A small whale reveals diversity of the Eocene cetacean fauna of Antarctica

Abstract Cetacean fossils have been recorded from middle and late Eocene deposits on Seymour Island since the beginning of the twentieth century and include fully aquatic Basilosauridae and stem Neoceti. Here, we report a small cetacean vertebra tentatively referred to as Neoceti from the late Eocene of Seymour Island. It shows a mosaic of traits, some of which are characteristic of early Neoceti (anteroposteriorly long transverse processes; a ventral keel on the ventral side of the centrum; thin pedicles of the neural arch), whereas others are shared with Basilosauridae (low-placed bases of the transverse processes). However, some traits are unique and may be autapomorphic: presence of separate prezygapophyses on the vertebra at the thoracic/lumbar boundary and a proportionally short centrum. Both traits imply a fast swimming style, which is characteristic of modern dolphins rather than Eocene cetaceans. Thus, this specimen can be identified as Neoceti indet., with some hypothetical odontocete affinities. Along with a few other Eocene whale taxa, it seems to be among the earliest known members of Neoceti on Earth. The finding of small and fast-swimming Neoceti in Antarctica also demonstrates early diversification of cetaceans and ecological niche partitioning by them dating back as early as the late Eocene.


Introduction
Seymour Island is rich in late Cretaceous and Palaeogene fossil records of terrestrial and aquatic tetrapods (Reguero 2019, Mörs et al. 2020. Palaeogene, secondarily aquatic tetrapods are represented by penguins (Jadwiszczak & Mörs 2017) and cetaceans, with all of the cetacean records coming from the La Meseta and Submeseta formations. Cetacean remains have been recorded from Seymour Island since the beginning of the twentieth century, when Wiman (1905) described two cetacean caudal vertebrae as Zeuglodon sp. (Fig. 1c). Eighty years later, vertebrae and sternum remains were described by Borsuk-Bialynicka (1988). Fordyce (1989) and Buono et al. (2016Buono et al. ( , 2019 reported isolated teeth, mandibles, vertebrae and ribs of Eocene whales and referred to them as Basilosauridae (the earliest fully aquatic cetaceans) or Cetacea indet. The toothed mysticete Llanocetus denticrenatus (Mitchell 1989, Fordyce & Marx 2018 has been found in the late Eocene Submeseta Formation of Seymour Island. It represents Neoceti, the crown group of cetaceans that includes baleen whales (Mysticeti) and toothed whales (Odontoceti). Another giant specimen of Llanocetus sp. is known from its isolated teeth ). In addition, a right innominate bone resembling that of another early mysticete, Mystacodon selenensis (Muizon et al. 2019), was found in the same formation . Therefore, all of the cetaceans known from Seymour Island are referred to as fully aquatic cetaceans (Pelagiceti, a group pooling together Basilosauridae and Neoceti).
Here, we report a small cetacean vertebra tentatively referred to as Neoceti from the late Eocene of Seymour Island and compare it with other early whale taxa.

Geological setting and palaeoenvironmental context
The Submeseta Formation is located on Seymour Island, 100 km east of the Antarctic Peninsula, within the James Ross Basin (Fig. 1a & b). The formation and the underlying La Meseta Formation consist of a succession of sedimentary marine beds of sandstone, siltstone and shell, which are divided into seven lithofacies units referred to as 'Tertiary Eocene La Meseta' stratigraphic units (TELMs), representing the 720 m-thick fill of a 7 km-wide incised valley system (Sadler 1988, Marenssi et al. 2002. The two Eocene formations form together Antarctic Science 33(1), 81-88 (2021) © The Author(s), 2020. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.  (Sadler 1988, Montes et al. 2013). Both formations have undergone minimal burial and diagenetic alterations (Marenssi et al. 2002). Ages based on 87 Sr/ 86 Sr chemostratigraphy from bivalve carbonates indicate that the Submeseta Formation is bound with TELM 6, being c. 41 Ma or younger, and the top of TELM 7 lying at the Eocene/Oligocene boundary (Douglas et al. 2014).

