temporal fossa ( p. 47). Indeed, several diagnoses of the genus Balaenula were reported by Bisconti [2]. One of the diagnoses for the genus, not having a protruding nuchal crest is not seen on FCCP 1049. FCCP 1049 shows a lateral expansion at the posterior part of the supraoccipital crest. One of the most recent phylogeny works found that FCCP 1049 did not form a clade with B. astensis [1].
To expand diversity and morphological information for understanding the earlier balaenid evolution, we update the identification of FCCP 1049 and its geological age, and also re-describe the specimen. [23]. The age of Ops tuff is 4.5 ± 0.7 Ma. Thus, the age of FCCP 1049 can be taken as 3.5-5.2 Ma (Zanclean, early Pliocene) with wider ranges. Very near to the locality (about 10-20 km away), Numataphocoena yamashitai and Herpetocetinae gen. et sp. indet. and some Mysticeti indet. materials from the early Pliocene (about 4.5-3.5 Ma) possibly be cetaceans of the same age [25][26][27][28][29]. From the same area, there are some late Miocene Mysticeti reports: type specimen of Miobalaenoptera numataensis (6.5-6.8 Ma) and a referred specimen Herpetocetus sp. (7.7-6.8 Ma), but are not simultaneous records of FCCP 1049 [30,31]. Etymology. Dosanko means people and things born in Hokkaido, northern Japan, originating from the native horse of Hokkaido. Diagnosis. Among the Balaenidae, A. dosanko uniquely has a deep promontorial groove of the pars cochlearis of the periotic (character 151). Archaeobalaena dosanko can be differentiated from other balaenids, except Morenocetus parvus by having a slender zygomatic process, and posteriorly oriented postorbital process in dorsal view (character 38). Archaeobalaena dosanko can be differentiated from M. parvus by having a slender and laterally tilted down of the supraorbital process. Archaeobalaena dosanko can be differentiated from all balaenids, except the Balaena and Eubalaena by having a weakly and laterally projected lateral margin of the nuchal crest.
Comparison with more basal balaenids (M. parvus and Peripolocetus vexillifer), A. dosanko can be differentiated by having a more or less same length of the anterior process and pars cochlearis of the periotic (character 139), hypertrophied and blade-like lateral tuberosity (character 144), posterior process of the periotic orienting a right angle to the axis of the anterior process in ventral view (character 170), laterally reduced involucral ridge in dorsal view (character 181), the dorsolateral surface of involucrum forming a continuous rim (character 190), and a flat anteromedial portion of the ventral surface of tympanic bulla (character 195). Comparison with more late diverging balaenids (B. astensis, Balaenella brachyrhynus and so on), A. dosanko can be differentiated by having the medial lobe of the tympanic bulla (character 187). Comparison with crown balaenids (Balaena spp. and Eubalaena spp.), A. dosanko can be differentiated by having a thickened and flat lateral surface of the orbital rim of the supraorbital process (character 40), pyramidal process (character 141), laterally exposed and distinct compound posterior process from the lateral skull wall (character 172), no crest on the parietosquamosal suture (character 93), no hypertrophied suprameatal fossa (character 162), and no transverse creases on the dorsal surface of the involucrum (character 191).

