Genetic Structure and Diversity in Relation to the Recently Reduced Population Size of the Rare Conifer, Pseudotsuga Japonica, Endemic to Japan

Genetic structure and diversity in relation to the recently reduced population size of the rare conifer, Pseudotsuga japonica, endemic to Japan

Satoshi Tamaki, Keiya Isoda, Makoto Takahashi, Hiroo Yamada & Yumiko Yamashita

Conservation Genetics

ISSN 1566-0621

Conserv Genet DOI 10.1007/s10592-018-1092-5

1 23 Your article is protected by copyright and all rights are held exclusively by Springer Nature B.V.. This e-offprint is for personal use only and shall not be self-archived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com”.

1 23 Author's personal copy

Conservation Genetics https://doi.org/10.1007/s10592-018-1092-5

RESEARCH ARTICLE

Genetic structure and diversity in relation to the recently reduced population size of the rare conifer, Pseudotsuga japonica, endemic to Japan

Satoshi Tamaki1 · Keiya Isoda2 · Makoto Takahashi2 · Hiroo Yamada3 · Yumiko Yamashita4

Received: 21 September 2017 / Accepted: 20 July 2018 © Springer Nature B.V. 2018

Abstract Rare species consisting of small populations are subject to random genetic drift, which reduces genetic diversity. Thus, determining the relationship between population size and genetic diversity would provide key information for planning a conservation strategy for rare species. We used six microsatellite markers to investigate seven extant populations of the rare conifer Pseudotsuga japonica, which is endemic to the Kii Peninsula and Shikoku Island regions that are geographi- cally separated by the Kii Channel in southwest Japan. The population differentiation of P. japonica was relatively high (FST = 0.101) for a coniferous species, suggesting limited gene flow among populations. As expected, significant regional differentiation (AMOVA; p < 0.05) indicated genetic divergence across the Kii Channel. A strong positive correlation between census population size and the number of rare alleles (r = 0.862, p < 0.05) was found, but correlations with major indices of genetic diversity were not significant (allelic richness: r = 0.649, p = 0.104, expected heterozygosity: r = 0.361, p = 0.426). The observed order of magnitude of correlation with three genetic diversity indices corresponded with the theoretically expected order of each index’ sensitivity (i.e., the rate of decline per generation) to the bottleneck event. Thus, features that exhibit a faster response, i.e., the number of rare alleles, would have been subject to deleterious effects of the recent decline in population size, which is presumably caused by the development of extensive artificial plantations of other tree species over the last several decades. Finally, we propose a conservation plan for P. japonica based on our findings.

Keywords Phylogeography · Genetic diversity · Number of rare alleles · Rare species · Effective population size · Pseudotsuga japonica

Introduction

Rare species, consisting of small naturally fragmented populations, are subject to random genetic drift, which reduces genetic diversity (Ellstrand and Elam 1993). Genetic diver- * Satoshi Tamaki sity is generally associated with fitness and the ability to [email protected] evolve in response to environmental change (Hamrick 2004; 1 Tohoku Regional Breeding Office, Forest Tree Breeding Willi et al. 2006) and, hence, long-term species survival Center, Forestry and Forest Products Research Institute, 95 (Reed and Frankham 2003; Leimu et al. 2006). Suscep- Oosaki, Takizawa, Iwate 020‑0621, Japan tibility to a reduction in genetic diversity by genetic drift 2 Genetic Resources Department, Forest Tree Breeding Center, depends on the harmonic mean of the effective population Forestry and Forest Products Research Institute, 3809‑1 Ishi, size over generations (Wright 1938). Thus, large popula- Juo, Hitachi, Ibaraki 319‑1301, Japan 3 tions are expected to have higher genetic diversity than small Kansai Regional Breeding Office, Forest Tree Breeding populations. Multiple studies have focused on this hypoth- Center, Forestry and Forest Products Research Institute, 1043, Uetsukinaka, Shoo, Katsuta, Okayama 709‑4335, esis and have generally revealed a positive relationship Japan between population size and genetic diversity (Gao 2005; 4 Wakayama Prefectural Forestry Experiment Station, 1504‑1 Ilves et al. 2013; Busch and Reisch 2016). A meta-analysis Ikuma, Kamitonda, Wakayama 649‑2103, Japan study by Leimu et al. (2006) elucidated that the strength

Vol.:(0123456789)1 3 Author's personal copy

Conservation Genetics and direction of the relationship depend on species char- between the effective in addition to the actual population acteristics, such as mating system, and possibly, on species size and genetic diversity indices would provide fundamen- rarity. Thus, broadening of the study to include a wide vari- tal information for conserving rare species that consist of ety of species makes it possible to clarify the relationships small populations. among them from the perspective of species characteristics. The genus Pseudotsuga (Pinaceae) consists of two spe- However, previous studies on plants have mainly focused cies that are distributed in North America and two (Farjon on short-lived and insect-pollinated species (Leimu et al. 1990) or six species (Hermann 1982) that are distributed in 2006; Ilves et al. 2013; Busch and Reisch 2016), and stud- East Asia. The classification of East Asian species is not yet ies on long-lived tree species and wind-pollinated species established, except for Pseudotsuga japonica. Phylogenetic are relatively limited (Billington 1991; del Castillo et al. relationships within the Pseudotsuga genus have been inves- 2011; Chybicki et al. 2012). Previous studies have tested tigated via molecular analysis (Strauss et al. 1990; Gernandt the above hypothesis mostly in artificially fragmented areas, and Liston 1999; Wei et al. 2010). Although widespread with the aim to detect deleterious effects of human-caused species, such as Pseudotsuga menziesii and Pseudotsuga fragmentation on the genetic diversity of local populations. sinensis, are included in this genus, P. japonica is very rare Although studying such areas is important from the perspec- species. The microhabitat of P. japonica is topographically tive of conservation, it may be difficult to test the hypothesis restricted to ridge conditions (Yamamoto 1992), and it in investigations of long-lived tree species. According to grows alongside other conifer species (Abies firma, Tsuga theoretical studies, it takes many generations following the sieboldii, Sciadopitys verticillata, and Pinus parviflora) and population size reduction for the loss of genetic diversity to broad-leaved trees (Hayashi 1952). Its altitudinal distribu- become apparent (Lowe et al. 2005; England et al. 2010). tion extends from 500 to 900 m above sea level (Hayashi Indeed, several empirical studies in tree species to test this 1960). The natural distribution of P. japonica in small, hypothesis revealed no signs of genetic depauperation in scattered groups and human activities such as logging have fragmented populations (Young et al. 1993; Victory et al. reduced the population size and eliminated local popula- 2006). The relatively high gene flow in tree species may also tions (Yamanaka 1975); therefore, P. japonica has been complicate establishing a relationship between population designated “vulnerable” in the Japanese Red Data Book size and genetic diversity as even relatively low-frequency (Environmental Agency of Japan 2000). To conserve the gene flow between remnant populations can mitigate the genetic variation of P. japonica, it is essential to understand loss of genetic diversity (Lowe et al. 2005). To overcome its genetic structure. However, molecular marker studies of these difficulties, it might be effective to select study species genetic variation in P. japonica have not been performed. consisting of spatially and temporarily isolated populations. Species such as P. japonica with fragmented distribution Species with naturally fragmented distribution patterns are tend to show strong population differentiation (Hamrick likely to fulfill this necessary condition. et al. 1992). In addition, the Kii Channel, which separates Pseudotsuga japonica (Shirasawa) Beissner is a threat- the distribution areas, may hamper gene flow between the ened, rare conifer species that is endemic to the Kii Penin- regions. Channels and straits often act as topographic bar- sula and the eastern part of Shikoku in Japan (Hayashi 1960, riers (Terrab et al. 2008; Jaramillo-Correa et al. 2010) and Fig. 1). P. japonica is naturally distributed over separate and have attracted the interest of evolutionary and conservation different-sized fragments. The heterogeneous fragment size biologists because observed genetic structure patterns can might represent a substitution for population size; therefore, be interpreted with the help of information regarding past this species is likely to consist of local populations of dif- land-bridge emergence (Arroyo et al. 2008). ferent sizes. Considering its rarity and fragmented distribu- In the current study, we used six microsatellite mark- tion, the effective population size of each local population ers for genetic analysis of seven remnant populations of P. is expected to be generally small because small populations japonica to answer the following questions. (i) Does the geo- are more prone to genetic drift (Ellstrand and Elam 1993; graphic location of each population, and especially, the topo- Frankham et al. 2010). For evaluating the extinction risk of graphic separation by the Kii Channel, dictate the genetic small populations, the effective population size is an impor- differentiation pattern of P. japonica? (ii) Are the census tant parameter for the prediction of the future persistence of and effective population sizes associated with genetic diver- populations. Adverse genetic consequences of a small popu- sity? Based on our findings, we propose conservation strate- lation size, such as inbreeding and loss of genetic diversity, gies to maintain the genetic diversity of local P. japonica have been predicted to depend on the effective rather than populations. the actual population size (Frankham 1995). Several studies have reported that the effective population size is considera- bly smaller than the actual population size (Frankham 1995; Palstra and Ruzzante 2008). Thus, clarifying the association

