7KLVLVDSRVWSHHUUHYLHZSUHFRS\HGLWYHUVLRQRI DQDUWLFOHSXEOLVKHGLQ+HUHGLW\7KHILQDO DXWKHQWLFDWHGYHUVLRQLVDYDLODEOHRQOLQHDW KWWSG[GRLRUJV K. Aoki et al. 1

1 Approximate Bayesian computation analysis of EST-associated

2 microsatellites indicates that the broadleaved evergreen tree

3 Castanopsis sieboldii survived the Last Glacial Maximum in

4 multiple refugia in Japan

5

6 K Aoki1, I Tamaki2, K Nakao3, S Ueno4, T Kamijo5, H Setoguchi6, N

7 Murakami7, M Kato6, and Y Tsumura4,5

8

9 1United Graduate School of Child Development, Osaka University,

10 Suita, Osaka 565-0871, Japan, 2Gifu Academy of Forest Science

11 and Culture, Mino, Gifu 501-3714, Japan, 3Kansai Research Center,

12 Forestry and Forest Products Research Institute, Forest Research

13 and Management Organization, Kyoto, Kyoto 612-0855, Japan,

14 4Department of Forest Molecular Genetics and Biotechnology,

15 Forestry and Forest Products Research Institute, Forest Research

16 and Management Organization, Tsukuba, Ibaraki 305-8687, Japan,

17 5Faculty of Life and Environmental Sciences, University of Tsukuba,

18 Tsukuba, Ibaraki 305-8572, Japan, 6Graduate School of Human and

19 Environmental Studies, Kyoto University, Kyoto, Kyoto 606-8501,

20 Japan, 7Makino Herbarium, Tokyo Metropolitan University, Hachioji,

21 Tokyo 192-0397, Japan.

22 K. Aoki et al. 2

23 Correspondence: Saneyoshi Ueno, Department of Forest Molecular

24 Genetics and Biotechnology, Forestry and Forest Products Research

25 Institute, Forest Research and Management Organization, Tsukuba,

26 Ibaraki 305-8687, Japan; Tel. +81 29 829 8261; Fax. +81 29 874

27 3720; E-mail: [email protected]

28

29 Running title: ABC-based demographic history of Castanopsis

30 K. Aoki et al. 3

31 Abstract

32 Climatic changes have played major roles in plants’ evolutionary

33 history. Glacial oscillations have been particularly important, but

34 some of their effects on plants’ populations are poorly understood,

35 including the numbers and locations of refugia in Asian warm

36 temperate zones. In the present study, we investigated the

37 demographic history of the broadleaved evergreen tree species

38 Castanopsis sieboldii (Fagaceae) during the last glacial period in

39 Japan. We used approximate Bayesian computation (ABC) for model

40 comparison and parameter estimation for the demographic modelling

41 using 27 EST-associated microsatellites. We also performed the

42 species distribution modelling (SDM). The results strongly support a

43 demographic scenario that the Ryukyu Islands and the western parts

44 in the main islands (Kyushu and western Shikoku) were derived from

45 separate refugia and the eastern parts in the main islands and the

46 Japan Sea groups were diverged from the western parts prior to the

47 coldest stage of the Last Glacial Maximum (LGM). Our data indicate

48 that multiple refugia survived at least one in the Ryukyu Islands, and

49 the other three regions of the western and eastern parts and around

50 the Japan Sea of the main islands of Japan during the LGM. The

51 SDM analysis also suggests the potential habitats under LGM

52 climate conditions were mainly located along the Pacific Ocean side K. Aoki et al. 4

53 of coastal region. Our ABC-based study helps efforts resolve the

54 demographic history of a dominant species in warm temperate

55 broadleaved forests during and after the last glacial period, which

56 provides a basic model for future phylogeographical studies using

57 this approach.

58

59 Keywords: approximate Bayesian computation (ABC), demography,

60 phylogeography, glacial refugia, Castanopsis, species distribution

61 modelling (SDM)

62

63 K. Aoki et al. 5

64 INTRODUCTION

65 Glacial cycles during the Quaternary period have played important

66 roles in shaping the diversity patterns of contemporary populations

67 (Hewitt, 2000; Hewitt, 2004), even in the warm temperate zone

68 where the land was not covered with ice sheets (Tsukada, 1974;

69 Minato and Ijiri, 1976). Severe climatic oscillations drove repeated

70 north–south and highland–lowland migrations of communities and

71 ecosystems, thereby strongly affecting the genetic structure and the

72 distribution of organisms they hosted (reviewed in Taberlet et al.,

73 1998; Hewitt, 1999; Hewitt, 2004). Thus, current geographic patterns

74 of genetic diversity within species reflect their responses to

75 expansions and contractions of their habitats associated with the

76 glacial cycles, particularly the numbers, sizes, and locations of

77 refugia; the timing and severity of bottlenecks; and the rates of

78 recolonization.

79 Molecular phylogeographic studies of the genetic structure of extant

80 populations have provided detailed insights into apparent glacial and

81 postglacial distributions of numerous species (Avise, 2000). Key

82 issues for phylogeographic analyses include the numbers and

83 locations of refugia, which are still uncertain, despite intensive recent

84 investigations, especially in Europe (Provan and Bennett, 2008b; K. Aoki et al. 6

85 Svenning et al., 2008; Stewart et al., 2010) and North America (Soltis

86 et al., 2006; Shafer et al., 2010). In Europe, Svenning et al. (2008)

87 found indications suggesting that temperate trees were probably

88 largely confined to the traditionally recognized southern refugia in

89 and around Mediterranean lowlands, but several boreal tree species

90 may have been widespread not only in these southern areas but also

91 in several areas of northern Europe during the Last Glacial Maximum

92 (LGM). Existence of previously unknown, or cryptic, northern refugia

93 have also been proposed (Birks and Willis, 2008; Provan and

94 Bennett, 2008a; Rull, 2009; Stewart et al., 2010; Parducci et al.,

95 2012; Tzedakis et al., 2013; de Lafontaine et al., 2014) , based on

96 the phylogeographic patterns of several species together with

97 evidence from pollen and plant macrofossils. The eastern edge of

98 East Asia, including East China, Japan, and the Korean Peninsula,

99 harbors diverse of the temperate flora, and the effects of the

100 locations of refugia throughout Quaternary glacial cycles on the

101 current genetic structure are important (Qiu et al., 2011). For East

102 Asian temperate plants, the population fragmentation occurred

103 caused by fluctuations in sea level throughout the Quaternary, and

104 these taxa restricted to distinct refugia (Qiu et al., 2011). Recent

105 phylogeographic studies of the warm temperate trees in this region

106 also detected the multiple localized refugia (Huang et al., 2002; K. Aoki et al. 7

107 Cheng et al., 2005; Cheng et al., 2006; Wu et al., 2006; Lee et al.,

108 2013; Lee et al., 2014; Xu et al., 2015), though most of them were

109 based on a single locus (chloroplast DNA). Tests of refugia models

110 using multi locus genetic data based on estimated impacts of

111 demographic events and the associated genetic processes to

112 patterns of species’ genetic variation can help to understand more

113 detailed history during and after the glacial periods.

114 Advances in population genetics have led to the development of

115 statistical approaches based on coalescent models for estimating

116 population genetic parameters, reconstructing demographic histories,

117 and testing related hypotheses using Bayesian or likelihood-based

118 frameworks (Knowles and Maddison, 2002). In these approaches, a

119 hypothesis is considered more probable than postulated alternatives

120 if all the available data are more likely under the applied model and

121 underlying assumptions (Hickerson et al., 2010). In addition, a more

122 versatile tool for phylogeographic analysis, termed the approximate

123 Bayesian computation (ABC), has been recently introduced. This

124 “likelihood-free” method relatively easily enables the formulation of

125 model-based inferences in a Bayesian setting for evaluating the

126 demographic or evolutionary scenarios (Pritchard et al., 1999).

127 Because it relies on only summary statistics (Beaumont et al., 2002)

128 rather than evaluations of the overall probability of the data, we can K. Aoki et al. 8

129 make models flexibly to conform to the situations and compare them

130 without much computational effort. ABC has been successfully used

131 very recently to estimate the demographic parameters relating to

132 various complex demographic scenarios and to assess the effects of

133 the past climatic oscillations (Gao et al., 2012; Budde et al., 2013; Li

134 et al., 2013; Liu et al., 2014a; Liu et al., 2014b; Louis et al., 2014;

135 Wang et al., 2016; Yoichi et al., 2016; Chen et al., 2017; Ren et al.,

136 2017).

137 Following biogeographic scenarios have been proposed to

138 describe the glacial and postglacial history of Japan’s warm

139 temperate ecosystems based on fossilized pollen data and

140 reconstructions of past climate. The pollen record indicates that

141 refugial populations of broadleaved evergreen forests existed in the

142 warmer southern end of Kyushu in the main islands as well as in the

143 Ryukyu Islands and migrated northward from these southern refugia

144 after the LGM (Tsukada, 1974; Matsuoka and Miyoshi, 1998).

145 However, based on historical climate data and tolerance of cool

146 temperature of the component species of these forests, some

147 ecologists have proposed that these forests might also have survived

148 in multiple refugia along the Pacific coasts of the main islands as well

149 as on the southern end of Kyushu during the glacial periods (Maeda,

150 1980; Kamei and Research group for the biogeography from Würm K. Aoki et al. 9

151 Glacial, 1981; Nakanishi, 1996; Hattori, 2002). However, it has been

152 difficult to resolve the detailed demographic history of the plant

153 species currently inhabiting the broadleaved evergreen forests,

154 especially Castanopsis-type forest, because of the limited fossilized

155 pollen records and of the relatively slow molecular evolution of

156 chloroplast DNA in plants (Wolfe et al., 1987; Chase et al., 1993;

157 Aoki et al., 2005) and the extremely low levels of intraspecific

158 variation in the chloroplast DNA of Japanese broadleaved evergreen

159 species (Aoki et al., 2003; Aoki et al., 2004a; Aoki et al., 2016).

160 Application of the ABC method to reconstruct the demography of the

161 dominant species in the type of the forest can help to understand

162 detailed history from the information of the locations of refugia, the

163 diverged time and the population size changes.

164 In the present study, we examined the demographic history of

165 Castanopsis sieboldii (Fagaceae), the dominant tree species in

166 Japan’s broadleaved evergreen forests, by conducting the ABC

167 method partially using the expressed sequence tag (EST)-associated

168 microsatellite genetic variation analyzed in Aoki et al. (2014). For

169 this, we first used an ABC framework to identify the most likely model

170 of demographic history among the several alternative scenarios

171 based on Bayesian clustering of current genetic variation. We then

172 applied the model to estimate the demographic parameters K. Aoki et al. 10

173 describing effective population sizes of the species at several points

174 in time and the recent migrations in their demographic histories.

175 Finally, we combined the results of the species distribution modelling

176 (SDM) with the ABC analyses and those of previous phylogeographic

177 studies of other organisms inhabiting the same climatic zone to

178 reconstruct the glacial and postglacial history of Japan’s broadleaved

179 evergreen forests.

