A long-forgotten ‘dinosaur’ bone from a museum cabinet, uncovered to be a Japan's iconic extinct mammal, Paleoparadoxia (Desmostylia, Mammalia)

Here, we report a new ‘discovery’ of a desmostylian fossil in the geological collection at a national university in Japan. This fossil was unearthed over 60 years ago and donated to the university. Owing to the original hand-written note kept with the fossil in combination with interview investigation, we were able to reach two equally possible fossil sites in the town of Tsuchiyu Onsen, Fukushima. Through the interviews, we learned that the fossil was discovered during construction of a debris flow barrier and that it was recognized as a ‘dinosaur’ bone among the locals and displayed in the Village Hall before/until the town experienced a fire disaster in 1954. As scientific findings, the fossil was identified to be a right femur of Paleoparadoxia (Desmostylia), which shows well-preserved muscle scars on the surface. The age was estimated to be 15.9 Ma or younger in zircon-dating. This study shows an excellent case that historical and scientific significances could be extracted from long-forgotten uncatalogued specimens as long as the original information is retained with the specimens.


Introduction
Scientific 'treasures' are housed in museum collections with many specimens waiting to be catalogued and described. These specimens sometimes attract the attention of museum scientists or visiting researchers who provide new insights for evolutionary history, taxonomy and palaeoecology. Similar discoveries have been made for marine mammals as well. Tsai et al. [1] reported the unexpected discovery of the first pygmy right whale fossil from the Northern Hemisphere based on a collection of the Smithsonian National Museum of Natural History. This specimen was collected from Okinawa, Japan in 1948 by the US Geological Survey and has been housed in the museum. Unfortunately, the exact locality was not properly recorded in the original finding of the fossil despite that it is located somewhere in the United States Marine Corps Base Camp. They are only a few examples that evidence the value of museum collections, which could lead to unexpected discoveries. The label information is critically important for these discoveries as fossil materials cannot be fully used in research if locality information is lost.
Here we report that a long-forgotten desmostylian fossil was uncovered from a museum cabinet, and that two equally possible locations were fortunately marked as the fossil locality based on the original hand-written note associated with the fossil and interview investigation.
In the spring of 2017, one of the authors (Y.K.) happened to find an old wooden box in the geological collection room at the University of Tsukuba. A desmostylian femur was in the box with an old label that contains a person's name, the date and a local address. According to the label information, this uncatalogued specimen was supposed to be discovered from the right bank of the Higashi Karasugawa River, Tategoshi, Tsuchiyu Onsen Town, Fukushima City in 1955. Despite that the local address (Tategoshi) is no longer used as the official postal address, the town, Tsuchiyu Onsen, still exists. In order to find the exact locality of the fossil and search for possible additional specimens, we interviewed local people and reviewed archived documents and photos along with the geological fieldwork. Based on comparative studies, we confirm that the fossil specimen (EESUT-PV-0001) found from Tsuchiyu Onsen town, which was known as a 'dinosaur' bone by locals, is desmostylian Paleoparadoxia and provide emended diagnosis of femoral morphologies for the Desmostylia. Desmostylia is a fossil group of Japanese iconic mammals, and a lot of skeletons and postcranial materials of the genus have been reported from the early to middle Miocene (e.g. [2]) in North Pacific Rim. In the northern part of Japan, many paleoparadoxiid specimens were reported because Miocene strata are well-exposed (Yanagawa [3]; Desmostylus japonicus; Iwaki [4,5]). This specimen shows the best-preserved femoral surface of all Paleoparadoxia specimens.

Material and methods
We compare EESUT-PV-0001 to all desmostylian genera for which the femur is known (Behemotops, Paleoparadoxia, Neoparadoxia, Ashoroa, Cornwallius, Desmostylus); observed specimens are listed below. The phylogenetic relationship of the Desmostylia to the closest living group is still under debate. Some researchers consider it to be a tethytherian (Afrotheria; [6,7]), and others included it in the Perissodactyla (Laurasiatheria; [8,9]

. Measurements
The measurements of the femora were made following Inuzuka [10,11]. Measurements were made in mm units by using a digital caliper (SHINWA Digital Vernier Caliper with Hold Function 19974) and ImageJ2 (Fiji) [12].

Comparative specimens
The following specimens and references were used in this study.  [13,14]. This specimen is the holotype of B. katsuiei. AMP 22 has an erupted M2 and is considered as an adult.

