bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Original article

2 Geographic and subsequent biotic isolations led to a diversity anomaly of Heterotropa

3 () in insular versus continental regions of the Sino-Japanese Floristic

4 Region

5

6

7 Author affiliations

8 Daiki Takahashi1*, Yu Feng2, Shota Sakaguchi1, Yuji Isagi3, Ying-Xiong Qiu2, Pan Li2, Rui-Sen

9 Lu2, Chang-Tse Lu4, Shih-Wen Chung5, Yang-Shan Lin6, Yun-Chao Chen6, Atsushi J. Nagano7,

10 Lina Kawaguchi7, Hiroaki Setoguchi1

11 1Graduate school of Human and Environmental studies, Kyoto University, Japan; 2Systematic &

12 Evolutionary Botany and Biodiversity Group, MOE Laboratory of Biosystem Homeostasis and

13 Protection, College of Life Sciences, Zhejiang University, China; 3Graduate school of

14 Agriculture, Kyoto University, Japan; 4Department of Biological Resources, National Chiayi

15 University, Taiwan; 5Herbarium of Taiwan Forestry Research Institute, Taiwan; 6Miaoli Distinct

16 Agricultural Research and Extension station, Taiwan; 7Faculty of Agriculture, Ryukoku

17 University, Japan

18 *Corresponding author: [email protected]

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

19

20 Abstract

21 The Sino-Japanese Floristic Region is highly diverse with respect to temperate .

22 However, the reasons for this diversity are poorly understood because most studies have

23 only considered geographic isolation caused by climatic oscillations. Heterotropa

24 ( ; Aristolochiaceae) diverges here and shows high species diversity in

25 insular systems (63 species) compared to continental areas (25 species). Heterotropa

26 shows low dispersal ability with small distribution ranges, implying diversification by

27 geographic events, and high floral diversity, implying pollinator-mediated

28 diversification. To reveal how abiotic and biotic factors have shaped the diversity

29 anomaly of Heterotropa, we conducted phylogenetic analysis using ddRAD-seq and

30 chloroplast genome datasets including 79 species, estimation of floral trait evolution,

31 and comparison of isolation factors within clades based on distribution range and floral

32 trait analysis. Phylogenetic analysis indicates that Heterotropa originated in mainland

33 China and expanded to the Japanese Archipelago in the Miocene, and the major clades

34 almost correspond to geographic distributions. Floral traits evolved repeatedly in the tip

35 nodes within the clades. Although the major clades include a high proportion of species

36 pairs showing isolation by floral traits, there are no conditional relationships between

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

37 two isolation factors, indicating that most species pairs with floral trait isolation are

38 distributed allopatrically. The repeated exposure and submergence of land-bridges

39 caused by climatic oscillations would have led to significant population fragmentation

40 in insular systems. Thus, the diversity anomaly of Heterotropa would have resulted

41 from geographic and climatic events during the Miocene, while subsequent repeated

42 floral trait evolution would have followed geographic isolation during the Pleistocene.

43

44 Keywords

45 Sino-Japanese Floristic Region, biotic and abiotic factors, ddRAD-seq, Asarum,

46 Heterotropa

47

48

49

50

51

52

53

54

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

55

56 Introduction

57 The Sino-Japanese Floristic Region (SJFR) extends from the eastern Himalayas to the

58 Japanese Archipelago through south and central China (Takhtajan, 1969; Wu & Wu,

59 1996). This region can boast one of the most diverse temperate floras anywhere in the

60 world and has high endemism (Wu & Wu, 1996). This diversity has been thought to be

61 linked to climatic and physiographical complexity and historical environmental changes

62 associated with the Pleistocene (< 2.6 Mya) climatic oscillations (Qian & Ricklefs,

63 2000). During glacial periods, when the climate of this region was cooler by ca. 4-6 °C

64 and sea level was approximately 130m lower than its present level, temperate plants in

65 this region retreated to refugia at lower altitudes or southern parts (Harrison, Yu,

66 Takahara, & Prentice, 2001; Tsukada, 1984). During the interglacial periods, expansion

67 to higher altitudes or northern parts would have occurred. In the eastern island systems,

68 sea level changes due to climatic oscillations have caused repeated formation and

69 division of land-bridges in the East China Sea (Ujiie, 1990), and these events have

70 provided opportunities for population expansion and fragmentation (Qiu, Fu, & Comes,

71 2011). Many phylogeographic studies have revealed that the present interspecific and

72 intraspecific genetic structures of temperate plants in this region reflected the range

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

73 shifts caused by climatic oscillations (Li, Yan, & Ge, 2012; Sakaguchi et al., 2012;

74 Setoguchi et al., 2006; Wang et al., 2015; Yang et al., 2017). It has been considered that

75 these climatic and associated environmental changes during the Pleistocene triggered

76 range fragmentation, vicariance, and population isolation (Qiu et al., 2011). It has also

77 been hypothesised that allopatric speciation would be a major mode of speciation in the

78 temperate plants of this region (Qian & Ricklefs, 2000).

79 Although the importance of geographic isolation as a major isolation

80 mechanism in plants has been addressed (Boucher, Zimmermann, & Conti, 2016;

81 Esselstyn, Timm, & Brown, 2009; Govindarajulu, Hughes, & Bailey, 2011; Verboom,

82 Bergh, Haiden, Hoffmann, & Britton, 2015), recent studies in other regions have

83 implied that biotic factors also promote species diversification (Lagomarsino,

84 Condamine, Antonelli, Mulch, & Davis, 2016; L. Lu et al., 2019). In particular, floral

85 trait evolution has been thought to promote speciation through segregation of gene flow

86 by pollinator shifts (Armbruster, 2014; Van der Niet, Peakall, & Johnson, 2014).

87 Previous studies have shown that the tempo and pattern of floral trait evolution varies

88 distinctively among lineages, and floral trait evolution has been implicated in shaping

89 patterns of species diversification (Givnish et al., 2015; Jaramillo & Manos, 2001). It

90 has been considered that biotic drivers can play complementary roles to abiotic factors

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

91 in reproductive isolation; that is, biotic factors can facilitate reproductive isolation even

92 without geographical isolation (Rundle & Nosil, 2005). To fully understand the

93 diversification process of species groups, it is essential to reveal the relative

94 contributions of biotic and abiotic drivers. However, many studies conducted in the

95 SJFR have only discussed the role of allopatric fragmentation due to geographic and

96 climatic events, and few studies have considered other factors as drivers of the

97 diversification of the temperate plants. In addition, although the high endemism in the

98 SJFR provides an outstanding opportunity for investigating the evolutionary history of

99 diversification (Yang et al., 2017), to date, most phylogeographic studies in the

100 region have focused on individual species or only small groups, including fewer than 10

101 taxa (But Mitsui et al., 2008; Yoichi, Jin, Peng, Tamaki, & Tomaru, 2017). Thus, our

102 knowledge of the diversification process of temperate plants in the SJFR remains

103 fragmentary, due to a lack of integrative multidimensional studies of morphology,

104 phylogeny, biogeography, and ecology with adequate sampling of diversified groups.

105 In this study, we focused on the Heterotropa (genus Asarum; Aristolochiaceae)

106 as a model group endemic to the SJFR, and one of the most speciose warm-temperate

107 plant groups (comprising approximately 90 species) in this region (Sugawara, 2006).

108 Taxa of Heterotropa are low-growing, rhizomatous herbs that grow in shaded

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

109 understories, and are distributed in mainland China (25 species: Huang, Kelly, & Gilbert,

110 2003), Taiwan (13 species: C. T. Lu & Wang, 2009; Chang Tse Lu & Wang, 2014), and

111 the Japanese archipelago, including the Ryukyu islands (52 species: Sugawara, 2006).

112 The species diversity of Heterotropa is uneven, and given the difference in areas,

113 Heterotropa shows higher diversity in the eastern insular region (from Taiwan to

114 mainland Japan; 2.7 × 10-4 species/km2 and mainland China; 1.3 × 10-5 species/km2, see

115 Results for details). Some taxa of Heterotropa have very limited geographic ranges (e.g.,

116 in only one island or mountain range), and the dispersal ability of Heterotropa is

117 estimated to be 1050 cm per year due to its myrmecochore seeds with elaiosome

118 (Hiura, 1978; F. Maekawa, 1953). Low dispersal ability promotes genetic differentiation

119 among populations and often leads to allopatric speciation (Petit et al., 2005). These

120 confined distribution ranges and the low dispersal ability led us to hypothesise an

121 allopatric speciation process for Heterotropa. On the other hand, Heterotropa taxa are

122 characterised by high divergence in floral traits, in terms of their shapes, sizes, and

123 colours of calyx tubes and lobes (Fig. 1a & S1), while their vegetative traits show

124 almost no differences (Sugawara & Ogisu, 1992). The sepals connect beyond

125 attachment to the ovary and form a calyx tube with calyx lobes (Sugawara, 1987), and

126 their flowers have been hypothesised to mimic fungi in order to attract fungus gnats

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

127 (Sinn, Kelly, & Freudenstein, 2015; Vogel, 1978). In addition to flower shape,

128 Heterotropa taxa are highly divergent in flowering time; most taxa have flowers in

129 spring, while others have flowers in autumn or winter (Huang et al., 2003; Sugawara,

130 2006). A genus-wide phylogenetic study of Asarum showed that diversification of

131 Heterotropa could have been triggered by the presence of putative fungal-mimicking

132 floral structures, loss of autonomous selfing, and loss of vegetative growth (Sinn et al.,

133 2015). Given these characteristics, we considered that Heterotropa would be an ideal

134 subject for investigating the relative importance of the abiotic and biotic effects on its

135 diversification in the SJFR. Our previous phylogenetic study using the ITS region

136 showed that Heterotropa was monophyletic and comprised two clades, which

137 corresponded to species distribution ranges, namely mainland China and the island arc

138 from Taiwan to mainland Japan, and the estimated time of divergence between these

139 two clades was approximately 9 million years ago (Mya) in the late Miocene (Takahashi

140 & Setoguchi, 2018). In order to obtain the taxonomic implications of insular

141 Heterotropa, Okuyama et al., (in press) conducted phylogenetic analysis using RAD-seq

142 datasets including 47 insular and 5 continental species, and they implied that insular

143 Heterotropa comprises nine groups that almost correspond to the geographic

144 distributions. However, due to the low resolution of the datasets or lack of inclusive

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

145 sampling around the SJFR, the formation mechanisms of the diversity anomaly of

146 Heterotropa in insular systems and their diversification history in terms of temporal and

147 spatial patterns of floral trait evolution (e.g., parallel evolution in each region or single

148 origin with subsequent expansion) remain unknown.