180

181 MATERIALS AND METHODS

182 Plant materials

183 We used fresh or silica-gel-dried leaves of 792 C. sieboldii trees

184 collected from 33 populations (Table S1, Figure 1) covering most of

185 the species’ natural distribution in Japan. Castanopsis sieboldii is

186 represented by two varieties, var. sieboldii and var. lutchuensis and

187 they are not distributed sympatrically; C. sieboldii var. sieboldii is

188 found in the main islands whereas C. sieboldii var. lutchuensis in the

189 Ryukyu Islands and Kyushu (southern regions from Amamioshima)

190 (Yamazaki and Mashiba, 1987b). These varieties have

191 morphologically different; seed size and shape of C. sieboldii var.

192 lutchuensis are larger and wider than C. sieboldii var. sieboldii,

193 however, intermediate morphologies are often observed (Yamazaki

194 and Mashiba, 1987a). K. Aoki et al. 11

195 C. sieboldii var. sieboldii is sometimes codistributed with its

196 congeneric species C. cuspidata var. cuspidata in the main islands of

197 Japan. Hybrids with intermediate morphology sometimes occur,

198 especially at sites where the two species coexist (Kobayashi and

199 Hiroki, 2003). Our previous study by Aoki et al. (2014) examined the

200 epidermis of the leaves from each sampled individual, since this is

201 the most effective way of discriminating the two species and hybrids

202 (Yamazaki and Mashiba, 1987b). In this study, we identified

203 individuals with double epidermal cell layers as C. sieboldii var.

204 sieboldii, with no ancestral contributions from C. cuspidata var.

205 cuspidata. In subsequent analyses, we used the six C. sieboldii var.

206 lutchuensis populations in the Ryukyu Islands and the 27 C. sieboldii

207 var. sieboldii populations in the main islands, which do not contain

208 any individuals of C. cuspidata var. cuspidata.

209 The value of minimum temperature of the coldest month of the

210 altitudinal distribution of Castanopsis was estimated to be 0Ѹ1 °C

211 (Hattori, 1985). Castanopsis is insect pollinated plant species, which

212 is difficult to detect the fossilized pollen in the past. Fossilized pollen

213 of Castanopsis trees at Ryukyu Islands existed at the LGM (Kuroda,

214 1998). The fossilized pollen records of Castanopsis-type broadleaved

215 evergreen tree genus (i.e., Myrica and, or Podocarpus) at the LGM

216 existed in southwestern Kyushu (at the mean frequency of several to K. Aoki et al. 12

217 10 %), northern Kyushu and Japan Sea (several %) as well as at

218 Ryukyu Islands (10 %) (Kuroda, 1998; Takahara and Takeoka, 1992).

219 There were several fossilized pollen records of Castanea-

220 Castanopsis (less than 1 % to several %) at the LGM in Japan Sea

221 (Takahara and Takeoka, 1992; Miyoshi et al., 1999) and along the

222 Pacific coasts (Kanauchi, 2005), however, we cannot determine

223 these pollen as Castanopsis.

224 Fagaceae acorns including Castanopsis are dispersed by

225 transporting and caching by animals, especially jay and rodents, and

226 jay participates in longer distance dispersals, for example, 100Ѹ1900

227 m (Darleyhill and Johnson, 1981), 250Ѹ1000 m (Gómez, 2003) and 3

228 Ѹ500 m (Pons and Pausas, 2007) than rodents, 4Ѹ37 m (Jensen and

229 Nielsen, 1986) and 30Ѹ40m (Miyaki and Kikuzawa, 1988).

230

231 EST-SSR analysis

232 Total DNA was extracted from each leaf sample using either the

233 CTAB (hexadecyltrimethylammonium bromide) method (Murray and

234 Thompson, 1980) or the method of Doyle and Doyle (1987) after

235 removing the polysaccharides using HEPES buffer (pH 8.0)

236 (Setoguchi and Ohba, 1995). We determined the genotypes of each

237 sample based on 32 pairs of nuclear microsatellite markers

238 (expressed sequence tags-simple sequence repeats, EST-SSRs) K. Aoki et al. 13

239 used in the study of Aoki et al. (2014). The DNA at each EST-SSR

240 locus was amplified using a QIAGEN Multiplex PCR Kit following the

241 manufacturer’s recommended protocol. PCR products were detected

242 using a PRISM 3100 sequencer in conjunction with GENESCAN

243 software, and genotypes were scored using GENOTYPER software

244 (both supplied by Applied Biosystems).

245

246 Choosing suitable loci for ABC analyses

247 Because we need to simulate neutral loci without null alleles in the

248 ABC analysis in order to estimate the past population demography,

249 we checked the neutrality and the frequencies of null alleles of each

250 locus. First, we converted the genotype data from size numbers to

251 repeat numbers; in the course of this procedure, indels were not

252 taken into consideration. For these alleles with indels other than

253 SSRs, we modified the allelic calls to the nearest alleles in size.

254 Second, we estimated the frequencies of null alleles at each SSR

255 locus within a population using INEst 2.0 (Chybicki, 2014). The

256 unique feature of this program is taking into account a possibility of

257 inbreeding within a population during the estimation of null allele

258 frequencies. We removed the two loci with the null allele frequencies

259 of values >0.05. Third, we compared the distribution of the FST values

260 over all loci to their expected distributions under an island model with K. Aoki et al. 14

261 the assumption of neutrality using the LOSITAN program (Antao et

262 al., 2008), based on fdist as described by (Beaumont and Nichols,

263 1996). To calculate the approximate P values for each locus, 10,000

264 independent loci were generated and the simulated FST distribution

265 was compared with the observed FST values. This made it possible to

266 identify outliers in a one-step process by defining them as observed

267 FST values falling outside the 99% confidence interval for the

268 simulated group. We removed three outlier loci in the process. In

269 total, we removed the five EST-SSR loci (two loci with high null allele

270 frequencies and three outlier loci) in these steps from the 32 loci of

271 which C. sieboldii genotypes were scored and used the remaining 27

272 loci in the subsequent analyses.

273

274 Genetic diversity

275 We calculated the total number of alleles (NA), the average gene

276 diversity within populations (HS) (Nei, 1978), the observed

277 heterozygosity (HO), and the FIS fixation indices (Weir and

278 Cockerham, 1984) across all populations at each locus to estimate

279 departures from the Hardy–Weinberg equilibrium (HWE). The

280 significance of deviations of FIS from zero was evaluated by

281 permutation tests with sequential Bonferroni correction. K. Aoki et al. 15

282 We calculated the average NA, the unbiased heterozygosity (Nei,

283 1978), and the allelic richness (El Mousadik and Petit, 1996), based

284 on data from a minimum sample size of 17 per population, using

285 MSA (Dieringer and Schlotterer, 2003) and FSTAT version 2.9.3.2

286 (Goudet, 2002) software.

287

288 Genetic structure

289 We used the genotype data, based on the 27 EST-SSR loci, of 792 C.

290 sieboldii trees collected from 33 sites. We estimated the genetic

291 structure among populations by constructing a neighbor-joining (NJ)

292 tree (Saitou and Nei, 1987) based on DA distances (Nei et al., 1983)

293 between all pairs of populations using MSA and PHYLIP

294 (Felsenstein, 1989). We used 10 C. cuspidata populations as

295 outgroups that do not contain any individual of C. sieboldii var.

296 sieboldii populations (the populations consisting of only C. cuspidata

297 type of leaf epidermal cells in the study of Aoki et al. (2014)).

298 We used the Bayesian clustering method to elucidate the genetic

299 structure among the C. sieboldii populations using TESS version 2.3

300 (François et al., 2006; Chen et al., 2007; Durand et al., 2009),

301 because TESS may detect less clusters than STRUCTURE and is

302 suitable for the data in the presence of clines and the sampling

303 covered the species' geographic distribution (François et al., 2006; K. Aoki et al. 16

304 Chen et al., 2007; Durand et al., 2009). A total of 100 simulations

305 were run for each value of K (2–10) with 50,000 Markov chain Monte

306 Carlo samplings after 10,000 burn-in iterations using the admixture

307 model. Following the manuals (François et al., 2006; Chen et al.,

308 2007; Durand et al., 2009), we filtered the candidate runs by

309 choosing to keep 20% of them with the lowest deviance information

310 criterion values and estimated Q-matrices of each K value from a

311 mean of the permuted matrices across replicates using CLUMPP

312 (Jakobsson and Rosenberg, 2007). The most appropriate cluster

313 number (K) was selected using the ΔK criterion of Evanno et al.

314 (2005).

315

316 Estimation of population demographic history

317 We used the four distinct genetic groups, the Ryukyu Islands, the

318 West (western parts of the main islands along the Pacific Coast), the

319 East (eastern parts of the main islands along the Pacific Coast), and

320 the Japan Sea (Japan Sea side of the main islands) shown by the NJ

321 and Bayesian clustering methods of C. sieboldii for the subsequent

322 ABC analyses. Four probable alternative scenarios for population

323 diversifications in four genetic groups of C. sieboldii were then

324 constructed (Figure 4). Model 1 assumed one refugia in the southern

325 Kyushu as well as another in the Ryukyu Islands. This Model K. Aoki et al. 17

326 corresponds to the most classic history supported by the fossilized

327 pollen record of the broadleaved evergreen trees at LGM (Tsukada

328 1974). Therefore, we assumed that the Ryukyu Islands and the West

329 groups were derived from the refugia and the other two groups were

330 recently diverged from the West. Model 2 assumed that as large

331 refugia existed in the Pacific Ocean side of coastal region in the East

332 as in the West and the Ryukyu Islands. This Model corresponds to

333 the alternative history supported by the past climatic conditions and

334 tolerance of cool temperatures of the broadleaved evergreen trees

335 (Hattori 2002, Kamei and Research group for the biogeography from

336 Würm Galacial, 1981). So we modeled that the Ryukyu Islands, the

337 West and the East groups were derived from each refugia, and the

338 Japan Sea group was recently diverged from the West group. Model

339 3 was based on the information of the phylogeographic tree that was

340 constructed by mitochondrial DNA sequence of Curculio hilgendorfi,

341 which is the host-specific predator of C. sieboldii acorns (Aoki et al.,

342 2008). The phylogeographic tree was chosen because the

343 association between the host plant C. sieboldii and its host-specific

344 insect is known to be very tight (Aoki et al., 2005). Moreover, the

345 geographic patterns of the genetic differentiation of these species

346 seemed to be similar (Aoki et al., 2008; Aoki et al., 2014). This model

347 assumes that first, the populations of the Ryukyu Islands diversified K. Aoki et al. 18

348 from the other three groups, then the East diversified from the other

349 two groups, and finally, the West and the Japan Sea diversified.

350 Model 4 was constructed based on the result of the TESS analysis at

351 K = 2, where the West and the Japan Sea showed genetic admixture

352 between ancestries dominated in the Ryukyu Islands and the East.

353 We assumed that all the four groups exponentially grew with gene

354 flow (except for the model 4) from the time after the four groups were

355 established (from the present to T1). We assumed no migration in the

356 model 4, because in this model, observed multiple ancestries in the

357 West and the Japan Sea groups were solely explained by genetic

358 admixture at T1. Symmetric migrations were considered only

359 between the Ryukyu Island and the West, between the West and the

360 Japan Sea, and between the West and the East groups, because

361 these three group pairs showed relatively low FST values (FST < 0.02;

362 Figure S1) and adjacently located. Moreover, when the source-sink

363 relationships existed, asymmetric migration from sink to source in the

364 direction of coalescence was assumed (e.g., from the Japan Sea to

365 the West, and from the East to the West in model 1, and from the

366 Japan Sea to the West in model 2). In the preliminary analyses, we

367 tested without exponential growth or without migration (other than

368 model 4) models, but supports of such models were much lower

369 (data were not shown). We conducted coalescent simulations for K. Aoki et al. 19

370 each four models, compared the four models, and estimated the

371 parameters using the ABC technique (Beaumont et al., 2002).