Skull
General features of the skull. A preserved left side of the skull has a slender supraorbital process with weakly projected pre and postorbital processes, slender zygomatic process and laterally weakly protruded nuchal crest. The temporal fossa is anteroposteriorly short and mediolaterally wide. Maxilla. A supposed ascending process is wide and flat (about 87 mm wide, figures 2 and 3). Its posterior end still has a matrix and does not show the posterior border. The infraorbital plate covers the anterior border of the supraorbital process. The plate has a transverse ridge on the dorsal surface.
Frontal. The supraorbital process is slender and tilts down laterally. The mid part of the supraorbital process is anteroposteriorly shorter, and the lateral part is anteroposteriorly longer, because the posterior border of the supraorbital process is anteriorly excavated at the middle ( figure 2). There is a weak transverse orbitotemporal crest on the dorsal surface of the supraorbital process. The lateral margin of the orbit in dorsal view is more or less straight. In lateral view (figure 4), an anteroposteriorly thin postorbital process projects posteroventrally, and it has a rounded outline in lateral view. The preorbital process has a rounded outline and is more robust than the postorbital process. Between the preorbital and postorbital processes, a dorsally excavated orbital rim is formed. The medial end of the frontal contacts with the parietal posteriorly at the anterior portion of the temporal fossa. The frontal forms a part of the vertex, and is partially covered by the ascending process of the maxilla. Medial to the ascending process, the frontal is exposed at a supposed centre of the vertex, but the interfrontal suture is not preserved. The frontal ridge might be absent, not like M. parvus [1]. In ventral view, the optic canal has a short and deep medial part and long and shallow lateral part, and is restricted by the pre and postorbital ridges running mediolaterally.   Exoccipital and supraoccipital. In dorsal and posterior views (figures 2 and 7a,b), the exoccipital is wide and occupies about 66% of the bizygomatic width. The exoccipital forms a rounded ventrolateral part of the occipital shield. In ventral view, the exoccipital is a mediolaterally long plate (about 27 mm long), but the medial end is strongly worn. The lateral part of the exoccipital forms a posterior part of the fossa for the posterior process. The occipital condyle is flat and only weakly projects posteriorly. There is a shallow dorsal condyloid fossa dorsal to the occipital condyle. The foramen magnum preserves dorsal side and its dorsal margin becomes thicker. The supraoccipital is a long triangle (figure 2). The lateral margin of the supraoccipital and exoccipital is incomplete, but shows a small protrusion slightly dorsal to the level of the supramastoid crest

Periotics
The periotics (figures 6 and 8-10 and table 1) have a robust anterior process, small globular pars cochlearis and large compound posterior process of the tympanoperiotic. The anterior process is short. The size and shape of the lateral tuberosities and also the posterior processes are slightly different on the right and left sides (table 1). The right one is a broad triangle with an extension that projects laterally. The left one is smaller than the right side. Anterior to the lateral tuberosity, there is a shallow notch, which might be the anteroexternal sulcus. Medial to the lateral tuberosity, the anterior pedicle of the tympanic bulla is anteroposteriorly long, rectangular, and is placed on the ventral surface of the anterior process. Between the anterior pedicle and lateral tuberosity, there is a large and weakly depressed plane for the sigmoid process of the tympanic bulla (figure 8). A shallow wide mallear fossa is located slightly posterior to the level of the lateral tuberosity. In dorsal view, the medial part of the periotic has two large processes: the pyramidal process anteriorly and the dorsal tuberosity posteriorly. Between these processes, is the anteroposteriorly long and strongly curved dorsal rim of the suprameatal fossa [35]. Ventral to the rim, the suprameatal fossa is large and shallow. royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 192182 An anteroposteriorly long globular pars cochlearis covers the body of the periotic anteriorly as a thin bone and it forms the anterior incisure. In the anterior incisure, there is a huge dorsoventrally long elliptical hiatus fallopii (4.7 mm high, 3.2 mm wide) opening anteriorly. The medial surface of the pars cochlearis has a deep, anteroposteriorly long and weakly curved promontorial groove. Just posterior to the hiatus fallopii, a large internal acoustic meatus (9.1 mm long, 5.2 mm high) opens medially and contains the proximal opening of the facial canal and dorsal vestibular area. The proximal opening of the facial canal is more or less circular (5.0 mm high, 4.3 mm long). Just posterior to the facial canal, and clearly separated from it by a low transverse crest, there is a circular dorsal vestibular area (5.2 mm high, 5.9 mm long). In the dorsal vestibular area, there is a small foramen singulare (1.0 mm in diameter) anterodorsally and two depressed areas, which might be the spiral cribriform tract and area cribrosa media. The spiral cribriform tract and area cribrosa media are connected and form a weakly curved depression. Posterior to these openings, there are two foramina:

