Subtle East–West Phylogeographic Break of Asteropyrum (Ranunculaceae) in Subtropical China and Adjacent Areas

Abstract

East–west phylogeographic break is common among plant species in subtropical China. However, the estimation time of east–west phylogeographic break has always relied on inferences of calibrated phylogenies, and the contribution of environmental heterogeneity to population differentiation has largely been ignored. In this study, we estimated the divergence time of Asteropyrum populations through coalescent-based approaches based on DNA sequences of ten nuclear loci and evaluated the contribution of environmental heterogeneity to population differentiation. The results showed that there were two chloroplast clades and nuclear groups within Asteropyrum, displaying a subtle pattern of east–west differentiation. The divergence time of the two nuclear groups was dated to ~1.2 Ma, which is associated with climate changes during the Mid-Pleistocene transition. A genetic admixture event between the two genetic groups happened at ~0.46 Ma, resulting in several admixed populations. Isolation by environmental distance (IBE) explained the majority (46.32%) of population differentiation, but that isolation by geographic distance (IBD) only contributed 4.66%. The results of this study suggest that climate changes during the Pleistocene may be a major cause for the east–west phylogeographic break in subtropical China. However, the complex terrain and high environmental heterogeneity in the west of subtropical China (and adjacent regions such as the Hengduan Mountains and the Himalayan Moutains) caused by strong geological uplift may have profoundly shaped the population structure of plant species in subtropical China.

1. Introduction

In a milestone paper concerning plant molecular phylogeography in China about one decade ago, the authors encouraged Chinese phylogeographers to explicitly state the time scale to which they refer when interpreting their results with respect to refugial isolation or range expansions [1]. Afterwards, a large number of plant phylogeographic studies have been implemented in China following Qiu’s et al. (2011) recommendation, which has substantially enhanced our understandings of the evolution and biogeographic history of the world’s most diverse temperate flora since the late Cenozoic [2,3,4,5,6,7,8]. For example, using DNA sequences of four chloroplast spacer regions, Sun et al. (2014) found that the time to the most recent common ancestor of all haplotypes in Tetracentron sinense (Trochodendraceae) was 9.6 million years ago (Ma) [9]. The extant distribution of T. sinense is likely to have been shaped by both pre-Quaternary and Pleistocene climate changes. However, the time estimates in most studies have been based on calibrated phylogenies [2,4,5,6,7,9,10,11], which may reflect the divergence history of genes (particularly chloroplast genes) rather than divergence history of populations [12] and thus may compromise the conclusions that have been made. During the last decade, several coalescent-based approaches such as Approximate Bayesian Computation (ABC) [13] and Isolation with Migration (IM) [14] have been developed to infer the population divergence time based on multi-locus or genomic variation, providing powerful arsenals for deciphering population history in a diverse array of organisms.

Subtropical China is located in eastern China between the Qinling Mountains-Huai River line (at ~34° N) and the tropical South (≦22° N), and bordered by the Hengduan Mountains (~98° E) in the west and the coastline in the east [15]. This region has never been directly affected by extensive and unified ice sheets [16], and thus served as one of the most important refuge areas for numerous Cenozoic plant genera [17], constituting one of the most diverse temperate flora in the world. Owing to the uplift of Qianghai-Tibetan Plateau since ~40 Ma [18], the western part of subtropical China is characterized by diverse topography (high plateaus, large mountains and deep gorges) but the eastern part by hills and low plains (0–500 m a.s.l.) [19]. The two parts of subtropical China are controlled by different summer monsoon systems, i.e., the eastern part by the Pacific monsoon and the western part by the Indian monsoon, which are divided approximately along longitude 105° E [20,21]. The distinct terrain and climate in subtropical China, coupled with strong uplift of mountains during the late Cenozoic and climatic oscillations during the (late) Pleistocene, have profoundly shaped the biome of this region. For example, the zonal vegetation of subtropical China is dominated by humid evergreen broad-leaved forests (EBLF) in the east but by semi-moist EBLF in the west [22]. At the species level, many plant species exhibit a pattern of conspicuous east–west phylogeographic break (hereafter EWPB, e.g., Dysosma versipellis [23], Sophora davidii [2], Davidia involucrata [24], Cephalotaxus oliveri [25], and Liriodendron chinense [8,26], also the review in [27]), which has been ascribed to strong geological uplift of the Qinghai-Tibetan Plateau and adjacent regions [24], the development of contrast summer monsoon systems [2], or the cold and dry climate during the glacial periods [23]. However, the time estimates for the population divergence and subsequent population demographic history have been either lacking or dependent on calibrated phylogenies [2,23], which may bias the inferences in those studies.