372 We used randomly selected 100 diploid individuals per group

373 to reduce the computational costs. To check the sampling biases, we

374 drew randomly selected 100 individuals in each group and calculated

375 the summary statistics 100 times. The average and standard

376 deviation among the 27 loci for the NA, the heterozygosity, and the

377 allele size range in each of the four groups were calculated. Pair-

378 wise FST and FST among the total four groups for overall loci were

379 also calculated. A total of 31 summary statistics for the observed

380 data were calculated using arlsumstat version 3.5.2 (Excoffier and

381 Lischer, 2010). We compared them and confirmed that sampling

382 biases were very small.

383 We assumed that current effective population size (NCUR) was

384 common among the four groups because the average and standard

385 deviation values of the NA, the heterozygosity, and the allele size

386 range were not so different among the four groups (Figure S1), and

387 this indicated that these four groups had experienced a similar recent

388 population history. Moreover, each group was assumed to have

389 shrunk at the rate G from the current to T1, i.e., to have grown from

390 T1 to the current [NT = NCUR exp (G × T); NT is an effective population

391 size at time T]. Prior for G was drawn from uniform distribution from K. Aoki et al. 20

392 í0.01 to 0. The population size for the ancestral population among all

393 the four groups (NANC) was also set. Priors for the NCUR, and NANC

394 were drawn from uniform distribution from 1 to 104 independently.

395 The unit of all the population sizes was scaled in the unit of the

396 number of diploid individuals and converted into the number of

397 haploid individuals by multiplying with two when applying a

398 coalescent simulator. The time when the demographic events

399 occurred (T) was scaled in the unit for generations. Priors for Ts were

4 400 drawn from uniform distribution from 1 to 10 simultaneously, and T1

401 < T2 < T3 was assumed. The number of migrants per generation

402 (Nm) was used as migration parameter and the value divided by

403 NCUR (i.e., migration rate) was passed to the coalescent simulator. A

404 single common value was used for all migration events. Prior for Nm

405 was drawn from uniform distribution from 0.1 to 20. In model 4, Pij

406 indicates the proportion of genes that coalesce from the group i to j

407 within the group i at T1. Priors for PWR and PJR were drawn from

408 uniform distribution from 0 to 1 and then PWE and PJE were 1 – PWR

409 and 1 – PJR, respectively. W, R, J and E denote the West, the

410 Ryukyu Islands, the Japan Sea and the East groups, respectively.

411 Generalized stepwise mutation (GSM) model was used for the

412 microsatellite mutation model (Estoup et al., 2002). The GSM model

413 has two parameters, mutation rate (ȝ) and geometric parameter K. Aoki et al. 21

414 (PGSM). The value of PGSM can range from 0 to 1, and PGSM = 0

415 indicates the strict stepwise mutation model. To accommodate the

416 variation among loci, locus-specific ȝ and PGSM were generated as a

417 random effect variable. We applied the slightly modified method used

418 by Excoffier et al. (2005). The value of ȝ for microsatellites is

419 generally in the range from 5 × 10í5 to 5 × 10í3, and several studies

420 have used 5 × 10í4 (Estoup and Angers, 1998; Estoup et al., 2002).

421 In this study, we set the mean value of ȝ to 5 × 10í4 and each locus

422 value was assumed to follow a Gamma distribution with shape and

423 rate parameters. Prior distribution of the shape parameter was drawn

424 from uniform distribution from 0.5 to 2.0 and that of the rate

425 parameter was calculated by shape / (mean value of ȝ). Prior of the

426 mean value of PGSM among the loci was drawn from uniform

427 distribution from 0 to 0.8, and each locus value was assumed to

428 follow beta distribution with Į and ȕ parameters. According to

429 Excoffier et al. (2005), Į = 0.5 + 199 (mean PGSM) and ȕ = Į (1 í

430 mean PGSM)/(mean PGSM) were used.

431 Coalescent simulations were replicated 1 × 104 times in each

432 model using program fastsimcoal2 version 2.5.2.21 (Excoffier and

433 Foll, 2011). A total of 31 summary statistics that were the same to the

434 ones calculated by the observed dataset were calculated using

435 arlsumstat. Since there were too many numbers of summary K. Aoki et al. 22

436 statistics, we consistently used machine learning approach to reduce

437 the effects of curse of dimensionality in model choice and parameter

438 estimation. ABC model choice via random forest (ABC-RF)

439 implemented in the abcrf package in R was conducted (Pudlo et al.,

440 2016). ABC-RF allows us to compare models with relatively small

441 number of simulations and thus greatly contributes to save time. One

442 thousand trees were constructed and the best model was selected by

443 classification votes of random forest. Posterior probability of the best

444 model and confusion matrix among compared models were also

445 calculated. We replicated 2 × 106 simulations for the best model and

446 posterior distributions of parameters were estimated by a neural

447 network regression method implemented in the abc package in R

448 (Blum and Francois, 2010; Csillery et al., 2012). Tolerance value was

449 set to 5 × 10-4 to obtain the closest 1,000 simulated datasets to the

450 observed dataset. The number of neural networks was set to 20.

451 When estimating the posterior distributions by the regression method,

452 to prevent the estimated posterior falling outside the lower or upper

453 bounds of the priors, the logit transformation option of the abc

454 function was used. Posterior mode and the 95% highest posterior

455 density (HPD) were calculated using density and HPDinterval

456 functions in the base and coda packages, respectively, in R

457 (Plummer et al., 2006). Time was converted to years, assuming that K. Aoki et al. 23

458 the generation time of C. sieboldii is 25 or 100 years per generation

459 for short and long expected values. Short and long generation time

460 values assume the average age to start reproduction (Avise, 2000)

461 and the intermediate age between the age at maturity and the

462 maximum life span (Petit and Hampe, 2006).

463 Finally, a posterior model check was performed to assess the

464 goodness-of-fit to the data estimated from the posterior distributions

465 (Csillery et al., 2012). A total of 1,000 new genotype datasets were

466 generated using parameters randomly sampled from the posterior

467 distributions. Summary statistics were calculated and compared with

468 the observed data. To validate the parameter estimation, posterior

469 quantiles for each parameter were calculated with 1,000 pseudo

470 observed data set (Cook et al., 2006; Wegmann et al., 2010). When

471 the parameter estimation is unbiased, distribution of posterior

472 quantiles can be approximated by a uniform distribution. To test the

473 deviation from a uniform distribution, a Kolmogorov-Smirnov test was

474 conducted using R.

475

476 Species distribution modelling

477 To assess the refugia of C. seiboldii under LGM (21,000 years BP)

478 climate conditions, the geographical distribution of potential habitats

479 were estimated using the existing species distribution model of the K. Aoki et al. 24

480 target species (Nakao et al., 2014). The model was constructed with

481 the presence/absence records of the species as a response

482 variables, and four climatic factors (Bio6, mean of daily minimum

483 temperature of winter; Bio10, mean temperature of warmest quarter;

484 Bio18, precipitation of warmest quarter; and Bio19, precipitation of

485 coldest quarter) as explanatory variables (Nakao et al., 2014).

486 Climatic variables at a spatial resolution of 2.5 minutes were used as

487 explanatory variables which extracted from the WorldClim database

488 (http://www.worldclim.org). We also predicted potential habitats

489 under LGM climate conditions using three LGM climate data (i.e.

490 MIROC, CCSM, MPI-ESM) from WorldClim.

491

492 RESULTS

493 Genetic diversity

494 The EST-SSR loci were highly polymorphic (Table S2). Across all

495 populations, the FIS values deviated significantly and positively from

496 zero at two loci. High levels of genetic diversity within populations

497 were also observed in each population (mean numbers of alleles,

498 expected heterozygosity, and allelic richness were 5.226, 0.574, and

499 4.888, respectively; Table S1). Of the four genetic groups of the ABC

500 analyses, the average allelic richness among each populations was

501 highest in the group of the Ryukyu Islands (5.337), followed by those K. Aoki et al. 25

502 in the West (5.191), the East (4.680), and the Japan Sea (4.659).

503 The amount of unique alleles to each group per population was also

504 highest in the group of the Ryukyu Islands (2.667), followed by those

505 in the West (1.000), the East (0.846), and the Japan Sea (0.625).

506

507 Genetic structure

508 The NJ tree of C. sieboldii with C. cuspidata populations as

509 outgroups contained four clusters corresponding to populations from

510 the Ryukyu Islands and the West and the East and the Japan Sea in

511 the main islands (Figure. 2).

512 The Bayesian clustering of the C. sieboldii populations indicated

513 that ΔK was highest when K = 2; however, fractions of ancestry from

514 three or four clusters (K = 3 or 4) were also found (Figure. 3). Black

515 cluster dominated in the populations of the Ryukyu Islands (97%),

516 while white cluster dominated in the populations of the East (98%)

517 but was also present in the populations around the Japan Sea (88%)

518 at K = 2. Light gray clusters at K = 3 and 4 dominated in the

519 populations around the Japan Sea. Populations of the West and

520 around the Japan Sea showed genetic admixture between three

521 clusters (the West, black (47%), white (15%), and light gray (36%)

522 and the Japan Sea, black (2%), white (6%), and light gray (75%) at K

523 = 4). K. Aoki et al. 26

524 The NJ tree and the admixture analyses using TESS of C.

525 sieboldii showed that there were almost shared four distinct genetic

526 groups (the Ryukyu Islands, the West, the East, and the Japan Sea).

527 Only one population of No. 20 was assigned to the Japan Sea in the

528 NJ tree, while to the West in the TESS analysis. We assigned the No.

529 20 population to the West genetic groups in the subsequent ABC

530 analyses because No. 20 geographically located around the Pacific

531 Ocean and was not suitable for including in the Japan Sea genetic

532 group. The populations of No. 13, 14, and 19, which located

533 geographically intermediated between the eastern and western parts

534 of the main islands and showed intermediate cluster between the

535 eastern and western clusters by the NJ tree, were assigned to the

536 East genetic groups in the subsequent ABC analyses because the

537 black cluster (8%, 9%, and 11%, respectively at No. 13, 14, and 19)

538 that dominated in the populations of the Ryukyu Islands was less

539 frequent and the white cluster (64%, 65%, and 42%, respectively)

540 that dominated in the populations of the East was more frequent in

541 the TESS analysis.

542

543 Model comparison and parameter estimation for the

544 demographic modelling K. Aoki et al. 27

545 Prior to the explanation for the results of model comparison and

546 parameter estimation, in this paragraph, we summarized the

547 validations of model comparison and parameter estimation.

548 Confusion matrix of model comparison among the four demographic

549 models was shown in Table S3. Classification error rates of each

550 model ranged 0.010 to 0.238 and overall classification error rate was

551 0.116 (Tables 1 and S3). These low classification errors indicate that

552 the wrong model is rarely selected with high confidence. Distributions

553 of predicted summary statistics using posterior distribution of the best

554 model were well congruent with the observed ones (Figure S2).

