Description of a new species of beaked whale (Berardius) found in the North Pacific

Two types of Berardius are recognised by local whalers in Hokkaido, Japan. The first is the ordinary Baird’s beaked whale, B. bairdii, whereas the other is much smaller and entirely black. Previous molecular phylogenetic analyses revealed that the black type is one recognisable taxonomic unit within the Berardius clade but is distinct from the two known Berardius species. To determine the characteristics of the black type, we summarised external morphology and skull osteometric data obtained from four individuals, which included three individuals from Hokkaido and one additional individual from the United States National Museum of Natural History collection. The whales differed from all of their congeners by having the following unique characters: a substantially smaller body size of physically mature individuals, proportionately shorter beak, and darker body colour. Thus, we conclude that the whales are a third Berardius species.

www.nature.com/scientificreports www.nature.com/scientificreports/ external characters. External appearance is mostly known from a male individual found stranded on 10 November 2012 in Sarufutsu, Hokkaido (Fig. 4). Most of the external characters of B. minimus are typical of medium-to large-sized ziphiids, with several discriminating characters, such as the narrow, straight, and longer beak; reverse V-shaped throat grooves; relatively smaller flippers (flipper length is 11.4% of body length on average; range, 7.7-13.4%); small dorsal fin (dorsal fin height is 3.7% of body length on average; range, 3.4-3.9%) located 70% of body length (on average; range, 66.7-71.8%); and tail flukes that lack the median notch. However, the posteriorly opened crescent-shaped blowhole slit indicates Berardius affinity. Additionally, B. minimus has a substantially smaller body size (maximum body length of 6.9 m in physically mature individuals, so far), more spindle-shaped body, and relatively shorter beak, which is approximately 4% of the body length and is not consistent with the morphology of either of the known Berardius species.
Body colour is almost black with a pale white portion on the rostrum; this is in contrast to B. bairdii, which is described as "slatish" 4 or "slate grey" 6,7 or B. arnuxii, which is described as black 30 or light grey 31 . The greyish tone of the B. bairdii body is mainly attributed to the dense healed scars that are probably caused by intraspecific conflicts and/or behaviour. At least in adult and subadult individuals of B. minimus, cookie-cutter shark bites are fairly conspicuous, but not to the extent as usually seen in some other species such as Ziphius cavirostris, Mesoplodon densirostris, and/or Balaenoptera borealis. The darker body colour with almost no scars produces a sharp contrast with the healed cookie-cutter shark bites, which are white and very conspicuous against the black body of B. minimus.
The beak is much shorter than in the other two Berardius species. In B. bairdii, the head proportions are extremely small, and are much smaller than that of B. minimus. Body colour is almost uniformly dark brown with a whiter portion at the tip of rostrum. No white patch on the belly was confirmed in B. minimus. An illustration of an adult male of B. minimus is shown as Fig. 5. At present, we do not know what adult females look like.    viduals and proportionately shorter beak are the most reliable characters which indicate that the population in question represents a species that was previously not known to science.
Regarding body length, a strong significant difference was found between the body length of male B. bairdii from the Okhotsk Sea (n = 34) 32 and mature male B. minimus (n = 4, Table 1) (Welch's t-test, t = 18.5, P < 0.001).
To confirm relative rostrum-to-body length, Welch's t-test was also conducted. For B. minimus, four samples in Table 1 were analysed. For B. bairdii, the mean and standard deviations for male B. bairdii in the Okhotsk Sea (n = 29) that appeared in Table 2 of Kishiro 32 were used. Rostrum length was standardised by body length, and was 3.62 ± 0.39 SD% (n = 4) for B. minimus and 5.81 ± 0.80 SD% (n = 29) for B. bairdii. Welch's t-test showed strong significant difference (P = 2.3 × 10 −5 ). Female B. bairdii relative length was 6.27, which is longer than that of males. Note this female was not physically mature. The difference between B. minimus and B. bairdii was obviously larger if the sex-pooled data were used. A strong significant difference was also found between B. minimus and B. bairdii in the Pacific Ocean and Sea of Japan (P < 0.001). Thus, the relative rostrum length of B. minimus was significantly shorter than that of B. bairdii. However, we note that the sample size for both B. minimus and B. arnuxii are extremely small, in contrast to B. bairdii. Skull morphology. The skull morphology resembles the skulls of both existing Berardius species, but B. minimus has a distinctly shorter rostrum if contrasted to the condylobasal length, and smaller bulla and periotic bone. In general, the sutures are more tightly closed in B. minimus than those in the other Berardius species. In the hyoid bone, thylohyal and basihyal are not fused at all (Fig. 6).
Superior aspect. The following characters are readily recognisable as species-specific. The relative beak length in B. minimus is clearly smallest among the three Berardius species. The B. minimus skull has much tighter sutures compared with those in both B. arnuxii and B. bairdii. The proportional distance of the anterior end of the maxillae from the tip of the rostrum (i.e. premaxillae) relative to condylobasal length of the skull is much smaller in B. minimus (6.93% in NSMT35131) than the two previously known Berardius species (which have a distance of approximately 10%). The inclination of the occipital bone is stronger in B. minimus, and the occipital plane is much wider compared with the other two species. The antorbital notch is proportionately narrower in B. minimus than in B. bairdii but similar to that in B. arnuxii. The B. minimus rostrum has simple tapering contour lines toward the tip, whereas both contour lines of the rostrum are parallel in B. bairdii and B. arnuxii. The lateral border of the orbit, which consists of the maxilla and frontal bones, is almost parallel to the sagittal plane in B. minimus, but is oblique in other two species.