Population differentiation can be a result of random genetic drift and/or natural selection [28]. The EWPB in subtropical China is generally thought to have resulted from allopatric isolation via genetic drift (coupled with restricted genetic exchange) across complex topography generated by tectonic movements and vegetation changes linked to the Plio-Pleistocene glacial/interglacial cycles [8,23,24]. Such an allopatric model of non-adaptive population differentiation essentially implies that mutation and random genetic drift, rather than habitat-mediated selection, will be the primary factors causing divergence of populations. However, this can be misleading because the substantial differences in their environments (such as different summer monsoon systems [2]) may reduce the rate of successful dispersal and gene flow (isolation by environmental distance, IBE [28,29]), having contributed to population differentiation in subtropical China.

In this study, we selected the genus Asteropyrum J. R. Drummond et Hutchinson in Ranunculaceae as a focal system. Asteropyrum is a small genus of perennial herbs widely distributed in subtropical China and also occurring in Bhutan and north Myanmar [30]. Originally, this genus contains two species: A. cavaleriei (Lévl. et Vant.) Drumm. et Hutch. and A. peltatum (Franch.) Drumm. et Hutch. [31]. By investigating the cytology, palynology, and leaf morphology, Yuan and Yang (2006) proposed that the two species should be treated as two subspecies [30]. Recently, we applied an integrative approach to delimit the species boundary of the genus and found that the two species cannot be differentiated in terms of leaf form, genetics and ecology and even the classification of two subspecies should be abandoned [32]. Asteropyrum always inhabits in humid and shady environments (personal observation) and plants with such ecological requirement may be sensitive to the variation of precipitation that is significant in subtropical China due to the differential influence of two summer monsoon systems [15]. Therefore, Asteropyrum may represent an ideal system to test the roles of historical events and environmental heterogeneity on population differentiation and demography of plant species in subtropical China. The main questions we would like to address in this study are as follows: (i) Does Asteropyrum display a pattern of EWPB? (ii) What are the historical factors related with the population divergence and demography? (iii) Are the geographical isolation and environmental heterogeneity responsible for the population differentiation?

2. Material and Methods

2.1. Population Sampling, DNA Amplification, and Sequencing

A total of 28 populations spanning the entire natural distribution of Asteropyrum were sampled in subtropical region of China and adjacent areas (Figure 1, Supplementary Table S1). One to 10 individuals were collected, spacing at least 10 m apart when the populations were large. All fresh leaves were immediately dried and stored in silica gel. All voucher specimens were deposited at Jiangxi Agricultural University.

Total genomic DNA was extracted from dried and healthy leaves using a modified cetyltrime-thylammonium bromide (CTAB) method [33]. Ten single-copy nuclear DNA (nDNA) loci developed from transcriptome sequences of Asteropyrum previously [32] and one chloroplast region (cpDNA, psbA-trnH) were amplified and sequenced (

Table 1. Information of ten nuclear loci and one chloroplast fragment used in this study.