Tympanic bulla
The tympanic bullae (figures 6, 8, 11 and 12 and table 1) are longer than their width, but with a swollen anterior portion in ventral view, and dorsoventrally compressed in anteroposterior view. The anterior lobe of the tympanic bulla is swollen laterally. The ventral to the lateral surface of the anterior lobe has a blunt ridge, which is a part of the main ridge running to the medial surface. Posterior to the anterior lobe, a large triangular sigmoid process projects laterally. Its dorsal most point bends posteriorly (figure 12a). The sigmoid process has thick margins, about 8.0 mm (anteroposterior length). Anterior to the sigmoid process, a deep lateral furrow runs transversely and separates the sigmoid process and anterior lobe. Anterior to the sigmoid process, the rim of the outer lip becomes thicker, which is the mallear ridge. The mallear ridge contains structures such as the sulcus for the chorda tympani and fossa for the malleus

Mandible
An incomplete right mandible is weakly laterally bowed, especially the anterior part ( figure 13 and table 2). The most anterior part has an anteroposteriorly long mental foramen, which also opens anteriorly. The   14k and table 3). A side is convex, which can be identified as the ventral side (figure 14k). On the lateral margin, there are a couple of thicker parts, which might be a surface for the ribs. Rib. Five ribs are preserved ( figure 15). The rib in figure 15b is wide, which might be the most anterior rib. The longest one, figure 15e, is about 560.0 mm long and 65.0 mm wide, and its restoration using clay might overestimate its length.

Phylogenetic analysis and comparison
The phylogenetic position of A. dosanko gen. et sp. nov. (FCCP 1049) was analysed using the matrix of Buono et al. [1], which was derived from the matrix of Marx & Fordyce [37]. This study modified codings of A. dosanko with the direct examination (electronic supplementary material, S1-S4) and contains 257 morphological characters and 43 taxa. Percentages of coded data of A. dosanko are 46% (includes soft tissue characters), 48% (excludes soft tissue) and 89% for the ear bones.
The matrix was managed using MESQUITE 2.75 [38]. Analysis was performed with TNT v. 1.5 [39]. All of the characters were treated as unweighted and unordered with backbone constraint of extant taxa, based on a topology of the molecular tree [40]. The analysis used New Technology Search with recover minimum length trees = 1000 times.
The phylogenetic analysis shows two shortest trees of 960 steps each. The strict consensus trees (figure 16; electronic supplementary material, S5) are slightly different from that of the equally weighted analysis of Buono et al. [1] in that the most basal balaenid is not P. vexillifer but M. parvus, and A. dosanko (Balaenula sp. in the previous study) is placed more basal to B. astensis. Indeed, the branch lengths of the previous phylogenetic hypothesis around them were relatively low [1].
Archaeobalaena dosanko is identical to the all named balaenids. Here, we compare A. dosanko with closely related B. astensis and Balaenella brachyrhynus (table 4). Comparison with Balaenella brachyrhynus, A. dosanko shows a more strongly curved posterior margin of the supraorbital process in dorsal view, much weaker supramastoid crest, more rounded postorbital process in dorsal view, ventrally almost closed frontal groove by the pre and postorbital ridges, much smaller lateral process royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 192182    As is mentioned above, the type species of the genus Balaenula: the holotype of B. balaenopsis is difficult to recognize as an individual [1]. Indeed, the long and triangular anterior process of the periotic on the type specimen probably does not belong to the Balaenidae, but Balaenopteroidea. Based on illustrations of previous studies [2,41], the periotic seems having some balaenopteriid diagnoses such as a narrow and triangular anterior process [14,42], but lacks other diagnoses for the family such as a well-defined fossa for the malleus [14] and transversely elongated pars cochlearis [43]. Thus, it might be a stem balaenopteroid. Its overall shape in ventral view is similar to that of Tiphyocetus temblorensis of Kellogg [44] by having a slander anterior process and globular pars cochlearis, but morphologies of the internal acoustic meatus and lateral tuberosities are different.
However, the holotype skull of B. balaenopsis belongs to the Balaenidae, and can we compare here. In comparison with B. astensis + B. balaenopsis, A. dosanko shows weaker pre and postorbital processes in lateral view, and weaker orbitotemporal crest only on the medial part of the supraorbital process. In comparison with B. balaenopsis, A. dosanko shows a more or less straight lateral margin of the supraorbital process, and much weaker supramastoid crest. Some Balaenula sp. specimens from the late Miocene to early Pliocene, and Late Pliocene of California have been mentioned, but not photographed and illustrated [45].