Neoparadoxia.
Two femora of Neoparadoxia were used for this study. One is the femur of Neoparadoxia repeninngi (UCMP 81302), a complete left femur from the middle Miocene Ladera Formation in California, USA. Epiphyses of the whole skeleton of this specimen were fused and the specimen is considered as an adult. The other femur is Neoparadoxia cecilialina (LACM 150150), a nearly complete left femur from the lower upper Miocene Monterey Formation in California, USA. Epiphyses of the femur of LACM 150150 are not fused and the specimen is thus considered as a juvenile [17].

2.1.3.1.4.
Ashoroa. AMP 21, a fragmentary left and a nearly complete right femur of Ashoroa laticosta from the late Oligocene Morawan Formation, Hokkaido, Japan, described by Inuzuka [11,13]. AMP 21 has an erupted M3 and is considered as an adult.  [18]. In this study, we treated these specimens as Cf. Cornwallius sookensis (see discussion part of Systematic Palaeontology). USNM 11075 is missing both epiphyses and is most likely a young individual.
2.1.3.1.6. Desmostylus. We used three femora of Desmostylus. Two are UHR 18466, a nearly complete set of right and left femurs of D. hesperus from the Middle Miocene Uchiboro coal-bearing Formation, Sakhalin, Russia, described by Inuzuka [19]. UHR 18466 shows epiphyseal fusion in the femur and is considered an adult. The other is USNM 26134, a right femur of D. hesperus from Astoria Formation, Oregon, USA. USNM 26134 also shows epiphyseal fusion in the femur and is considered an adult.

Zircon U-Pb dating methods
We separated host pebbly medium-grained sandstone (B110) from the EESUT-PV-0001 specimen for chronological studies. Sample B110 contains angular to subangular clasts of quartz, plagioclase, potassium feldspar, biotite, rock fragments and opaque minerals. Mineral separation was conducted at the Kyoto Fission-Track Co., Ltd, Japan. Zircons were separated from the sample (B110) using crushing, sieving, panning, magnetic separation and heavy liquid (sodium polytungstate) techniques. The separated 100 zircon grains were mounted in a PFA Teflon™ sheet and polished. To check the internal structure of zircons, they were observed in a JEOL JSML-6600LV scanning electron microscope with the cathodoluminescence (CL) and backscattered electron (BSE) detectors at the Geological Survey of Japan, AIST (Tsukuba, Japan). We selected measurement spots for the zircon U-Pb analysis using these images. The analysis spots were selected from the rim part of the zircons with no cracks or inclusions (figure 1). Zircon U-Pb analysis was conducted with a Nu Plasma II multiple collector-ICP-mass spectrometer (MC-ICPMS; Nu Instruments, Wrexham, UK) combined with the IFRIT 260 nm femtosecond laser system (Cyber Laser Inc., Tokyo, Japan) at the Geochemical Research Center, the University of Tokyo. A baffled-type stabilizer [20] was inserted between the ablation chamber and the argon gas inlet to obtain stable isotopic ratios. Prior to the U-Pb isotopic analysis, pre-ablation procedure was adopted to eliminate possible surface contaminations of U and Pb [21]. The measurement spot size was about 15 µm in diameter. Monitored peaks were of 202 Hg, 204 (Hg + Pb), 206 Pb, 207 Pb, 208 Pb, 232 Th, and 238 U. Detailed analytical conditions are shown in electronic supplementary material, table S1. The Nancy 91500 zircon ( 206 Pb/ 238 U age of 1062.4 ± 0.4 Ma; [22]) was used as the primary standard zircon. In addition, the OD-3 zircon ( 206 Pb/ 238 U age of 33.0 ± 0.1 Ma; [23]) was measured as the secondary standard sample during the measurement of unknown samples. Data were acquired in sequences of 54 analyses, consisting of three analyses of NIST (National Institute of Standards and Technology, USA) SRM 612 glass standard, three Nancy 91500 zircon, one OD-3 zircon, 14 unknown, three SRM 612 standard, and three Nancy 91500 zircon, with one gas blank before each analysis.
After the analyses, we processed the data with Microsoft Excel and Isoplot 4.15 [24] and plotted all the data on a concordia diagram. Then we chose concordant zircon data with the 2σ error-ellipse overlapping the concordia curve (a plot of 206 Pb/ 238 U (ordinate) against 207 Pb/ 235 U (abscissa) for concordant samples of various ages should define a single curve; the curve is named concordia curve and is the locus of all concordant U-Pb ages) and drew a probability density plot with a histogram (data interval of 10 Myr) using the 238 U-206 Pb data.