Primers	Sequences(5′–3′)	N	L	S	π	θw	Nh	Hd	D	Rm
psbA- trnH	F:GTTATGCATGAACGTAATGCTC	323	219	13	0.011	0.009	10	0.849	0.491	1
	R:CGCGCATGGTGGATTCACAAATC
X14	F:GTTTCGGGTGTTCTTGTT	162	388	19	0.005	0.008	27	0.807	−0.837	4
	R:CATCTTCTTGGCTCGTAG
X17	F:AACCTTAGAACTCCGATTA	146	368	49	0.016	0.021	40	0.953	−0.747	7
	R:CCTGATGATACCCTACTTT
X23	F:TGAGCGGTTCCTTGTCCC	156	411	33	0.007	0.013	28	0.914	−1.316	1
	R:CAGCGTGATGTCTTACGATGC
X35	F:CCGCTTTGCCACAGATTA	162	253	14	0.009	0.009	15	0.820	−0.047	1
	R:TGCTTTACCAGCCGTTGA
X47	F:ACAACATCCCAATCAGCA	160	231	23	0.017	0.017	23	0.881	−0.267	0
	R:ACAACCCACAACACCAGA
X63	F:CGTCGCCCAGTAGTATCTT	159	389	23	0.004	0.009	16	0.807	−1.485	0
	R:ACATTCATCGTTCGCTTG
X82	F:ATGAGAAGGGCACCAAAT	159	308	17	0.007	0.009	18	0.867	−0.431	1
	R:AAGCAAACAGAGGGAAGC
X89	F:CGAGTAGCCTTTACGACG	157	432	39	0.010	0.014	35	0.927	−0.881	3
	R:GAGTGGAACCCAACCTTT
X125	F:CAGGTGCGGTCATAGTTG	134	206	26	0.008	0.020	23	0.794	−1.802 *	1
	R:CGAGTCGCCTTGAGTTTT
X130	F:GGGAAGCCGTAGACTCAC	160	275	21	0.006	0.012	21	0.780	−1.419	0
	R:CCCGACAAGGCATAGAAC

N, sample size; L, length in base pair; S, number of polymorphic sites; π, nucleotide diversity; θ_w, Watterson’s parameter; N_h, number of haplotype; H_d, haplotype diversity; D, Tajima’s D statistic; R_m, minimum number of recombinant event. The asterisk indicates statistical significances (p < 0.05).

). Polymerase chain reactions (PCRs) were performed in a 20 μL volume containing 10 μL 2 × Taq PCR MasterMix (Tiangen, Shanghai, China), 7 μL ddH₂O, 1 μL template DNA (c. 50–100 ng), and 1 μL each primer (5 μM). All reactions were carried out in a Bioer XP cycler (Bioer, Hangzhou, China) programmed for an initial step for 5 min at 94 °C, followed by 36 cycles of 94 °C for 50 s, 50–53 °C for 50 s, and 72 °C for 1 min or 1.5 min, with a final extension for 10 min at 72 °C. The PCR products were checked on 1% agarose gel and then were directly sequenced with the corresponding PCR primers in both directions commercially by Sangon Biotech Co., Ltd. (Shanghai, China). The haplotype phases of ten nuclear sequences were determined by applying the phase algorithm in DnaSP 5.10. Then, DNA sequences were edited with Sequencher 4.1.4 (Gene Codes Corporation, Ann Arbor, MI, USA), aligned by BioEdit 7.2 [34], and finally adjusted manually in MEGA v. 5.05 [35]. All obtained sequences were deposited into GenBank database with accession numbers (MV033551-MV033560, OL871580-OL871750, OL901373-OL901434).

2.2. Population Genetic Analyses

To measure the level of genetic variation, the number of segregating sites (S), nucleotide diversity (π), the number of haplotypes (N_h), and haplotype diversity (N_d) were calculated for cpDNA and each nuclear locus in DnaSP v5.10 [36]. For cpDNA dataset, a median-joining network was constructed to view intraspecific relationships of cpDNA haplotypes (chlorotypes) by coalescent simulations in Network 10.3 (https://www.fluxus-engineering.com/sharenet.htm (accessed on 25 November 2021)) and the geographic distribution of chlorotypes were displayed on the relief map of China using ArcGIS 10.7 (ESRI, Inc., Redlands, CA, USA, Figure 1).

For the nuclear dataset, eight loci were used to infer population structure owing to the difficulty of amplification and sequencing of two loci (X17 and X125). We assessed population genetic structure under an admixture model using the Bayesian method in Structure 2.3.4 [37]. Segregating sites with significant linkage disequilibrium were excluded from this analysis. The number of clusters (K) ranged from 1 to 10 using 20 independent runs for each value of K. Burn-in was set to 20,000 generations and Markov chain Monte Carlo (MCMC) run length to 200,000. The optimal K was estimated using ΔK statistics [38] implemented in Structure Harvester [39]. The population clusters were visualized by the program Distruct 1.1 [40]. The amount of variation among groups (Group_1 and Group_2), population within groups and within populations were calculated by the analysis of molecular variance (AMOVA) performed in Arlequin 3.01 [41].