Materials and methods
Specimen NRM-PZ M8154 is a single torso vertebra. The specimen described here was collected by TM in January 2012 at 64°14'39.42"S, 56°36'52.62"W in a gully on the slope below and south-east of the Argentinian Marambio Base (Locality 9 in Fig. 1c). Here, marine sediments of the Submeseta Formation (Montes et al. 2013), or TELM 7 according Sadler (1988), are exposed ( Fig. 1c). Because the vertebra was not found in situ, it is unclear whether its provenance is level 38 or 39 of the Submeseta Formation (Fig. 1d). There is agreement on this part of the depositional succession on Seymour Island being late Eocene (Priabonian) in age (e.g. Douglas et al. 2014, Buono et al. 2016, Kriwet et al. 2016. The specimen is housed in the palaeozoological collections of the Swedish Museum of Natural History, Stockholm. Measurements were taken point to point on the anterior side of the specimen if not specially mentioned. Anatomical terminology follows Uhen (2004). The anatomical abbreviations used in these measurements are as follows: H = centrum height and L = centrum length.
Analysis of the geometric morphometry of the anterior surface of the centra of the anterior lumbar vertebrae was conducted for the specimen under study and those of an additional 20 cetacean taxa, including 2 members of Protocetidae, 8 members of Basilosauridae and 10 members of Neoceti (4 mysticetes and 6 odontocetes). For M. selenensis, the only available posterior thoracic vertebra was used for the analysis. Ten landmarks were identified in all analysed specimens and digitized using TpsDig software (Supplemental Fig. S3). After the

Results
Specimen NRM-PZ M8154 is shown in (Fig. 2). This new specimen represents a single vertebra of an adult cetacean close to physical maturation. It is tentatively identified as the posterior-most thoracic or anterior-most lumbar vertebra due to the large and triangular neural canal ( Fig. 2a & b) and low, wide bases of transverse processes (Fig. 2c). The anterior and posterior articular surfaces of the centrum are roughly trapezoid and the lateral surfaces of the centrum are slightly transversely concave. In lateral view, the centrum is rectangular, and a ventral keel is seen on its ventral surface (Fig. 2e). The neural arch is best preserved near the anterior surface of the vertebra. The anteriorly directed prezygapophyses are short and robust. A laterally directed left metapophysis is preserved on the left lateral surface of the neural arch; it is slightly longer and thinner than the prezygapophysis (Fig. 2d). Transverse processes are attached to the ventral part of the centrum, with the bases situated approximately at the level of a quarter of the centrum height. The base of the right transverse process is relatively well preserved. It is anteroposteriorly elongated, being almost as long as the centrum. There is no significant cortical bone on the vertebral surface, and the whole centrum is composed of tight spongious bone, as is seen from natural fractures. Both epiphyses are completely fused to the centrum, and the fusion suture of the posterior epiphysis is only poorly visible on the right lateral surface of the centrum (Fig. 2c).
The length of centrum is 62 mm and its width at its anterior end is 78 mm. The heights of centrum at its anterior and posterior ends are 78 and 76 mm, respectively. The total height of the vertebra (with the preserved part of the neural arch) is 161 mm.

Comparison
NRM-PZ M8154 is distinct among early extinct cetaceans with known posterior-most thoracic or anterior-most lumbar vertebrae (Supplemental Table SI) by the presence of relatively large, separate and anteriorly directed prezygapophyses (Fig. 3). Such an anatomy can be seen in the mid-posterior thoracic vertebrae of Protocetidae (anterior to T12 in Aegicetus gehennae, with the total count of thoracics as 15 (Gingerich et al. 2019)) and Basilosauridae (anterior to T13 of 17 thoracics in Dorudon atrox (Uhen 2004) and to T14 of 20 thoracics in Cynthiacetus peruvianus (Martínez-Cáceres et al. 2017)). However, all such vertebrae, unlike NRM-PZ M8154, also have anteroposteriorly and proximodistally short, highly placed bases of transverse processes. Unlike NRM-PZ M8154, protocetids have the centra constricted in the middle part and with highly placed bases of transverse processes in all posterior thoracic and anterior lumbar vertebra (Hulbert 1998, Gingerich et al. 2019. By contrast, the posterior-most thoracic vertebrae of basilosaurids bear relatively low-placed bases of transverse processes, as in NRM-PZ M8154 (Figs 3 & 4); however, their prezygapophyses are reduced as concave facets that are far smaller than metapophyses, and the neural canal is semi-circular (rather than triangular, as in NRM-PZ M8154). The triangular neural canal is typical of the lumbar vertebrae of many cetaceans (as well as the low-placed bases of the transverse processes) (Fig. 4), which, however, have only rudimentary prezygapophyses capped by metapophyses. Therefore, NRM-PZ M8154 may be identified either as the posterior-most thoracic vertebra with an unusually large neural canal or an anterior lumbar vertebra.