1 3 Author's personal copy

Conservation Genetics

Fig. 1 Geographical distribution of the population-level membership of gene pools using STRU CTURE analysis (Pritchard et al. 2000). Pie charts represent the proportion of memberships assigned to each gene pool (K = 2, 4) in each population. The relative difference in chart size corresponds to the number of sampled individuals in each population. The distribution range of Pseudotsuga japonica is indicated by gray shading; relatively abundant areas are indicated in dark gray, and less abundant areas in light gray (Hayashi 1960). The reason for the disagreement in the location of the sampled population in the Shikoku Island region and published distribution rage is explained in the “Materials and methods” section. The relationships among four gene pools are illustrated using the neighbor- joining tree. FST values on the right side of the colored circles represent the genetic drift of each gene pool from the ancestral population (Falush et al 2003)

Materials and methods we selected six preserved populations and one non-preserved population (Fig. 1; Table 1). Four sampled popu- Study populations and sampling design lations were located in the Kii Peninsula region (K-SAN, K-OMA, K-KAW, and K-OTO), and three populations There are seven sites of P. japonica, which is preserved occurred in the Shikoku Island region (S-SEN, S-YAS, by national or local governments, that include most of the and S-NIS; Fig. 1). The distribution range of P. japonica remnant natural stands of this species. To sample popula- is now much smaller than previously published (Hayashi tions that cover as wide a distribution range as possible, 1960) due to extensive artificial plantation of the conifer

1 3 Author's personal copy

Conservation Genetics

Table 1 Description of the sampled populations of Pseudotsuga japonica, including the location, size of the preservation area, start year of preservation, preservation area of the sampling plot, population size and sample size Region Population Code Latitude, longitude Preserva- Start year of Area of Population size Sample size tion area preservation sampling plot (ha) (ha)

Kii Peninsula Sannokou K-SAN 34°15′N, 136°4′E 22.0 1929 0.70 704 66 Oomata K-OMA 34°1′N, 136°5′E 7.1 1993 0.77 266 90 Kawamatakannon K-KAW 33°54′N, 135°22′E 3.9 1976 0.42 556 101 Ootousan K-OTO 33°45′N, 135°41′E – – 0.19 250 55 Shikoku Island Senbonnyama S-SEN 33°38′N, 134°5′E 16.0 1968 0.38 288 30 Yasudagouyama S-YAS 33°37′N, 134°2′E 4.3 1973 0.09 274 59 Nishinokou S-NIS 33°36′N, 133°57′E 7.9 1916 0.20 168 50 species Cryptomeria japonica and Chamaecyparis obtusa DNA extraction and microsatellite analysis in southwest Japan during the latter half of the twentieth century (Mori and Kumazaki 1990). We conducted a field Total DNA was extracted using a modified cetyltrimethyl- survey in the Shikoku Island region and a database search to ammonium bromide method (Shiraishi and Watanabe 1995). identify remnant stands in the relatively abundant and less We first tested whether 72 microsatellite markers devel- abundant areas shown on the distribution map of Hayashi oped for the congener species P. menziesii (Amarasinghe (1960). However, we could not find any remnant stands of P. and Carlson 2002; Slavov et al. 2004) could be used to test japonica, except in the northern edge area of the distribution P. japonica DNA. Considering marker polymorphism and range. Accordingly, we had no choice but to select a sample stability, we selected eight markers for the microsatellite population from the northern edge of the distribution range analysis: BCPsmAG12, BCPsmAG14, BCPsmAG23, BCPs- in this region. Population size was estimated in different mAG37, PmOSU_2D4, PmOSU_4G2, PmOSU_3H4, and ways, depending on the size of the preservation area. Popu- PmOSU_4E9. The PCR protocol was as follows: 95 °C for lations covering less than 5 ha (including the non-preserved 15 min, followed by 30 cycles of 94 °C for 30 s, X °C for population, K-OTO) were estimated by direct counting in 90 s, and 72 °C for 90 s, and then 72 °C for 10 min, where X the field survey, and those occupying more than 5 ha were was 53 °C for PmOSU_4G2, PmOSU_3H4, PmOSU_4E9, estimated using published census data. Since census data BCPsmAG12, and BCPsmAG14, and 57 °C for the remain- encompassing the total preserved area were available only ing markers. The PCR products were sequenced on an ABI for the S-SEN and S-NIS populations, the sizes of the other PRISM model 3100 (Applied Biosystems). Genotypes were populations were estimated by extrapolating the number determined using the Genotyper 3.7 software (Applied Bio- of individuals in the sampling plot to the total preserva- systems). The frequency of null alleles was estimated with tion area. The spatial distance between the populations was the second method of Brookfield (1996), as implemented 5–208 km, with an average of 113 km. Sampling was car- in the Micro-Checker 2.2.3 software (van Oosterhout et al. ried out between 2006 and 2009. Since P. japonica is rare, 2004). it was not feasible to maintain a consistent distance between sampled individuals. Therefore, one quadrat was established Data analysis within the most densely distributed area for each population, and all trees located within that quadrat were sampled. Population genetic structure However, because the distribution of P. japonica is strongly topography-dependent (i.e., ridge condition), in some quad- The level of polymorphism at each microsatellite locus rats, the number of individuals was insufficient. Therefore, was evaluated by calculating the following: the number of multiple quadrats were established within the K-SAN (three alleles (A); the observed heterozygosity (HO); the expected quadrats), K-OTO (two quadrats), and S-SEN populations heterozygosity within a population (HS); the expected het- (two quadrats). As the population size and density varied erozygosity in the total population (HT); and the inbreed- among the populations, the established quadrat sizes var- ing coefficient (fixation index (FIS) = 1 − HO/HE). Departure ied from 0.09 ha to 0.77 ha (Table 1). The locations of the from Hardy–Weinberg equilibrium at each locus and linkage sampled trees were determined by a compass survey. Leaf disequilibrium between pairs of loci within each popula- samples were collected from 451 individuals (30–101 per tion were tested by randomization tests with 1000 permu- population; Table 1) and were stored at − 20 °C until DNA tations followed by Bonferroni correction to correct for extraction. multiple testing. Statistical analyses and permutation tests