Geological setting and age
As the EESUT-PV-0001 specimen is a historical collection, the field records are lacking. According to memories of local people, the fossil was discovered during the construction of a debris flow barrier (sabo dam) in Higashi Karasugawa River. Based on the local address on the label and interviews with local people, we speculate the exact locality to be within two possibilities, which are only 350 m apart. In either case, the exact locality is covered by the concrete foundation of the sabo dams. One possibility

Geological setting
The lower to middle Miocene Tsuchiyutoge Formation [25] is distributed in the Tsuchiyutoge Area, west of Fukushima City, Fukushima Prefecture. This formation, approximately 600 m in thickness, unconformably covers pre-Neogene rocks and consists mainly of mudstone intercalated with tuff layers [25,26]. Marine fossils (such as bivalves and calcareous nannofossils) and plants were reported. These calcareous nannofossils (such as Cylicargolithus floridanus and Sphenolithus heteromorphus) indicate CN4 zone (13.5-14.9 Ma; [26,27]).

Potential locality no. 1-the Sabo Dam No. 3 of Higashi Karasugawa River
Most of the outcrops were covered by the foundation. However, the outcrop is exposed on the left bank of the river. The sediments are composed of medium-grained volcanoclastic sandstone containing many rock fragments (figure 2).

Potential locality no. 2-confluence point of Higashi Karasugawa River and Arakawa River
The outcrop is completely covered by concrete foundation of the sabo dam, but the upper strata with several metres in thickness are exposed above the level of the foundation. The lithofacies of the strata are massive sandstone mainly composed of volcanic ash, and coloured minerals and rock fragments are very few (figure 2).

Host rock of this specimen
A small portion of the host rock was still attached to EESUT-PV-0001, it is medium sandstone with small gravels (2-3 mm in size). These gravels were composed of coloured minerals and rock fragments. The character of this host rock was almost corresponding to lithofacies of potential locality no. 1, but it was impossible to specify an accurate horizon only by the characteristics of host rock because lithofacies at potential locality no. 2 are impossible to observe. Discordant ages usually imply that one or both of the isotopic systems has been disturbed by some geologic event, such as metamorphism and weathering, following the initial crystallization of the zircon grain (modified from [28]). YZ: youngest zircon age. All the data to generate figure 1 are listed in electronic supplementary material, table S2. The concordant age of the youngest zircon was 15.1 ± 0.8 Ma (2σ). The Th/U ratio of each analysis was 0.14-0.99 and fell in the range of igneous zircon, Th/U > 0.1 [29].

