Taxonomic Revision of Pinus fujiii (Yasui) Miki (Pinaceae) and Its Implications for the Phytogeography of the Section Trifoliae in East Asia

Pinus trifolia Miki 1939 (Pinaceae) was originally proposed based on seed cones from the upper Miocene of Aichi and Gifu Prefectures, central Japan. However, before the publication of P. trifolia, a different name (Pinus fujiii (Yasui) Miki) was given to a female cone with the same morphology. On the other hand, P. fujiii auct. non (Yasui) Miki has been used for seed cones with different morphologies from Yasui’s holotype, i.e., apophyses arranged in 5:8 parastichies and a perexcentromucronate slightly-pointed umbo. As a result of re-examination on the Miki and Yasui specimens, we concluded that P. trifolia was a synonym for P. fujiii and proposed here Pinus mikii sp. nov. for cones assigned to P. fujiii auct. non (Yasui) Miki. We also emended the diagnosis of P. fujiii based on these specimens. Pinus fujiii is characterized by a large female cone in which the apophyses with a centromucronate prickle-like umbo are arranged in 8:13 parastichies, and deciduous seed wings. These characters suggest that P. fujiii belongs to the section Trifoliae of the subgenus Pinus, which is now restricted to North and Central America and the Caribbean islands. Fossil data suggest that the P. fujiii lineage firstly appeared in Japan around the Eocene/Oligocene boundary. We speculate that the P. fujiii lineage might have moved southward to Japan from a refugium located elsewhere in high-latitude areas in response to the late Eocene cooling event, as occurred with other Trifoliae species in North America.


Introduction
Extant species of the section Trifoliae Duhamel [1] (subgenus Pinus, genus Pinus L. [2], Pinaceae Spreng. ex F.Rudolphi [3]) are placed into three subsections [4], Contortae Little and Critchfield [5], Australes Loudon [6], and Ponderosae Loudon [7], and are restricted in their distribution to North and Central America and the Caribbean islands [4,[7][8][9]. The Trifoliae are characterized by two to five leaves clustered in a persistent fascicle sheath and deciduous other specimens were used in this study. Specimens were photographed using a D200 camera (Nikon, Tokyo, Japan) with an AF MICRO NIKKOR 60 mm lens (Nikon) under fluorescent illumination.
The holotype of Pinites fujiii stored in TI consists of a female cone, a replica of the cone, and four microscope slides mounting sectioned parts of the cone. No specimen number is assigned to the holotype, while it is registered as "holotype of Pinites fujiii". The holotype was collected in Seto-shi, Aichi Pref., Japan (Fig 1), from the Seto Formation, but the exact locality is not available.
Other specimens, including Miki's [30] specimens, are stored in OSA F. These specimens were slightly compressed mummifications collected from the Seto or Tokiguchi Formations (Fig 1).

Geological setting
The Seto and Tokiguchi Formations are fluvial deposits consisting of lignites, claystones, siltstones, sandstones and conglomerates [36]. These two formations were formed almost coevally in different basins close to each other (Fig 1) [36]. Their age is estimated as the early late Miocene (10.5 ± 0.4-9.7 ± 0.4 Ma), based on fission-track dating of intercalated tuff layers [37,38]. 83 plant species were reported from these two formations such as Glyptostrobus pensilis Koch, Metasequoia disticha (Heer) Miki [27,38]. Among them, 40 species are extinct from present Japan and extant species of the 9 genera are confined to China and North America. This composition suggests floristic ties between Japan and these regions before the early late Miocene [27].
The holotype was collected from a lignite bed consisting mainly of wood and bark fragments [34]. Although we could not know from what facies Miki's [30] specimens were collected, it is reported that "P. trifolia" cones densely occur in semiautochthonous plant litters which contained in fine-to coarse-grained sandstone beds of channel bar deposits [38]. Miki [30] collected other organs of genus Pinus along with cones from the same horizons, supporting that these Pinus remains were not transported for a long distance from their living sites.