1 3 Author's personal copy

Conservation Genetics were conducted with FSTAT ver. 2.9.3 software (Goudet results and the results with all individuals revealed very sim- 1995). Significant linkage disequilibrium between pairs of ilar data (HE: r = 0.97, AR: r = 0.98). Therefore, we decided loci within populations was detected in 11 out of 196 tests to use all individuals for the subsequent analysis. (α = 0.05). All the detected pairs were combinations of only To examine genetic differentiation among populations, three loci, BCPsmAG12, BCPsmAG14, and BCPsmAG23, Wright’s FST values (Wright 1951) according to Weir and which indicates that these loci are linked. Since linked mark- Cockerham (1984) were calculated for each locus and ers can produce biased results in population genetics, we across all loci, using FSTAT. The significance of popu- eliminated two causal markers, BCPsmAG12 and BCPs- lation differentiation was tested by randomization tests mAG23, from further analyses. To assess the discrimina- with 1000 permutations. Since FST values are influenced tion power of the six remaining markers, the probability of by marker polymorphisms, we also calculated a standard- identity (PID) assuming random mating and (PIDsib) given ized measure of population differentiation, G′ST (Hedrick the similarity between siblings was calculated according to 2005). Analysis of molecular variance (AMOVA; Waits et al. (2001), using GenAlEx ver. 6.1 software (Peakall Excoffier et al. 1992) was performed with a two-factorial and Smouse 2006). design using Arlequin ver.3.0 (Excoffier et al. 2005). To To evaluate within-population genetic diversity, the fol- elucidate the degree of population differentiation within lowing parameters were calculated for each population using each region (i.e., the Kii Peninsula and Shikoku Island FSTAT: HO; expected heterozygosity (HE; Nei 1987); FIS regions), AMOVA was performed on two levels: among = 1 − HO/HE; and allelic richness (AR) calculated using populations within the regions, and within populations. rarefaction (Hurlbert 1971) and standardized to the mini- Then, an overall AMOVA was performed on three lev- mum sample size of 30 diploid individuals. Departures from els: among regions, among populations within regions, Hardy–Weinberg equilibrium were tested by randomization and within populations. The relationships among popula- tests with 1000 permutations followed by Bonferroni cor- tions were inferred by drawing a phylogenetic tree based rection. To evaluate allelic characteristics, the numbers of on Nei’s genetic distance (DA distance; Nei et al. 1983). private (unique) alleles (PA) and rare alleles (RA, defined The POPULATIONS 1.2.30 software (Langella 1999) as alleles with frequencies < 5% in their total populations) was used to calculate genetic distances and to construct a were calculated using the GenAlEx ver. 6.1 software. As neighbor-joining tree (Saitou and Nei 1987). To test the our intensive sampling approach may have resulted in the relative stability of groups within the phylogenetic tree, oversampling of related individuals, we conducted a simula- bootstrap analysis was performed with 1000 replicates, tion test comparing the genetic diversity calculated on the using the same software. To further examine the genetic basis of data for all individuals with that calculated on the structure of populations, a model-based Bayesian cluster- basis of data from randomly sampled individuals, as follows. ing approach was applied using STRUCTU RE ver. 2.3.4 Random sampling was conducted under the condition that (Pritchard et al. 2000) to detect underlying patterns of the pair-wise distance was not shorter than the distance class genetic structure. This program assigns each individual for which significant deviation from random spatial genetic to a putative ancestral population based on the assump- structure (SGS) was observed in each population. To test tion that each putative ancestral population conforms to for the significance of SGS, the kinship coefficient (Loiselle Hardy–Weinberg equilibrium and linkage equilibrium. et al. 1995) for each distance class was calculated using the As ancestry models, we selected the admixture model software SPAGeDi 1.2 (Hardy and Vekemans 2002). As a and the correlated allele frequency model. These models result, five out of seven populations showed a significant assume that all individuals have mixed ancestry and that positive deviation within a certain distance class, which allele frequencies in ancestral subpopulations correlate. ranged from 15.9 m for the S-NIS population to 27.3 m for No a priori population information was used. Analysis the K-OMA population. Next, we randomly sampled 350 with the correlated allele frequency model can estimate individuals with replacement from each population and we parameters (FST) that represent the amount of genetic examined whether these individuals had more than one pair drift undergone for each lineage from a putative ancestral of individuals located closer than the distance at which sig- population (Falush et al. 2003). Lower FST values indicate nificant SGS was detected. Through the process of eliminat- less divergence among inferred gene pools and greater ing those individuals, we obtained the unit of individuals shared genetic frequency with a common ancestor. We with no genetic relatedness. This process was repeated 20 performed 20 independent runs for each K (ranging from times, and the average number of individuals per unit ranged 1 to 10), which included a burn-in period of 250,000 steps from 10.4 for the S-NIS population to 16.5 for the K-SAN and 750,000 Markov chain Monte Carlo steps, to obtain population. Genetic diversity statistics were then calculated parameter estimates. We used the software CLUMPP from the genotyped data of each unit of sampled individuals. ver. 1.1.2 (Jakobsson and Rosenberg 2007) to calculate A correlation analysis of the mean value of the simulation the average matrix of the membership proportion of