84
However, the presence of large prezygapophyses in the vertebral column is unique for both interpretations. Furthermore, NRM-PZ M8154 differs in its anatomy and proportions from all known lumbar vertebrae of Eocene cetaceans. Its centrum is similar in proportions to protocetids (Hulbert 1998, Gingerich et al. 2019; however, it greatly differs from those in its derived, rectangular shape in lateral view, presence of the ventral keel and long transverse processes. Its centrum is slightly shorter and higher than the anterior lumbar vertebrae of D. atrox (Uhen 2004). The overall shape of the centrum resembles the lumbar vertebrae of C. peruvianus show that its posterior-lying vertebrae could be even longer than those of NRM-PZ M8154. Another fossil whale vertebra from Seymour Island, ZPAL M-VII/1, described by Borsuk-Bialynicka (1988), has similarly elongated transverse processes. However, it is larger than NRM-PZ M8154, and its centrum is proportionally longer. Finally, NRM-PZ M8154 lacks two vascular foramina separated by a bony septum, which are present on the ventral side of the lumbar vertebrae in all of the basilosaurids mentioned here.
A few derived traits are shared by NRM-PZ M8154 and Neoceti. First, anteroposteriorly long, transverse processes are common within Neoceti; they are particularly long in mysticetes (Fig. 3). Second, there is a distinct ventral keel on the centrum of NRM-PZ M8154 (Fig. 2): such a keel is observed in many extinct and extant Neoceti species (Emlong 1966, Buchholtz 2011 and some protocetids (Gingerich et al. 2019), but not in basilosaurids (Uhen 2004, Martínez-Cáceres et al. 2017. Compared with Oligocene cetaceans, NRM-PZ M8154 is similar to the early odontocetes Mirocetus riabinini, Albertocetus meffordorum (Boessenecker et al. 2017) and Ankylorhiza tiedemani (Boessenecker et al. 2020) in the triangular shape of the neural canal (Fig. 4). However, all of these odontocetes have proportionally longer vertebrae than NRM-PZ M8154: their L/H ratios are 1.07 (L1) and 1.34 (LA), respectively. In addition, the bases of their transverse processes are higher placed than in NRM-PZ M8154. Furthermore, these odontocetes share relatively thin pedicles of the neural arch with NRM-PZ M8154. This state differs from the Oligocene mysticetes Aetiocetus cotylalveus (Emlong 1966) and Fucaia buelli (Marx et al. 2015), which have similar neural canals but more robust pedicles.
The similarity of NRM-PZ M8154 and Neoceti is supported by the principal component analysis of its geometric morphometry traits. In the shape of the vertebra, NRM-PZ M8154 is pooled together with Neoceti (rather than Basilosauridae) in the morphospace of the first two principal components. It is found to be the most similar to the Oligocene odontocetes M. riabinini and A. meffordorum (Fig. 5).
Finally, NRM-PZ M8154 has another unusual derived trait. Its L/H ratio is different from whales with Pattern 1 of cetacean vertebral anatomy, which is typical of Mysticeti and early Odontoceti (Buchholtz 2001(Buchholtz , 2007. It is more similar to Pattern 3 of cetacean vertebral anatomy, which is usual for Neogene and modern Delphinida (e.g. the Miocene phocoenid Piscolithax longirostris (de Muizon 1984) has an L/H ratio of 0.68 (MNHN SAS 957, anterior lumbar vertebra)) and many modern delphinids (Buchholtz 2001, Gillet et al. 2019.

NEW EOCENE NEOCETI FROM SEYMOUR ISLAND
In conclusion, NRM-PZ M8154 shows a mosaic of traits, some of which are derived and characteristic of early Neoceti (anteroposteriorly long transverse processes, a ventral keel on the ventral side of the centrum and thin pedicles of the neural arch), whereas others are shared with Basilosauridae (low-placed bases of the transverse processes) or Pelagiceti at most or in general (a rectangular centrum from the lateral view and a triangular neural canal), and some traits are unique and can be autapomorphic (presence of relatively large, separate, anteriorly directed prezygapophyses on the posterior-most thoracic or anterior-most lumbar vertebra and a proportionally short centrum, L/H ratio = 0.79). Thus, NRM-PZ M8154 can be identified as Neoceti indet., with some hypothetical odontocete affinities. Along with Mystacodon, Llanocetus and undescribed specimens , it seems to be one of the earliest known members of Neoceti on Earth.