Lateral aspect. The relative rostrum length is obviously shorter in B. minimus, and the B. minimus rostrum also looks much shorter than those of the other two species in side view. The skull height relative to condylobasal length is much larger (0.41-0.44) in B. minimus than those in B. bairdii (0.35-0.40) and B. arnuxii (0.40-0.41). There is stronger inclination of the higher portion of the occipital plane in B. minimus, and the convexity of the occipital plane is stronger in B. minimus. The temporal fossa is the shallowest in B. minimus and the medial wall of the fossa is convex, but is concave in B. bairdii and B. arnuxii.
Posterior aspect. The structure above the temporal fossa is proportionately much larger and higher in B. minimus than those in B. bairdii and B. arnuxii, which gives the impression that the B. minimus skull is rather triangular in the posterior view, whereas those of the other two species are pentagonal.
Anterior aspect. In the frontal view, lateral expansion of the premaxillae at the posterior is prominent, and the posterior margins of both maxillae are clearly visible in B. minimus. www.nature.com/scientificreports www.nature.com/scientificreports/ In B. minimus, the height of skull relative to the width is much higher than those of the other Berardius species. The prominential notch and related structure are much higher, more distinct and more rugged in B. bairdii and B. arnuxii.
Teeth. As in the other two Berardius species, B. minimus has two pairs of teeth only at the tip of the lower jaw. The anterior tooth is much larger than the posterior tooth. Teeth dimensions of the holotype are shown in Table 3 (57-1 and 2, 58-1 and 2). In the holotype specimen of B. minimus the pulp cavities are almost closed in all teeth other than the right 2 nd tooth, where the pulp cavity is open.  www.nature.com/scientificreports www.nature.com/scientificreports/ Post cranial skeleton. The vertebral column has proportionately high spinous processes, which is observed in most ziphiid species (Fig. 7). The bone matrix is coarse and porous, and they float on the processing water after internal soft tissue was removed. In the holotype specimen, the vertebral formula is C. 7, Th. 10, L. 10, Ca. 19, making the total count as 46. Among 7 cervical vertebrae, C1-C3 were fused. L4 and L5 are the tallest vertebrae. Ca10 and 11 are so-called ball vertebrae. Ten chevrons were counted. Ribs are in 10 pairs, among which seven pairs are dual-headed with both costovertebral and costotransversal articulations. The remaining three pairs have  www.nature.com/scientificreports www.nature.com/scientificreports/ only one articulation facet which articulate with "transverse" processes of the caudal thoracic vertebrae. No ossified cervical ribs were found. The sternum is composed of five segments.
Paired ossified pelvic bones have a lateral surface which is fairly smooth; however, in the medial surface, approximately two-thirds of the total length is an elevated area where the corpus cavernosum penis attaches. Viewed from the dorsal side, the pelvic bones show a very gentle s-shape. No rudimental femur or any additional appendicular bone was collected.
Pectoral appendage. Regrettably, we could not secure all phalangeal bones of the left flipper. On the right side, there are three carpal bones in the proximal row, possibly the Ossa radiale, centrale, and ulnare. In the distal row are another three carpal bones. All five digits have one each metacarpal; the phalangeal formula is 0-5-4-3-2.
Multi-measurement comparison. Table 2 shows the mean, standard deviation, and range of each measurement by species. PCA showed that the contribution of the first principal component (PC1) was 73.9%, and the cumulative contribution reached 90% for PC1-6. Thus, linear discriminant analysis was conducted using PC1-6. Table 4 shows the linear discriminant coefficients obtained by linear discriminant analysis (LDA). The linear discriminants coefficients of each sample are plotted in Fig. 8. The distribution of the linear discriminants variates was very clearly separated by species.
Genetic considerations. Molecular phylogenetic relationships among three Berardius species were examined using nucleotide sequence variation of the mitochondrial (mt)DNA control region (CR). The 879-bp complete CR sequence data from eight B. minimus specimens (Table 5) (Acc. Nos AB572006-AB572008 from Kitamura et al. 6 , Acc. Nos LC175771-LC175773 in this study, and Acc. Nos KT936580-KT936581 from Morin et al. 7 ) showed five haplotypes with only 1-4 nucleotide differences without gaps after multiple alignment. Using the CR sequences aligned with 430-bp B. arnuxii sequences (Acc. Nos AF036229 and AY579532 from Dalebout et al. 16 ) excluding gaps, the number of nucleotide differences between B. minimus and its congeners was 18-22 for B. bairdii and 25-29 for B. arnuxii. Thus, the mtDNA nucleotide difference between B. minimus and any of its congeners was much greater than the difference between B. bairdii and B. arnuxii, which is 12-16 nucleotides. The observed CR nucleotide differences supported the distinct position of B. minimus in the Berardius tree constructed from 430-bp sequences using the maximum likelihood method, where B. bairdii and B. arnuxii formed a sister clade (Fig. 9).
Known distribution. As is indicated by the map of localities where B. minimus was found (Fig. 10), their known distribution is very limited and occurs between 40°N and 60°N, and 140°E and 160°W. Kasuya 5,18 summarised Hokkaido whalers' traditional knowledge. The whalers recognised two types of tsuchi-kujira: the ordinary "tsuchi-kujira" (Berardius bairdii) and the darker and smaller "kuro-tsuchi" (black Baird's beaked whale) or "karasu" (crow). However, it is unclear whether "kuro-tsuchi" and "karasu" are used to describe the same type of whales or each notion represents the different population.