1 3 Author's personal copy

Conservation Genetics independent runs (using the Greedy algorithm with 1000 Results random input orders). To detect the true value of K, ΔK (Evanno et al. 2005) was calculated using log likelihood Genetic variation and demographic history values for each run. To examine the genetic relationships among inferred gene pools, a neighbor-joining tree was Seventy-six alleles were discovered through analysis using constructed based on the allele frequency divergent esti- the six microsatellite markers. The number of alleles per mates using the NEIGHBOR components of the PHYLIP locus ranged from 3 to 33, with an average of 12.7 (Table 2). software ver. 3.69 (Felsenstein 2010). The average expected heterozygosity within populations (HS) was moderate (mean = 0.568), ranging from 0.416 for PmOSU_4E9 to 0.865 for PmOSU_2D4. As significant Demographic history heterozygosity excess was observed for three out of the six markers, the possibility of balancing selection was sus- The effective population size (Ne) was estimated from the pected. Thus, we conducted a Ewens–Watterson neutrality genetic data for the seven populations using the linkage test. The results revealed no deviation from neutral allele disequilibrium method (Hill 1981) as implemented in the distributions within populations (data not shown), support- software NeEstimator ver. 2.01 (Do et al. 2014). To avoid ing the appropriateness of these markers for the population bias in estimating Ne, the lowest allele frequency permit- genetic study. The estimated null allele frequencies did not ted was established at 0.02 following the recommendation exceed 0.07, according to Micro-Checker analysis. The of Waples and Do (2010) as an appropriate critical value probability of identity using the six markers combined was −6 for our sample size. Recent reduction in effective popula- 7.1 × 10 for unrelated individuals (P ID) and 0.011 for full tion size was assessed using BOTTLENECK ver. 1.2.02 sibling individuals (PIDsib). (Piry et al. 1999). Following the recommendations of Piry Measures of genetic diversity generally did not greatly et al. (1999), we used a two-phased model assuming 95% vary among populations (Table 3). The expected mean het- single-step mutation and a variance of 12 among multi- erozygosity (HE) ranged from 0.472 for the S-SEN popula- ple steps. The significance of heterozygosity excess was tion to 0.655 for the K-OTO population (Table 3). Allelic evaluated using a one-sided Wilcoxon’s signed-rank test. richness (AR) averaged across loci ranged from 4.669 for Finally, to explore the effects of census and effec- the K-OMA population to 6.263 for the K-SAN population. tive population sizes on the genetic diversity statistics, Although the number of rare alleles (RA) showed a trend namely, AR, average HE, and the number of RA, a correla- similar to that of the allelic richness, the difference was more tion analysis was conducted with R 3.3.2 (R Core Team pronounced, and the greatest number (18: K-SAN popula- 2016). Prior to analysis, the census and effective population) was two times higher than the lowest number (9: S-NIS tion sizes were log-transformed. population). The number of private alleles (AP) ranged from 2 for the K-OMA population to 7 for the K-SAN population.

Table 2 Characteristics of the Locus A HO HS HT FIS FST PID PIDsib Null allele six microsatellite loci used in frequency this study BCPsmAG14 15 0.708 0.701 0.752 − 0.010** 0.068*** 0.098 0.405 0.051 BCPsmAG37 12 0.661 0.633 0.721 − 0.044** 0.122*** 0.109 0.416 0.046 PmOSU_2D4 33 0.889 0.865 0.912 − 0.028* 0.051*** 0.016 0.300 0.029 PmOSU_4G2 7 0.355 0.365 0.418 0.026** 0.128*** 0.335 0.612 0.068 PmOSU_3H4 6 0.437 0.430 0.465 − 0.014 0.074*** 0.342 0.597 0.046 PmOSU_4E9 3 0.396 0.416 0.489 0.049** 0.148*** 0.387 0.612 0.057 Meana 12.7 0.574 0.568 0.626 − 0.004** 0.099*** 7.4 10−6 0.011 0.051

A total number of alleles observed, HO observed heterozygosity, HS expected heterozygosity within populations, heterozygosity in the total populations, FIS fixation indices, FST coefficient of genetic differentiation among populations, PID probability of identity, PIDsib probability of identity among siblings. Significant departures from Hardy–Weinberg equilibrium and genetic differentiation were tested by randomization tests with 1000 permutations (*p < 0.05; **p < 0.01; ***p < 0.001). The null allele frequency was estimated using the second method of Brookfield (1996) a In the columns of PID and P IDsib, the values represent cumulative probabilities, which are the products of the PID or P IDsib of individual loci

1 3 Author's personal copy

Conservation Genetics

Table 3 Measurements of a Region Population Code N A AP RA AR HO HE FIS genetic diversity based on six SSR loci in seven populations Kii Peninsula Sannokou K-SAN 66 7.333 7 18 6.263 0.657 0.647 − 0.011 of Pseudotsuga japonica Oomata K-OMA 90 5.333 2 12 4.669 0.547 0.538 − 0.017 Kawamatakannon K-KAW 101 6.833 6 15 5.785 0.564 0.566 − 0.004 Ootousan K-OTO 55 6.667 4 14 5.939 0.647 0.655 − 0.012 Shikoku Island Senbonnyama S-SEN 30 5.167 3 10 5.167 0.467 0.472 − 0.011 Yasudagawayama S-YAS 59 5.333 3 10 4.791 0.558 0.548 − 0.020 Nishinokou S-NIS 50 5.333 4 9 5.110 0.582 0.552 − 0.054

N number of samples, A average number of alleles per locus, AP number of private alleles, RA number of rare alleles, AR allelic richness, HO average observed heterozygosity, HE average expected heterozygosity, FIS fixation indices a Rare alleles correspond to a frequency < 0.05

Table 4 Bottleneck test results using Wilcoxon’s signed-rank test, Table 5 Analysis of molecular variance (AMOVA) estimates of the effective population size (Ne) based on the linkage disequilibrium method and the ratio of the effective population size Source of variation d.f. Sum of squares Percentage and census population size (Ne/Nc) of variation Code Wilcoxon test Ne (CI 95%) Ne/Nc Kii Peninsula region Among populations within 3 71.02 7.71*** K-SAN 0.422 376.4 (63.7–∞) 0.535 regions K-OMA 0.781 30.1 (22.8–40.3) 0.113 Within populations 620 1066.63 92.2 K-KAW 0.961 66.8 (41–127.3) 0.120 Shikoku Island region K-OTO 0.344 35.7 (17.8–98.9) 0.143 Among populations within 2 23.04 6.65*** S-SEN 0.945 18.6 (10.1–41.3) 0.065 regions S-YAS 0.719 25.7 (16.1–44.1) 0.094 Within populations 275 429.81 93.35 S-NIS 0.578 63.3 (36.8–149.6) 0.377 Overall Among regions 1 57.65 5.43* 95% confidence intervals of the estimates of effective population size Among populations within 5 94.24 7.04 *** are indicated in parentheses regions Within populations 895 1496.44 87.52 Estimates of the effective population size were highly ***p < 0.001; *p < 0.05 heterogeneous among populations (Table 4), with the highest value (376.4) noted for the K-SAN population and the lowest value (18.6) for the S-SEN population. The average were regarded as moderately differentiated. The three- value of the ratio of the effective population size and census level AMOVA revealed a significant differentiation among population size (Ne/Nc) was 0.207, ranging from 0.065 in regions (5.43%, p < 0.05); therefore, we concluded that the the S-SEN population to 0.535 in the K-SAN population. Kii Channel is a significant determinant of the genetic struc- BOTTLENECK analysis to examine departures from muta- ture of P. japonica. The neighbor-joining tree of the seven tion drift equilibrium did not reveal any distortion in the populations generally reflected the geographical proximity seven populations (p = 0.344–0.961, Table 4), suggesting (Fig. 2). Although the bootstrap support was weak, the tree that none of the populations had experienced a significant consisted mainly of two clusters, with each cluster consist- recent reduction in size, at least within the period encom- ing of populations from the same region. The most west- passing the theoretically detectable number of generations. ern population of the Kii Peninsula, K-KAW, was clustered separately. The shortest branch length and highest bootstrap Population genetic structure value (93%) were found in the K-SAN and K-OTO populations, indicating that these two populations are genetically The coefficient of genetic differentiationF ( ST) was 0.101, close. and the standardized measure of genetic differentiation, Using the Bayesian clustering approach in STRUCUR G′ST, was 0.233. At the regional level, the AMOVA showed E, two peaks for K, K = 2 and K = 4, were detected as the that the percentage of variation among populations was best estimates of the number of genetic groups (Fig. 3). At 7.71% in the Kii Peninsula region and 6.65% in the Shikoku K = 2, all Kii Peninsula populations, except the K-KAW Island region (Table 5). The populations in both regions population, were consistently assigned to gene pool E. The