555 Therefore, we concluded that the goodness-of-fit of the best model to

556 the observed data was well. Posterior quantiles of each parameter

557 were not significantly deviated from a uniform distribution except for

558 PGSM (Figure S3). Distribution of posterior quantile of PGSM was

559 biased toward 1.0. These results suggest that the parameters except

560 for PGSM can be estimated without bias, but PGSM may be

561 underestimated. However, as the PGSM is a nuisance parameter, it

562 will not become a big problem when we consider the following results.

563 Model 1 was selected as the best model among the compared

564 four candidate models with a posterior probability of 0.906 (Table 1).

565 Posterior distributions of all parameters for the best model were

566 clearly different from prior ones (Table 2; Figure S4). All posterior K. Aoki et al. 28

567 distributions of parameters showed a clear single peak. Posterior

568 modes (95% HPD) of NCUR and NANC were 5,463 (2,349–9,239) and

569 1,517 (13–8,809), respectively. Posterior mode (95% HPD) of the

570 effective population size at T1 estimated using G was 1,468 (989–

571 3,228; Figure S5).

572 Posterior modes (95% HPD) of T1 and T2 were 431 (95–2,453)

573 and 1,184 (110–8,311) generations ago, respectively. When the

574 short generation time (25 years per generation) was applied,

575 posterior modes of T1 and T2 were 10,775 and 29,600 years ago,

576 respectively. When the long generation time (100 years per

577 generation) was applied, posterior modes of T1 and T2 were 43,100

578 and 118,400 years ago, respectively.

579 Posterior mode (95% HPD) of Nm was 9.70 (6.08–17.82).

580 According to Sewall Wright’s one migrant per generation theory, one

581 migrant is sufficient to prevent complete differentiation of the

582 populations (Wright, 1931). As Nm was significantly larger than 1.0,

583 we considered that migrations were effective.

584

585 Species distribution modelling

586 Potential habitats under the present climate conditions were mainly

587 located in the lowland around the coastal regions. Potential habitats

588 under LGM climate conditions were mainly located in the Pacific K. Aoki et al. 29

589 Ocean side (southern to northwestern Kyushu, southern Shikoku to

590 western Kii Pen. and around Izu Pen.), including the area of below

591 sea level under current climate conditions (Figure 5).

592

593 DISCUSSION

594 The numbers and the locations of the glacial refugia in several areas

595 of the world, including Japan, are key phylogeographic concerns and

596 have been intensely debated. The ABC analysis of the C. sieboldii

597 genetic dataset provided much stronger support for the scenario that

598 the Ryukyu Islands and the West groups were derived from separate

599 refugia and the East and the Japan Sea groups were diverged from

600 the West. Available fossil pollen data for broadleaved evergreen

601 trees during the LGM (Matsuoka and Miyoshi, 1998) suggest that

602 refugia for Castanopsis-type broadleaved evergreen forests were

603 present in the Ryukyu Islands and the southwestern Kyushu in the

604 main islands. The higher levels of genetic diversity of C. sieboldii, the

605 ancient diversification from ABC analysis and higher probability of

606 potential habitats under LGM climate conditions from SDM analysis

607 are observed in the populations of the Ryukyu Islands and the West

608 group containing Kyushu, suggesting that these populations have

609 remained sufficiently large for ancestral polymorphism to be retained

610 from the glacial periods up to the present day. K. Aoki et al. 30

611 The time when the East and the Japan Sea groups diverged from

612 an ancestral population of the West were estimated to be 431

613 generations ago. Under the coalescent theory, basic demographic

614 parameters related to population size, timescale and migration must

615 be scaled by the rate of genetic drift or by the mutation rate (Hey and

616 Nielsen, 2004; Wakeley, 2008). In the case of timescale, the

617 parameter is expressed in terms of mutations as the number of

618 generations × mutation rate (Hey and Nielsen, 2004). In this study,

619 we thus fixed the value of mutation rate at the moderate level (5 ×

620 10í4) and estimated the number of generations. As we used EST-

621 SSR and in generally mutation rate for EST-SSR is considered to be

622 lower than that for genomic-SSR (Cubry et al., 2014), the estimated

623 values of T1 and T2 might be underestimated. Moreover, lacking

624 accurate information on the generation time for Castanopsis, we

625 assumed generation times of 25 and 100 years in ABC analyses.

626 When we consider the generation time, it is important to distinguish

627 among age at maturity, generation time and lifespan (Petit and

628 Hampe, 2006). For instance, according to the first complete lifetable

629 for a tree, the estimated generation time of the palm tree Enterpe

630 globosa is 101 years, intermediate between its age at maturity (50

631 years) and maximum observed lifespan (156 years) (van Valen,

632 1975). Thus, we choose more probable 100-year generation time, K. Aoki et al. 31

633 which provide that the East and the Japan Sea groups were diverged

634 from the West at 43,100 years ago. The time predates the coldest

635 stage around 18,000 years ago (Maeda, 1980; Tsukada, 1983;

636 Martinson et al., 1987) of the LGM. These results suggest that the

637 current populations of C. sieboldii descended from at least four

638 isolated populations that were established prior to the LGM, however,

639 the groups of the eastern parts and around the Japan Sea were more

640 recently established from ancient southern refugia. Thus, these

641 results provide evidence for the presence of LGM refugia in the

642 regions of the eastern parts and around the Japan Sea, as well as

643 the Ryukyu Islands and the western parts in Japan.

644 There was clear evidence of about four times expansion in the

645 historical effective sizes from the glacial periods to the present

646 (Figure S5). Chronological data of the earth’s climate (Martinson et

647 al., 1987) and pollen records from the Japanese forests (Maeda,

648 1980; Tsukada, 1983) suggest that the Japanese climate was coldest

649 around 18,000 years ago, started to warm about 12,000 years ago,

650 and warmest about 6,000–8,000 years ago. The result suggest that

651 the smaller populations of the four genetic groups may have survived

652 in each glacial refugium for several glacial and interglacial periods in

653 the Quaternary period, and when the climate began to get warm after

654 the LGM, these populations regained and expanded their population K. Aoki et al. 32

655 sizes from the coastal regions to the inland areas. Migrations also

656 occurred between adjacent populations from the southwestern to the

657 eastern regions in the main islands of Japan from the LGM to the

658 present.

659 In the main islands of Japan, a similar genetic differentiation

660 between the western and eastern populations that was observed in C.

661 sieboldii has also been detected in several other species (both plants

662 and animals mentioned later) of the broadleaved evergreen forests.

663 Geographical boundaries between the western and eastern

664 populations are often reportedly located in a region spanning the

665 main islands of Shikoku and Chugoku, the western tip of Honshu to

666 Kii Peninsula, for example, Elaeocarpus sylvestris var. ellipticus (Aoki

667 et al., 2004b), Photinia glabra (Aoki et al., 2006), Myrsine seguinii

668 (Aoki et al., 2011) and Camellia japonica (Ueno, 2015) for plant

669 species, C. hilgendorfi; a specific predator of C. sieboldii seeds (Aoki

670 et al., 2008) and Rhynchaenus dorsoplanatus; a specific leaf miner of

671 C. sieboldii leaves (Aoki et al., 2010) for insect species. Genetic

672 boundaries between several species of warm to cool temperate

673 zones in Japan also appear to be located in this region, for example,

674 3LQXV 0L\DWDDQG8EXNDWD,ZDL]XPLHWDO Cerasus

675 (Tsuda et al., 2009), Zanthoxylum

676 Carpinus and Magnolia (Iwasaki et al., 2010; Iwasaki et al., 2012)

677 and Acer (Yoshimaru and Matsumoto, 2015) for plant species;

678 Curculio (Aoki et al., 2009) for insect species; and Cervus 1DJDWDHW

679 DO , Macaca .DZDPRWRHWDO , Petaurista 2VKLGDHWDO

680  , Ursus

681 2010) for mammal species. 7KHFRPPRQJHRJUDSKLFGLIIHUHQWLDWLRQ

682 DPRQJPXOWLSOHFRGLVWULEXWHGWD[DPD\KHOSHIIRUWVWRHOXFLGDWHWKH

683 UHODWLYHLQIOXHQFHRIPDMRUKLVWRULFDOHYHQWVVXFKDVFOLPDWHFKDQJHV

684 GXULQJWKHJODFLDODQGLQWHUJODFLDOSHULRGV $YLVH 7KHFRPPRQ

685 JHQHWLFXQLTXHQHVVRIWKHHDVWHUQSRSXODWLRQVREVHUYHGLQYDULRXV

686 FRPSRQHQWVSHFLHVRIEURDGOHDYHGHYHUJUHHQIRUHVWVDOVRVXJJHVWV

687 WKHHDVWHUQUHIXJLDLQWKHPDLQLVODQGV

688 There is no clear palynological evidence at the LGM that

689 considerable number of Castanopsis-type broadleaved evergreen

690 trees survived the LGM in the eastern regions of the main islands

691 (Figure S7). However, frequencies of fossilized pollen of broadleaved

692 evergreen trees in the eastern parts of Japan began to markedly K. Aoki et al. 34

693 increase at the southern ends of the Kii, Izu, and Boso Peninsulas

694 (located sequentially eastward along the Pacific Coast, close to the

695 route of the warm Kuroshio Current) from 7,500, 8,500, and 7,500

696 years ago, respectively, approximately 500–1,500 years after similar

697 increases at the southern ends of Kyushu (Matsushita, 1992).

698 Paleontological evidence also indicates that several other

699 broadleaved evergreen trees were present on the Boso Peninsula

700 about 9,000 years ago (Okazaki et al., 2011). Moreover, the genetic

701 diversity and the distinct chloroplast DNA haplotypes of these types

702 of trees suggest the existence of refugia around the Kii Peninsula

703 (Aoki et al., 2004b; Liu et al., 2013). Potential habitats of C. sieboldii

704 under LGM climate conditions also located in the Pacific Ocean side

705 including the area of below sea level under current climate conditions.

706 Thus, some places in the southern tips of these peninsulas along the

707 Pacific Coast are possible candidate locations of the cryptic eastern

708 refugia in Japan.

709 Genetic boundaries between the Japan Sea and the Pacific

710 Ocean populations of various plant species of temperate zones in

711 Japan have been observed, for example, Cryptomeria (Tsumura et

712 al., 2007), Fagus (Fujii et al., 2002; Hiraoka and Tomaru, 2009), and

713 Euonymus (Iwasaki et al., 2012). Fossil pollen evidence of

714 Cryptomeria at the LGM also existed in the Japan Sea region K. Aoki et al. 35

715 (Kawamura, 1977). There were fossilized pollen records of

716 Castanea-Castanopsis (less than 1 % to several %) at the LGM in

717 the Japan Sea (Takahara and Takeoka, 1992; Miyoshi et al., 1999).

718 These information and the ABC-based demographic analyses of C.

719 sieboldii suggest that Castanopsis forests could have survived in the

720 small refugia along the Japan Sea coasts, though lack of SDM

721 support for the refugia of this region because of the SDM model is

722 too coarse to take into account topographic effects on microclimate.