Phylogenetic analyses
Phylogenetic analyses were conducted based on the morphological character matrix of Gernandt et al. [4] with some modifications. The number of vascular bundles (character #1) and resin duct position in leaves (charcter #4) were removed from the matrix because these are not available for P. fujiii. Distribution (character #9) was also removed from the matrix. Dissection of the basal cone scales was added to the matrix in which character states were coded as present (1) or absent (0) after Klaus [10] and Farjon and Styles [11].
A batch file for parsimony ratchet analysis was generated by PRAP2 [39] with settings of 1000 ratchet replicates, weight 2 and 25% weighted. Parsimony ratchet analysis was conducted by PAUP Ã version 4.0b10 [40] based on the batch file. Bootstrap supports were calculated by making 1000 replicates, with 10 trees held for each of 100 multiple tree-bisection-reconnection (TBR) search replications. In both parsimony ratchet and bootstrap analyses, molecular phylogeny [4] was used as a backbone constraint.
The character matrix and constraint tree used for these analyses were available as S1 and S2 Datasets.

Nomenclature
The electronic version of this article in Portable Document Format (PDF) in a work with an ISSN or ISBN will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants, and hence the new names contained in the electronic publication of a PLOS ONE article are effectively published under that Code from the electronic edition alone, so there is no longer any need to provide printed copies.
The online version of this work is archived and available from the following digital repositories: PubMed Central, LOCKSS.
Description. The female cone is ovoid or oblong ovoid with an asymmetrical base, 6.5-13 cm long, and 4.9-10 cm wide when it is fully grown (Figs 2, 3B and 3C and Figs 4-6). The basalmost cone scales usually remain on the branch when the cone has fallen off ( Fig 3D). Cone scales are arranged in 8:13 parastichies. Up to 11 vascular bundles are coplanarly arranged in the middle part of a cone scale (Fig 2F. see also [34]). Apophyses are thick, 10-13 mm high, 15-20 mm wide, and pyramidally risen (Figs 2D, 3B and 3C). The risen part of an apophysis is rolled downward on the abaxial (away from the branch) side of a cone, but the bend is not as distinct on the adaxial side (Figs 2D, 3B and 3C). The umbo is centromucronate and the mucro forms a distinct prickle (Figs 2E, 3E and 3F). The seed is articulated to the wing, elliptic, 9-10 mm long, and 4-5 mm wide ( Fig 3H). The seed wing is 13-16 mm long and 8 mm wide (Fig 3G. See also [30]).
Comparison. Ten fossil species of subgenus Pinus are known that have large female cones (greater than 10 cm long) with centromucronate umbos (Table 1). Among them, P. fujiii is similar in female cone characteristics to P. engelhardtii (Menzel) Mai from the lower Miocene of the Czech Republic [22,23], P. lawsoniana from the middle Pliocene of California (CA), U. S.A. [14], P. piperi Dorf from the upper Miocene to Pliocene of CA [18,49], and P. truckeensis Axelrod from the upper Miocene of Nevada, U.S.A. [15], i.e., ratios of width to length are greater than 0.5 and they exhibit pyramidally-risen apophyses. However, the bending of the risen part in the abaxial apophyses, a characteristic of P. fujiii, is not distinct in the former three species. Pinus engelhardtii is also different from P. fujiii in the indistinct mucro [22,23]. Pinus truckeensis has the most similar appearance to P. fujiii among the four species, but apophyses of P. truckeensis are larger than those of P. fujiii [15]. In addition, apophyses have a greater width than height in P. truckeensis [15], while the width is almost the same as the height in apophyses of P. fujiii.
Etymology. Commemorating late Prof. Shigeru Miki who conducted the first comprehensive study on the Japanese fossil Pinaceae and greatly contributed to the clarification of the Neogene vegetation of Japan.
Diagnosis. Female cones conical to ovoid, with a cordate base; cone scales in 5:8 parastichies; apophyses rhombic, moderately swollen, with transverse keel and radiating ridges, vallate; umbos present at the upper one-third to one-half of dorsal apophyses, weakly depressed on basal half, perexcentromucronate; mucros slightly hooked, prominent even in basal cone scales.
Description. Female cones are long elliptic to oblong ovoid, 40-70 mm long, and 18-37 mm wide (Fig 7A-7D). The cone base is cordate with a short peduncle (Fig 7A-7C). Cone scales are arranged in 5:8 parastichies (Fig 7A-7D). The apophysis is rhombohedral to hexagonal, moderately swollen, 8-12 mm wide, and 5-7 mm high at the middle of the cone (Fig 7A-7E). A weak transverse keel and radiating ridges are developed on the apohyses (Fig 7E). The umbo is rhombic to elliptic, 1.6-2.6 mm wide by 1.5-2 mm high, and located one-half to onethird from the upper corner of the apophysis (Fig 7E). The lower half of the umbo is depressed (Fig 7E). An obtusely pointed mucro develops at the upper-center (perexcentromucronate) of the umbo (Fig 7E).
Remarks. Some leaves occur in association with female cones of P. mikii described above (For details, see S1 Table). Leaves are linear, in bundles of two, and up to 1.5 mm wide. A single ridge is present on the abaxial side. The leaf sheath is persistent (Fig 7F).
Yamada et al. [55] compared P. fujiii auct. non (Yasui) Miki (P. mikii in this study) to other fossil and extant species. We inferred that this species has an affinity with extant P. thunbergii Parl. mainly distributed in Japan, as well as P, hwangshanensis W.Y. Hsia and P. tabuliformis Carrière which are distributed in China [55]. This inference suggests phytogeographyic tie between Japan and China during the Miocene [29].