1 3 Author's personal copy

Conservation Genetics

gene pool 4 (0.063). K-SAN and K-OTO were predominated by the lowest FST-value gene pool (gene pool 4; 0.063), suggesting that these two populations were less divergent from their ancestral population than the other populations.

Relationship between population size and genetic diversity

Fig. 2 Neighbor-joining tree based on the DA genetic distance (Nei Strong, positive relationships were consistently found et al. 1983) for seven populations of Pseudotsuga japonica. Bootstrap between effective population size and the three genetic values based on 1000 replications are represented as percentages diversity indices, although only the correlation with the number of rare alleles was significant (r = 0.761, p = 0.047, Fig. 4). The number of rare alleles also showed a strong A and significant correlation with the census population size (r = 0.862, p = 0.013). In contrast to the case of effective population size, the strength of the correlation with census population size varied among the two other indices; that is, the correlation coefficient for expected heterozygosity (r = 0.361) was less than half of that for the number of rare alleles (r = 0.862).

Discussion B Geographic pattern of genetic differentiation and diversity

FST (0.101) and G′ST (0.233) values for P. japonica in the current study were much higher than those estimated using SSR markers in the same-genus species, P. menziesii var. menziesii (θ = 0.003, θS = 0.044; Krutovsky et al. 2009), which is distributed along the Pacific coast of North Amer- ica from central British Columbia to central California. Lower FST values have also been reported in other conifers, Fig. 3 Estimates of the population structure for Pseudotsuga japon- such as Picea abies (FST = 0.029, F′ST = 0.071; Tollefsrud ica using the STRUCTU RE program. Posterior probability [Ln P a et al. 2009), Tsuga canadensis (FST = 0.077; Potter et al. K (D)] values represented by 20 replicates per cluster ( ). b Ad hoc sta- Chamaecyparis obtusa G tistics: ΔK, which is calculated from Ln P (D) values, according to 2012), ( ST = 0.040; Matsumoto Evanno et al. (2005) et al. 2010), Cryptomeria japonica (FST = 0.028; Takahashi et al. 2005), and Pinus densiflora (FST = 0.013, G′ST = 0.122; Iwaizumi et al. 2013). Low levels of population differentia- Shikoku populations were consistently assigned to gene pool tion are expected in conifers because outcrossing and wind W, while the K-KAW population showed a mixture of these pollination facilitate gene flow among populations (Hamrick two gene pools. At K = 4, K-SAN and K-OTO were similarly et al. 1992). This, in turn, homogenizes within-population assigned to each gene pool, confirming the genetic close- genetic components. The relatively high among-population ness among these two populations. In the Shikoku Island differentiation of P. japonica may be explained by the spe- region, S-YAS and S-NIS were mostly assigned to gene pool cies’ localized distribution pattern. Several studies have 2, whereas S-SEN consisted of gene pools 2 and 3, with revealed greater differentiation in species with a narrow similar ratios. Gene pool 3 was the primary composition geographical range than in widespread congeneric species of the K-KAW population, while gene pool 1 was highly (Gustafsson and Sjögren-Gulve 2002; Hamrick 2004; Khasa dominant in the K-OMA population. The FST values for the et al. 2006; Chung et al. 2014). Thus, the geographic iso- four gene pools were highly variable (Fig. 1), with that for lation of the P. japonica populations may have facilitated gene pool 1 (0.188) being three times higher than that for genetic differentiation among the populations.

1 3 Author's personal copy

Conservation Genetics

A was detected, which suggests that the Kii Channel forms a genetic barrier between the two regions. Although this barrier effect was largely supported by the two main clusters corresponding to each regional location in the neighbor- joining tree, incompleteness of the barrier effect was also suggested. The westernmost population in the Kii region, K-KAW, did not belong to either of the two clusters. A similar trend was also indicated by the STRUCTU RE analysis at K = 2, revealing a regional difference in predominant gene pools; however, admixture was observed in the K-KAW population. This evidence of admixture of linages could be explained by a distribution shift because of global climate oscillation. During the last glacial maximum, the Kii Peninsula and Shikoku Island were connected by low sea levels (Nasu 1981). During this period, the distribution of B P. japonica shifted toward southern and lower elevations according to palynological studies (Nakamura et al. 1972). As shown on the distribution map (Hayashi 1960, Fig. 1), the three sampled populations are biasedly located in the northern edge of a relatively large patch of P. japonica that existed at the time the map was published. Therefore, our sampled Shikoku populations may not provide a whole pic- ture of the original gene pool and genetic features in this region. In addition, we must acknowledge that, although the discriminant power of the six markers was supported by the reasonably low probabilities of identity, the low number of SSR markers used in this study might have limited the reliability of inferences of population genetics and structure. Koskinen et al. (2004) assessed the influence of the number C of microsatellites on the reliability of phylogenetic construc- tion and reported that the topological similarity index (using DA genetic distance) of twelve markers was 18% higher than that of six markers.