723

724 Conclusion

725 The results of this study may help efforts to resolve the demographic

726 dynamics for species of warm temperate ecosystems during the

727 Quaternary period in Japan. Our ABC-based simulations of

728 demographics and species distribution modelling of the dominant

729 tree Castanopsis (Fagaceae), serving here as an indicator for the

730 behavior of species in broadleaved evergreen forests, indicate that

731 refugia were located at least one in the Ryukyu Islands and the other

732 three regions of the western and eastern parts and around the Japan

733 Sea of the main islands in Japan at the LGM. The results also

734 provide ABC-based indications of demographic expansions driven by

735 glacial oscillations. Our analyses provide foundations for further

736 ABC-based explorations of the demographic history of species in K. Aoki et al. 36

737 warm temperate broadleaved forests during and after the last glacial

738 period and a basic model for future phylogeographical studies using

739 this approach.

740

741 DATA ARCHIVING

742 Data available from the Dryad Digital Repository.

743

744

745 CONFLICT OF INTEREST

746 The authors declare no conflict of interest.

747

748 ACKNOWLEDGMENTS

749 The authors are grateful to Hiroshi Yoshimaru, Yoshiaki Tsuda,

750 Akihiro Seo, Kentaro Kimura, Hiroshi Takada, Sota Matsunaga,

751 Ryouichi Kusano, Tsai-Wen Hsu, and Ho-Ming Chang for collecting

752 the samples. We also thank Yuriko Taguchi and Yasuyuki Komatsu

753 for the laboratory work and Lerma San Jose Maldia for the valuable

754 comments on ABC analysis. This research was partly supported by a

755 grant for research on Genetic Guidelines for Restoration Programs

756 using Genetic Diversity Information from the Ministry of Environment,

757 Japan, Grants-in-Aid from the Japan Society for the Promotion of K. Aoki et al. 37

758 Science (nos. 1701416, 21770087, 2240030, 15J40011 and

759 18K06381 to K. A.), and the Research Project “A new cultural and

760 historical exploration into human-nature relationships in the

761 Japanese archipelago” of the Research Institute for Humanity and

762 Nature, Kyoto, Japan.

763

764 Supplementary Information accompanies this paper on Heredity

765 website

766 K. Aoki et al. 38

767 REFERENCES

768 Antao T, Lopes A, Lopes RJ, Beja-Pereira A, Luikart G (2008). LOSITAN:

769 a workbench to detect molecular adaptation based on a Fst-outlier method.

770 BMC Bioinformatics 9: 323.

771

772 Aoki K, Hattori T, Murakami N (2004a). Intraspecific sequence variation of

773 chloroplast DNA among the component species of evergreen broad-leaved

774 forests in Japan II. Acta Phytotax Geobot 55: 125-128.

775

776 Aoki K, Kato M, Murakami N (2005). Mitochondrial DNA of phytophagous

777 insects as a molecular tool for phylogeographic study of host plants. Acta

778 Phytotax Geobot 56: 55-69.

779

780 Aoki K, Kato M, Murakami N (2008). Glacial bottleneck and postglacial

781 recolonization of a seed parasitic weevil, Curculio hilgendorfi, inferred from

782 mitochondrial DNA variation. Mol Ecol 17: 3276-3289.

783

784 Aoki K, Kato M, Murakami N (2009). Phylogeographical patterns of a

785 generalist acorn weevil: insight into the biogeographical history of

786 broadleaved deciduous and evergreen forests. BMC Evol Biol 9: 103.

787 K. Aoki et al. 39

788 Aoki K, Kato M, Murakami N (2011). Review: Phylogeography of

789 phytophagous weevils and plant species in broadleaved evergreen forests: a

790 congruent genetic gap between western and eastern parts of Japan. Insects

791 2: 128-150.

792

793 Aoki K, Matsumura T, Hattori T, Murakami N (2006). Chloroplast DNA

794 phylogeography of Photinia glabra (Rosaceae) in Japan. Am J Bot 93:

795 1852-1858.

796

797 Aoki K, Murakami N, Kato M (2010). Phylogeography of a specialist leaf-

798 mining weevil, Rhynchaenus dorsoplanatus (Coleoptera: Curculionidae),

799 associated with Castanopsis species. Ann Entomol Soc Am 103: 379-388.

800

801 Aoki K, Suzuki T, Hsu T-W, Murakami N (2004b). Phylogeography of the

802 component species of broad-leaved evergreen forests in Japan, based on

803 chloroplast DNA. J Plant Res 117: 77-94.

804

805 Aoki K, Suzuki T, Murakami N (2003). Intraspecific sequence variation of

806 chloroplast DNA among the component species of evergreen broad-leaved

807 forests in Japan. J Plant Res 116: 337-344.

808 K. Aoki et al. 40

809 Aoki K, Ueno S, Kamijo T, Setoguchi H, Murakami N, Kato M et al (2014).

810 Genetic differentiation and genetic diversity of Castanopsis (Fagaceae), the

811 dominant tree species in Japanese broadleaved evergreen forests, revealed

812 by analysis of EST associated microsatellites. PLOS ONE 9: e87429.

813

814 Aoki K, Ueno S, Kamijo T, Setoguchi H, Murakami N, Kato M et al (2016).

815 Detecting east–west genetic differentiation in Castanopsis (Fagaceae) on the

816 main islands of Japan and north–south on the Ryukyu Islands, based on

817 chloroplast haplotypes. Plant Syst Evol 302: 1093-1107.

818

819 Avise JC (2000). Phylogeography: the history and formation of species.

820 Harvard University Press: Cambridge, Massachusetts.

821

822 Beaumont MA, Nichols RA (1996). Evaluating loci for use in the genetic

823 analysis of population structure. Proc R Soc Lond B Biol Sci 263: 1619-

824 1626.

825

826 Beaumont MA, Zhang WY, Balding DJ (2002). Approximate Bayesian

827 computation in population genetics. Genetics 162: 2025-2035.

828

829 Birks HJB, Willis KJ (2008). Alpines, trees, and refugia in Europe. Plant

830 Ecology & Diversity 1: 147-160. K. Aoki et al. 41

831

832 Blum M, Francois O (2010). Non-linear regression models for Approzimate

833 Bayesian Computation. Statistics and Computing: 63-73.

834

835 Budde KB, Gonzalez-Martinez SC, Hardy OJ, Heuertz M (2013). The

836 ancient tropical rainforest tree Symphonia globulifera L. f. (Clusiaceae) was

837 not restricted to postulated Pleistocene refugia in Atlantic Equatorial Africa.

838 Heredity 111: 66-76.

839

840 Chase MW, Soltis DE, Olmstead RG, Morgan D, Les DH, Mishler BD et al

841 (1993). Phylogenetics of seed plants: an analysis of nucleotide sequences

842 from the plastid gene rbcL. Ann Mo Bot Gard 80: 528-580.

843

844 Chen C, Durand E, Forbes F, François O (2007). Bayesian clustering

845 algorithms ascertaining spatial population structure: a new computer

846 program and a comparison study. Mol Ecol Notes 7: 747–756.

847

848 Chen C, Lu RS, Zhu SS, Tamaki I, Qiu YX (2017). Population structure and

849 historical demography of Dipteronia dyeriana (Sapindaceae), an extremely

850 narrow palaeoendemic plant from China: implications for conservation in a

851 biodiversity hot spot. Heredity 119: 95-106.

852 K. Aoki et al. 42

853 Cheng YP, Hwang SY, Chiou WL, Lin TP (2006). Allozyme variation of

854 populations of Castanopsis carlesii (Fagaceae) revealing the diversity

855 centres and areas of the greatest divergence in Taiwan. Ann Bot 98: 601-608.

856

857 Cheng YP, Hwang SY, Lin TP (2005). Potential refugia in Taiwan revealed

858 by the phylogeographical study of Castanopsis carlesii Hayata (Fagaceae).

859 Mol Ecol 14: 2075-2085.

860

861 Chybicki IJ (2014). INEST 2.0.

862 wwwukwedupl/pracownicy/strona/igor_chybicki/software_ukw/.

863

864 Cook SR, Gelman A, Rubin DB (2006). Validation of software for Bayesian

865 models using posterior quantiles. Journal of Computational and Graphical

866 Statistics 15: 675–692.

867

868 Csillery K, Francois O, Blum MGB (2012). abc: an R package for

869 approximate Bayesian computation (ABC). Methods in Ecology and

870 Evolution 3: 475-479.

871

872 Cubry P, Pujade-Renaud V, Garcia D, Espeout S, Le Guen V, Granet F et al

873 (2014). Development and characterization of a new set of 164 polymorphic K. Aoki et al. 43

874 EST-SSR markers for diversity and breeding studies in rubber tree (Hevea

875 brasiliensis Müll. Arg.). Plant Breeding 133: 419-426.

876

877 Darleyhill S, Johnson WC (1981). Acorn dispersal by the blue jay

878 (Cyanocitta cristata). Oecologia 50: 231-232.

879

880 de Lafontaine G, Amasifuen Guerra CA, Ducousso A, Petit RJ (2014).

881 Cryptic no more: soil macrofossils uncover Pleistocene forest microrefugia

882 within a periglacial desert. New Phytol 204: 715-729.

883

884 Dieringer D, Schlotterer C (2003). Microsatellite analyser (MSA): a

885 platform independent analysis tool for large microsatellite data sets. Mol

886 Ecol Notes 3: 167-169.

887

888 Doyle JJ, Doyle JL (1987). A rapid DNA isolation procedure for small

889 quantities of fresh leaf material. Phytochemistry 19: 11-15.

890

891 Durand E, Chen C, François O (2009). TESS version 2.3: free computer

892 program that implements a Bayesian clustering algorithm for spatial

893 population genetic studies.

894 K. Aoki et al. 44

895 El Mousadik A, Petit RJ (1996). High level of genetic differentiation for

896 allelic richness among populations of the argan tree [Argania spinosa (L.)

897 Skeels] endemic to Morocco. Theor Appl Genet 92: 832-839.

898

899 Estoup A, Angers B (1998). Microsatellites and minisatellites for molecular

900 ecology: theoretical and empirical considerations. In: Carvalho G (ed)

901 Advances in molecular ecology. NATO Press: Amsterdam, Holland, pp 55-

902 86.

903

904 Estoup A, Jarne P, Cornuet J-M (2002). Homoplasy and mutation model at

905 microsatellite loci and their consequences for population genetics analysis.

906 Mol Ecol 11: 1591-1604.

907

908 Evanno G, Regnaut S, Goudet J (2005). Detecting the number of clusters of

909 individuals using the software STRUCTURE: a simulation study. Mol Ecol

910 14: 2611-2620.

911

912 Excoffier L, Estoup A, Cornuet J-M (2005). Bayesian analysis of an

913 admixture model with mutations and arbitrarily linked markers. Genetics

914 169: 1727-1738.

915 K. Aoki et al. 45

916 Excoffier L, Foll M (2011). fastsimcoal: a continuous-time coalescent

917 simulator of genomic diversity under arbitrarily complex scenarios.

918 Bioinfomatics 27: 1332-1334.

919

920 Excoffier L, Lischer H (2010). Arlquin suite ver 3.5: a new series of

921 programs to perform population genetics analyses under Linux and

922 Windows. Molecular Ecology Resources 10: 564-567.

923

924 Felsenstein J (1989). PHYLIP - Phylogeny Inference Package (version 3.2).

925 Cladistics 5: 164-166.

926

927 François O, Ancelet S, Guillot G (2006). Bayesian clustering using hidden

928 Markov random fields in spatial population genetics. Genetics 174: 805–

929 816.

