The taxonomic status of extinct Japanese or Honshu wolves (Canis lupus hodophilax) has been disputed since the name hodophilax was first proposed by Temminck in 1839 on the basis of specimens stored in Leiden, the Netherlands. Points of controversy include whether the type specimen of hodophilax (Jentink c: RMNH.MAM.39181) and the other two specimens from Leiden (Jentink a: RMNH.MAM.39182 and Jentink b: RMNH.MAM.39183) represent different varieties or subspecies of Japanese wolves or not. Two Japanese names, ookami and jamainu, used to describe wild Canis species, further complicate the issue. In this study, the taxonomic status of Japanese wolves was clarified using mitochondrial DNA of the three specimens stored at the Naturalis Biodiversity Center in Leiden, in addition to three Japanese wolf specimens stored at the Museum für Naturkunde in Berlin and five new samples from Japan. The mitochondrial genomes of the type specimen of hodophilax (Jentink c) and another sample from Leiden (Jentink b) as well as Berlin specimens were included in the cluster of Japanese wolves distinct from other grey wolves. However, the other sample from Leiden (Jentink a) was identified as a domestic dog. A mitochondrial genome analysis suggested that Japanese wolves could be categorized into two distinct clusters. Studies of nuclear genomes are needed to further clarify the taxonomic status, divergence time, and population genetic structure of Japanese wolves.
INTRODUCTION
The taxonomic status of wolves in Japan, which became extinct about 100–120 years ago, has long been disputed because a quite limited number of bone specimens and stuffed specimens are available for study (Ishiguro et al., 2009; Ishiguro et al., 2010; Miyamoto, 2001). Generally, two lineages of wolves from Japan are recognized: Japanese or Honshu wolves (Canis lupus hodophilax Temminck, 1839) and Ezo or Hokkaido wolves (Canis lupus hattai Kishida, 1931). The former inhabited the central and southern islands of Honshu, Shikoku, and Kyushu, whereas the latter was limited to the northern island of Hokkaido and neighboring islands. Recent phylogenetic studies based on mitochondrial genomes have clearly demonstrated that the two lineages are phylogenetically distinct sub-species of grey wolves; Japanese wolves are among the earliest diverged groups within grey wolves (Matsumura et al., 2014), and Ezo wolves are closely related to grey wolves in North America (Ishiguro et al., 2010; Matsumura et al., 2014).
However, there remain unresolved issues about the taxonomy of Japanese wolves in Honshu, Shikoku, and Kyushu. In Honshu, Shikoku, and Kyushu, two different words have historically been used to describe Canis species in the wild: ookami and jamainu (Walker, 2005; Funk, 2015). (These words are sometimes spelled okami, ookame, or yamainu; the names used by Philipp Franz von Siebold (Siebold: 1796–1866) are adopted in this manuscript.) Ookami means wolf, whereas the Japanese word yama-inu (jamainu) literally translates to mountain dog. Therefore, most agree that ookami represented Japanese wolves C. l. hodophilax, but there is a controversy about what jamainu represented. Some believe that jamainu refers to free-ranging dogs (i.e., Canis lupus familiaris) which breed and range in the mountains, some regard jamainu as a synonym of ookami (i.e., Japanese wolves), while others maintain that a variety of wolves distinct from Japanese wolves inhabited Honshu, Shikoku and Kyushu until recently (Hiraiwa, 1992; Walker, 2015).
One essential step to clarify the taxonomic status of wolves in Japan is to investigate specimens stored in museums in Europe because the scientific description of wild Canis in Japan has been made on the basis of those specimens. In Europe, wolf specimens from Japan are stored in three locations: (1) Naturalis Biodiversity Center (NBC), Leiden, the Netherlands; (2) Museum für Naturkunde, Berlin, Germany; and (3) Natural History Museum, London, United Kingdom. Temminck (1839) first described Japanese wolves as a distinct wolf species, C. hodophilax, based on the specimens at Leiden, which Siebold sent from Japan. In Fauna Japonica edited by Siebold, Temminck referred to this new species C. hodophylax (hodophilax) as jamainu, and suggested that it is distinct from ookame (Walker, 2015). Nehring (1885) examined another specimen, ZMB_Mam_22326 (An 25546), in Berlin, and described it as Lupus japonicus or Canis lupus var. japonica, different from C. hodophilax.