Pinus trifolia Miki is a synonym for Pinites fujiii Yasui
In 1928, Pinites fujiii was established as a new species based on a single female cone from the upper Miocene Seto Formation reported by Yasui [34]. This species is characterized by cone scales arranged in 8:13 parastichies (Fig 1C), apophyses bending downwardly on one side of the cone (Fig 1C and 1D), and umbos with a centrally positioned prickle-like mucro (Fig 2E) [34]. Later, in 1939, a new combination Pinus fujiii was legitimately proposed for Pinites fujiii based on Yasui's [34] holotype [30]. At the same time, Pinus trifolia was established as a new species by Miki, based on female and male cones, leaves, pollen, and seeds from the Tokiguchi and Seto Formations (Fig 3) [30]. However, Miki's [30] female cones of P. trifolia (Fig 3B, 3C, 3E and 3F) clearly exhibit the diagnostic characteristics of the Pinites fujiii specimen reported by Yasui [34].
Although we do not know the exact length in Yasui's holotype because it is compressed vertically [34], it is half the size of Miki's female cones of P. trifolia [30], judging from the width. To evaluate whether P. trifolia differs from P. fujiii in the size of the female cone, we examined variations in female cone sizes within a population by using specimens collected from Akadu or Hatagoya. The size of cones varied continuously in both localities, i.e., 52-85 mm wide and 75-128 long in Akadu (Figs 4 and 6), and 54-105 mm wide and 65-130 long in Hatagoya (Figs  5 and 6), and the minimum width in each population was slightly wider than that of Yasui's holotype [34] (49 mm ; Figs 1 and 4-6). Therefore, it is reasonable to conclude that Yasui's holotype [34] and Miki's cones [30] could be small and large cones of a single species, respectively.
Yasui's holotype [34] would be an immature cone that was not spontaneously detached from a branch. In P. fujiii, the basal cone scales remained on the branch (Fig 3D) when the cones were excised, as seen in Miki's female cones [30] (Figs 3B, 3C, 4 and 5). However, such excision is not observed in the holotype (Fig 2C). The holotype was found in a lignite bed mainly composed of large branches, wood, and bark remains [34], while the other specimens used in this study were collected from clay or sandy siltstone beds. The holotype might have been trapped in the sediment along with the branch bearing it, while excised cones were transported to the depositional place in a different manner from the stunted cone.
In conclusion, Pinus trifolia Miki is a synonym for Pinus fujiii (Yasui) Miki. However, Yasui's diagnosis [34] was solely based on a putatively immature female cone. Thus, we propose here an emended specific diagnosis by integrating diagnostic features stated in both studies [30,34]. It is suggested that one Miki's [30] specimen is designated as an epitype that represents a fully-grown female cone.