930

931 Fujii N, Tomaru N, Okuyama K, Koike T, Mikami T, Ueda K (2002).

932 Chloroplast DNA phylogeography of Fagus crenata (Fagaceae) in Japan.

933 Plant Syst Evol 232: 21-33.

934

935 Gao J, Wang B, Mao I-F, Ingvarsson P, Zeng Q-Y, Wang X-R (2012).

936 Demography and speciation history of the homoploid hybrid pine Pinus

937 densata on the Tibetan Plateau. Mol Ecol 21: 4811-4827. K. Aoki et al. 46

938

939 Gómez JM (2003). Spatial patterns in long-distance dispersal of Quercus

940 ilex acorns by jays in a heterogeneous landscape. Ecography 26: 573-584.

941

942 Goudet J (2002). FSTAT, a program to estimate and test gene diversities

943 and fixation indices (version 2.9.3.2)

944 http://www2.unil.ch/popgen/softwares/fstat.htm. University of Lausanne,

945 Lausanne, Switzerland.

946

947 Hattori T (1985). Synecological study on the lucidophyllous forest of

948 Castanopsis-Persea type in Japan proper. Bull Kobe Geobot Soc 1: 1-98.

949

950 Hattori T (2002). Chapter 7. In: Yahara T and Kawakubo N (eds) Biology of

951 conservation and restoration. Bun-ichi: Tokyo, pp 203-222.

952

953 Hewitt G (2000). The genetic legacy of the Quaternary ice ages. Nature

954 405: 907-913.

955

956 Hewitt GM (1999). Post-glacial re-colonization of European biota. Biol J

957 Linn Soc 68: 87-112.

958 K. Aoki et al. 47

959 Hewitt GM (2004). Genetic consequences of climatic oscillations in the

960 Quaternary. Philos Trans R Soc Lond, Ser B: Biol Sci 359: 183-195.

961

962 Hey J, Nielsen R (2004). Multilocus methods for estimating population

963 sizes, migration rates and divergence time, with applications to the

964 divergence of Drosophila pseudoobscura and D. persimilis. Genetics 167:

965 747-760.

966

967 Hickerson MJ, Carstens BC, Cavender-Bares J, Crandall KA, Graham CH,

968 Johnson JB et al (2010). Phylogeography's past, present, and future: 10

969 years after Avise, 2000. Mol Phylogen Evol 54: 291-301.

970

971 Hiraoka K, Tomaru N (2009). Population genetic structure of Fagus

972 japonica revealed by nuclear microsatellite markers. Int J Plant Sci 170:

973 748-758.

974

975 Huang SSF, Hwang SY, Lin TP (2002). Spatial pattern of chloroplast DNA

976 variation of Cyclobalanopsis glauca in Taiwan and east Asia. Mol Ecol 11:

977 2349-2358.

978

979 Iwaizumi MG, Tsuda Y, Ohtani M, Tsumura Y, Takahashi M (2013).

980 Recent distribution changes affect geographic clines in genetic diversity and K. Aoki et al. 48

981 structure of Pinus densiflora natural populations in Japan. For Ecol Manage

982 304: 407-416.

983

984 Iwasaki T, Aoki K, Seo A, Murakami N (2012). Comparative

985 phylogeography of four component species of deciduous broad-leaved

986 forests in Japan based on chloroplast DNA variation. J Plant Res 125: 207-

987 221.

988

989 Iwasaki T, Tono A, Aoki K, Seo A, Murakami N (2010). Phylogeography of

990 Carpinus japonica and Carpinus tschonoskii (Betulaceae) growing in

991 Japanese deciduous broad-leaved forests, based on chloroplast DNA

992 variation. Acta Phytotax Geobot 61: 1-20.

993

994 Jakobsson M, Rosenberg NA (2007). CLUMPP: a cluster matching and

995 permutation program for dealing with label switching and multimodality in

996 analysis of population structure. Bioinformatics 23: 1801–1806.

997

998 Jensen TS, Nielsen OF (1986). Rodents as seed dispersers in a heath - oak

999 wood succession. Oecologia 70: 214-221.

1000 K. Aoki et al. 49

1001 Kamei T, Research group for the biogeography from Würm Glacial (1981).

1002 Fauna and flora of the Japanese Islands in the last glacial. The Quaternary

1003 Res (Japan) 20: 191-205.

1004

1005 Kanauchi A (2005). Pollen analysis of the Jaishi oike Moordeposit, in the

1006 southern part of the Izu Peninsula. Sundai Shigaku (Sundai Historical

1007 Review) 125: 119-130.

1008

1009 Kawamoto Y, Shotake T, Nozawa K, Kawamoto S, Tomari K, Kawai S et

1010 al (2007). Postglacial population expansion of Japanese macaques (Macaca

1011 fuscata) inferred from mitochondrial DNA phylogeography. Primates 48:

1012 27-40.

1013

1014 Kawamura T (1977). Pollen analytical studies on the distribution of

1015 Cryptomeria japonica D. Don I. Akita Prefecture. Pollen Science 11: 8-20.

1016

1017 Knowles LL, Maddison WP (2002). Statistical phylogeography. Mol Ecol

1018 11: 2623-2635.

1019

1020 Kobayashi S, Hiroki S (2003). Patterns of occurrence of hybrids of

1021 Castanopsis cuspidata and C. sieboldii in the IBP Minamata Special K. Aoki et al. 50

1022 Research Area, Kumamoto Prefecture, Japan. J Phytogeogr Taxon 51: 63-

1023 67.

1024

1025 Kuroda T (1998). Chapter II-9. In: Yasuda Y and Miyoshi N (eds) The

1026 illustrated vegetation history of the Japanese Archipelago. Asakura-shoten:

1027 Tokyo, pp 162-175.

1028

1029 Lee JH, Lee DH, Choi IS, Choi BH (2014). Genetic diversity and historical

1030 migration patterns of an endemic evergreen oak, Quercus acuta, across

1031 Korea and Japan, inferred from nuclear microsatellites. Plant Syst Evol 300:

1032 1913-1923.

1033

1034 Lee JH, Lee DH, Choi BH (2013). Phylogeography and genetic diversity of

1035 East Asian Neolitsea sericea (Lauraceae) based on variations in chloroplast

1036 DNA sequences. J Plant Res 126: 193-202.

1037

1038 Li L, Abbott RJ, Liu BB, Sun YS, Li LL, Zou JB et al (2013). Pliocene

1039 intraspecific divergence and Plio-Pleistocene range expansions within Picea

1040 likiangensis (Lijiang spruce), a dominant forest tree of the Qinghai-Tibet

1041 Plateau. Mol Ecol 22: 5237-5255.

1042 K. Aoki et al. 51

1043 Liu B, Abbott RJ, Lu Z, Tian B, Liu J (2014a). Diploid hybrid origin of

1044 Ostryopsis intermedia (Betulaceae) in the Qinghai-Tibet Plateau triggered

1045 by Quaternary climate change. Mol Ecol 23: 3013-3027.

1046

1047 Liu C, Tsuda Y, Shen H, Hu L, Saito Y, Ide Y (2014b). Genetic structure

1048 and hierarchical population divergence history of Acer mono var. mono in

1049 south and northeast China. PLoS ONE 9.

1050

1051 Liu H-Z, Takeichi Y, Kamiya K, Harada K (2013). Phylogeography of

1052 Quercus phillyraeoides (Fagaceae) in Japan as revealed by chloroplast DNA

1053 variation. Journal of Forest Research 18: 361-370.

1054

1055 Louis M, Fontaine MC, Spitz J, Schlund E, Dabin W, Deaville R et al

1056 (2014). Ecological opportunities and specializations shaped genetic

1057 divergence in a highly mobile marine top predator. Proc R Soc Lond B Biol

1058 Sci 281: 20141558.

1059

1060 Maeda Y (1980). Jomon no umi to mori. Soju-shobo: Tokyo.

1061

1062 Martinson DG, Pisias NG, Hays JD, Imbrie J, Moore TC, Shackleton NJ, Jr.

1063 (1987). Age dating and the orbital theory of the Ice Ages: development of a

1064 high-resolution 0 to 300,000-year chronostratigraphy. Quatern Res 27: 1-29. K. Aoki et al. 52

1065

1066 Matsuoka K, Miyoshi N (1998). Chapter III-4. In: Yasuda Y and Miyoshi N

1067 (eds) The illustrated vegetation history of the Japanese Archipelago.

1068 Asakura-shoten: Tokyo, pp 224-236.

1069

1070 Matsushita M (1992). Lucidophyllous forest development along the Pacific

1071 Coast of the Japanese Archipelago during the Holocene. The Quaternary

1072 Res (Japan) 31: 375-387.

1073

1074 Minato M, Ijiri S (1976). The Japanese archipelago, 3rd edn.

1075 Iwanamishoten: Tokyo.

1076

1077 Miyaki M, Kikuzawa K (1988). Dispersal of Quercus mongolica acorns in a

1078 broadleaved deciduous forest 2. scatterhoarding by mice. For Ecol Manage

1079 25: 9-16.

1080

1081 Miyata M, Ubukata M (1994). Genetic variation of allozymes in natural

1082 stands of Japanese black pine. J Jpn For Soc 76: 445-455.

1083

1084 Miyoshi N, Fujiki T, Morita Y (1999). Palynology of a 250-m core from

1085 Lake Biwa: a 430,000-year record of glacial-interglacial vegetation change

1086 in Japan. Rev Palaeobot Palynol 104: 267-283. K. Aoki et al. 53

1087

1088 Murray MG, Thompson WF (1980). Rapid isolation of high molecular

1089 weight plant DNA. Nucleic Acids Res 8: 4321-4432.

1090

1091 Nagata J, Masuda R, Tamate HB, Hamasaki S, Ochiai K, Asada M et al

1092 (1999). Two genetically distinct lineages of the sika deer, Cervus nippon, in

1093 Japanese islands: comparison of mitochondrial D-loop region sequences.

1094 Mol Phylogen Evol 13: 511-519.

1095

1096 Nakanishi H (1996). Plant species with northbound distribution in western-

1097 Kyushu, Japan: definition, composition and origin. Acta Phytotax Geobot

1098 47: 113-124.

1099

1100 Nakao K, Higa M, Tsuyama I, Lin C-T, Sun S-T, Lin J-R et al (2014).

1101 Changes in the potential habitats of 10 dominant evergreen broad-leaved

1102 tree species in the Taiwan-Japan archipelago. Plant Ecol 215: 639-650.

1103

1104 Nei M (1978). Estimation of average heterozygosity and genetic distance

1105 from a small number of individuals. Genetics 89: 583-590.

1106

1107 Nei M, Tajima F, Tateno Y (1983). Accuracy of estimated phylogenetic

1108 trees from molecular data. II. Gene frequency data. J Mol Evol 19: 153-170. K. Aoki et al. 54

1109

1110 Nunome M, Torii H, Matsuki R, Kinoshita G, Suzuki H (2010). The

1111 Influence of Pleistocene refugia on the evolutionary history of the Japanese

1112 hare, Lepus brachyurus. Zool Sci 27: 746-754.

1113

1114 Okazaki H, Kobayashi M, Momohara A, Eguchi S, Okamoto T, Yanagisawa

1115 S et al (2011). Early Holocene coastal environment change inferred from

1116 deposits at Okinoshima archeological site, Boso Peninsula, central Japan.