NBC at Leiden now registers one skeleton with a skull (Jentink a), two skulls (Jentink b and Jentink c), and one mounted skin as hodophilax. According to Jentink (1887), the type specimen of hodophilax is the skull “Jentink c” and the skin (Funk, 2015). Peculiarly, the condylobasal length of the skull “Jentink c” (174 mm) is small compared with those of the skull “Jentink b” (206 mm) and other Japanese wolf skulls found in Japan (Saito, 1964; Imaizumi, 1970; Nakamura, 1998). For example, the estimated range of Japanese wolf skull length was 193.1–235.9 mm (P = 99%) based on the measurement of 14 specimens from the Tanzawa population in Honshu (Nakamura, 1998). Imaizumi (1970) concluded that the skull “Jentink a” is a domestic dog based on the position of the external auditory meatus and the form of the anterior border of the mesopterygoid fossa. Based on the size and description in Fauna Japonica, some believe that the skull “Jentink b” is the only Japanese wolf (i.e., ookame) specimen at Leiden, and that the skull “Jentink c” is jamainu, an undescribed variety of wolves that once lived in Japan.
The objectives of the present study were to clarify the taxonomic status of the specimens registered as C. hodophilax at Leiden based on a mitochondrial genome sequence analysis. Specifically, the enigmatic “Jentink c” is expected to be phylogenetically different from “Jentink b” and other Japanese wolf specimens if it represents a distinct variety or a subspecies of Japanese wolves. Specimens of wolves from Japan stored at Museum für Naturkunde in Berlin, Germany, including a “type specimen” of Lupus japonicus described by Nehring (1885), and five additional DNA samples collected in Japan by Matsumura et al. (2014) and Ishiguro et al. (2016) were also examined.
MATERIALS AND METHODS
Specimens
To the best of our knowledge, only eight specimens of wolves from Japan are stored in three museums (NBC in Leiden, Museum für Naturkunde in Berlin, and Natural History Museum in London) in Europe. Permission to analyze ancient DNA was obtained for six out of the eight specimens: Jentink a: RMNH.MAM.39182(52995), Jentink b: RMNH.MAM.39183, and Jentink c: RMNH. MAM.39181(52994) in Leiden (Imaizumi, 1970; Miyamoto, 2001; Obara, 2002), and ZMB_Mam_22326 (ZMB22326), ZMB_ Mam_42983 (ZMB42983), and ZMB_Mam_48817 (ZMB48817) in Berlin (Miyamoto, 2001; Yagi, 2016). In particular, the ventral nasal concha fragments were kindly provided by NBC in Leiden and the Museum für Naturkunde in Berlin (Table 1). Five Japanese wolf bone samples from personal collections throughout Japan were also examined (JW275, JW286, JW289, JW290, and JW292; Table 1).
Complete mtDNA sequencing
DNA was extracted from bones according to the methods described by Okumura et al. (1999), Ishiguro et al. (2009, 2016), and Matsumura et al. (2014). Bone powder (0.02–0.1 g) was obtained from fragments of the ventral nasal concha of the specimens from the NBC in Leiden and the Museum für Naturkunde in Berlin using a Multi-beads Shocker (Yasui Kikai, Osaka, Japan), and from five Japanese wolf bone samples in Japan using an electric drill. The powder was then suspended in 10 ml of 0.5 M ethylenediaminetetraacetic acid (EDTA) at pH 7.0 and rotated for decalcification. A pellet of bone powder was collected by centrifugation and decalcified by repeated washes with 10 ml of 0.5 M EDTA until the supernatant became clear. The bone powder sample was then treated for 24 h with 5 ml of 0.5 M EDTA with proteinase K (300 µg/ml) and N-lauryl sarcosine (0.5%). The samples were centrifuged at 3000 rpm for 10 min, and the supernatant containing ancient DNA was extracted twice with phenol, once with chloroform:phenol (1:1), and once with chloroform for protein removal. The supernatant was concentrated using an Amicon Ultra Centrifugal Filter 30K (Merck Millipore Ltd., Tullagreen, Ireland) and washed with distilled water. The extracted DNA samples (about 1–3 µl) were used for direct PCR or next-generation sequencing. Precautions described by Okumura et al. (1999) were taken to prevent contamination with DNA from modern dogs.