149 Our primary purpose in this study was to reveal the diversification history of

150 Heterotropa taxa in the SJFR, especially focusing on the relative contribution of abiotic

151 and biotic drivers and diversity anomalies in insular systems. For this purpose, we first

152 constructed a highly resolved and time-calibrated phylogenetic tree including most

153 Heterotropa taxa using genome-wide single nucleotide polymorphisms (SNPs) and

154 chloroplast genomes. We also estimated the patterns of floral trait evolution using

155 macroevolutionary modelling and indices of phylogenetic signals based on the obtained

156 phylogenetic tree. In addition, to test whether the geographic isolations would be major

157 forces for speciation or whether biotic factors were more likely to facilitate the

158 speciation in Heterotropa, we compared patterns of geographic and morphological

159 isolation within the clades and between sister taxa pairs and non-sister taxa pairs. The

160 implications for the relative roles of biotic and abiotic drivers and the temporal and

161 spatial patterns of their influences would help to understand the diversification process

162 of temperate plants in the SJFR.

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

163

164

165

166

167 Materials and methods

168 Taxon sampling

169 Here, we present a comprehensive sample collection of sect. Heterotropa, increasing the

170 number of species from 64 species (Takahashi & Setoguchi, 2018) or 52 species

171 (Okuyama et al., in press) to 79 species (Fig 1b, Table S1), including three undescribed

172 but morphologically distinct taxa (Asarum kiusianum var. tubulosum nom. nud.

173 [Maekawa, 1983], Asarum tarokoense nom. nud. [Lu et al., in prep.], and Asarum

174 titaense nom. nud. [Setoguchi et al., in prep]) in our analysis. Asarum satsumense has

175 been considered to be distributed in both Taiwan and Kyushu islands (C.-T. Lu, Chiou,

176 Liu, & Wang, 2010), and our preliminary examination suggested that the Taiwanese

177 entity should be distinguished from the species (Lu and Takahashi, personal

178 observation). Thus, in this study, we treated A. satsumense collected from the Taiwan

179 and Kyushu islands as a different taxon (A. satsumense W; Taiwanese and A.

180 satsumense K; Japanese). The sample set included 46 Japanese species (52 taxa), 13

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

181 Taiwanese species (13 taxa), and 20 mainland Chinese species (20 taxa). As an outgroup

182 species, we used one sect. Hexastylis species (Asarum shuttleworthii) according to our

183 previous study (Takahashi & Setoguchi, 2018).

184

185 Chloroplast genome sequencing and phylogenetic analysis

186 For chloroplast genome construction, we conducted genome sequences of five Asarum

187 species, including A. shuttleworthii (Tables S2 & S3). Details of library preparation,

188 sequencing methods, and phylogenetic analysis are described in Appendix 1.

189

190 Double-digest restriction-associated DNA sequencing

191 Genomic DNA was extracted from silica-dried leaf tissues using the CTAB method

192 (Doyle & Doyle, 1987). For all collected samples, a double-digest restriction-associated

193 DNA (ddRAD) library was prepared using Peterson’s protocol with slight modifications

194 (Peterson, Weber, Kay, Fisher, & Hoekstra, 2012). Genomic DNA was digested with

195 BglII and EcoRI, ligated with Y-shaped adaptors, amplified by PCR with KAPA HiFi

196 HS ReadyMix (KAPA BIOSYSTEMS) and size-selected with the E-Gel size select

197 (Life Technologies, CA, USA). Approximately 350 bp of library fragments were

198 retrieved. Further details of the library preparation method were described in a previous

11 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

199 study (Sakaguchi et al., 2015). Sequencing was performed with paired-read 101bp +

200 100bp mode of HiSeq2500 (Illumina, CA, USA).

201

202 Data analysis

203 The raw reads obtained from ddRAD-seq were trimmed by Trimomomatic v. 0.32

204 software (Bolger, Lohse, & Usadel, 2014) with the following settings: HEADCRAP:10,

205 LEADING:30, TRAILING:30, SLIDINGWINDOW:4:30, AVGQUAL:30, and

206 MINLEN:50. The program ipyrad (http://github.com/dereneaton/ipyrad) was used to

207 process the ddRAD-seq reads and detect SNPs. The parameters that influenced the

208 assembly were set as follows: the minimum depth coverage for base calling at each

209 locus was set at 6 and the similarity threshold for clustering reads within/across samples

210 was set at 0.85. Potential paralogous loci were filtered out based on the number of

211 samples with shared heterozygous sites (more than 15 sites). We explored a range of

212 thresholds for the minimum genotyped samples (30, 51, and 70 samples; equivalent to

213 50%, 75%, and 90% of samples were genotyped, respectively). All three data sets were

214 examined in the phylogenetic analysis, and we selected the 50% genotyped data set as

215 the primary data set for all other analyses (see Results section).

216

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

217 Phylogeny construction and dating

218 To construct a phylogenetic tree of ddRAD-seq data and estimate divergence times

219 within the Heterotropa clade, we used Bayesian inference (BI) in BEAST v. 1.10.4

220 (Drummond & Rambaut, 2007). We calibrated the crown age of the Heterotropa clade

221 (with lower limit of 4.77 Mya and upper limit of 14.54 Mya) following the results of the

222 chloroplast genome phylogenetic analysis (see Results section). The Markov Chain

223 Monte Carlo method was performed using two simultaneous independent runs with four

224 chains each (one cold and three heated), saving one tree every 1000 generations for a

225 total 30,000,000 generations with 10% burn-in for each run. The convergence of the

226 chains was checked using the program Tracer v. 1.5 (Rambaut & Drummond, 2013).

227

228 Ancestral area reconstruction

229 To infer the ancestral areas and phylogeographic history of Heterotropa taxa, we

230 performed statistical dispersal-vicariance analysis (S-DIVA) using RASP v3.2 (Yu,

231 Harris, & He, 2010). The analysis was conducted using a phylogenetic tree of the

232 ddRAD-seq dataset with a default setting. We divided the distribution range of

233 Heterotropa into seven regions: (A) Sichuan basin and surrounding mountains, (B)

234 other parts of mainland China, (C) Taiwan and southern Ryukyu islands, (D) central

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

235 Ryukyu islands, (E) northern Ryukyu islands and Kyushu island, (F) southern part of

236 mainland Japan including Shikoku island, and (G) northern part of mainland Japan (see

237 Fig. 1c). The boundaries of these regions were defined with reference to biogeographic

238 studies (e.g., C and D; Kerama gap, D and E; Tokara gap [Kimura, 1996], F and G;

239 Itoigawa-Shizuoka tectonic line [Okamura et al., 2017]) and phylogeographic studies

240 (Landrein, Buerki, Wang, & Clarkson, 2017).

241

242 Analysis of trait evolution

243 To estimate the evolutionary patterns of floral traits concerning reproductive barriers,

244 we focused on flowering time, which we defined as a month when the taxon starts

245 flowering, and calyx tube width, defined as a median value between maximum and

246 minimum calyx tube widths. Flowering time is related to the local environment,

247 including pollinator fauna, and its differences play a role in the reproductive barrier

248 among taxa. Calyx tube width would be linked to pollinator size selection and its

249 difference could affect the difference in pollinator fauna, which leads to reproductive

250 isolation. The trait values were obtained from field observations and literature (Huang et

251 al., 2003; C. T. Lu & Wang, 2009; Sugawara, 2006). The evaluated trait values are

252 summarised in Table S1.