1117 Quat Int 230: 87-94.

1118

1119 Oshida T, Masuda R, Ikeda K (2009). Phylogeography of the Japanese giant

1120 flying squirrel, Petaurista leucogenys (Rodentia: Sciuridae): implication of

1121 glacial refugia in an arboreal small mammal in the Japanese Islands. Biol J

1122 Linn Soc 98: 47-60.

1123

1124 Parducci L, Jorgensen T, Tollefsrud MM, Elverland E, Alm T, Fontana SL

1125 et al (2012). Glacial Survival of Boreal Trees in Northern Scandinavia.

1126 Science 335: 1083-1086.

1127

1128 Petit RJ, Hampe A (2006). Some evolutionary consequences of being a tree.

1129 Annual Review of Ecology, Evolution, and Systematics 37: 187-214.

1130 K. Aoki et al. 55

1131 Plummer M, Best N, Cowles K, Vines K (2006). CODA: Convergence

1132 Diagnosis and Output Analysis for MCMC. R News 6: 7-11.

1133

1134 Pons J, Pausas JG (2007). Acorn dispersal estimated by radio-tracking.

1135 Oecologia 153: 903-911.

1136

1137 Pritchard JK, Seielstad MT, Perez-Lezaun A, Feldman MW (1999).

1138 Population growth of human Y chromosomes: A study of Y chromosome

1139 microsatellites. Mol Biol Evol 16: 1791-1798.

1140

1141 Provan J, Bennett KD (2008a). Phylogeographic insights into cryptic glacial

1142 refugia. Trends Ecol Evol 23: 564-571.

1143

1144 Provan J, Bennett KD (2008b). Phylogeographic insights into cryptic glacial

1145 refugia. Trends Ecol Evol 23: 564-571.

1146

1147 Pudlo P, Marin J-M, Estoup A, Cornuet J-M, Gautier M, Robert CP (2016).

1148 Reliable ABC model choice via random forests. Bioinfomatics 32: 859–866.

1149

1150 Qiu YX, Fu CX, Comes HP (2011). Plant molecular phylogeography in

1151 China and adjacent regions: Tracing the genetic imprints of Quaternary K. Aoki et al. 56

1152 climate and environmental change in the world's most diverse temperate

1153 flora. Mol Phylogen Evol 59: 225-244.

1154

1155 Ren GP, Mateo RG, Liu JQ, Suchan T, Alvarez N, Guisan A et al (2017).

1156 Genetic consequences of Quaternary climatic oscillations in the Himalayas:

1157 Primula tibetica as a case study based on restriction site-associated DNA

1158 sequencing. New Phytol 213: 1500-1512.

1159

1160 Rull V (2009). Microrefugia. J Biogeogr 36: 481-484.

1161

1162 Saitou N, Nei M (1987). The neighbor-joining method; a new method for

1163 reconstructing phylogenetic trees. Mol Biol Evol 4: 406-425.

1164

1165 Setoguchi H, Ohba H (1995). Phylogenetic relationships in Crossostylis

1166 (Rhizophoraceae) inferred from restriction site variation of chloroplast DNA.

1167 J Plant Res 108: 87-92.

1168

1169 Shafer ABA, Cullingham CI, Cote SD, Coltman DW (2010). Of glaciers

1170 and refugia: a decade of study sheds new light on the phylogeography of

1171 northwestern North America. Mol Ecol 19: 4589-4621.

1172 K. Aoki et al. 57

1173 Soltis DE, Morris AB, McLachlan JS, Manos PS, Soltis PS (2006).

1174 Comparative phylogeography of unglaciated eastern North America. Mol

1175 Ecol 15: 4261-4293.

1176

1177 Stewart JR, Lister AM, Barnes I, Dalen L (2010). Refugia revisited:

1178 individualistic responses of species in space and time. Proc R Soc Lond B

1179 Biol Sci 277: 661-671.

1180

1181 Svenning JC, Normand S, Kageyama M (2008). Glacial refugia of

1182 temperate trees in Europe: insights from species distribution modelling. J

1183 Ecol 96: 1117-1127.

1184

1185 Taberlet P, Fumagalli L, Wust-Saucy A-G, Cosson J-F (1998). Comparative

1186 phylogeography and postglacial colonization routes in Europe. Mol Ecol 7:

1187 453-464.

1188

1189 Takahara H, Takeoka M (1992). Vegetation history since the last glacial

1190 period in the Mikata lowland, the Sea of Japan area, western Japan. Ecol

1191 Res 7: 371-386.

1192

1193 Tsuda Y, Kimura M, Kato S, Katsuki T, Mukai Y, Tsumura Y (2009).

1194 Genetic structure of Cerasus jamasakura, a Japanese flowering cherry, K. Aoki et al. 58

1195 revealed by nuclear SSRs: implications for conservation. J Plant Res 122:

1196 367-375.

1197

1198 Tsukada M (1974). Paleoecology. II. Synthesis. Kyoritsu: Tokyo.

1199

1200 Tsukada M (1983). Vegetation and climate during the last glacial maximum

1201 in Japan. Quatern Res 19: 212-235.

1202

1203 Tsumura Y, Kado T, Takahashi T, Tani N, Ujino-Ihara T, Iwata H (2007).

1204 Genome scan to detect genetic structure and adaptive genes of natural

1205 populations of Cryptomeria japonica. Genetics 176: 2393-2403.

1206

1207 Tzedakis PC, Emerson BC, Hewitt GM (2013). Cryptic or mystic? Glacial

1208 tree refugia in northern Europe. Trends Ecol Evol 28: 696-704.

1209

1210 Ueno S (2015). Camellia japonica. In: Tsumura Y and Suyama Y (eds)

1211 Chizu de wakaru jumoku no shubyo ido gaidorain. Bunichi: Tokyo.

1212

1213 van Valen L (1975). Life, Death, and Energy of a Tree. Biotropica 7: 259-

1214 269.

1215 K. Aoki et al. 59

1216 Wakeley J (2008). Coalescent theory. Roberts and Company Publishers:

1217 Colorado, U.S.

1218

1219 Wang Q, Liu JQ, Allen GA, Ma YZ, Yue W, Marr KL et al (2016). Arctic

1220 plant origins and early formation of circumarctic distributions: a case study

1221 of the mountain sorrel, Oxyria digyna. New Phytol 209: 343-353.

1222

1223 Wegmann D, Leuenberger C, Neuenschwander S, Excoffier L (2010).

1224 ABCtoolbox: a versatile toolkit for approximate Bayesian computations.

1225 BMC Bioinformatics 11: 116.

1226

1227 Weir BS, Cockerham CC (1984). Estimating F-statistics for the analysis of

1228 population structure. Evolution 38: 1358-1370.

1229

1230 Wolfe KH, Li W-H, Sharp PM (1987). Rates of nucleotide substitution vary

1231 greatly among plant mitochondrial, chloroplast, and nuclear DNAs. Biol J

1232 Linn Soc 84: 9054-9058.

1233

1234 Wright S (1931). Evolution in Mendelian populations. Genetics 16: 97-159.

1235

1236 Wu SH, Hwang CY, Lin TP, Chung JD, Cheng YP, Hwang SY (2006).

1237 Contrasting phylogeographical patterns of two closely related species, K. Aoki et al. 60

1238 Machilus thunbergii and Machilus kusanoi (Lauraceae), in Taiwan. J

1239 Biogeogr 33: 936-947.

1240

1241 Xu J, Deng M, Jiang X-L, Westwood M, Song Y-G, Turkington R (2015).

1242 Phylogeography of Quercus glauca (Fagaceae), a dominant tree of East

1243 Asian subtropical evergreen forests, based on three chloroplast DNA

1244 interspace sequences. Tree Genet Genom 11.

1245

1246 Yamazaki T, Mashiba S (1987a). A taxonomical revision of Castanopsis

1247 cupidata (Thunb.) Schottky and the allies in Japan, Korea and Taiwan. 1. J

1248 Jap Bot 62: 289-298.

1249

1250 Yamazaki T, Mashiba S (1987b). A taxonomical revision of Castanopsis

1251 cupidata (Thunb.) Schottky and the allies in Japan, Korea and Taiwan. 2. J

1252 Jap Bot 62: 332-339.

1253

1254 Yasukochi Y, Nishida S, Han SH, Kurosaki T, Yoneda M, Koike H (2009).

1255 Genetic structure of the asiatic black bear in Japan using mitochondrial

1256 DNA analysis. J Hered 100: 297-308.

1257

1258 Yoichi W, Tamaki I, Sakaguchi S, Song JS, Yamamoto S, Tomaru N (2016).

1259 Population demographic history of a temperate shrub, Rhododendron K. Aoki et al. 61

1260 weyrichii (Ericaceae), on continental islands of Japan and South Korea.

1261 Ecology and Evolution 6: 8800-8810.

1262

1263 Yoshida T, Nagai H, Yahara T, Tachida H (2010). Genetic structure and

1264 putative selective sweep in the pioneer tree, Zanthoxylum ailanthoides. J

1265 Plant Res 123: 607-616.

1266

1267 Yoshimaru H, Matsumoto A (2015). Acer palmatum. In: Tsumura Y and

1268 Suyama Y (eds) Chizu de wakaru jumoku no shubyo ido gaidorain. Bunichi:

1269 Tokyo.

1270

1271

1272 K. Aoki et al. 62

1273 Table 1 Fraction of votes, classification error rate and posterior probability

1274 (PP) of the best model, which were estimated by random forest composed of

1275 1,000 trees based on a trained set of 10,000 simulations. Best model was

1276 shown in bold

Fraction of votes Classification PP of the Model 1 Model 2 Model 3 Model 4 error rate best model 0.634 0.212 0.152 0.002 0.116 0.906 1277

1278 K. Aoki et al. 63

1279 Table 2 Prior and posterior distributions of parameters in the best model

1280 (model 1)

㻌㻌 Posterior 95% HPD a Parameter Prior Mode Lower Upper 4 NCUR Unif (1, 10 ) 5463 2349 9239 4 NANC Unif (1, 10 ) 1517 13 8809 G (×10-3) Unif (-0.01, 0) -1.89 -4.65 -0.39 4 b T1 Unif (1, 10 ) 431 95 2453 4 b T2 Unif (1, 10 ) 1184 110 8311 Nm Unif (0.1, 20) 9.70 6.08 17.82 shape Unif (0.5, 2) 0.88 0.50 1.67

mean PGSM Unif (0, 0.8) 0.217 0.077 0.352 1281 a 95% highest posterior density.

b 1282 T1 < T2 was assumed.

1283 NCUR and NANC, current and ancestral effective population sizes,

1284 respectively; G, growth rate; T, the time when the demographic events

1285 occurred; Nm, the number of migrants per generation; shape and mean PGSM,

1286 parameters for the mutation model of microsatellite.

1287 Units of the effective population size and time parameters are the numbers

1288 of diploid individuals and generations, respectively.

1289

1290 K. Aoki et al. 64

1291 Titles and legends to figures

1292

1293 Figure 1 Locations of the 33 sampled Castanopsis sieboldii populations.

1294 Numbers correspond to the population numbers in Table S1. The dotted line

1295 indicates the coastline during the LGM, about 18,000–24,000 years ago.