Different magnitudes of correlation of the census and effective population sizes with three genetic diversity indices

Although effective population size was significantly correlated only with the number of rare alleles, it was consistently correlated with the three genetic diversity indices used (Fig. 4). Similarly, high correlation coefficients were observed for census population size and allelic richness, and the number of rare alleles, whereas the correlation with the expected heterozygosity was weaker than the other two Fig. 4 Relationships of the census and effective population size with allelic diversity measures. This observed lack of consist- a average expected heterozygosity, b allelic richness and c number of ency in correlation magnitude would be relevant to the rare alleles rapid decline of the natural stand of P. japonica over the last several decades (Yamanaka 1975) and its long generation time, which is typical for conifer species. Theoretical Although the partitioned regional variation (5.43%, studies examining the bottleneck effect and genetic variabil- AMOVA) was the lowest of the three hierarchical factors, ity have reported that allelic diversity declines faster than significant genetic differentiation between the two regions average heterozygosity during a bottleneck (Nei et al. 1975),

1 3 Author's personal copy

Conservation Genetics and rare alleles are rapidly lost compared with intermedi- populations that differed in size by five orders of magnitude. ate frequency alleles (Fuerst and Maruyama 1986). Based In contrast, the largest population in the present study was on this theory, the speed of response to a bottleneck event only 4.2 times larger than the smallest population. In addi- (i.e., the rate of decline per generation) in the three diversity tion, the cited studies sampled small populations comprising indices tested in our study should be ranked in the order of fewer than 50 individuals, while the smallest population in the number of rare alleles, allelic richness, and expected our study was 168. Our narrow range of sampled population heterozygosity (i.e., the order of the magnitude of correla- size and lack of smaller populations, which are more prone tion coefficients with the census population size), suggesting to fragmentation (Cruzan 2001), seem to have been disad- that the observed pattern would be explained by differences vantageous for detecting significant associations between in sensitivity to the bottleneck event in the three indices. population size and major genetic diversity indices. Thus, a stronger correlation with more sensitive indices, such as the number of rare alleles, is reasonable because Implications for conservation sensitive (i.e., fast response) indices are expected to rapidly reestablish a relation to the changed population size, even We observed different magnitudes of correlation between in the case of a limited number of generations following the census population size and the three genetic diversity indi- population decline. ces, corresponding with their sensitivity to the bottleneck This scenario is particularly well fitted to the K-OTO event. This result seems to represent a signature of the begin- population. Although BOTTLENECK analysis did not ning of the adverse effect of the human-induced recent popu- reveal a significant signature of fluctuation in population lation decline on genetic diversity (i.e., loss of rare alleles) in size in any population, the lowest p-value was detected in each population, rather than the natural population history. this population (Table 4). Similar to the K-SAN population, Thus, maintenance of population size should be prioritized the K-OTO population is located in a large distribution area for the conservation of P. japonica. The prominently highest and consists mainly of a gene pool with a low FST value estimates of effective population size and allelic diversity (FST = 0.063) according to STRUCTURE analysis. This were found in the K-SAN population, which is located in the suggests that a relatively large effective population size has largest preservation area and is the second oldest (1929) of been historically maintained in this population (Faulsh et al. the seven populations. This population could be regarded as 2003). However, actual estimates of the effective population a representative natural stand of P. japonica and is expected size by the LD method revealed that the value of K-OTO to contain a large portion of the genetic diversity within this (35.7) was far smaller than that of K-SAN (376.4, Table 4). species. In contrast, several populations showed lower esti- Based on our field survey, the K-OTO population seemed mates of effective population size, even lower than the crite- to be derived from a relatively small number of founders or rion value of Ne = 50, which is sufficient to prevent inbreed- parents, as the stand area of K-OTO was narrow, isolated ing on short term (Franklin 1980). To maintain, or preferably by artificial plantations, and consisted of small trees with to increase, the effective population size, transplantation a diameter of < 50 cm at breast height (data not shown). of seedlings originating from different population sources This evidence suggests that the K-OTO population rep- would not be recommended, considering the occurrence of resents a typical case of recent population size reduction. significant differentiation even among populations separated Another noteworthy feature of this population is that it had by less than 15 km (AMOVA results for the Shikoku popula- the highest expected heterozygosity (0.655, Table 3), which tions, Table 5). In the field survey, we often observed small, was disproportionate to the intermediate size of the census fragmented stands of P. japonica around the preserved P. and effective population sizes. This, together with the recent japonica populations, although these were surrounded and population decline, is consistent with a theoretical study that isolated by artificially planted trees of other species. Cut- showed that changes in expected heterozygosity following ting the surrounding artificial plantation area (restriction to a population reduction require many generations to become ridge topographic areas might be sufficient) and replanting apparent (Lowe et al. 2005). or reseeding of the cut areas with P. japonica originating The lack of significant associations of census population from locally preserved populations would benefit effective size with expected heterozygosity and allelic richness in population size maintenance by enhancing gene flow among this study is not in accordance with previous reports show- the fragmented populations. To predict the efficiency of this ing significant positive relationships between these features approach for increasing the effective population size, genetic (del Castillo et al. 2011; Ilves et al. 2013; Busch and Reisch analysis of the fragmented population is needed to acquire 2016). This inconsistency might be clarified by a focus on basic information to estimate future contributions when iso- the range of population sizes included in this study. The lation is eliminated. cited studies used a populations ranging in size by at least There is concern that the loss of rare alleles in smaller two orders of magnitude, and del Castillo et al. (2011) used populations, as suggested by correlation analysis, might