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

253 We first plotted each trait value on the tips of the phylogenetic tree to visualise

254 trait-phylogeny relationships using “plotTree.wBars” function in “phytools” package

255 (Revell, 2012) for R v. 3.5.4 (R Core Team, 2013). Next, to estimate the tempo and

256 mode of floral trait evolution on the phylogenetic tree of Heterotropa, we conducted

257 Bayesian macroevolutionary analysis for flowering time and calyx tube width

258 implemented in BAMM v. 2.5.2 (Rabosky, 2014). BAMM models shift in

259 macroevolutionary regimes across a phylogenetic tree using reversible-jump Markov

260 chain Monte Carlo (rjMCMC) sampling. To treat the flowering times as continuous

261 traits, we transformed the flowering times of Heterotropa onto a circular scale, where

262 the difference between December and January is equivalent to one month. The prior

263 values were set using the package “BAMMtools” (Rabosky et al., 2014) for R. To ease

264 model complexity, we adopted the time-invariant Brownian motion model of trait

265 evolution. The analysis involved a rjMCMC run of 10,000,000 generations sampled

266 every 10,000 steps, and the initial 3,000,000 generations were discarded as burn-in. The

267 rjMCMC convergence was confirmed using BAMMtools. For each trait, we reported

268 the mean scaled tree from the outputs of BAMM, in which each branch length is

269 shortened or stretched proportional to the mean evolutionary rates of the trait using the

270 “getMeanBranchLengthTree” function in BAMMtools. In addition, to infer the temporal

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

271 change in traits’ evolution rates, we plotted the evolutionary rate variation through time

272 using “plotRateThroughTime” function for each trait and clade.

273 To determine the evolutionary pattern of the two floral traits within

274 Heterotropa taxa, we also measured phylogenetic signals by estimating к and λ Pagel’s

275 statistics (Pagel, 1999). These values provide insights into temporal patterns of trait

276 evolution. Values of к and λ close to 1 indicate that the trait showed a strong

277 phylogenetic signal and evolved in a Brownian motion-like manner. Values к and λ

278 close to 0 indicate that closely related taxa showed relatively row trait similarity,

279 implying that the trait evolved punctually and that the trait changes occurred late in

280 evolutionary history. Estimation of Pagel’s statistics was conducted by using the

281 “phylosig” function in the “phytools” package. We tested for the significance of

282 phylogenetic signal values using 999 randomisation tests.

283

284 Geographic overlaps within each clade

285 The potential geographic range of each taxon was set by creating a convex polygon

286 from the distribution data. Distribution data for all taxa were based on the specimen

287 records of the herbarium of Kyoto University (KYO), S-Net data portal

288 (http://science-net.kahaku.go.jp/, accessed on 2019/1/26), and the Chinese Virtual

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

289 Herbarium (http://www.cvh.org.cn/, accessed on 2019/1/26), and the results of personal

290 observations. Because the distribution range of many Heterotropa taxa tends to be

291 confined to a small area, we did not set any threshold values of the number of records to

292 make distribution data. In total, we collected 1396 occurrences, with a mean number of

293 16 records per taxon (minimum 1, maximum 197). We were not able to find available

294 records for Asarum nobillisimum and excluded this taxon from the analysis. A single

295 convex polygon for each taxon was created by connecting the outline of the occurrence

296 point(s) by placing a 1 km round buffer and masking by a costal line. Following

297 Anacker and Strauss (2014), we calculated range overlap as the area occupied by both

298 taxa divided by that of the smaller ranged taxa. The range overlap values ranged from 0

299 (no overlap) to 1 (complete overlap) and were calculated for all pairs of taxa within the

300 three major clades (mainland China, Ryukyu-Taiwan, and mainland Japan clades, see

301 Results section). All geographic analysis were conducted by using “sf” package

302 (Pebesma, 2018) in R.

303

304 Geographic and morphological isolation

305 To infer the patterns of reproductive isolation in Heterotropa, we measured the

306 geographic and morphologic isolations between OTU pairs. The geographic isolation

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

307 index was calculated by transforming the geographic overlap values into binary data (0;

308 overlap, 1; non-overlap). We set the values of the geographic isolation index of partially

309 overlapping pairs to 0. We also calculated the morphological isolation indices for two

310 floral traits (flowering time and calyx tube width) between OTU pairs. To estimate the

311 differentiation of the floral traits, we set the intervals of the traits using the first and last

312 flowering months, and the maximum and minimum values of calyx tube width,

313 respectively, according to the trait data. For each trait, when a pair of taxa had an

314 overlap of the intervals, they were scored as 0 for “overlapping”; if they showed no

315 overlaps, they were scored as 1 for “isolation”. We calculated these isolation indices for

316 all pairs of OTUs and conditional relationships among the attributes within the three

317 major clades.

318

319 Comparison of sister and non-sister taxon pairs

320 If allopatric speciation was common in Heterotropa, we can expect that closest relatives

321 would exhibit less geographic overlap compared with non-sister pairs. Likewise, we can

322 speculate whether the floral traits are concerned with speciation by examining how

323 floral traits differentiate between close relatives. We compared the geographic and

324 morphological isolation values of sister taxa with those of non-sister taxa by using

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

325 Bayesian generalised linear mixed models (GLMM) analysis. GLMM analyses were

326 conducted using “brms” package (Burkner, 2017) in R. We set each isolation index as a

327 response variable following the Bernoulli distribution, and the relationship of pair

328 (sister or non-sister taxa) was used as the fixed effect. We selected sister pairs with

329 posterior probabilities of nodes higher than 80% in the phylogenetic analysis of

330 ddRAD-seq data sets as sister taxa and excluded pairs with less than 80% support in the

331 analysis. We included a nested random effect: OTU pairs nested within their clades

332 (Mainland Japan, Ryukyu-Taiwan, or mainland China). For OTU pairs, to account for

333 non-independence resulted from the use of pairwise matrices of individual-level metrics,

334 we used “multi-membership” function implemented in brms. We run the four chains for

335 10,000 iterations with a warm-up length 5,000 of iterations with default priors. We also

336 conducted the same analyses using the data set adopting the sister taxa with posterior

337 probabilities > 99% to account for phylogenetic uncertainty in our analyses.

338

339

340

341

342

19 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

343

344

345

346

347

348

349

350

351 Results

352 ddRAD-seq data

353 The ddRAD-seq reads generated by Illumina sequencing were deposited in GenBank

354 (BioProject ID: PRJDB8943). After filtering low-quality reads and bases, the number of

355 reads of each sample ranged from 510,135 to 2,558,859 reads and the average number

356 of reads was 1,348,526 (Table S1). Our 50% genotyped matrix consisted of 469 loci,

357 which contained 3,415 parsimony-informative SNPs. Our 70% and 90% matrices

358 included 117 and 46 loci with 710 and 266 parsimony informative SNPs, respectively.

359

360 Phylogenetic inference and ancestral area reconstruction

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

361 Phylogenetic analysis based on 59 chloroplast CDS regions supported all clades within

362 the tree with posterior probabilities > 0.99 and showed that the Heterotropa was

363 monophyletic (Fig. S2). The estimated divergence time between a mainland China taxon

364 (A. wulingense) and the insular clade including Asarum forbesii was 9.16 Mya (95%

365 HPD: 4.66 - 13.55 Mya).

366 Phylogenetic analysis with 50% genotyped ddRAD-seq matrix yielded strongly

367 supported clades within Heterotropa (Fig. 1c). Within the Heterotropa clade, the

368 mainland China clade diverged early, followed by a splitting off of A. forbesii, which is

369 also distributed in mainland China, and the clade that consisted of insular taxa with only

370 one mainland Chinese taxa (Asarum ichangense). Within each insular and mainland

371 Chinese clade, there were subclades that almost corresponded with the geographic

372 structures. The mainland China clade was divided into two subclades, including the taxa

373 distributed around the Sichuan basin and in other parts of mainland China. The insular

374 clade consisted of two subclades, including taxa distributed in the southern parts of the

375 Japanese island arc (from Taiwan to the Amami islands; Ryukyu-Taiwan clade) and

376 northern parts (from Tokara islands to Honshu; mainland Japan clade). Asarum

377 ichangense was included in the Ryukyu-Taiwan clade. Within the mainland Japan clade,

378 almost all taxa found on Kyushu island (except for Asarum minamitanianum and

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

379 Asarum asperum) formed a clade with Asarum lutchuense and Asarum curvistigma,

380 which are found on Amami islands and Honshu island, respectively. The other taxa in

381 mainland Japan clade split into two subclades, both of which included southern and

382 northern Japanese taxa. All clades mentioned above showed high support (posterior

383 probabilities > 99%) and diverged during pre-Pleistocene periods (> 2.6 Mya; Fig. S3).

384 The topologies within the insular clade of ddRAD-seq data (this study) and ITS

385 (Takahashi & Setoguchi, 2018) were different from those of the chloroplast data. This

386 discordance could be due to incomplete lineage sorting within the recently divergent

387 clades and low mutation rates of chloroplast CDS regions (Wolfe, Li, & Sharp, 1987).

388 Our BI trees inferred from 75% and 90% genotyped ddRAD-seq matrices also

389 supported the hypothesis that Heterotropa split into the insular and mainland China

390 clades, while the nested structures were not resolved (Fig. S4). The tree obtained from

391 the 50% genotyped matrix showed relatively high support of nodes (mean posterior

392 probabilities = 91.0%), while other matrices showed lower supports (75% genotyped

393 matrix; 77.7%, and 90% genotyped matrix; 63.2%). Thus, we adopted the 50%

394 genotyped tree for further analysis.