PCR products were purified using the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany), and firstly sequenced by Sanger sequencing to obtain a mitochondrial (mt) DNA fragment of 215–223 bp (Ishiguro et al., 2016). Successfully obtained samples were then subjected to next generation sequencing (NGS) to obtain full-length mtDNA genome sequences. Before NGS, DNA samples were repaired using NEBNext FFPE DNA Repair Mix (New England Bio Labs, Ipswich, MA, USA), and DNA libraries were constructed using the NEBNext Ultra II DNA Library Prep Kit (New England Bio Labs) following the manufacturer's instructions, with 11–15 PCR cycles. For each library, 5 Gb of short DNA sequences (paired-end, 125 bp) were determined using the Illumina HiSeq2500 platform (Illumina, Inc., San Diego, CA, USA). Sequence reads were mapped to a reference mtDNA genome (JW229: AB499818.1), and consensus sequences were extracted using CLC Genomics Workbench ver. 11 ( https://www.qiagenbioinformatics.com/). The consensus sequences for the control region were manually corrected based on the re-aligned short sequences. A part of the entire mtDNA genome sequences (consensus sequences of the mapped reads), i.e., the sequences of the control regions of JW255 and JW271, have already been published.
Phylogenetic and demographic analyses
In addition to mitochondrial genome sequence data of wolves from central and southern Japan (Honshu, Shikoku, and Kyushu, 13 samples, Table 1), all wolf mitochondrial genome sequences available in GenBank (85 wolf samples except for our data, Supplementary Table S1 (zs200019_TableS1.pdf)) were used for analyses. The dataset for the phylogenetic and demographic analyses was created following the methods of Matsumura et al. (2014). The total length of the four concatenated files (tRNA, rRNA, protein-coding regions, and control region) was 15,477 bp. Phylogenetic analyses were performed using MEGAX (Kumar et al., 2018), MrBayes 3.2 (Ronquist et al., 2012), and BEAST v2.5.2 (Bouckaert et al., 2019). Details of the methods for tree reconstruction using the neighbor-joining method (Saitou and Nei, 1987) and Bayesian methods, including parameter settings, were described by Matsumura et al. (2014). A maximum likelihood tree was constructed using MEGAX; the Hasegawa-Kishino-Yano model (Hasegawa et al., 1985) was used as a substitution model, and a gamma distribution was assumed to model evolutionary rate differences among sites (four categories). Two coyote (Canis latrans) sequences (DQ480510 and DQ480511) were used as outgroups. Eighteen mitochondrial genome sequences from ancient canid samples (Thalmann et al., 2013; Supplementary Table S1 (zs200019_TableS1.pdf)) were included to calibrate the ages of nodes in the phylogenetic trees.
RESULTS
Full-length mtDNA genome sequences were newly assembled for ancient DNA samples from seven putative Japanese wolves (JW255, JW271, JW275, JW292, Jentink b, Jentink c, and ZMB48817). In addition, only partial mtDNA control region sequences (215–240 bp) were obtained for five putative Japanese wolves (JW286, JW289, JW290, Jentink a, and ZMB22326), as their DNA was highly degraded. These DNA sequences have been deposited in DDBJ (LC520089–LC520095, LC519904–LC519906). Based on these nucleotide sequences, all ancient DNA samples other than the Jentink a specimen (RMNH. MAM.39182) had nucleotide substitutions specific to Japanese wolves (Table 1). Jentink a appeared to be a domestic dog based on its sequence similarity to a Japanese dog (Shibainu breed). No mtDNA sequences were amplified from ZMB42983.
Table 1.
Characterization of Japanese wolf and dog specimens, and variation of mtDNA control region.
Fig. 1.