Discussion about depositional age of the specimen
The sample contained many Cretaceous zircons and a Miocene zircon. The Tsuchiyutoge Formation, which the zircon samples come from, unconformably overlies the Abukuma plutonic rocks [30]. Recently, 100-118 Ma zircon U-Pb ages were obtained from the Abukuma plutonic rocks (e.g. [31]). Hence, the Early Cretaceous zircons from the sample (104-120 Ma) were presumably derived from the Abukuma plutonic rocks having the similar range of U-Pb ages. On the other hand, the youngest zircon age of 15.1 ± 0.8 Ma of the sample constrains the depositional age of the Tsuchiyutoge Formation to be 15.9 Ma or younger, considering the error, although the Miocene age was obtained only from one zircon.
Paleoparadoxiidae Reinhart, 1959     Emended diagnosis for the genus. Relatively larger femoral head with respect to the proximal width than Neoparadoxia and Ashoroa; lesser trochanter is bulged to inside; lesser angle of intertrochanteric crest than Behemotops, Neoparadoxia, Ashoroa and Desmostylus; lesser trochanter closer to femoral head than Desmostylus, Cornwallius and Behemotops. For diagnosis of other parts, see [2,34].
The fractured surface of EESUT-PV-0001 shows a thick cortical bone and seems to lack a medullary cavity. In contrast, Cornwallius (USNM 11076) has a medullary cavity [18], and Desmostylus (GSJ F07745, 07748) has a spongy organization and lacks a medullary cavity [35]. There is no available published information of the other desmostylians, but this feature of EESUT-PV-0001 is different from Desmostylus (GSJ F07745, 07748) and Cornwallius (USNM 11075 and 11076).
Discussions. EESUT-PV-0001 was identified as Paleoparadoxia based on the characteristics described above. In addition, we propose new diagnostic characters in femoral morphology. Several previous studies included analyses of femoral characteristics of Desmostylia [10,11,14,36]. According to Inuzuka [36], the differences between Desmostylus and Paleoparadoxia (note that Paleoparadoxia sensu Inuzuka [36] included specimens of genus Archaeoparadoxia, Paleoparadoxia and Neoparadoxia) are: 1. femoral head of Desmostylus is larger than that of Paleoparadoxia, 2. femoral neck of Paleoparadoxia is longer and thinner than Desmostylus, 3. lesser trochanter and its extension of Desmostylus is longer than that of Paleoparadoxia. Inuzuka [36] only used deformed Desmostylus femora. In this study, we used undeformed femora for comparisons. We found only the character 3 to be valid to distinguish Paleoparadoxia from Desmostylus. Inuzuka [14] made a simple comparison list for femora of Behemotops, Paleoparadoxia (possibly the specimens currently referred to as the holotype of Neoparadoxia) and Desmostylus, yet femoral morphology is not yet characterized for official identification at the generic level. Our study is the first report in which morphological characteristics were compared for all desmostylian genera, whose  Table 1. Measurements (in mm) of EESUT-KZ-0001 and comparative specimens. Measurement data are from the original data of this study and [10,19]. Neoparadoxia cecilialina (LACM 150150) is currently mounted in display and accurate measurements have not been taken. Therefore, we take measurements based on photographic images in ImageJ2 (Fiji) [12]. EESUT-PV-0001 has the best-preserved femoral surface of all Paleoparadoxia specimens. In the specimen, muscle scars are easily observable, which makes the specimen useful for future studies that rely on accurate muscle maps for modelling studies of locomotion of the hind limb. We note that some past studies did not clearly show the diagnostic characters of desmostylians (e.g. [37,38]). In these papers, diagnosis is not explicitly written, or comparative analyses are not enough. Some specimens were assigned to desmostylians heavily relying on their occurrence in a formation or its concordant age without careful assessment of the anatomical basis of their identifications. Many of them are included in databases (e.g. PBDB http://paleodb.org/), and then they are used in big data analyses [39][40][41][42][43][44]. For example, the femora described as those of Cornwallius [18] were primarily identified based on little evidence that its morphology is similar to those of other desmostylians but is smaller and different from Behemotops proteus [6], the only other desmostylian known from the synchronous age and general location at the time. Since then, evidence of at least one other smaller desmostylian, Seuku, has been found in the same age and region [45] suggesting that there is greater diversity than previously expected and we need more caution in making assumptions about identities of unassociated specimens.
From the above, we should be more cautious in making assumptions about identities of fossil specimens unassociated with primary locality information.