Discussion
In this study, we described a new species, B. minimus, which corresponds to "kuro-tsuchi". If "karasu" exists as a third type, it could be a species that is not yet recognised or a Mesoplodon species found in Hokkaido (either M. stejnegeri or M. carlhubbsi). Recognition of these Mesoplodon species around Hokkaido is rather recent; the  Table 4. Linear discriminant coefficients obtained by linear discriminant analysis (LDA). As was also pointed out by Kasuya 18 , Fig. 364 and 366 of Heptner 35 hinted at the possibility of a Hyperoodon-like whale in the northern Pacific. The animal in the photo was definitely not Berardius. This could be a species of probably about 10-m long with a beak almost like that of Hyperoodon. We suspect this could be an example of an extralimital occurrence of H. ampullatus. Considering the recent sightings of the gray whales in the Mediterranean or in Namibia 36,37 , the possibility of vagrant individual navigate through the Northwest passage during summer should be studied.
The species we described is rather readily recognisable by people with whale taxonomy experience based on the external characters. The species has an obviously smaller body size, which is 6.3-6.9 m in physically mature individuals, so far we confirmed (Morin et al. 7 reported an adult male with 7.3 m body size). Their body size  www.nature.com/scientificreports www.nature.com/scientificreports/ ranges from 9.1-11.1 m in B. bairdii and 8.5-9.75 m in B. arnuxii 38 . They have a relatively short beak that is approximately 4% of the body length. They have a dark body colour, which is almost uniformly black with noticeable healed cookie-cutter shark bites forming white dots; this impressively contrasts with the much lighter colouration of B. bairdii and likely result from healed scratches and scars that were probably caused by intra-specific struggling and bottom-feeding behaviour.
As was mentioned above, when comparing the skull sutures in similarly mature individuals of different species of cetaceans, there is a general tendency that the larger the adult form is the less rigid skull composition is observed. Cetacean facial skull consists loosely articulated bones, including the maxillae, premaxillae and frontals, which are adhered to the mesorostral cartilage pillar on the vomer by connective tissue. It is a physically significant principle where cetaceans swing their rostrum in the water for foraging. It requires tremendous power and the flexibility of the skull structure must ease the stress given to the skull structure. In this context it is quite reasonable that the skull of B. minimus is far more rigidly composed compared to those of the far larger species,  The molecular biology of B. minimus was previously discussed by Kitamura et al. 6 , and specific genome characters were only identified in individuals collected from Hokkaido. However, we found a skull with B. minimus characters in the collection of the USNM which was collected from the Unalaska Island in 1943. Additional individuals were detected among the samples collected in the Aleutian area, and further analyses and considerations were conducted and discussed by Morin et al. 7 . Further detailed analyses on Berardius species in both the northern and southern hemispheres are needed to explain Berardius speciation processes.
The currently recorded B. minimus distribution is very limited and occurs between 40°N and 60°N, and 140°E and 160°W. They have fairly dense cookie-cutter shark (Isistius brasiliensis) bites. The cookie-cutter shark is understood to be a tropical to warm-temperate species and their northern limit in the western North Pacific is reported to be 30°N to 43°N 39 . However, the southern limit of the B. minimus distribution might extend further south.
Although species identities of B. arnuxii and B. bairdii have been previously debated, we described another species of this genus. However, it is unclear whether B. minimus speciation occurred before or after the antitropical split of B. arnuxii and B. bairdii. Additionally, the area where Berardius speciation took place should be examined in the future.