(A) Neighbor-joining (NJ) tree (Saitou and Nei, 1987) of wolf mitochondrial genome sequences, rooted by coyote sequences. Scale bar indicates 0.005 substitutions per site. Among wolves from Japan, previously published sequences (Matsumura et al., 2014) and newly determined sequences are shown by solid and open red circles, respectively. (B) NJ tree of mitochondrial genome sequences of Japanese wolves, rooted by a sequence of a grey wolf from the United States. In both trees, red stars indicate bootstrap support values of ≥ 95%.
Thirteen full-length mtDNA genome sequences of Japanese wolves were used for phylogenetic analyses. In our previous analysis (Matsumura et al., 2014), all six Japanese wolf samples formed a distinct monophyletic cluster. In this study, the seven newly assembled putative Japanese wolf samples, including the type specimen of C. l. hodophilax (Jentink c), formed a distinct cluster with the previous six samples (Fig. 1A, see Supplementary Figure S1 (zs200019_FigS1.pdf) online). The two samples from Leiden (Jentink b and Jentink c) and one sample from Berlin (ZMB48817) were phylogenetically closely related.
Based on their analysis of the control region, Ishiguro et al. (2016) identified two groups of Japanese wolves: one has an 8 bp deletion in the control region (Group B) and the other lacks this deletion (Group A). When this criterion was applied to the 13 full-length mtDNA genome sequences, seven were assigned to Group B. Three samples housed outside Japan (Jentink b, Jentink c, and ZMB48817) were included in Group B. ZMB22326, for which only a partial mtDNA sequence was obtained, was assigned to Group A. Group B shared nine unique substitutions [positions 1552, 1777, 3209, 4754, 8998, 10313, 11141, 12975, and 13944 in JW229 (AB499818)] (Fig. 1B) in addition to the 8 bp deletion (Table 1). Taking the 13 full-length mtDNA genome sequences together with 10 Japanese wolf samples for which only the control region sequences were known, the 8 bp deletion was evaluated in 23 samples (Table 1); 11 and 12 samples were assigned to Group A and B, respectively. Based on the sample locations, no clear geographical structure was detected in the distributions of Groups A and B (Fig. 2).
Based on the 13 complete mtDNA sequences, the estimated divergence time of Japanese wolves from other wolf lineages was 108,000 years ago, assuming that the dog-wolf lineage separated from the coyote lineage 1 million years ago (Fig. 3). The estimated time to the most recent common ancestor (MRCA) of Japanese wolves was 42,000 years ago. However, the estimated divergence time and MRCA were 45,000 and 17,000 years ago, respectively, if the calibration was based on the age of archaeological samples (see Supplementary Figure S2 (zs200019_FigS2.pdf)).
Fig. 2.
Geographic distribution of samples assigned to Group A (●) and B (○) based on the absence or presence, respectively, of an 8-bp deletion in the control region.
Fig. 3.
Age estimates for the Japanese wolf cluster. The phylogenetic tree was inferred by a Bayesian analysis (Bouckaert et al., 2019) of mitochondrial genomes. Blue bars represent 95% highest posterior densities of nodal age estimates.
DISCUSSION
Our study clearly reveals the phylogenetic status of putative Japanese wolf samples stored in two museums outside of Japan. The mitochondrial genomes of the type specimen of C. hodophilax (Jentink c), another sample from Leiden (Jentink b), and one sample from Berlin (ZMB48817) clustered with the mitochondrial genome sequences of other Japanese wolf samples from Japan. Only a portion of the mtDNA control region was successfully sequenced for another sample from Berlin (ZMB22326) as well as the remaining sample from Leiden (Jentink a); the former had several nucleotide substitutions specific to Japanese wolves, but the latter did not. We conclude that Jentink a is likely to be a domestic dog because it had a sequence similar to a Japanese dog, although we are also aware of the limitation involving the first-evolving control region. Imaizumi (1970) concluded that the skull Jentink a at Leiden is a domestic dog based on key morphological features, including the position of the external auditory meatus and the form of the anterior border of the mesopterygoid fossa. Our mtDNA sequence analysis supported this conclusion.