2.3. Population Divergence and Demographic Analyses

The Approximate Bayesian Computational (ABC) approach was used to decipher the divergence time, historical demography, and genetic admixture event among different population groups of Asteropyrum. The cpDNA sequences were not included in these analyses because of their low variation. Three groups (Group_1, N₁; Group_2, N₂; and mixed populations, N₃) were set based on above-mentioned structure analysis. We tested five alternative scenarios (Figure 2) to estimate the origin of N₃ with ten nuclear loci using DIYABC 2.1.0 [42]. In the first model, N₁, N₂, and N₃ derived from an ancestral population at the same time (T₂), and the next three scenarios modeled the possible hypotheses about the divergence patterns among the three groups. In the last model, N₃ was formed from a genetic admixture event between N₁ and N₂ at time T₁. We gave a uniform prior probability and ran 10⁶ simulations for each scenario, of which 10% was used to estimate the relative posterior probability with 95% confidence intervals via logistic regression and posterior parameter distribution (Supplementary Figure S1, Table S2) according to the most likely scenario [43]. The time parameters were calculated in generations and converted into years by considering generation time, which was set to 1 year based on filed observation.

In addition, we also used isolation with a migration (IM) model [14,44] to estimate divergence time and gene flow among genetic groups. The demographic parameters including effective population size (θ), divergence time (t), and migration rate (m) between two groups of Asteropyrum were estimated using MCMC coalescent simulation implemented in IMa2 software [14,45]. We ran 6 × 10⁶ simulations with a burn-in of 10% under the HKY mutation model for each of the ten loci. These parameters from IM model were further scaled by a mean mutation rate. Initially, two rates (3.46 × 10⁻⁹ or 8.69 × 10⁻⁹ per site per year (s/s/y)) were adopted to estimate lineage divergence according to the range of substitution rate of herbaceous angiosperms [46,47]. However, because the former rate is close to the result in ABC simulation (3.30–3.47 × 10⁻⁹ s/s/y), we finally chose this rate (3.46 × 10⁻⁹ s/s/y) to scale the divergence time and population demographic parameters. We also applied the coalescence-based skyline-plot method (i.e., Bayesian skyline plot (BSP)) to infer population size change over time using BEAST v1.10.4 software [48] and visualized the results using Tracer v1.7.1 program [49].

2.4. Isolation by Distance and Isolation by Environment

To understand the effect of geographic distance and environmental heterogeneity on spatial genetic variation, we quantified the contributions of isolation by geographic distance (IBD) and isolation by environmental distance (IBE) to genetic distance using multiple matrix regression with randomization (MMRR) function in R [29]. In the case of MMRR, three distance matrices of genetic, geographic, and environmental are needed. The genetic distance (F_st) among each pair of populations was regarded as the response variable (Y), and geographic (X₁) and environmental distances (X₂) were used as explanatory variables. First, 19 bioclimatic variables of each population were extracted from WorldClim dataset (current climatic data; http://www.worldclim.org (accessed on 25 November 2021)) with 30 arc-second resolution using DIVA GIS [50]. An environmental distance matrix was constructed based on Euclidean distances of pairwise populations using IBM SPSS Statistics 22 (IBM, New York, NY, USA). Next, we calculated the genetic and geographical distances using Arlequin 3.01 and GenAlEx 6.5, respectively [51]. The regression coefficients between genetic distance and geographic distance (β_D)/environmental distance (β_E) and their significance values (P) were obtained after standardizing the three distance matrices using R software. Finally, the average values of 19 bioclimatic variables and elevation between the two genetic groups were compared by two-tailed t tests with IBM SPSS Statistics 22.

3. Results

3.1. Sequence Variation and Distribution of cpDNA Haplotypes

We obtained 219 bp cpDNA sequences across the 323 individuals of 28 populations of Asteropyrum. At the species level, nucleotide diversity (π) and haplotype diversity (H_d) were high with values of 0.011 and 0.849, respectively (Table 1). Genetic differentiation was one owing to monomorphism in each population, which was not shown in this study. A total of 10 chlorotypes were identified by 13 polymorphic sites (Supplementary Table S3). The phylogenetic network can be divided into two clades: Clade A includes H1 to H4, while the others belonged to Clade B. Haplotypes of Clade A were largely distributed in east and southeast of the range. Most haplotypes of Clade B were distributed in the west and southwest, however, some central (YL and ZX) and northeastern populations (PL, SNJ, XE, and SZ) were also featured by haplotypes of Clade B.