1 3 Author's personal copy

Conservation Genetics lead to a decline in the total genetic diversity in subsequent England PR, Luikart G, Waples RS (2010) Early detection of popu- generations. Furthermore, it might lead to a reduction in lation fragmentation using linkage disequilibrium estimation of effective population size. Conserv Genet 11:2425–2430 reproductive fitness, including seed viability and germina- Environment Agency of Japan (2000) Threatened wildlife of japan-red tion rate, in future generations. Thus, in future studies, fit- data book, vol 8 vascular plants. Japan Wildlife Research Center, ness traits should be evaluated to predict the long-term per- Tokyo (in Japanese) sistence of local populations of P. japonica. In addition, for Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clus- P. japonica ters of individuals using the software STRUCTURE: a simulation the long-term survival of , the adaptive potential study. Mol Ecol 14:2611–2620 of each population to global climate change should be evalu- Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular ated since a lack of suitable refugia has been predicted for variance inferred from metric distance among DNA haplotypes: P. japonica based on geographic restrictions on migrating application to human mitochondrial DNA restriction data. Genet- ics 131:479–491 north or to higher altitudes (Tanaka et al. 2009). Excoffier L, Laval G, Schneider S (2005) ARLEQUIN version 3.0: an integrated software package for population genetics data analysis. Acknowledgements We would like to thank the staff of the Kansai Evol Bioinform Online 1:47–50 Regional Breeding Office of the Forest Tree Breeding Center for col- Falush D, Stephens M, Pritchard JK (2003) Inference of population lecting leaf samples. We are particularly grateful to Koji Hashimoto structure using multilocus genotype data: linked loci and corre- for support with sampling at most sites. We also thank Dr. Naoyuki lated allele frequencies. Genetics 164:1567–1587 Nishimura (Gunma University) for advising us to collect leaf samples Farjon A (1990) Pinaceae: drawings and descriptions of the genera from a permanent plot managed by his research group. We are also Abies, Cedrus, Pseudolarix, Keteleeria, Nothotsuga, Tsuga, grateful to the education board of Kawakami village for permitting us Cathaya, Pseudotsuga, Larix and Picea. Koeltz Scientific Books, to collect sampling materials, and to Mayumi Shimoyama for DNA Konigstein extraction from all samples. Finally, we are grateful to the anony- Felsenstein J (2010) PHLIP (Phylogeny Inference Package), version mous reviewers for the insightful and constructive comments, which 3.69. Department of Genome Sciences, University of Washing- improved the manuscript greatly. ton, Seattle Frankham R (1995) Effective population size/adult population size ratios in wildlife: a review. Genet Res 66:95–107 Frankham R, Ballou JD, Briscoe DA (2010) Introduction to conserva- References tion genetics. Cambridge University Press, Cambridge Franklin IR (1980) Evolutionary change in small populations. In: Soulé Amarasinghe V, Carlson JE (2002) The development of microsatel- ME, Wilcox BA (eds) Conservation biology: an evolutionary–eco- lite markers for genetic analysis in Douglas-fir. Can J For Res logical perspective. Sinauer Associates, Sunderland, pp 135–149 32:1904–1915 Fuerst PA, Maruyama T (1986) Considerations on the conservation of Arroyo J, Aparicio A, Albaladejo RG, Muñoz J, Braza R (2008) alleles and of genic heterozygosity in small managed populations. Genetic structure and population differentiation of the Mediter- Zoo Biol 5:171–179 ranean pioneer spiny broom Calicotome villosa across the Strait Gao LZ (2005) Microsatellite variation within and among populations of Gibraltar. Biol J Linn Soc 93:39–51 of Oryza officinalis (Poaceae), an endangered wild rice from Billington HL (1991) Effect of population size on genetic variation in China. Mol Ecol 14:4287–4297 a dioecious conifer. Conserv Biol 5:115–119 Gernandt DS, Liston A (1999) Internal transcribed spacer region evolu- Brookfield J (1996) A simple new method for estimating null allele tion in Larix and Pseudotsuga (Pinaceae). Am J Bot 86:711–723 frequency from heterozygote deficiency. Mol Ecol 5:453–455 Goudet J (1995) FSTAT (Version 1.2): a computer program to calculate Busch V, Reisch C (2016) Population size and land use affect the F-statistics. J Hered 86:485–486 genetic variation and performance of the endangered plant spe- Gustafsson S, Sjögren-Gulve P (2002) Genetic diversity in the rare cies Dianthus seguieri ssp. glaber. Conserv Genet 17:425–436 orchid Gymnadenia odoratissima and a comparison with the more Chung MY, López-Pujol J, Chung MG (2014) Comparative genetic common congener, G. conopsea. Conserv Genet 3:225–234 structure between Sedum ussuriense and S. kamtschaticum (Cras- Hamrick JL (2004) Response of forest trees to global environmental sulaceae), two stonecrops co-occurring on rocky cliffs. Am J Bot changes. For Ecol Manage 197:323–335 101:946–956 Hamrick JL, Godt MJW, Sherman-Broyles SL (1992) Factors influ- Chybicki IJ, Oleksa A, Kowalkowska K (2012) Variable rates of ran- encing levels of genetic diversity in woody plant species. New dom genetic drift in protected populations of English yew: impli- For 6:95–124 cations for gene pool conservation. Conserv Genet 13:899–911 Hardy OJ, Vekemans X (2002) SPAGeDi: a versatile computer pro- Cruzan MB (2001) Population size and fragmentation thresholds for gram to analyse spatial genetic structure at the individual or popu- the maintenance of genetic diversity in the herbaceous endemic lation levels. Mol Ecol Resour 2:618–620 Scutellaria montana (Lamiaceae). Evolution 55:1569–1580 Hayashi Y (1952) The natural distribution of important trees, indig- del Castillo RF, Trujillo-Argueta S, Sánchez-Vargas N, Newton AC enous to Japan: conifers reports 2. Bull For For Prod Res Inst (2011) Genetic factors associated with population size may 55:1–251 (in Japanese) increase extinction risks and decrease colonization potential in a Hayashi Y (1960) Taxonomical and phytogeographical study of Japa- keystone tropical pine. Evol Appl 4:574–588 nese conifers. Norin-Shuppan, Tokyo (in Japanese) Do C, Waples RS, Peel D, Macbeth GM, Tillett BJ, Ovenden JR (2014) Hedrick PW (2005) A standardized genetic differentiation measure. NeEstimator v2: re-implementation of software for the estimation Evolution 59:1633–1638 of contemporary effective population size (Ne) from genetic data. Hermann RK (1982) The genus Pseudotsuga: historical records and Mol Ecol Resour 14:209–214 nomenclature. Forest Research Laboratory, Special Publication Ellstrand NC, Elam DR (1993) Population genetic consequences of 2a Oregon State University, Corvallis small population size: implications for plant conservation. Annu Hill WG (1981) Estimation of effective population size from data on Rev Ecol Syst 24:217–242 linkage disequilibrium. Genet Res 38:209–216