Methods
Specimens examined. The specimens of this unknown species, which were collected in Hokkaido, are listed in Table 1. No live animals were used for the current research. Observations on the external appearance and morphometrics, observations on the skeleton especially of the skull, skull morphology and measurements and molecular phylogenetic analysis were conducted. external morphology and measurements. External observations of the five individuals of Berardius minimus (three physically mature males, one subadult female, and a head of one neonate female) were made, and the external morphometrics following previous studies 32,40 (Tables 6 and 7) were conducted on four B. minimus (all physically mature males; Table 1). Raw data examination revealed that body length and the ratio of beak length-to-body length significantly differed, and Welch's t-test was applied to these variables.
Skeletal morphology and measurements of the skull. Observations of the skeleton, especially of the skull, and skull measurements were made for 21 specimens (10 B. bairdii, seven B. arnuxii, and four B. minimus) ( Table 2). Specimens are stored at the USNM, NMNS, MNHN, Natural History Museum of London (BMNH), and Museo Acatushún (MA).

Multivariate analysis.
To examine the difference between the morphological features among species, a multivariate analysis was conducted. To describe the effect of the difference of body size by species, a principal component analysis (PCA) was conducted for 27 measurements shown in Table 5   nucleotide sequence analysis and molecular phylogeny. The 18 mtDNA control region (CR) sequences analysed (Table 5) included sequences from three B. minimus specimens (Acc. Nos LC175771-LC175773 for SNH12044, SNH12054, and SNH14016, respectively) and 15 previously reported sequences, which included seven for B. bairdii (Kitamura et al. 6 , AB571999-AB572005), five B. minimus (Kitamura et al. 6 , AB572006-AB572008, updated complete sequences August 2016; and Morin et al. 7 , Acc. Nos KT936580-KT936581), two B. arnuxii (Dalebout et al. 16 , Acc. Nos AF036229 and AY579532), and one Indopacetus pacificus (Kitamura et al. 6 , AB572012) as an outgroup. I. pacificus was selected because it belongs to the same family but is in a rather distant genus, which was inferred by a previous CR phylogenetic tree 6 . All the newly collected samples for the nucleotide sequence analysis and molecular phylogeny were officially transferred to the authors from the original sample holder, the Stranding Network Hokkaido. Nucleotide sequencing of the complete mtDNA CR in the three B. minimus was performed using primer pairs CRL (5′-CAA CAC CCA AAG CTG GAA TTC T-3′) 6 and CRH2 (5′-TAG ACA TTT TCA GTG TCT TGC-3′, which was newly designed for this study) for PCR amplification, and CRH (5′-CCA TCG AGA TGT CTT ATT TAA G-3′) 6 and LCR (5′-GAC ATC TGG TTC TTA CTT CAG G-3′) 42 as internal sequencing primers.
CR sequence alignment was performed using CLUSTAL X 43 , and the output was inspected by eye following the application of multiple alignment parameters in the program. All CR sequences were adjusted to the short length of the B. arnuxii sequence, 430 bp (Dalebout et al. 16 , Acc. Nos AF036229 and AY579532), for multiple sequence comparison and molecular phylogenetic analysis.
A molecular phylogenetic tree was constructed with 430-bp mitochondrial CR sequences of all analysed species using the maximum likelihood algorithm in MEGA version 7 44 based on the Tamura 3-parameter model 45 with gamma distribution (parameter = 0.2001), which was suggested to be the best nucleotide substitution model based on a model test in this program. Bootstrap values were calculated by 1,000 replicates 46 .

Data Availability
Genbank Accession Numbers for sequences used in molecular phylogenetic analysis are listed in Table 5. Materials examined in this study and associated museum number are listed in Table 1.

Mean (cm) SD
Measurement items listed in Table 6 Table 7. Measured external morphometrics characters for B. bairdii as described in previous studies 32,40 .