Chloroplast population genetics reveals low levels of genetic variation and conformation to the central–marginal hypothesis in Taxus wallichiana var. mairei, an endangered conifer endemic to China

Abstract The central–marginal hypothesis predicts that geographically peripheral populations should exhibit reduced genetic diversity and increased genetic differentiation than central populations due to smaller effective population size and stronger geographical isolation. We evaluated these predictions in the endangered conifer Taxus wallichiana var. mairei. Eight plastid simple sequence repeats (cpSSRs) were used to investigate plastid genetic variation in 22 populations of Taxus wallichiana var. mairei, encompassing nearly its entire distribution range. Low levels of plastid genetic variation and differentiation were detected in the populations, and the findings were attributed to low mutation rates, small population sizes, habitat fragmentation and isolation, and effective pollen or seed dispersal. Hunan and Hubei were identified as major refugia based on the number of private haplotypes and species distribution modeling. Trends in plastid genetic diversity and genetic differentiation from central to peripheral populations supported the predictions of the central–marginal hypothesis. In scenarios wherein the future climate becomes warmer, we predict that some peripheral populations will disappear and southern and southeastern regions will become significantly less habitable. Factors that include the levels of precipitation during the driest month, annual precipitation level, and annual temperature range will be decisive in shaping the future distribution of these populations. This study provides a theoretical basis for the conservation of T. wallichiana var. mairei.

that geographically peripheral populations should exhibit reduced genetic diversity and increased genetic differentiation than central populations due to smaller effective population size and stronger geographical isolation (Eckert et al., 2008;Trumbo et al., 2016). If this is true, peripheral populations will have limited evolutionary potential to adapt to habitat conditions beyond current range limits.
However, ecological processes such as phenotypic plasticity and migration and their interplay with evolutionary changes may play equally important roles in the persistence of species in changing environments (Aitken, Yeaman, Holliday, Wang, & Curtis-McLane, 2008;Anderson, Panetta, & Mitchell-Olds, 2012).
It has been noted that climate change can affect the fate of peripheral populations by the following mechanisms: (a) occurrence of significant evolutionary changes; (b) an increase in the risk of extinction; (c) peripheral populations becoming the leading edge of migrations; (d) genetic novelty being created and perhaps reinforcing standing genetic variation; and (e) fluctuations in the extent and scale of local adaptations being generated (Fady et al., 2016;Marschalek & Berres, 2014). Pfeifer et al. (2009) pointed out that conservation of peripheral populations is potentially valuable. In this context, a test of the CMH with rare species is relevant to forecast range shifts under climate change as well as to evaluate the conservation concern of peripheral populations.
Empirical tests of the CMH have found support in a range of plant taxa Myking, Vakkari, & Skrøppa, 2009), but not others (Dixon, Herlihy, & Busch, 2013;Garner, Pearman, & Angelone, 2004;Munwes et al., 2010). Theoretically, it is quite challenging to disentangle the causes of a central-marginal structure of genetic diversity, because factors such as biological characters, persistent historical influences, effects of current demographic and ecological variables, population sampling strategies, and molecular markers applied for analysis may all exert impacts (Loveless & Hamrick, 1984;Schiemann, Tyler, & Widén, 2000;Wagner, Durka, & Hensen, 2011). In this study, we tested the genetic predictions of the CMH by using plastid simple sequence repeats (cpSSRs) to assess populations of the endangered conifer Taxus wallichiana var. mairei (Taxaceae).
Global warming may alter environments suitable for many species of conifer, forcing them to migrate north (Ettinger & HilleRisLambers, 2017;Quiroga, Premoli, & Kitzberger, 2018;Wang, Wang, Xia, & Su, 2016). Climate is known as the most important factor shaping the natural distribution of T. wallichiana var. mairei. As the plant prefers moist and shady habitats, its widespread distribution particularly makes it sensitive to climatic changes (Wu & Wen, 2017). Therefore, understanding the distribution of T. wallichiana var. mairei in terms of past, present, and future climate conditions would provide a good basis for conservation strategies. It would be valuable to identify the climatic variables that make the most significant impact on its distribution.
In this study, we used cpSSRs to investigate T. wallichiana var. mairei populations from across China. Our goals were (a) to reveal the level of plastid genetic diversity and genetic differentiation; (b) to test the central-marginal hypothesis by examining the plastid genetic variation between central and marginal populations; and (c) to define the ecologically suitable area for these populations and determine the decisive climatic factors limiting their distribution.