Discussion
As we mentioned above, EESUT-PV-0001 was known as a 'dinosaur' bone by locals. This 'dinosaur' specimen was long-forgotten but was luckily kept with associated information written on a piece of note, and thus we were able to recover two equally possible localities even more than 60 years after the first discovery. On the note, a person's name (Tadayasu Azuma), year/date (14 June 1955) and the geographical location are written (figure 2). The geographical location reads 'the right bank of the Higashi Karasugawa River, Tatenokoshi, Tsuchiyu Onsen Town, Fukushima City'. Thus, as a part of this research, we interviewed local people and reviewed archived documents and photos in order to specify a possible locality of the fossil and acknowledge why the specimen was brought to the University of Tsukuba.
Tsuchiyu Onsen (Hot spring) is a mountainous town famous till now for hot springs with more than 1400 years of history. This area is, however, also known for destructive flooding due to the steep slopes of the Arakawa river system (Arakawa means a tyrant river in Japanese, implying that this area has suffered from flooding for a long time). Only in the Arakawa river system, there are more than 35 sabo dams (debris flow control structures) to trap debris flow easily caused by heavy rains. Because the river system flows towards the capital city of Fukushima Prefecture, the erosion control of the region has been directly managed by the Ministry of Land, Infrastructure and Transport. As a result, documents about the historical sabo dams and photographs taken during the construction projects are well-archived in the local bureau, the Azuma Mountains Erosion Control Branch Office. According to the available documents, the labelled address corresponds to the Higashi Karasugawa River 3rd dam (figure 2), for which the construction was launched in July, 1952 and completed in March, 1953. Thus, if the piece of note truly contains the original information, the specimen must have been discovered during the construction.
Over the course of the study, we found two people who knew about the fossil and/or Mr Tadayasu Azuma. One is Mr Kazusuke Tanno, a local who worked with Mr Azuma for multiple sabo dam projects, and the other is Mr Yukio Sugeno, a retired high school science teacher. Both of them have known the fossil as a 'dinosaur' bone. In addition, we were also able to communicate with the wife of Mr Azuma's oldest son to get information from her husband.
When we showed the femur to Mr Tanno, he immediately recognized it. According to his memory, a large 'dinosaur' bone with a rib bone and some fragments were found by Mr Tadayasu Azuma during the construction of the Higashi Karasugawa River 1st Dam (figure 2), which is only about 350 m north of the 3rd Dam. Mr Tanno says that the fossils were discovered at the intersection between the Higashi Karasugawa River and its mainstream river, Arakawa River, while they were digging the bedrock to make steps towards the mainstream. It is said that Mr Azuma kept the fossils in his house for a while, but the 'dinosaur' bones were later displayed in the Village Hall. Then, at some point, a university teacher (but not Tokyo Educational University, predecessor of the University of Tsukuba) took the well-preserved ones (presumably, a femur and a rib) with him. Mr Tanno   Tsuchiyu Onsen village of Shinobu County was merged into Fukushima City as Tsuchiyu Onsen town on 1 April 1955. During the incident, the Village Hall also caught on fire. However, the small bones survived from it and were in town until a high school teacher brought them away. We further checked newspapers issued in 1954 and before, but we could not find any clue about the fossils, probably because scientific articles were not common in the newspaper at that time. However, we confirmed that the discovery of 'dinosaur' bones was at least known by local science teachers. Mr Sugeno heard about it in the 1960s from his colleague, Mr Minoru Katagiri, who was also a geology teacher/a geologist. Mr Tadayasu Azuma's oldest son gave us another story. According to him, he is the one who found the fossils from the 3rd Dam when he was helping his father. He also knows that the fossils belong to Desmostylus. Thus, some communication might have been made between a scientist and him at least until the initial identification, but the scientist did not formally report it, and the specimen was long-forgotten in a museum cabinet. Nevertheless, we need to emphasize that Mr Tadayasu Azuma's oldest son did not see the femur for the indirect interview.
As described above, there is a discrepancy in the fossil locality between Mr Tanno's memory and the note associated with the specimen. With current information, we cannot weigh either possibility due to the following reasons. First, because sedimentary facies in the outcrops of both dams are lithologically compatible with a portion of sediment which remained to be attached to the fossil, we cannot determine the specific locality based on geology. Second, based on our historical investigation, we know that the label contains geographical information after 1 April 1955, and then we must presume that the piece of the note was written with some time lag perhaps by a third person, which can result in partially incorrect information. Third, the construction year of the 1st Dam, beginning in 1947 and ending in 1951, is not contradictory to Mr Tanno's memory and the labelled information.  Although we could not identify the specific locality of EESUT-PV-0001, we were able to mark two equally possible localities, which were only 350 m apart. This could be done only because of the written record kept with the fossil in combination with local interviews. Fossils become museum specimens only when they are documented with associated records. Otherwise, they become uninformative rocks. We believe that this specimen is a good lesson not only for vertebrate palaeontologists but also for all museum curators and researchers ( figure 7).
Many studies increasingly use the database like the PBDB database (http://paleodb.org/) to extract information for big data analyses of fossils. This is one of the many reasons for the importance of specimen occurrence information. In marine mammal studies, some recent examples for the use of PBDB include ecological analysis of marine mammals versus diatoms as primary producers and that of marine mammals versus other animals [39,47], the evolution of gigantism in whales and sirenians [48,49], and reconstruction of desmostylian habitats [50]. Correct provenance data (information of locality and occurrence) are requisites for these studies. It is very important to share the information of primary identification so that specimens are not forgotten like the case of this study. However, when using primary identification data and its occurrence records for research, it is very important to pay attention because there is a possibility that it will cause large change in interpretation of the result.

Conclusion
A 'dinosaur' femur was long-forgotten over 60 years and was discovered from the geological collection room at the national university in Japan. The morphological features of this femur correspond to those of Paleoparadoxia. As a result of interviews, we find two potential localities of this forgotten Paleoparadoxia. Combining the zircon data of the host rock and the upper limit age of the strata, the age of this 'dinosaur' is 16-13.53 Ma. Rediscovery of this specimen tells us about the importance of recording secondary information of museum specimens accurately, especially as it is difficult to know how informative or rare a specimen may turn out to be when examined within a comparative context that allows for their proper identification.
Permission to carry out fieldwork. We were not required to get permission prior to conducting our fieldwork. Permission to write interview contents has been obtained from all interviewees.