Affinity of Pinus fujiii to extant species
Pinus fujiii has leaves clustered in groups of three or four, rarely in five (Fig 3A). This character is commonly found in most extant species of the Trifoliae (subgenus Pinus), especially in species of the subsection Ponderosae [4,11]. Exceptionally in the section Pinus, two species of the subsection Pinaster also have three leaves per fascicle, i.e., P. canariensis C. Sm. and P. roxburghii Sarg. [4,59]. However, cones of P. canariensis and P. roxburghii have excentromucronate umbos [10], while cones of P. fujiii have centromucronate umbos (Fig 3D and 3E), like many Trifoliae species [4,11,15]. In addition, seeds of P. canariensis and P. roxburghii are adnate to the wing [10], contrary to the articulated seeds found in both P. fujiii (Fig 3H) and the Trifoliae species [4,11].
Yasui [34] inferred that P. fujiii is closely related to P. pinaster Aiton of the subsection Pinaster because the vascular bundles of the cone scales are arranged in a single plane in both species (Fig 2F) [34]. However, similar arrangements are also found in some Trifoliae species, such as P. palustris Mill. and P. rigida Mill. (M. Yamada, unpublished data). The two leaves per sheath in P. pinaster [7,10] are also different from the three to five leaves found in P.fujiii.
These morphological characters seem to support Miki's [30] inference that "P. trifolia" is a species of the section Taeda (sensu Pilger [31]) which is now included in the section Trifoliae (sensu Gernandt et al. [4]). To test objectively if these characters support the affinity of P. fujiii with the Trifoliae, we conducted a phylogenetic analysis based on them by using molecular phylogeny of the extant species as a backbone constraint. As a result, P. fujiii was included in a clade consisting of extant Australes and Ponderosae species within the Trifoliae clade with 79% bootstrap supports (Fig 8). Pinus coulteri D. Don, P. fujiii and P. jeffreyi Balf. and P. sabineana formed a clade within the Ponderosae clade, but support for this clade was not sufficient (Fig  8). Since small number of characters was used for this analysis, the obtained result is a preliminary one. However, the result clearly shows that Miki's [30] inference is reasonable as long as the morphological characters available for P. fujiii are concerned. Some molecular clock-based studies suggested that the Trifoliae was originated in North America in the early Miocene [9,26], contrary to the Paleogene fossil records [15,21] and molecular dating suggesting pre-Eocene divergence [25]. If this would be the case, our phylogenetic inference implies that the ancestor of P. fujiii was dispersed from North America to Japan during the Miocene in spite of a huge geographic barrier between them [32]. Alternatively, the result of our phylogenetic analysis would be artifact despite many characters shared between P. fujiii and the extant Trifoliae. Phylogenetic analysis incorporating more morphological characters, as well as molecular dating based on other methods and/or markers, would be helpful to resolve this discrepancy between paleobotanical and neobotanical data.

Implications of Pinus fujiii on history of pines during Cenozoic
Extant species of the section Trifoliae are restrictively distributed in North and Central America and the Caribbean islands [4,[7][8][9]. Many fossil species of this section have also been reported from the Eocene to Pleistocene in North America [12][13][14][15][16][17][18][19][20][21]; thus North America is considered to be the cradle of this section [15,24]. In the traditional paleobotanical scenario on North American pines, the distribution of the Trifoliae is considered to be retracted to refugia that were located in the high and low latitudes and middle latitude uplands during the Eocene [15,24] when the climate was globally humid and warm [60]. The Trifoliae expanded their distribution from the refugia around the Eocene/Oligocene boundary [15,24] as the climate became drastically cooler and drier in this period [60].
A possible ancestor of P. fujiii [52] appeared in Japan around the late Eocene to early Oligocene (ca. 37-31 Ma [51]) when the backbone of the Japanese Archipelago was still connected to the Eurasian Continent [32]. Since Trifoliae-like fossils are extremely rare in eastern Eurasia, it is not clear when and where a lineage giving rise to P. fujiii first appeared there. However, it is possible to hypothesize that the Trifoliae retreated to the high-latitudes of North America [24] and migrated to the high-latitudes of eastern Eurasia by passing through high-latitude corridors in clockwise or counterclockwise directions during the Eocene [61,62]. Ancestor of P. fujiii might have originated from this migrated population. Alternatively, the ancestor might have arrived directly in Japan from high-latitude refugia of North America through the Beringian Corridor around the Eocene/Oligocene boundary. Whichever is the case, the late Eocene cooling event [60] enabled the P. fujiii lineage to move southward to Japan, as was the case with other Trifoliae species in North America [15,24]. The identification of Paleogene fossils with Trifoliae-like appearance in eastern Eurasia would help greatly in tracing the migration history of P. fujiii.