Body size
There is evidence of the existence of gigantic Neoceti in the late Eocene (Priabonian) in Antarctica: L. denticrenatus (9 m long) (Fordyce & Marx 2018) and Llanocetus sp. (12 m long) . The specimen described here is smaller: it is impossible to estimate the size of NRM-PZ M8154 due to its incompleteness and unusual vertebral proportions. However, as can be seen from the height and width of the centrum, it is comparable in size to D. atrox (∼5 m) or M. selenensis (∼4 m). Therefore, this is the evidence for coexisting small and large members of Neoceti in the late Eocene of Antarctica.
Importantly, among the undetermined late Eocene cetaceans reported from Seymour Island, there is a skull fragment of a small cetacean  containing the skull vertex and the supraorbital process of the frontal bone : fig. 5a & b). Its anterolaterally directed supraorbital process with a rounded lateral margin and a distinct notch at the orbit resembles the anatomy of Eocene mysticetes (Fordyce & Marx 2018, de Muizon et al. 2019, whereas the trace on its dorsal surface can be hypothetically explained as a trace of the overlying maxilla (although this is somewhat speculative), making it similar to odontocetes. Therefore, its plausible interpretation is as Neoceti indet. The vertebra described here could belong to the same or a similar taxon as MLP 83-V-20-386. In addition, the innominate bone MLP 84-II-1-568 is notably small, and it could belong to a small mysticete similar to M. selenensis in size. Thus, the diversity of Neoceti from the late Eocene of Seymour Island could comprise at least three or even five taxa, including mysticetes and forms with some odontocete affinities.

Functional morphology
The unusual anatomy of the prezygapophyses of NRM-PZ M8154 can be related to their functional morphology. They extend beyond the anterior margin of the centrum but are too short for articulation with the postzygapophyses of anterior-lying vertebra. Uhen (2004) suggests that large metapophyses of the thoracic, lumbar and anterior caudal vertebrae of the archaeocete D. atrox were insertion points for multifidus muscles. This suggestion is probably applicable to other early Pelagiceti that retain well-developed metapophyses. In this case, the long prezygapophyses of NRM-PZ M8154 could serve as reinforced places of insertion of multifidus muscles (Fig. 6a). In modern cetaceans, multifidus muscles provide a stiff platform for the longissimus muscle, which provides force to the caudal a. Prezygapophyses and metapophyses as reinforced places of insertion of multifidus muscles. b. Prezygapophyses as bases for interspinous ligaments. lg = ligament, mlt = multifidus muscle, mtp = metapophysis, prz = prezygapophysis, sp = spinal process.

86
spine (Berta et al. 2006). Another interpretation of such a morphology is for providing a base for interspinous ligaments (Fig. 6b) between the lateral surfaces of the neural spine of cranial vertebra and the medial surfaces of the articular processes of caudal vertebra (Long Jr et al. 1997). Both functional interpretations of the prezygapophyses, as well as the short and high centrum, imply an increase of stiffness of the torso (Buchholtz 2001). This is evidence of a tail-driven lift-based high-speed locomotion style with a rigid torso and high-amplitude oscillations of the caudal region. Such a locomotion style is common for modern odontocetes but is unusual for early Pelagiceti (Fish 1996, Buchholtz 2001. However, any detailed and robust conclusions regarding NRM-PZ M8154 swimming mode, morphology and phylogenetic relations require more complete specimens.

Initial divergence of Neoceti in Antarctica
Fordyce (1989) suggested Antarctica as a region where Neoceti initially evolved into mysticetes and odontocetes in response to increasing isolation and cooling of Antarctica at the Eocene-Oligocene boundary. In fact, the discovery of late Eocene gigantic mysticetes Llanocetus spp. strongly supports the idea of Antarctic waters as an important region for the early evolution of Neoceti. The finding of small and fast-swimming Neoceti specimens different from Llanocetus, including NRM-PZ M8154, shows the early divergence and diversification of cetaceans and ecological niche partitioning by them dating back as early as the late Eocene in Antarctica.

Supplemental material
Three supplemental figures and two supplemental tables will be found at https://doi.org/10.1017/S0954102020000516.

Acknowledgements
TM is very thankful to the Argentine Antarctic Institute (IAA-DNA), especially M. Reguero and the Argentine Air Force for logistical support and for the hospitality at the Marambio Base, to the Swedish Polar Research Secretariat (SPFS) for logistical support and to J. Moly and J. O'Gorman for assistance in the field. The authors sincerely thank Mark D. Uhen for comments on an early draft of the paper and for providing photographs of the lumbar vertebrae of Basilotritus wardii, which were important for comparison. The authors also thank Yoshihiro Tanaka for additional comments on an early draft of the paper.

Author contributions
TM performed the fieldwork and photographed the specimen. SD and PG analysed the material. All authors discussed the results, wrote the manuscript, provided illustrations and agreed upon the final version of the manuscript.

Financial support
Financial support from the Swedish Research Council (VR Grant 2009-4447) to TM is gratefully acknowledged.