| Study species and study range
Taxus wallichiana var. mairei (Taxaceae) is a slow-growing tree that varies in height from 2.5 to 20 m (Fu et al., 1999). Its needle leaves with white stomatal bands are arranged in a spiral along the stem, and brightly colored arils that are attractive to birds facilitate seed dispersal. As a dioecious and wind-pollinated coniferous species, T. wallichiana var. mairei occurs as part of the understory of mixed forests located in mountains and valleys at an altitude of 1,000-1,500 m (Fu et al., 1999;Zheng & Fu, 1978). The plant is rich in the anticancer agent Taxol and can also be used to make high-quality redwood furniture (Fan, Tang, & Shu, 1996).

| Plant materials
A total of 339 individuals were sampled from 22 populations, which covered almost the entire range of T. wallichiana var. mairei's distribution across China ( Figure 1, Table 1). For each population, 4-20 individuals, located at least 20 m apart, were randomly selected. Fresh young leaves were dried immediately in silica gel for DNA extraction.

| DNA extraction
Genomic DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB) procedure (Su et al., 2005). The DNA was quantified using a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific) and stored at −20°C at a final concentration of 50 ng/μl.
Eight primer pairs that produced clear bands were selected for further analysis (Table 2). Of these, the markers cpSSR11, cpSSR18, and cpSSR20 were described by Vendramin et al. (1996).
PCR amplification was performed in a 25 μl final volume containing 1 × PCR buffer, 2.5 mM MgCl 2 , 0.2 mM of each dNTP, 0.2 μM of each primer, 1 U of Taq DNA polymerase, and 50 ng of template DNA. The PCR conditions were initial denaturation at 94°C for 5 min; followed by 25 cycles of 94°C for 60 s, 55°C for 70 s, and 72°C for 70 s; then a final extension step at 72°C for 8 min. Amplification products were separated using a 6% (w/v) denaturing polyacrylamide gel that contained 7 M urea and 1 × TBE. The gel was run at a constant voltage of 200 V for 120 min with a 20-bp DNA ladder and silver-stained (Tixier, Sourdille, Röder, Leroy, & Bernard, 1997).

| Data analysis
Seven genetic parameters were calculated using FSTAT v. 2.9.3 (Goudet, 2001). These were the percentage of polymorphic bands The mean gene diversity within populations (H S ) and total gene diversity (H T ) were also estimated.
Analysis of molecular variance (AMOVA) was conducted to assess population differentiation within and among populations with 1,000 random permutation replicates using Arlequin software (ver. 3.0; Excoffier, Laval, & Schneider, 2005;Excoffier & Smouse, 1994). The values of pairwise fixation index (F ST ) among populations were obtained by the same software. Because of the lack of recombination in the plastid genome, alleles identified at polymorphic loci were combined to generate the plastid haplotype for each individual (Wang, Guo, & Zhao, 2011). Using Arlequin, the haplotype diversity (H d

| Comparison of genetic diversity between the central and marginal populations
The central and marginal T. wallichiana var. mairei populations were defined according to their geographical locations, based on our sampling and herbarium information (Table 1). We divided the range into two equal areas, peripheral and central (Channell & Lomolino, 2000).  climate scenarios (IPCC-CMIP5 RCP 4.5/8.5) was selected to ensure predictive accuracy. The difference between these scenarios is the concentration of greenhouse gases. The RCP 8.5 scenario includes a higher concentration of emissions than the RCP 4.5 scenario. Model performance was evaluated using the area under the (receiver operating characteristic) curve (AUC) calculated by MAXENT. Values greater than 0.9 indicate good discrimination (Swets, 1988). The key parameters were set as follows: random test percentage = 25, maximum iterations = 500, and replicates = 15. Ultimately, the importance of environmental variables was determined by a jackknife test during modeling.