395 The results of S-DIVA analysis (Fig. 1c) showed that the origin of Heterotropa

396 was in the mainland China region (B) and that dispersal to insular systems occurred

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

397 subsequently. From insular systems, only one back-dispersal to mainland China was

398 estimated (A. ichangense). The common ancestral area of the Ryukyu-Taiwan clade is

399 presumed to be Taiwan and the southern Ryukyu islands (C) or Taiwan with southern

400 and central Ryukyu islands (CD). In the mainland Japan clade, several taxa colonised

401 the northern part of Japan (G) from the southern part (E and F).

402

403 Analysis of trait evolution

404 Most Heterotropa taxa (55 taxa) analysed in this study start flowering in spring (March

405 to June), whereas 11 taxa flower in autumn (September to November) and 19 taxa in

406 winter (December to February). These autumn-flowering and winter-flowering taxa

407 were scattered across all three major clades, and many tip nodes did not share the

408 flowering time (Fig. 2a), implying that the flowering time of Heterotropa changed

409 several times in each of the three major clades, especially in tip nodes or branches. With

410 respect to the calyx tube width (Fig. 2b), the taxa that have more than 15 mm of calyx

411 width represented two clades (mainland Japan and mainland China clades), and the

412 changes would have evolved in parallel.

413 The results of BAMM analysis showed that for both traits, a single

414 macroevolutionary rate was unlikely to fit our genetic data (Fig. S5), indicating that

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

415 several rate shifts of trait evolution would have occurred in Heterotropa. From the

416 ddRAD-seq trees with branch lengths drawn proportional to their marginal evolutionary

417 rates (Fig. 2ab), the rate shifts were distributed across all three major clades within

418 Heterotropa for both traits. Most of these shifts accelerated the evolutionary rates, and

419 the shifts were estimated at more terminal nodes containing only several taxa, rather

420 than an early burst of evolutionary rate shift in crown nodes. As an exception, the clade

421 containing half the mainland Japanese taxa showed a substantial increase in the rate of

422 flowering time evolution. Rate variation through time plots indicated that the

423 evolutionary rates of both traits increased in all clades, and sudden increases were

424 observed in the flowering time of the mainland Japan and Ryukyu-Taiwan clades and in

425 the calyx tube width of the Ryukyu-Taiwan clade during the Pleistocene (Fig. 2cd).

426 For both traits, к and λ values were relatively low (к for flowering time; 0.389,

427 for calyx tube width; 0.390 and λ for flowering time; 0.433, calyx tube width; 0.521),

428 while all values were significantly different from 0 (p < 0.05), except for λ at flowering

429 time (p = 0.311; Table S4). This indicates that for both traits, trait evolution would have

430 occurred later within the regional clades and punctually in the evolutionary history.

431

432 Geographic and floral trait isolation

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

433 The mean values of the distribution areas of the taxa within mainland China,

434 Ryukyu-Taiwan, and mainland Japan clades were 172,168 km2, 570 km2, and 6,524 km2,

435 respectively, (Table S1; Fig S6). Within these clades, most of the taxa pairs were

436 distributed allopatrically (Table 1; Fig. S7abc). The mainland China clade contained a

437 relatively low proportion of geographically isolated pairs (69.85%), followed by the

438 Ryukyu-Taiwan clade (88.60%), and mainland Japan clade (90.36%). Within the

439 mainland Japan clade, 52.44% of taxon pairs showed flowering time isolation, whereas

440 the proportion of taxon pairs showing isolation in calyx tube width was lower (36.28%).

441 Conversely, both Ryukyu-Taiwan and mainland China clades contained relatively high

442 proportions of taxa pairs showing calyx tube width isolation (56.13% and 68.63%,

443 respectively), while only 21.37% and 28.76% of taxa pairs showed flowering time

444 isolation, respectively. All three clades included a high proportion of taxa pairs that

445 differentiated in either flowering time or calyx tube width (mainland Japan: 71.02%,

446 Ryukyu-Taiwan: 66.76%, and mainland China: 76.47%). In all three clades, most pairs

447 showing floral trait differentiation were geographically isolated, that is, there were no

448 conditional relationships between geographic overlaps and floral trait differentiation

449 (Fig. 3).

450

25 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

451 Sister and non-sister taxa pair comparisons

452 Eleven sister pairs out of 24 showed geographic overlaps to various degrees (17-100%),

453 and seven sister-pairs showed more than 80% range overlap (Table 2 & Fig. S7d). We

454 found 14 sister pairs showing the isolation of either flowering time or calyx tube width.

455 Within five pairs out of the 14 pairs, both traits were differentiated. In addition, six

456 sister-pairs showed floral trait isolation with geographic overlap (for example, Asarum

457 blumei vs. Asarum nipponicum, Asarum kinoshitae vs. Asarum rigescens var. rigescens,

458 Asarum gelasinum vs. Asarum monodoriflorum, Asarum celsum vs. Asarum gusk,

459 Asarum yaeyamense vs. Asarum taipingshanianum, A. wulingense vs. Asarum insigne,

460 and Asarum splendens vs. Asarum inflatum). Our GLMM analysis showed that, of the

461 three attributes (geographic isolation, flowering time, and calyx tube width), only

462 geographic isolation showed higher values between non-sister taxa pairs than sister

463 pairs (sister pairs; estimated median = 0.523, 95%CI = 0.255 - 0.7659, non-sister pairs;

464 estimated median = 0.918, 95%CI = 0.873 - 0.954: Table S5). The other attributes

465 showed no differentiation between sister and non-sister taxa pairs (Fig. 4). The datasets

466 containing sister-pairs with >99% supports showed the same tendencies (Fig. S8).

467

468

26 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

469

470

471

472

473

474

475

476

477 Discussion

478 Phylogeographic history of Heterotropa in the SJFR

479 Our phylogenetic analysis using ddRAD-seq data with comprehensive sampling

480 revealed the phylogeographic history of Heterotropa in the SJFR. The section originated

481 in mainland China (Fig. 1), and the divergence time between mainland China and

482 insular clades estimated from the chloroplast phylogenetic analysis was 9.16 Mya

483 (95%HPD 4.66 - 13.55 Mya), corresponding to the late Miocene (Fig. S2). These results

484 were concordant with the estimation of our previous study using the average

485 substitution rate of the ITS region (9.3 Mya; Takahashi & Setoguchi, 2018). During the

486 late Miocene, regressions and transgressions of the East China Sea (Haq, Hardenbol, &

27 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

487 Vail, 1987) caused land-bridge formation that allowed the migration of temperate plants

488 between the mainland China and insular systems (Cao, Comes, Sakaguchi, Chen, & Qiu,

489 2016; Yang et al., 2017). Furthermore, our study revealed that besides the insular clade,

490 the mainland China clade was also composed of subclades, which corresponded to taxa

491 geographic distribution (Fig. 1), and all subclades diverged during the late Miocene or

492 the Pliocene (Fig. S3). During this period, the establishment of a monsoon climate

493 caused by the uplift of the Himalayas and the Tibetan Plateau led to vegetational shifts

494 in the SJFR, and frequent glacial-regressions and inter-/after-glacial transgressions

495 (Kimura, 1996; Zheng, Powell, Rea, Wang, & Wang, 2004). We considered that

496 geographic and climatic events would have allowed colonisation of insular systems and

497 formation of regional lineages of Heterotropa, as shown in other studies (R.-S. Lu et al.,

498 2019; Mitsui et al., 2008).

499 As an exception, two Chinese species were not included in the mainland China

500 clade: A. ichangense and A. forbesii were included in or sister to the insular clade,

501 respectively. The phylogenetic placement of the two species is consistent with a

502 previous phylogenetic study (Okuyama et al., in press) and previous chromosomal

503 studies that showed they have the same chromosome numbers (2n = 24) as the insular

504 taxa, which is different from the other mainland China species (2n = 26) (Kelly, 1997;

28 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

505 Sugawara & Ogisu, 1992). Our study using exclusive sampling of Chinese taxa

506 demonstrated that only one back dispersal event from the insular systems to mainland

507 China would have occurred. In addition, colonisation to other regions after formation of

508 the regional lineages was observed only in several taxa (e.g., A. minamitanianum from

509 mainland Japan to Kyushu island, and A. lutchuense from Kyushu island to Amami

510 islands). This indicated that the regional lineages would have remained separated even

511 during the Pleistocene climatic oscillations. One of the isolation mechanisms would be

512 glacial isolation. In the SJFR, the existence of multiple refugia of temperate plants is

513 implied by phylogeographic studies (Qiu et al., 2011). In mainland China, in addition to

514 southern areas (< 30°N), the region around the Sichuan basin is inferred refugia and

515 isolated due to its complex topography, including high mountains and the Yangtze River

516 (Qiu, Guan, Fu, & Comes, 2009; Wang et al., 2015). In mainland Japan, during the

517 glacial periods, most parts were covered by mixed (boreal and cool temperate) forests or

518 boreal forests (Harrison et al., 2001), and warm temperate plants would have been

519 forced to retreat southward and survive separately in narrow glacial refugia on the

520 southern coasts of Kyushu, Shikoku, and Honshu islands (Aoki et al., 2019; Liu,

521 Takeichi, Kamiya, & Harada, 2013). Another isolating factor would be the seaways

522 barriers. In the Ryukyu islands, two deep-water passages (Tokara Tectonic Strait and

29 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

523 Kerama gap, currently > 1000m in depth), were formed during the Pliocene (Kimura,

524 1996). These deep-water passages act as isolation barriers for plant expansion

525 (Nakamura, Suwa, Denda, & Yokota, 2009). We considered that these glacial and

526 geographic isolations would prevent the colonisation of most Heterotropa taxa.