The supramastoid crest
The supramastoid crest of balaenids are generally developed, but morphological change of the supramastoid crest cannot be thought simple, like developing from incipient to large through balaenid evolution. The oldest known balaenid M. parvus shows a small supramastoid crest on the holotype, but just a ridge along the dorsal surface of the zygomatic process on the referred specimen (MPL 5-15) [1].
As mentioned above, having a low and rounded supramastoid crest was considered a diagnosis of the genus Balaenula. However, early Pliocene balaenids (A. dosanko, Balaenula spp. and Balaenella brachyrhynus) have relatively large supramastoid crests among the Balaenidae. Among them, the conditions of the supramastoid crest are varied as compared with above (see also table 4), especially B. balaenopsis has a larger and dorsally strongly projected supramastoid crest than those of A. dosanko, B. astensis and Balaenella brachyrhynus.
The rounded and low supramastoid crest had emerged already among the Balaenidae during the early Miocene as an incipient condition, then enlarged by the early Pliocene, and some kept the condition or others had the crest as a smaller secondary. In short, the morphological change of the supramastoid crest is not simple. The supramastoid crest is attached to the temporal fascia [50,51]. Thus, these conditions of the supramastoid crest might be related to the size and/or orientation of the temporal muscle. At least, from M. parvus to Pliocene balaenids (including A. dosanko), the width of the skull was increased twice, which was the result of lateral elongation of the supraorbital and zygomatic processes, and a mediolaterally expanded temporal fossa. This size and shape change of the skull is effected to shape the modification of the temporal muscle to control the mandibles effectively. However, the possibility of anteroposterior expansion of the temporal muscle might be limited owing to the elongation of the rostrum and whole body proportion. Thus, the temporal muscle earned a larger physically effective attachment point: the supramastoid crest. Many other factors such as width of the supraorbital process, degree of telescoping, arching of the rostrum and orientation of the skull can be thought of. However, the number of rostrum records is not enough to consider.

Conclusion
Archaeobalaena dosanko gen. et sp. nov. (FCCP 1049) represents an archaic balaenid from the early Pliocene, Zanclean (3.5-5.2 Ma) lower part of the Chippubetsu Formation, Hokkaido, Japan. Archaeobalaena dosanko is distinguishable from another balaenids by having a deep promontorial groove of the pars cochlearis of the royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 192182 periotic. Archaeobalaena dosanko can be differentiated from other balaenids, except M. parvus by having a slender zygomatic process, and posteriorly oriented postorbital process in dorsal view. The result of phylogenetic analysis places A. dosanko more basal to B. astensis. Archaeobalaena dosanko adds detailed skull, periotic and bulla morphologies for the earlier balaenids.
Data accessibility. The data for the phylogenetic analysis are available as electronic supplementary material. The LSID for this publication is: urn.lsid:zoobank.org:pub:322CDB25-4971-46A1-B40F-6200A8CCAF03.
Authors' contributions. Y.T. carried out the phylogenetic analysis and wrote the paper. H.F. and M.K. revised the manuscript. Competing interests. The authors declare that they have no competing interests. Funding. There is no funding for this paper.