| CpSSR variation, genetic diversity, and population structure
The eight plastid SSRs generated 176 bands from 339 individuals taken from the 22 populations surveyed. In total, 66 of these bands (37.50%) TA B L E 2 Summary of the eight cpSSR loci used to study the population genetics among 22 populations of The hierarchical AMOVA revealed that 61.01% of the total variation in cpSSRs was attributable to differences among individuals within populations, 16.31% to differences among populations within regions, and 22.68% to differences among regions (Table 3).
Genetic differences among and within populations were highly significant (G ST = 0.3592, p < .01) ( Table 2)

| Genetic comparison of the central and peripheral populations
We

| Prediction of distribution ranges
The    Table 4). The cumulative contribution of these five factors was 98%. As shown in Figure 5, the response curve indicated that T. wallichiana var.
mairei was most likely to be present when the precipitation levels during the driest month (Bio 14) were 50-150 mm, the annual precipitation level (Bio 12) was 1,250-1,800 mm, and the annual temperature range (Bio7) was 25-35°C.

| D ISCUSS I ON
In this study, we used cpSSRs to investigate genetic diversity and population genetic structure in 22 populations of T. wallichiana var.
mairei. Overall, T. wallichiana var. mairei had low levels of genetic variation although its PPB was slightly higher than other conifers ( Table 5). The other genetic parameters such as the observed num- 3.2 × 10 -5 to 7.9 × 10 -5 .
We found that the genetic diversity of T. wallichiana var. mairei based on cpSSR markers was significantly lower than previously estimated using nuclear SSR and ISSR markers (Zhang & Zhou, 2013;Zhang et al., 2009). Similar observations have been reported for other species . One reason for this may be the lower mutation rate of cpSSRs compared with nuclear markers (Provan et al., 1999). Nonetheless, the small population sizes may have an effect as well. The QL, CQ, and AH populations have only 4, 8, and 11 individuals, respectively. Such small sizes could cause the reduction of genetic diversity due to genetic drift.
Taxus wallichiana var. mairei has relatively low genetic differentiation compared with other conifers, such as Pinus tabuliformis, Pinus henryi, and Taiwania cryptomerioides ( Table 5). The UPGMA tree also showed that there was a considerable admixture among its populations ( Figure 2). The cpSSR-based AMOVA analysis indicated that most T. wallichiana var. mairei genetic differentiation occurred within populations, which is consistent with the nuclear SSR data but differs from the ISSR data (Zhang & Zhou, 2013;Zhang et al., 2009 (Templeton & Levin, 1979;Xu, Tremblay, Bergeron, Paul, & Chen, 2012). This involves using such traits as effective pollen or seed dispersal to elevate the level of gene flow and reduce population differentiation (Hamrick, Godt, & Sherman-Broyles, 1992;Heuertz et al., 2004;Xu et al., 2012). However, we would not suggest that the low genetic differentiation revealed here implies there exists efficient gene flow between the extant populations of T. wallichiana var.
mairei. Instead, it may reflect a time-lag effect of historical influences, for example, the plant was distributed more contiguously before the Quaternary glaciation (Yue, Chen, Guo, & Wang, 2011).
It is generally believed that refugial populations have high levels of genetic diversity and a large number of private haplotypes (Bhagwat & Willis, 2008;Comes & Kadereit, 1998;Stewart, Lister, Barnes, & Dalén, 2010). The large number of cpSSR private haplotypes and greater genetic diversity of T. wallichiana var. mairei detected in the populations from Hunan and Hubei provinces suggest that these regions probably serve as major refugia. Gao et al. (2007) previously argued that central China, including Hunan and Hubei, provided important refugia for T. wallichiana populations with ancient haplotypes, and this suggestion is supported by our MAXENT results. A spatial expansion from a stable population was also detected by mismatch analysis. In addition, our MAXENT analysis showed that T. wallichiana var. mairei had experienced two expansions into potentially suitable habitats: one between the LIG and the LGM and the other between the MH and the present. Moreover, the star-like shape of the minimum spanning network further suggests that T. wallichiana var. mairei experienced a relatively recent range expansion ( Figure S1). As a temperate plant, T. wallichiana var. mairei may not be directly affected by periods of glaciation. Its refugia in central China represent distribution centers from which peripheral populations may colonize previously glaciated regions Hewitt, 1999;Lesica & Allendorf, 1995).
We found that genetic diversity in T. wallichiana var. mairei decreased from the core to the peripheral populations for all genetic parameters except two. We also found that levels of genetic differentiation were greater in the peripheral populations. These findings from cpSSR data support the predictions of the CMH. After reviewing 134 nuclear marker-based studies, Eckert et al. (2008) concluded that 64.2% of them detected reduced genetic diversity and 70.2% found increased genetic differentiation between peripheral and central populations. However, plastid genome data were not included. The central regions, which included the proposed refugia of Hunan and Hubei provinces, represented the distribution core and were defined as the geographical center (Micheletti & Storfer, 2015). In particular, the TZ population was identified as the most centrally located population due to its high genetic diversity and large number of private alleles in comparison with other populations (Figure 1). Central populations may have a significant influence on the migration of plant populations, and the shape of the minimum spanning network suggests an asymmetric migration pattern ( Figure S1). The causes of reduced cpSSR diversity and increased differentiation from the central to the peripheral T. wallichiana var. mairei populations may include increased isolation, reduced effective population size and a corresponding increase in genetic drift, and historical demographic events (Christiansen & Reyer, 2011;Pfeifer et al., 2009). Importantly, strong adaptation to local conditions may also have a critical effect (Cahill & Levinton, 2016).
The factors that govern the range limits of T. wallichiana var.
mairei should be investigated as a priority, especially in the context of global warming (Trumbo et al., 2016;Ursenbacher et al., 2015).
Ranges are not simply limited by physiological tolerances or obvious dispersal barriers. Instead, they may be defined by the inability of peripheral populations to adapt to the prevailing conditions beyond their current range (Micheletti & Storfer, 2015 (Wu & Wen, 2017;Zhang et al., 2009). The mean monthly temperature range (Bio 2) and the mean temperature of the wettest quarter (Bio 8) are also important factors.
Future projections reveal that a greater reduction in areas suitable for T. wallichiana var. mairei habitation would occur under the RCP 8.5 compared with the RCP 4.5 scenario, which suggests that this species is sensitive to a high concentration of greenhouse gases. As the future temperature is predicted to increase 2.3-2.7°C by 2070 (Solomon et al., 2007), we speculate that global warming will have a significant impact on the distribution of T. wallichiana var. mairei. A substantial loss of area suitable for habitation will occur and the plant will shift northward. Our data may provide a starting point from which to devise a conservation strategy for T. wallichiana var. mairei in the face of climate change.