527

528 Diversity anomaly and its driving forces

529 There were no conditional relationships between geographic isolation and floral trait

530 isolation (Fig. 3), indicating that most pairs showing floral trait isolation were

531 distributed allopatrically. The results of trait evolutionary analyses indicated that

532 repeated trait evolutions would have occurred after the formation of regional lineages

533 during the Pleistocene period and that higher evolutionary rates of flowering time in

534 mainland Japan clade and of calyx tube width in the Ryukyu-Taiwan clade were

535 detected (Fig. 2). During the Pleistocene period, the repeated exposure and submergence

536 of the land-bridges would have led to significant population declines and isolations of

537 temperate plants in insular systems of the SJFR, while the populations in mainland

538 China would have been relatively stable (Qiu et al., 2011). Considering the current

539 smaller distribution ranges (Fig. S6) and a higher proportion of pairs showing

540 geographic isolation (Table 1), insular Heterotropa taxa would have experienced

30 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

541 repeated range fragmentations and contractions during the Pleistocene. The random

542 genetic drifts in the range contractions could be one of the mechanisms of trait

543 evolution in plants (Lande, 2000), and a simulation study also implied that a small

544 population size would promote floral evolution, including flowering time without

545 selective agents (Devaux & Lande, 2008). The neutral drift caused by the range

546 fragmentations and contractions due to climatic oscillations would have played a role in

547 generating the floral diversity of insular Heterotropa. In addition, we propose that

548 pollinator adaptation would have also occurred through range fragmentations and

549 contractions. The pollinator-mediated diversification of Heterotropa has been proposed

550 in a previous study (Sinn et al., 2015). Empirical studies have implied that the various

551 Diptera species are pollinators of insular Heterotropa taxa with specialisation (e.g.,

552 fungus gnus in A. tamaense [Sugawara, 1988], sciarid flies in A. costatum [Kakishima

553 and Okuyama, 2018], and Calliphoridae flies in Asarum fudsionoi [Maeda, 2013]).

554 These Diptera species differed in seasonal occurrence and would be attracted by

555 different cues (Funamoto, 2019). Although we cannot deny the possibility that the

556 pollinator difference results from the difference in the regional fauna, we consider that

557 the floral traits of Heterotropa would be related to the attraction of Diptera species.

558 Previous studies implied that geographic isolations often preceded ecological

31 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

559 divergence, especially in low mobility organisms (Boucher et al., 2016; Heinicke,

560 Jackman, & Bauer, 2017). In the SJFR, it was reported that the morphological

561 heterogeneity was facilitated by geographic isolations due to Pleistocene climatic

562 oscillations (Gao, Zhang, Gao, & Zhu, 2015), and the diversification within the insular

563 systems was due to genetic drifts caused by range fragmentations (Yoichi et al., 2017).

564 Thus, we considered that both selective and neutral agents facilitated by the range

565 fragmentations during the Pleistocene period would have led to the parallel evolution of

566 floral traits within the regional lineages and the formation of diversity anomaly of

567 Heterotropa in insular systems.

568

569 Biotic and abiotic drivers of sister taxa divergence

570 The results of sister and non-sister taxa comparison showed that the number of

571 geographically overlapping pairs was higher in sister-taxa pairs (Fig. 4), indicating that,

572 besides allopatric speciation, speciation on a small spatial scale would have also

573 occurred in Heterotropa taxa. Geographical overlap between close relatives requires

574 some kind of reproductive isolation to maintain the species boundary (Weber & Strauss,

575 2016). Six geographically overlapping sister taxa pairs showed floral trait isolation

576 (Table 2). In Heterotropa, most taxa inhabit almost the same environments (understory

32 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

577 of warm temperate forests) and the floral difference and/or geographic isolation would

578 act as reproductive barriers rather than habitat differences. This insight was

579 corroborated in a study of nine closely related Heterotropa taxa in the Amami islands,

580 which are distributed in sympatry and/or close parapatry and morphologically different

581 in floral traits (Matsuda, Maeda, Nagasawa, & Setoguchi, 2017). Thus, we considered

582 that the flowering time and calyx tube width would have acted as one of the possible

583 reproductive barriers in the six sister taxa pairs, and there are possibilities that the

584 speciation on a small spatial scale triggered by trait differentiations may have also

585 occurred in Heterotropa.

586

587 Conclusion

588 Our results implied that the diversity anomaly in Heterotropa in the SFJR would have

589 been formed by multiple drivers, including the geographic isolation and complemental

590 floral trait evolution with different temporal scales. Phylogenetic analysis showed that

591 in Heterotropa, the regional lineages would have been formed by geographic events

592 during the Miocene, and geographic and refugial isolation would have prevented most

593 chances of colonisation of the regional lineages. Traits evolution analysis implied that

594 the parallel floral trait divergence associated with reproductive isolation would have

33 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

595 occurred during the Pleistocene and that repeated range fragmentations in insular

596 systems would have promoted the evolution of floral traits, especially in insular systems.

597 Furthermore, the speciation on a small spatial scale with floral trait divergence was also

598 implied in at least six sister taxa-pairs. Hence, the diversification of insular Heterotropa

599 would have been triggered by the geographic and climatic events during the Miocene,

600 and subsequent repeated floral trait evolution with and without geographic isolations in

601 the regional lineages during the Pleistocene. Our study demonstrated the importance of

602 multidimensional studies to understand the diversification process of temperate plants in

603 the SJFR, where geographic isolation has been considered to play a major role in the

604 diversification.

605

606

607 Acknowledgements

608 We thank S. Gale, Y. Inoue, S. Liao, K. Maeda, J. Nagasawa, S. Nemoto, T. Teramine, Y.

609 Wakita, M. Yamamoto, and S. Zhou for their help with sampling and taxonomic

610 determination. We are also grateful to J. R. P. Worth, R. Imai, and M. Yamasaki for their

611 valuable comments on the genetic and statistical analyses and manuscript writing. This

612 work was supported by Grants-in-Aid for Scientific Research from the Japanese Society

34 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

613 for the Promotion of Science (Nos. 24247013, 26304013 and 18J22919), the

614 Environment Research and Technology Development Fund (grant no. 4-1702 and

615 4-1902), and the Environmental Research and Technology Development Fund of the

616 Ministry of the Environment SICORP Program of the Japan Science and Technology

617 Agency (“Spatial-temporal dimensions and underlying mechanisms of lineage

618 diversification and patterns of genetic variation of keystone plant taxa in

619 warm-temperate forests of Sino-Japanese Floristic Region”) (grant no. 4-1403). We

620 would like to thank Editage (www.editage.com) for English language editing.

621

622

623

624

625 Data accessibility

626 The obtained reads for chloroplast genome construction and ddRAD-seq analysis are

627 available in NCBI (GenBank BioProject no. PRJDB9302 and PRJDB8943) The

628 alignment sequences, morphological data, and distribution records were deposited in the

629 Dryad® digital repository under doi: 10.5061/dryad.xwdbrv1b4.

630

35 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

631 Author contributions

632 D. T., S. S., Y. Q., Y. I., and H. S. conceptualised and designed the study. Sample

633 collection was performed by D. T., S. S., Y. F., Y. Q., Y. I., P. L., R. L., C. L., S. C., Y. L.,

634 Y. C., and H. S. The molecular experiments were conducted by D. T., S. S., L. K., and A.

635 N., Data were generated, analysed, and visualised by D. T. and S. S. Manuscript writing

636 was led by D. T., with contributions from all authors.

637

638 Conflict of interest

639 The authors declare that they have no conflicts of interest.

640

641

642

643 Reference list 644 Anacker, B. L., & Strauss, S. Y. (2014). The geography and ecology of plant speciation: range 645 overlap and niche divergence in sister species. Proceedings of the Royal Society 646 B-Biological Sciences, 281(1778). doi:10.1098/rspb.2013.2980 647 Aoki, K., Tamaki, I., Nakao, K., Ueno, S., Kamijo, T., Setoguchi, H., . . . Tsumura, Y. (2019). 648 Approximate Bayesian computation analysis of EST-associated microsatellites 649 indicates that the broadleaved evergreen tree Castanopsis sieboldii survived the 650 Last Glacial Maximum in multiple refugia in Japan. Heredity, 122(3), 326. 651 Armbruster, W. S. (2014). Floral specialization and angiosperm diversity: phenotypic 652 divergence, fitness trade-offs and realized pollination accuracy. Aob Plants, 6. 653 doi:10.1093/aobpla/plu003 654 Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: a flexible trimmer for Illumina