The Jentink c sample has been controversial because it is relatively small, even among Japanese wolves (Saito, 1964; Imaizumi, 1970; Nakamura, 1998). In Japan, the terms ookami and jamainu have both been used to describe wild canids (Hiraiwa, 1992; Walker, 2005). The two terms translate to wolf and mountain dog, respectively. Unfortunately, there is no consensus regarding the identity of jamainu. It is possible that (1) both ookami and jamainu refer to the Japanese wolf C. l. hodophilax, (2) jamainu refers to free-ranging domestic dogs C. l. familiaris, or (3) jamainu is a third canid species that once inhabited Japan. This lack of clarity derived in part from the first description of C. hodophilax by Temminck (1839), which apparently ignored information from Siebold, who sent the specimens from Japan to Leiden (Funk, 2015). In fact, Siebold described in his diary that he collected ookami and jamainu in Osaka, Japan (Holthius and Sakai, 1970) and stated that the specimen Temminck described as C. hodophilax was jamainu (Funk, 2015). The stuffed specimen at Leiden is actually labeled jamainu. Therefore, those who believe that jamainu is a distinct species suspect that Jentink b and Jentink c are ookami and jamainu, respectively.
Our results clearly show that at least the maternal (mitochondrial) lineages of Jentink c as well as Jentink b are Japanese wolves. The reason why Jentink c is small compared to other Japanese wolves remains unresolved. The difference in size and morphological features (Miyamoto, 2001) might have resulted from hybridization between Japanese wolves and domestic dogs on an isolated archipelago (Naora, 1965). Since only mtDNA was considered in this analysis, further studies of nuclear DNA or entire nuclear genomes may resolve this issue. Such studies will also provide more precise estimates of the divergence time and past population dynamics of Japanese wolves.
Ishiguro et al. (2016) identified two groups of Japanese wolves that can be distinguished by a deletion in the control region. Phylogenetically, Group B (having the deletion) is derived from Group A (lacking the deletion), because the lack of the deletion is shared by all other wolves (85 samples) investigated here. We detected nine additional unique substitutions in the mtDNA genomes of Group B, supporting the distinct phylogenetic status. This suggests that Group B evolved in a geographically isolated area. Surprisingly, no clear geographical structure was detected. Admixture is possible based on the high mobility of wolves after the divergence of the two distinct lineages. Group B samples tend to be relatively recent in terms of age and tend to be from the central and eastern/northern parts of Japan in terms of geographic location. It is possible to hypothesize that the newly emerged Group B is dominant in the central and eastern/northern parts of Japan. Further studies of ancient DNA from Japanese wolves, ideally including samples from several hundred or thousands of years ago, are needed to confirm this hypothesis.
ACKNOWLEDGMENTS
We are very grateful to Dr. Pepijn Kamminga for providing the opportunity to analyze wolf specimens stored at the Naturalis Biodiversity Center in Leiden, the Netherlands. We also thank Mr. Steffen Bock for providing the opportunity to analyze Japanese wolf samples stored at the Museum für Naturkunde, Berlin, Germany. We would like to express our gratitude to Dr. Holger Funk for providing valuable advice, timely suggestions, and support. We appreciate the following individuals for enabling us to analyze Japanese wolf skeletons stored in museums or private collections: Mr. Tadashi Fujii (Iwate Prefectural Museum); Mr. Wataru Matsuda (Ohyodocho Education Board); and Mr. Kimio Endo, Mr. Kaoru Kikuchi, Mr. Saichi Senda, and Ms. Kimiko Nagai (personal collections). This study was partially supported by a Grant-in-Aid (17K01193) from the Ministry of Education, Science, Sports, and Culture of Japan.
AUTHOR CONTRIBUTIONS
SM, YT, HH, and NI designed the study, YT and NI conducted experiments, SM and YT analyzed the data, and SM, YT, HH, and NI wrote the manuscript.
SUPPLEMENTARY MATERIALS
Supplementary materials for this article are available online. (URL: https://doi.org/10.2108/zs200019)
Supplementary Table S1 (zs200019_TableS1.pdf). Mitochondrial genome sequence data of wolves included in the analysis.
Supplementary Figure S1 (zs200019_FigS1.pdf). Phylogenetic trees constructed by (A) maximum likelihood method and (B) Bayesian method, rooted by coyote sequences.
Supplementary Figure S2 (zs200019_FigS2.pdf). Age estimates for the Japanese wolf cluster calibrated with the archaeological samples.