| CON CLUS IONS
We examined the Taxus wallichiana var. mairei plastid genome and found it had low levels of genetic variation and genetic differentiation. The cpSSR-based genetic data complement previously published results based on nuclear SSRs and ISSRs. Our results support the CMH, which provides a theoretical basis for the conservation and management of peripheral populations. The suitable range of T. wallichiana var. mairei was predicted to decrease and shift northward in response to global warming. In addition, the most important climatic factors that limit the distribution of T. wallichiana var. mairei were identified. These included the levels of precipitation during the driest month (Bio14), annual precipitation level (Bio12), and annual temperature range (Bio7). It would be interesting in a future study to further compare phenotypic plasticity and epigenetic variation between the central and marginal populations of T. wallichiana var. mairei in the climate change context (Anderson et al., 2012).

ACK N OWLED G M ENTS
The authors thank Binghui Wei, Mingzhao Yang, Zhanming Ying, Zhenyu Li, and Junsheng Yu of the School of Life Sciences, Sun Yatsen University for assistance in collecting plant materials.

CO N FLI C T O F I NTE R E S T
The authors declare there are no competing interests.
Consent for publication: Not applicable.

AUTH O R CO NTR I B UTI O N S
TW designed the research and was involved in writing the manuscript; LL and ZW conducted data analysis and were involved in writing the manuscript and checking English grammar; LH performed the cpSSR experiment and conducted data analysis; and YS helped to supervise the research and was involved in writing the manuscript. All authors read and approved the final version of the manuscript.

E TH I C S A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
The acquisition of plant material used in this study complies with institutional, national, and international guidelines. No specific permits were required for the described field studies. No specific permissions were required for the locations/activities described in this study.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used for this study are available through Dryad at the time of publication (https ://doi.org/10.5061/dryad.c8c79h7).