36 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

655 sequence data. Bioinformatics, 30(15), 2114-2120. doi:10.1093/bioinformatics/btu170 656 Boucher, F. C., Zimmermann, N. E., & Conti, E. (2016). Allopatric speciation with little niche 657 divergence is common among alpine Primulaceae. Journal of Biogeography, 43(3), 658 591-602. doi:10.1111/jbi.12652 659 Burkner, P. C. (2017). brms: An R Package for Bayesian Multilevel Models Using Stan. 660 Journal of Statistical Software, 80(1), 1-28. doi:10.18637/jss.v080.i01 661 Cao, Y. N., Comes, H. P., Sakaguchi, S., Chen, L. Y., & Qiu, Y. X. (2016). Evolution of East 662 Asia's Arcto-Tertiary relict Euptelea (Eupteleaceae) shaped by Late Neogene 663 vicariance and Quaternary climate change. Bmc Evolutionary Biology, 16. 664 doi:10.1186/s12862-016-0636-x 665 Devaux, C., & Lande, R. (2008). Incipient allochronic speciation due to non-selective 666 assortative mating by flowering time, mutation and genetic drift. Proceedings of the 667 Royal Society B-Biological Sciences, 275(1652), 2723-2732. 668 doi:10.1098/rspb.2008.0882 669 Doyle, J., & Doyle, J. L. (1987). Genomic plant DNA preparation from fresh tissue-CTAB 670 method. Phytochem Bull, 19(11), 11-15. 671 Drummond, A. J., & Rambaut, A. (2007). BEAST: Bayesian evolutionary analysis by 672 sampling trees. Bmc Evolutionary Biology, 7, 8. doi:10.1186/1471-2148-7-214 673 Esselstyn, J. A., Timm, R. M., & Brown, R. M. (2009). Do geological or climatic processes 674 drive speciation in dynamic archipelagos? The tempo and mode of diversification in 675 Southeast Asian shrews. Evolution: International Journal of Organic Evolution, 676 63(10), 2595-2610. 677 Funamoto, D. (2019). Plant-pollinator interactions in east Asia: A review. Journal of 678 Pollination Ecology, 25(6), 46-68. 679 Gao, Y. D., Zhang, Y., Gao, X. F., & Zhu, Z. M. (2015). Pleistocene glaciations, demographic 680 expansion and subsequent isolation promoted morphological heterogeneity: A 681 phylogeographic study of the alpine Rosa sericea complex (Rosaceae). Scientific 682 Reports, 5. doi:10.1038/srep11698 683 Givnish, T. J., Spalink, D., Ames, M., Lyon, S. P., Hunter, S. J., Zuluaga, A., . . . Cameron, K. 684 M. (2015). Orchid phylogenomics and multiple drivers of their extraordinary 685 diversification. Proceedings of the Royal Society B-Biological Sciences, 282(1814), 686 171-180. doi:10.1098/rspb.2015.1553 687 Govindarajulu, R., Hughes, C. E., & Bailey, C. D. (2011). Phylogenetic and population 688 genetic analyses of diploid Leucaena (Leguminosae; Mimosoideae) reveal cryptic 689 species diversity and patterns of divergent allopatric speciation. American Journal 690 of Botany, 98(12), 2049-2063.

37 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

691 Haq, B. U., Hardenbol, J., & Vail, P. R. (1987). Chronology of fluctuating sea levels since the 692 Triassic. Science, 235(4793), 1156-1167. 693 Harrison, S. P., Yu, G., Takahara, H., & Prentice, I. C. (2001). Palaeovegetation - Diversity of 694 temperate plants in east Asia. Nature, 413(6852), 129-130. doi:10.1038/35093166 695 Heinicke, M. P., Jackman, T. R., & Bauer, A. M. (2017). The measure of success: geographic 696 isolation promotes diversification in Pachydactylus geckos. Bmc Evolutionary 697 Biology, 17. doi:10.1186/s12862-016-0846-2 698 Hiura, I. (1978). The histories traced by the butterfly (in Japanese). In. Tokyo: Aoki shobou. 699 Huang, S., Kelly, L. M., & Gilbert, M. G. (2003). Aristolochiaceae (Vol. 5): Science Press. 700 Jaramillo, M. A., & Manos, P. S. (2001). Phylogeny and patterns of floral diversity in the 701 genus Piper (Piperaceae). American Journal of Botany, 88(4), 706-716. 702 doi:10.2307/2657072 703 Kakishima, S., & Okuyama, Y. (2018). Floral scent profiles and flower visitors in species 704 of Asarum Series Sakawanum (Aristolochiaceae). J Bulletin of the National 705 Museum of Nature and Science. Series B, Botany, 44(1), 41-51. 706 Kelly, L. M. (1997). A cladistic analysis of Asarum (Aristolochiaceae) and implications for the 707 evolution of Herkogamy. American Journal of Botany, 84(12), 1752-1765. 708 doi:10.2307/2446475 709 Kimura, M. (1996). Quaternary paleogeography of the Ryukyu Arc. Journal of Geography 710 Tokyo, 105, 259-285. 711 Kinoshita, S. (1973). Obituary of the Late Mr. Y. Kato. Journal of Geobotany, 26(4), 109-111. 712 Lagomarsino, L. P., Condamine, F. L., Antonelli, A., Mulch, A., & Davis, C. C. (2016). The 713 abiotic and biotic drivers of rapid diversification in Andean bellflowers 714 (Campanulaceae). New Phytologist, 210(4), 1430-1442. doi:10.1111/nph.13920 715 Lande, R. (2000). Quantitative genetics and phenotypic evolution. In R. C. Lewontin & K. V. 716 Krimpas (Eds.), Evolutionary Genetics: From Molecules to Morphology (Vol. 1) (pp. 717 335-350). Cambridge, UK: Cambridge University Press. 718 Landrein, S., Buerki, S., Wang, H. F., & Clarkson, J. J. (2017). Untangling the reticulate 719 history of species complexes and horticultural breeds in Abelia (Caprifoliaceae). 720 Annals of Botany, 120(2), 257-269. doi:10.1093/aob/mcw279 721 Li, Y., Yan, H. F., & Ge, X. J. (2012). Phylogeographic analysis and environmental niche 722 modeling of widespread shrub Rhododendron simsii in China reveals multiple 723 glacial refugia during the last glacial maximum. Journal of Systematics and 724 Evolution, 50(4), 362-373. doi:10.1111/j.1759-6831.2012.00209.x 725 Liu, H. Z., Takeichi, Y., Kamiya, K., & Harada, K. (2013). Phylogeography of Quercus 726 phillyraeoides (Fagaceae) in Japan as revealed by chloroplast DNA variation.

38 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

727 Journal of Forest Research, 18(4), 361-370. doi:10.1007/s10310-012-0357-y 728 Lu, C.-T., Chiou, W.-L., Liu, S.-C., & Wang, J.-C. (2010). Asarum satsumense F. Maekawa 729 (Aristolochiaceae), a newly recorded species in Taiwan. Taiwania, 55(4), 396-401. 730 Lu, C. T., & Wang, J. C. (2009). Three new species of Asarum (section Heterotropa) from 731 Taiwan. Botanical Studies, 50(2), 229-240. 732 Lu, C. T., & Wang, J. C. (2014). Asarum ampulliflorum (Aristolochiaceae), a new species 733 from Taiwan. Phytotaxa, 184(1), 46-52. 734 Lu, L., Fritsch, P. W., Matzke, N. J., Wang, H., Kron, K. A., Li, D. Z., & Wiens, J. J. (2019). 735 Why is fruit colour so variable? Phylogenetic analyses reveal relationships between 736 fruit-colour evolution, biogeography and diversification. Global Ecology and 737 Biogeography, 28(7), 891-903. doi:10.1111/geb.12900 738 Lu, R.-S., Chen, Y., Tamaki, I., Sakaguchi, S., Ding, Y.-Q., Takahashi, D., . . . Qiu, Y.-X. 739 (2019). Pre-Quaternary diversification and glacial demographic expansions of 740 Cardiocrinum (Liliaceae) in temperate forest biomes of Sino-Japanese Floristic 741 Region. Molecular phylogenetics and evolution, 106693. 742 Maeda, Y. (2013). Distribution and life cycle of the wild gingers (Asarum , Aristolochiaseae) 743 on Amami-oshima, South Japan. PhD Dissertation of Kagoshima University. 744 Maekawa, F. (1953). Relationship between geological history and the variation in plants. 745 Democratic Scientists Association of Japan ed., Variability of Organism, 35-47. 746 Maekawa, F. (1983). Heterotropa kiusiana. In R. Ishii & Y. Inoue (Eds.), Encyclopedia of 747 horticulture (Vol. 5). Tokyo, Japan: Seibundo-Shinkosha. 748 Matsuda, J., Maeda, Y., Nagasawa, J., & Setoguchi, H. (2017). Tight species cohesion among 749 sympatric insular wild gingers (Asarum spp. Aristolochiaceae) on continental 750 islands: Highly differentiated floral characteristics versus undifferentiated 751 genotypes. Plos One, 12(3). doi:10.1371/journal.pone.0173489 752 Mitsui, Y., Chen, S. T., Zhou, Z. K., Peng, C. I., Deng, Y. F., & Setoguchi, H. (2008). Phylogeny 753 and biogeography of the genus () in the Sino-Japanese region 754 based on nuclear rDNA and plastid DNA sequence data. Annals of Botany, 101(1), 755 111-124. doi:10.1093/aob/mcm267 756 Nakamura, K., Suwa, R., Denda, T., & Yokota, M. (2009). Geohistorical and current 757 environmental influences on floristic differentiation in the Ryukyu Archipelago, 758 Japan. Journal of Biogeography, 36(5), 919-928. 759 doi:10.1111/j.1365-2699.2008.02057.x 760 Okaura, T., Quang, N. D., Ubukata, M., & Harada, K. (2007). Phylogeographic structure and 761 late Quaternary population history of the Japanese oak Quercus mongolica var. 762 crispula and related species revealed by chloroplast DNA variation. Genes & Genetic