1296

1297 Figure 2 Neighbor-joining tree based on Nei’s genetic distances (DA) among

1298 the 33 Castanopsis sieboldii populations over 27 EST-SSR loci. Numbers after

1299 the outgroup species, C. cuspidata, correspond to the population numbers in the

1300 study of Aoki et al. (2014). Values in italics are percentages of 1,000 bootstrap

1301 replicates supporting the respective nodes.

1302

1303 Figure 3 Genetic relationships among the 33 Castanopsis sieboldii

1304 populations estimated using the Bayesian clustering method, TESS (Durand

1305 et al., 2009).

1306

1307 Figure 4 The compared four candidate models for the population history of

1308 Castanopsis sieboldii. R, W, J and E correspond to the Ryukyu Islands, the

1309 West, the Japan Sea and the East groups, respectively, in Figs. 2 and 3.

1310

1311 Figure 5 Potential habitats of Castanopsis sieboldii estimated by species

1312 distribution models at the present and the Last Glacial Maximum (LGM). The K. Aoki et al. 65

1313 potential habitats under LGM were shown by the average of the three Global

1314 Climate Models (i.e. MIROC, CCSM, MPI-ESM) (Figure S6). 130° 135° 40° Honshu Japan Sea

1

Hokuriku 2 Kanto 17 16 15 8 7 4 3 18 Chugoku 9 35° Kinki-Tokai 5 10 6 21 Seto InlandShikoku Sea 11 Kyushu 13 Boso Pen. 14 Izu Pen. 19 22 12 20 Kii Pen.

23 24

25 Pacific Ocean 26 27 30°

28 Main Islands 30 29

31

33 Ryukyu Islands 32 25° C. cuspidata_57 _53 C. cuspidata_59 C. cuspidata_51 C. cuspidata_54 C. cuspidata

59 C. cuspidata_48 57 C. cuspidata_47 Ryukyu Islands C. cuspidata_50

33 60 C. cuspidata_63 32 30 99 31 28 29 C. cuspidata_55 West 26 98 24 100 27 53 25 93 23 74 19 13 61 15 14 60 17 11 73 16 18 3 12 21 7 20 5 4 10 22 C. sieboldii 6 8 9 2 East Japan Sea 1

0.03 100% K=2 80%

60%

40%

20%

0% 100% K=3 80%

60%

40%

20%

0% 100%

80% K=4

60%

40%

20%

0% 33 323130 29 28 2726 25 24 23 20 19 14 13 12 11 109 8 76543 2221181716152 1 Ryukyu Islands West East Japan Sea




    


 


   
 




  
 



 


 







   



 
 



   
 


! 

 
 
  

   



  


 
"!#$
  

!

 

%&'''

 
 


  


%'&'''
  
  
  

"
   #$

 
 

!


 

 $
! 

 
  
 
 




 $


(  

   



 
$
%&

&
 &

)



&

*  &

 &


+
 &
  $
,

 

 

 
"
 


 #

 
  



 &
  $


  
 
  


 
 
" 
%#$

-  

 

 

%&'''
  

  

 &
  $
,


 
  $
.&
-&




 


 


  &

/ 
0 &


 
 0
 &

  $
1 
%2)
 


&

*  &





+
 
&
  &

! $


$


  
3  


 $
, 
 
  

/ 



 
    $
 

4  



 5


 


 $
$$
  

    $

* 

 

  
5 
 &

    

  

3 


/  


 
    $


  

 
    

 


 
 

" 
%6
 

 
  &
  #$


  
    


 
 0
 


 
"78#&
%
"%#


".17#$


  
  


   
  

 

    
 
 
9:;
 
 
 

9:;
 

 
"#
;(7&
"#
77;

" #
;;$



  
  


9:;
  


  

 
  

+
 


.<7&
,;


  0 
 

 

 


  
 
 $
. 


 

 5


9:;
  
  

,;
 $
< 

5
  

 

/  


  0 
 
 


 

  
 

 
" $ $&

&

 #


9:;


    
  
 
  
" $ $&

 
 $

   #&
  
"4 &
''=6
; 

  &
%>>>6
 

 &
%>>6
?

;  &
%>>@#$

(


 


  
 

  0 
 
 


 
 3  

 

%'
A

 

%
A

 
A&

  $

 

4
.
"''=#$
 
 


 
;  &


 



0
  $
     

  !
 
%%>% '$

; 
1&
!B 
&
; 
?
"%>>>#$



='



9
< 5C

) '&''' 
 

    
  



+$
      

D@ $

 

 
-&
 
;
"%>>#$
E   
 
 


 

  


;  
5&

 

+
 &
5  
+$
"

 

% @D$

?
?&
; 
1
"%>>@#$
      # 

  $ $
.  C
&
+$
 
 


 


   
 




  
 



 

1$


.      1$
 /  . 

% 1$    9  9  "#      0     %

%  ., % @$=
$>>
)=$'  $')
'$='
$=>
%$'''

  ?; % $''=
D$@>=
%%$ ) )$>D
'$)>'
)$'D%
%$'''

 :17 %)'$ D>
=$ =%
==$% ) =$%%%
'$=D)
)$@%%
%$'''

) 4   7-< %)'$% D
=$%=D
D$' ) =$==D
'$=>'
=$% )
%$'''

= 45 4? %)'$%=%
=$%D'
%$  =$@=
'$D'
=$>
'$)=

D  1 % >$=>
)$>D@
)$> ) )$@=
'$= )
)$=%
%$'''

 -  0 +1; % >$D'D
=$ '=
@@$' %> )$>D
'$=@
)$@D
%$'''

@ -  ;: % >$)
=$D=
%%$'  )$')
'$==
$>@>
%$'''

> 0 F % >$'@
=$%%
@$) ) )$)@%
'$=>
)$ '=
%$'''

%' ;  0 ( % @$@)
)$D')
=D$' % =$' 
'$=@
=$' 
%$'''

%% 40  4(, % >$%)
)$%>
='$ ' )$
'$='
)$'>@
%$'''

% - B -7+ % >$@
$%'D
)>$@ ) )$@
'$= D
)$) '
'$' G

% 8 84 % =$>@@
$DD
$' ) =$
'$=D
)$@ D
%$'''

%)  8 % =$=>'
$='
%)$D @ )$@%=
'$=@D
)$=D'
%$'''

%= ; 0 ;F % =$)'
=$)
$= ) =$%%%
'$=>'
)$@ =
%$'''

%D 4  (1 % )$@''
=$D=
%>$' ) =$=>
'$=@
=$%@
%$'''

% ;  ;1 % $''
=$='>
$' ) =$ '
'$=>%
=$'
%$'''

%@   ,, % $%'D
)$>@
D=$' ) =$%@=
'$=@
)$@DD
%$'''

%> ;  ; % )$%D
$)@
%'$>  =$DD
'$=>
=$''=
%$'''

' . 0 .F % $'''
$)
$' = =$>D
'$=D>
)$@DD
'$'' G

%   ; %>$'
)$%=@
''$@ ) )$>D
'$=)'
)$D =
'$' )G

 4   * % '$%=D
$D=
DD$' ) =$)@%
'$=D>
=$' >
%$'''

 1 1(; % '$%D'
%$)''
=$' = =$==D
'$=@D
=$'
%$'''

) -  -4 % '$@)D
%$%%=
= $ ) =$==D
'$='
=$%)
'$'=%

=    .1 % %$')>
'$')
%%$' ) D$')
'$D')
=$DD
'$ %

D ?  ?48 % '$)
'$@
%$' D =$DD
'$D''
=$%'
%$'''

 1  4 %>$@@
>$@)D
D$' ) D$'''
'$==
=$%>
'$>

@ . .;; %>$ %
@$>
D@)$= ) D$' 
'$D'
=$==@
'$ ='

 
 

>   14 %@$>@=
$@
D'=$D ' =$@%=
'$=>
=$=D
%$'''

' 4   (4 %@$>)
$D@
%%D$  )$@=
'$=D>
)$D'
%$'''

% ( 5 ?( %@$%@
D$%@
)@$D ) =$@@>
'$D>
=$='
%$'''

   (; %)$%@D
)$)%@
$ % =$=>
'$D'
=$ 
%$'''

   % $@'D
)$ >
%$)  =$@=
'$D
=$)=
'$'@%

%

   

%
 
 
 3  
< 
 


  

  
"G
H
'$'=#$

 


 







   



 
 



   
 

% 
) = D 1$
9  .
( 
%
   
% .+=D=
'$=%%
'$=%=
%$'''
  ,1>='))D )
'$))=
'$)=@
%$'''
 ,1>='D ) 
'$ @
'$ >
%$'''
 ) ,1>='@@ 
'$'D
'$ 
'$=')
 = ! 7'%@ )
'$==
'$=@'
%$'''
 D ! 7' '>= %'
'$@
'$@)@
'$'@G   ! 7' =@= %
'$'=
'$'>
%$'''
 @ I7''@@
'$=''
'$=%@
%$'''
 > I7'')%> 
'$D'
'$D@
%$'''
E %' I7''@>@
'$) @
'$=%)
'$'@G E %% I7''>D
'$)D)
'$)>>
'$D)
E % I7'%=@ =
'$DD
'$DD)
%$'''
E % I7''= %
'$@@
'$@%
%$'''
E %) I7'D> )
'$==D
'$D'
'$'
E %= 7 7'''@ 
'$)D
'$=D
%$'''
E %D 7 7''D%' >
'$@ =
'$@ 
%$'''
E % 7 7''DD'
'$)%
'$ >=
%$'''
E %@ 7 7''> =
'$=>)
'$=>
%$'''
E %> 7 7'%)% @
'$
'$=
%$'''
E ' 7 7'%=% 
'$%)%
'$% %
%$'''
E % 7 7''%) 
'$=
'$'
'$'='
E  7 7'' 
'$@%
'$D@
%$'''
E  7 7''D> =
'$
'$%@
%$'''
E ) 7 7''= =
'$==
'$= )
%$'''
E = 7 7'>%
'$%>>
'$%>'
%$'''
E D 7 7' = )
'$ %
'$ D=
%$'''
E  7 7'= = 
'$)'
'$@'
'$'
E

 

%
;  /
7
  
  
5 

5C

&
%%&
% &
%D&
%&


'
"
%#6

%'&
%=&
%@&
 &
)&
>&

'
"
#6

&
=&
D&
&

%

"
#6

%&
>&
%&
%)&
=&

D
"
)#6


)&
@&
%>&
&
&

@
"

=#$

 )
.&
 
 

  6

(&
  
0 6

&
 
 

  
5 
 $

=

   

%
 
 
 3  
< 
 


  

  
"G
H
'$'=#$

D

"8 

&
''@#6

"8 
  &
''>#6

".
  &
'%)#6

E
"8 
  &
''@#6
E
"8 
  &
''>#$




! 

 
 
  

   



  


 
"!#$
  

!

 

%&'''

 
 


  


%'&'''
  
  
  

"
   #$

 
 

!


 

 $
! 

 
  
 
 




 $

     
  7      ; 
% ; 
 ; 
; 
) 
 ; 
%  '$' '$'%D '$'%= '$%' ; 
 '$'D>  '$%D '$''D '$ @ ; 
'$'%= '$'>  '$'') '$%% ; 
) '$''= '$'') '$''%  '$'%'