1 3 Author's personal copy

Conservation Genetics

Hurlbert SH (1971) The nonconcept of species diversity: a critique Peakall ROD, Smouse PE (2006) GENALEX 6: genetic analysis in and alternative parameters. Ecology 52:577–586 excel: population genetic software for teaching and research. Ilves A, Lanno K, Sammul M, Tali K (2013) Genetic variability, Mol Ecol Resour 6:288–295 population size and reproduction potential in Ligularia sibirica Piry S, Luikart G, Cornuet JM (1999) BOTTLENECK: a computer (L.) populations in Estonia. Conserv Genet 14:661–669 program for detecting recent reductions in the effective popula- Iwaizumi MG, Tsuda Y, Ohtani M, Tsumura Y, Takahashi M (2013) tion size using allele frequency data. J Hered 90:502–503 Recent distribution changes affect geographic clines in genetic Potter KM, Jetton RM, Dvorak WS, Hipkins VD, Rhea R, Whit- diversity and structure of Pinus densiflora natural populations tier WA (2012) Widespread inbreeding and unexpected geo- in Japan. Forest Ecol Manage 304:407–416 graphic patterns of genetic variation in eastern hemlock (Tsuga Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching canadensis), an imperiled North American conifer. Conserv and permutation program for dealing with label switching and Genet 13:475–498 multimodality in analysis of population structure. Bioinformat- Pritchard JK, Stephens M, Donnelly P (2000) Inference of population ics 23:1801–1806 structure using multilocus genotype data. Genetics 155:945–959 Jaramillo-Correa JP, Grivet D, Terrab A, Kurt Y, De-Lucas AI, R Core Team (2016) R: a language and environment for statistical Wahid N, Vendramin GG, González-Martínez SC (2010) The computing. R Foundation for Statistical Computing, Vienna. Strait of Gibraltar as a major biogeographic barrier in Mediter- https://www.R-project.org/ ranean conifers: a comparative phylogeographic survey. Mol Reed DH, Frankham R (2003) Correlation between fitness and Ecol 19:5452–5468 genetic diversity. Conserv Biol 17:230–237 Khasa DP, Jaramillo-Correa JP, Jaquish B, Bousquet J (2006) Con- Saitou N, Nei M (1987) The neighbor-joining method: a new method trasting microsatellite variation between subalpine and western for reconstructing phylogenetic trees. Mol Biol Evol 4:406–425 larch, two closely related species with different distribution pat- Shiraishi S, Watanabe A (1995) Identification of chloroplast genome terns. Mol Ecol 15:3907–3918 between Pinus densiflora Sieb. et Zucc. and Pinus thunbergii Koskinen MT, Hirvonen H, Landry PA, Primmer CR (2004) The Parl. based on the polymorphism in rbcL gene. J Jpn For Soc benefits of increasing the number of microsatellites utilized in 77:429–436 (in Japanese with English summary) genetic population studies: an empirical perspective. Hereditas Slavov GT, Howe GT, Yakovlev I, Edwards KJ, Krutovskii KV, Tus- 141:61–67 kan GA, Carlson JE, Strauss SH, Adams WT (2004) Highly Krutovsky KV, St. Clair JB, Saich R, Hipkins VD, Neale DB (2009) variable SSR markers in Douglas-fir: mendelian inheritance and Estimation of population structure in coastal Douglas-fir [Pseu- map locations. Theor Appl Genet 108:873–880 dotsuga menziesii (Mirb.) Franco var. menziesii] using allozyme Strauss SH, Doerksen AH, Byrne JR (1990) Evolutionary relation- and microsatellite markers. Tree Genet Genomes 5:641–658 ships of Douglas-fir and its relatives (genus Pseudotsuga) from Langella O (1999) Populations 1.2.30: a population genetic software. DNA restriction fragment analysis. Can J Bot 68:1502–1510 CNRS UPR9034 Takahashi T, Tani N, Taira H, Tsumura Y (2005) Microsatellite Leimu R, Mutikainen PIA, Koricheva J, Fischer M (2006) How general markers reveal high allelic variation in natural populations of are positive relationships between plant population size, fitness Cryptomeria japonica near refugial areas of the last glacial and genetic variation? J Ecol 94:942–952 period. J Plant Res 118:83–90 Loiselle BA, Sork VL, Nason J, Graham C (1995) Spatial genetic struc- Tanaka N, Nakazono E, Tsuyama I, Matsui T (2009) Assessing ture of a tropical understory shrub, Psychotria officinalis (Rubi- impact of climate warming on potential habitats of ten conifer aceae). Am J Bot 82:1420–1425 species in Japan. Glob Environ Res 14:153–164 (in Japanese) Lowe AJ, Boshier D, Ward M, Bacles CFE, Navarro C (2005) Genetic Terrab A, Schönswetter P, Talavera S, Vela E, Stuessy TF (2008) resource impacts of habitat loss and degradation; reconciling Range-wide phylogeography of Juniperus thurifera L., a pre- empirical evidence and predicted theory for neotropical trees. sumptive keystone species of western Mediterranean vegetation Heredity 95:255–273 during cold stages of the Pleistocene. Mol Phyl Evol 48:94–102 Matsumoto A, Uchida K, Taguchi Y, Tani N, Tsumura Y (2010) Tollefsrud MM, Sønstebø JH, Brochmann C, Johnsen Ø, Skrøppa T, Genetic diversity and structure of natural fragmented Chamaecy- Vendramin GG (2009) Combined analysis of nuclear and mito- paris obtusa populations as revealed by microsatellite markers. J chondrial markers provide new insight into the genetic structure Plant Res 123:689–699 of North European Picea abies. Heredity 102:549–562 Mori I, Kumazaki M (1990) Sensasu ni Miru Nihon no Ringyou (Cen- van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) sus of Japanese forestry). Zenkoku-Nourin-Toukei-Kyoukai- MICRO-CHECKER: software for identifying and correct- Rengoukai, Tokyo (in Japanese) ing genotyping errors in microsatellite data. Mol Ecol Notes Nakamura J, Mitsushio H, Kuroda T, Yoshikawa O (1972) Palynostrati- 4:535–538 graphical study, part i the quaternary system in Kochi prefecture. Victory ER, Glaubitz JC, Rhodes OE Jr, Woeste KE (2006) Genetic Res Rep Kochi University 21:87–113 homogeneity in Juglans nigra (Juglandaceae) at nuclear micros- Nasu T (1981) Fauna and flora of the Japanese islands in the last atellites. Am J Bot 93:118–126 glacial time. Quat Res 20:191–205 (in Japanese with English Waits LP, Luikart G, Taberlet P (2001) Estimating the probability of summary) identity among genotypes in natural populations: cautions and Nei M (1987) Molecular evolutionary genetics. Columbia University guidelines. Mol Ecol 10:249–256 Press, New York Waples RS, Do C (2010) Linkage disequilibrium estimates of con- Nei M, Maruyama T, Chakraborty R (1975) The bottleneck effect and temporary Ne using highly variable genetic markers: a largely genetic variability in populations. Evolution 29:1–10 untapped resource for applied conservation and evolution. Evol Nei M, Tajima F, Tateno Y (1983) Accuracy of estimated phylogenetic Appl 3:244–262 trees from molecular-data II: gene-frequency data. J Mol Evol Wei XX, Yang ZY, Li Y, Wang XQ (2010) Molecular phylogeny and 19:153–170 biogeography of Pseudotsuga (Pinaceae): insights into the floristic Palstra FP, Ruzzante DE (2008) Genetic estimates of contemporary relationship between Taiwan and its adjacent areas. Mol Phylo- effective population size: what can they tell us about the impor- genet Evol 55:776–785 tance of genetic stochasticity for wild population persistence? Mol Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis Ecol 17:3428–3447 of population-structure. Evolution 38:1358–1370

1 3 Author's personal copy

Conservation Genetics

Willi Y, Van Buskirk J, Hoffmann AA (2006) Limits to the adap- Pseudotsuga japonica forest at the Sannoko on the Kii Peninsula, tive potential of small populations. Annu Rev Ecol Evol Syst southwestern Japan. Jpn J For Environ 34:50–58 37:433–458 Yamanaka T (1975) Ecology of Pseudotsuga japonica and other conif- Wright S (1938) Size of population and breeding structure in relation erous forests in eastern Shikoku. Mem Natl Sci Mus 8:119–136 to evolution. Science 87:430–431 (in Japanese with English summary) Wright S (1951) The genetical structure of populations. Ann Eug Young AG, Merriam HG, Warwick SI (1993) The effects of forest frag- 15:323–354 mentation on genetic variation in Acer saccharum Marsh (sugar Yamamoto S (1992) Preliminary studies on the species composition, maple) populations. Heredity 71:277–289 stand structure and regeneration characteristics of an old-growth

1 3