39 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

763 Systems, 82(6), 465-477. doi:10.1266/ggs.82.465 764 Okuyama, Y., Goto, N., Nagano, J. A., Yasugi, M., Kokubugata, G., Kudoh, H., . . . Sugawara, 765 T. (in press). Radiation history of Asian Asarum (sect. Heterotropa , 766 Aristolochiaceae) resolved using a phylogenomic approach based on double-digested 767 RAD-seq data. Annals of Botany. 768 Pagel, M. (1999). Inferring the historical patterns of biological evolution. Nature, 401(6756), 769 877-884. doi:10.1038/44766 770 Pebesma, E. (2018). Simple features for R: standardized support for spatial vector data. The 771 R Journal, 10(1), 439-446. 772 Peterson, B. K., Weber, J. N., Kay, E. H., Fisher, H. S., & Hoekstra, H. E. (2012). Double 773 Digest RADseq: An Inexpensive Method for De Novo SNP Discovery and Genotyping 774 in Model and Non-Model Species. Plos One, 7(5). doi:10.1371/journal.pone.0037135 775 Petit, R. J., Duminil, J., Fineschi, S., Hampe, A., Salvini, D., & Vendramin, G. G. (2005). 776 Comparative organization of chloroplast, mitochondrial and nuclear diversity in 777 plant populations. Molecular Ecology, 14(3), 689-701. 778 doi:10.1111/j.1365-294X.2004.02410.x 779 Qian, H., & Ricklefs, R. E. (2000). Large-scale processes and the Asian bias in species 780 diversity of temperate plants. Nature, 407(6801), 180-182. 781 Qiu, Y. X., Fu, C. X., & Comes, H. P. (2011). Plant molecular phylogeography in China and 782 adjacent regions: Tracing the genetic imprints of Quaternary climate and 783 environmental change in the world's most diverse temperate flora. Molecular 784 Phylogenetics and Evolution, 59(1), 225-244. doi:10.1016/j.ympev.2011.01.012 785 Qiu, Y. X., Guan, B. C., Fu, C. X., & Comes, H. P. (2009). Did glacials and/or interglacials 786 promote allopatric incipient speciation in East Asian temperate plants? 787 Phylogeographic and coalescent analyses on refugial isolation and divergence in 788 Dysosma versipellis. Molecular Phylogenetics and Evolution, 51(2), 281-293. 789 doi:10.1016/j.ympev.2009.01.016 790 R Core Team. (2013). R: A language and environment for statistical computing. 791 Rabosky, D. L. (2014). Automatic Detection of Key Innovations, Rate Shifts, and 792 Diversity-Dependence on Phylogenetic Trees. Plos One, 9(2). 793 doi:10.1371/journal.pone.0089543 794 Rabosky, D. L., Grundler, M., Anderson, C., Title, P., Shi, J. J., Brown, J. W., . . . Larson, J. G. 795 (2014). BAMMtools: an R package for the analysis of evolutionary dynamics on 796 phylogenetic trees. Methods in Ecology and Evolution, 5(7), 701-707. 797 doi:10.1111/2041-210x.12199 798 Rambaut, A., & Drummond, A. J. (2013). Tracer v1. 5 Available from http://beast. bio. ed. ac.

40 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

799 uk/Tracer. In: Accessed. 800 Revell, L. J. (2012). phytools: an R package for phylogenetic comparative biology (and other 801 things). Methods in Ecology and Evolution, 3(2), 217-223. 802 doi:10.1111/j.2041-210X.2011.00169.x 803 Rundle, H. D., & Nosil, P. (2005). Ecological speciation. Ecology Letters, 8(3), 336-352. 804 doi:10.1111/j.1461-0248.2004.00715.x 805 Sakaguchi, S., Qiu, Y. X., Liu, Y. H., Qi, X. S., Kim, S. H., Han, J., . . . Isagi, Y. (2012). 806 Climate oscillation during the Quaternary associated with landscape heterogeneity 807 promoted allopatric lineage divergence of a temperate tree Kalopanax septemlobus 808 (Araliaceae) in East Asia. Molecular Ecology, 21(15), 3823-3838. 809 doi:10.1111/j.1365-294X.2012.05652.x 810 Sakaguchi, S., Sugino, T., Tsumura, Y., Ito, M., Crisp, M. D., Bowman, D., . . . Isagi, Y. (2015). 811 High-throughput linkage mapping of Australian white cypress pine (Callitris 812 glaucophylla) and map transferability to related species. Tree Genetics & Genomes, 813 11(6). doi:10.1007/s11295-015-0944-0 814 Setoguchi, H., Yukawa, T., Tokuoka, T., Momohara, A., Sogo, A., Takaso, T., & Peng, C. I. 815 (2006). Phylogeography of the genus Cardiandra based on genetic variation in 816 cpDNA sequences. Journal of Plant Research, 119(4), 401-405. 817 doi:10.1007/10265-006-0283-1 818 Sinn, B. T., Kelly, L. M., & Freudenstein, J. V. (2015). Putative floral brood-site mimicry, loss 819 of autonomous selfing, and reduced vegetative growth are significantly correlated 820 with increased diversification in Asarum (Aristolochiaceae). Molecular 821 Phylogenetics and Evolution, 89, 194-204. doi:10.1016/j.ympev.2015.04.019 822 Sugawara, T. (1987). Taxonomic studies of Asarum sensu lato III. comparative floral 823 anatomy. Botanical Magazine-Tokyo, 100(1060), 335-348. doi:10.1007/bf02488853 824 Sugawara, T. (1988). Floral biology of Heterotropa tamaensis (Aristolochiaceae) in Japan. 825 Plant Species Biology, 3(1), 7-12. 826 Sugawara, T. (2006). Aristorochiaceae. In K. Iwatsuki, D. E. Boufford, & H. Ohba (Eds.), 827 Flora of Japan (Vol. 2a). Tokyo: Kodansha. 828 Sugawara, T., & Ogisu, M. (1992). Karyomorphology of 11 species of Asarum 829 (Aristolochiaceae) from Taiwan and mainland China. Acta Phytotaxonomica et 830 Geobotanica, 43(2), 89-96. 831 Takahashi, D., & Setoguchi, H. (2018). Molecular phylogeny and taxonomic implications of 832 Asarum (Aristolochiaceae) based on ITS and matK sequences. Plant Species Biology, 833 33(1), 28-41. doi:10.1111/1442-1984.12189 834 Takhtajan, A. (1969). Flowering plants: origin and dispersal. Washington DC: Smithsonian

41 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

835 Institution Press. 836 Tsukada, M. (1984). A vegetation map in the Japanese Archipelago approximately 20, 000 837 years B. P. Japanese Journal of Ecology, 34(2), 203-208. 838 Ujiie, H. (1990). Geological history of the Ryukyu Island arc. Nature of Okinawa; 839 geomorphology and geology. Hirugisha, Naha, 251-255. 840 Van der Niet, T., Peakall, R., & Johnson, S. D. (2014). Pollinator-driven ecological speciation 841 in plants: new evidence and future perspectives. Annals of Botany, 113(2), 199-211. 842 doi:10.1093/aob/mct290 843 Verboom, G. A., Bergh, N. G., Haiden, S. A., Hoffmann, V., & Britton, M. N. (2015). 844 Topography as a driver of diversification in the Cape Floristic Region of South Africa. 845 New Phytologist, 207(2), 368-376. doi:10.1111/nph.13342 846 Vogel, S. (1978). Pilzmiickenblumen als Pilzmimenten I.[fungus-gnat flowers mimicking 847 fungi] Flora 167. 329-66. Asarum, Aristolochia. Pollination by fungus-gnats. Floral 848 morphology, Aristolochiaceae, Incompatibility Pollination (PMBD, 185610740). 849 Wang, Y. H., Jiang, W. M., Comes, H. P., Hu, F. S., Qiu, Y. X., & Fu, C. X. (2015). Molecular 850 phylogeography and ecological niche modelling of a widespread herbaceous climber, 851 Tetrastigma hemsleyanum (Vitaceae): insights into Plio–Pleistocene range dynamics 852 of evergreen forest in subtropical China. New Phytologist, 206(2), 852-867. 853 Weber, M. G., & Strauss, S. Y. (2016). Coexistence in Close Relatives: Beyond Competition 854 and Reproductive Isolation in Sister Taxa. Annual Review of Ecology, Evolution, and 855 Systematics, Vol 47, 47, 359-381. doi:10.1146/annurev-ecolsys-112414-054048 856 Wolfe, K. H., Li, W.-H., & Sharp, P. M. (1987). Rates of nucleotide substitution vary greatly 857 among plant mitochondrial, chloroplast, and nuclear DNAs. Proceedings of the 858 National Academy of Sciences, 84(24), 9054-9058. 859 Wu, Z., & Wu, S. (1996). A proposal for a new floristic kingdom (realm)-the E. Asiatic 860 kingdom, its delineation and characteristics. Beijing, China/Berlin, Heidelberg, 861 Germany: Beijing: China Higher Education Press/Springer-Verlag. 862 Yang, L. Q., Hu, H. Y., Xie, C., Lai, S. P., Yang, M., He, X. J., & Zhou, S. D. (2017). Molecular 863 phylogeny, biogeography and ecological niche modelling of Cardiocrinum (Liliaceae): 864 insights into the evolutionary history of endemic genera distributed across the 865 Sino-Japanese floristic region. Annals of Botany, 119(1), 59-72. 866 doi:10.1093/aob/mcw210 867 Yoichi, W., Jin, X. F., Peng, C. I., Tamaki, I., & Tomaru, N. (2017). Contrasting diversification 868 history between insular and continental species of three-leaved azaleas 869 (Rhododendron sect. Brachycalyx) in East Asia. Journal of Biogeography, 44(5), 870 1065-1076. doi:10.1111/jbi.12924

42 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

871 Yu, Y., Harris, A. J., & He, X. J. (2010). S-DIVA (Statistical Dispersal-Vicariance Analysis): A 872 tool for inferring biogeographic histories. Molecular Phylogenetics and Evolution, 873 56(2), 848-850. doi:10.1016/j.ympev.2010.04.011 874 Zheng, H. B., Powell, C. M., Rea, D. K., Wang, J. L., & Wang, P. X. (2004). Late Miocene and 875 mid-Pliocene enhancement of the East Asian monsoon as viewed from the land and 876 sea. Global and Planetary Change, 41(3-4), 147-155. 877 doi:10.1016/j.gloplacha.2004.01.003 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 Figure legends 898 899 Fig. 1. The diversification of Heterotropa in the Sino-Japanese Floristic region. (a) 900 Photographs of the flowers of Heterotropa taxa (a-1; Asarum takaoi, a-2; Asarum unzen, 901 a-3; Asarum dissitum, and a-4; Asarum maximum). (b) Map of the Sino-Japanese 902 Floristic Region. Red circles indicate sampling points. (c) Time-calibrated molecular 903 phylogenetic tree and the ancestral areas estimated from Bayesian analysis and 904 statistical dispersal-vicariance analysis (S-DIVA). Posterior probabilities of < 100% are 905 indicated above or below branches, and the branches without posterior probabilities are 906 those with 100% support. The colours of pie charts reflect the estimated distribution

43 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

907 areas according to the biogeographic delimitation as in map (A; Sichuan basin and 908 surrounding mountains, B; other parts of mainland China, C; Taiwan and southern 909 Ryukyu islands, D; central Ryukyu islands, E; northern Ryukyu islands and Kyushu 910 island, F; southern part of mainland Japan including Shikoku island, and G; northern 911 part of mainland Japan). 912 913 Fig. 2. The results of macroevolutionary modelling analysis of floral trait evolution 914 estimated from BAMM. Phylogenetic trees with branch lengths were proportional to 915 their marginal phenotypic evolution rate of flowering time (a) and calyx tube width (b). 916 The topology of the trees is the same as in Figure 2. Colours and lengths of bars across 917 the tips of the phylogenetic trees represent flowering time and mean calyx tube width 918 (mm), respectively. Flowering times are classified into three types: autumn (flowering at 919 September to November; purple), winter (flowering at December to February; yellow), 920 and spring (flowering at March to June; green). The median value with 95% CI of 921 evolutionary rate of flowering time (c) and calyx tube width (d) through time in each 922 clade (black; whole clades, red; mainland China clade, blue; Ryukyu-Taiwan clade, and 923 light green; mainland Japan clade). Dashed vertical lines indicate the boundary between 924 the Pliocene and the Pleistocene (2.6 Mya). 925 926 Fig. 3. Venn diagrams showing the number of taxa pairs with isolation of distribution 927 range (GEO; red), and flowering time (FT; blue) and calyx tube width (CTW; green) 928 within mainland Japan clade (a), Ryukyu-Taiwan clade (b), and mainland China clade 929 (c). 930 931 Fig. 4. The results of sister and non-sister pairs comparisons of geographic, flowering 932 time, and calyx tube width isolations estimated from the GLMM analysis. The mean 933 values and 95% CIs of sister pairs and non-sister pairs are shown in red and blue, 934 respectively. The included sister-pairs had posterior probabilities >80%. 935 Table 1. The proportions of taxa pairs showing geographic isolation and floral trait 936 isolation in the three major clades (mainland Japan, Ryukyu-Taiwan, and mainland 937 China). The values within parentheses show the number of pairs showing 938 overlap/differentiation and number of all pairs. 939 940 Table 2. Geographic overlaps and morphological differences in the 24 sister-taxa pairs 941 with more than 80% posterior support. For each sister-taxa pair, columns indicate the 942 posterior probability that the two taxa are sister, proportion of range overlap, flowering

44 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

943 time differentiation (month), calyx tube width differentiation (mm), estimated 944 divergence time, and the clade name that includes two taxa. 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 Supplemental Figures and Tables 972 973 Fig. S1. Part of the floral diversity of Heterotropa taxa. The font and the side views of 974 the flowers of Asarum takaoi (a), A. savatieri var. iseanum (b), A. nipponicum (c), A. 975 hexalobum var. perfectum (d), A. rigescens (e), A. costatum (f), A. trigynum (g), A. 976 sakawanum var. stellatum (h), A. satsumense K (i), A. dissitum (j), A. pellucidum (k), A. 977 gusk (l), A. senkakuinsulare (m), A. yaeyamense (n), A. tokarense (o), A. villisepalum 978 (p), A. hypogynum (q), A. chatienshanianum (r), A. macranthum (s), A. forbesii (t), A.

45 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

979 ichangense (u), A. delavayi (v), A. petelotii (w), A. inflatum (x), and A. maximum (y). 980 The taxa were ordered according to their distributions: mainland Japan (a-h), the 981 Ryukyu Islands (i-o), Taiwan (p-s), and mainland China (t-y). The colours of the 982 alphabet indicate the flowering time of the taxa (red; autumn, blue; winter, and green; 983 spring). 984 985 Fig. S2. Phylogenetic tree based on 59 CDS regions (26,786bp) of the chloroplast 986 genomes of the and Chloranthales species. The red circle indicates the 987 calibration point (169 – 180 Mya) followed by Zeng et al. (2014). The bars indicated 988 95% highest posterior density (HPD) intervals of estimated divergence times of nodes. 989 The posterior probabilities of all the blanches were > 0.999. 990 991 Fig. S3. A majority-rule consensus tree inferred from Bayesian analysis of ddRAD-seq 992 data (50% genotyped data) showing the divergence times of major nodes and 95% 993 highest posterior density (HPD) intervals. 994 995 Fig. S4. Molecular phylogenetic tree of Heterotropa taxa estimated from Bayesian 996 analysis using 75% genotyped (a) and 90% genotyped (b) matrices. Values above or 997 below branches indicate the posterior probabilities of branches. 998 999 Fig. S5. Histograms of supported number of macroevolutionary rate regimes for each 1000 trait evolution; (a) flowering time and (b) calyx tube width. The number of regimes = 1 1001 indicated there were no shifts in trait evolution rate throughout the tree. 1002 1003 Fig. S6. Histograms of distribution area of taxa within mainland Japan clade (a), 1004 Ryukyu-Taiwan clade (b), and mainland China clade (c). 1005 1006 Fig. S7. Histograms of proportion of geographic range overlap between taxa pairs 1007 within mainland Japan clade (a), Ryukyu-Taiwan clade (b), mainland China clade (c), 1008 and sister-taxa pairs (d). 1009 1010 Fig. S8. The results of sister and non-sister taxa pairs comparisons of the three attributes 1011 (geographical, flowering time, and calyx width isolations), including sister taxa pairs 1012 supported by more than 99% posterior probabilities. The mean values and 95% CIs of 1013 sister pairs and non-sister pairs are shown in red and blue, respectively. 1014

46 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1015 1016 1017 Table S1. Sample list used in ddRAD-seq analysis. The columns indicate the 1018 distribution region used in S-DIVA, the median value of flowering time (month), 1019 median value of calyx tube width (mm), distribution area calculated from the polygon 1020 formed by the distribution record(s), number of records used in the geographic analysis, 1021 the number of cleaned RAD-seq reads, and the number of loci in the 50% matrices. 1022 1023 Table S2. Information of newly obtained reads and chloroplast genomes of Asarum spp. 1024 1025 Table S3. The sample lists used in phylogenetic analysis of chloroplast genome. 1026 1027 Table S4. The values of к and λ Pagel’s statistics for flowering time and calyx tube 1028 width calculated from the phytools package. 1029 1030 Table S5. The estimated median values and 95% CIs of each attribute (geographic, 1031 flowering time, and calyx tube width isolations) in sister pairs with posterior 1032 probabilities of nodes higher than 80% and 99%, and non-sister pairs. 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043

47 bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Fig. 1

bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Fig. 2

bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Fig. 3

bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Fig. 4

bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Table 1. The proportions of taxa pairs showing geographic isolation and floral trait isolation in the three major clades (mainland Japan, Ryukyu-Taiwan, and mainland China). The values within parentheses show the number of pairs showing overlap/differentiation and number of all pairs. Clade

Attribute Mainland Japan Ryukyu-Taiwan Mainland China

Geographic isolation 90.36% (704/780) 88.60% (285/325) 69.85% (95/136)

Flowering time isolation 52.44% (409/780) 21.37% (75/351) 28.76% (44/153)

Calyx width length isolation 36.28% (283/780) 56.13% (197/351) 68.63% (105/153)

bioRxiv preprint doi: https://doi.org/10.1101/2020.04.25.060632; this version posted April 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.