Of the 10 chlorotypes, six were shared by at least two populations, while the other four (H2, H7, H9, and H10) were only found in a single population (WLS, MT, YL, and ZX, respectively) (Figure 1). The most common chlorotypes were H4 (found in eight populations and 84 individuals) and H6 (found in five populations and 54 individuals), and all the populations were monomorphic (Figure 1).

3.2. Population Diversity and Structure for Nuclear DNA

The length of ten nuclear DNA sequences ranged from 206 to 432 bp, and basic population genetic parameters for each locus were summarized in Table 1. Within-population nucleotide diversity (π) per locus ranged from 0.004 (X63) to 0.017 (X47) and haplotype diversity (H_d) from 0.780 (X130) to 0.953 (X17). All the nuclear loci did not deviate strongly from the neutral model except for X125. The minimum number of recombination events were also detected for all loci, ranging from zero to seven. Given the difficulty of amplification and sequencing and results of neutral test of X17 and X125, eight low-copy nuclear genes were jointed for 144 individuals from 28 populations. The aligned matrix was 2687 bp in length, with 121 variable nucleotide sites after excluding sites in significant linkage disequilibrium and recombinant nucleotides. Genetic diversity at species was very high (π = 0.007; H_d = 0.988,

Table 2. Genetic diversity, differentiation and neutrality test of two groups and Asteropyrum as a whole.

Group	π	Hd	Hs	Fst	Nm	D	Fs
Group_1	0.005	0.976	0.704	0.531	0.630	0.332	−1.222
Group_2	0.005	0.957	0.634	0.730	0.490	0.617	−0.840
Asteropyrum	0.007	0.988	0.689	0.872	0.570	0.599	−0.927

π, nucleotide diversity; H_d, haplotype diversity; H_s, mean genetic diversity within populations; F_st, genetic differentiation among populations within group/species; N_m, calculated by using G_st; D, Tajima’s D statistic; F_s, Fu’s F_s.

).

The STRUCTURE analysis indicated that the most likely number of genetic groups was K = 2 (Figure 1). Consistent with the distribution of two chloroplast clades, Group_2 (N₂) was distributed in the east and southeast of the range, while Group_1 (N₁) in the west and southwest but extending to the central (YL and ZX) and northeastern parts (PL and SNJ) of the range. Meanwhile, three populations (WS, XE, and SZ) showed signals of genetic admixture, and therefore were defined as N₃. The AMOVA analysis revealed significant genetic differentiation (F_st = 0.334, p < 0.001), with 0.18% of the variation partitioned among groups, 33.36% among populations within groups, and 66.46% within populations (

Table 3. Results of AMOVA analysis of Asteropyrum based on eight nuclear loci.

Source of Variation	d.f.	Sum of Squares	Variance Components	Percentage of Variation	Fst
Among groups	1	2.3370	0.0001	0.18
Among populations within groups	23	46.5690	0.1668	33.36
Within populations	231	76.7670	0.3323	66.46
Total	255	125.6730	0.5000		0.3342 *

d.f., degree of freedom; *, p < 0.001.

).

3.3. Population Divergence and Demographic History

The ABC results showed that the model that N₃ derived from historical admixture between N₁ and N₂ provided the best fit to our observed data (Scenario 5, 73.33% posterior probability, Figure 2). Posterior parameter estimates for Scenario 5 indicated that N₁ and N₂ diverged from each other at 1.23 Ma (95% HPD: 0.48–1.90 Ma), followed by a genetic admixture c. 0.47 Ma (

Table 4. Posterior estimates of demographic parameters for the best model of population divergences based on Approximate Bayesian Computation.

Group	Parameter	Mean	Median	Mode	95% HPD
	Na	1.57 × 106	1.59 × 106	1.59 × 106	0.47–2.61 × 106
	N1	2.32 × 106	2.40 × 106	2.55 × 106	1.48–2.86 × 106
	N2	1.61 × 106	1.63 × 106	1.85 × 106	0.70–2.51 × 106
3	N3	3.18 × 105	2.21 × 105	1.65 × 105	0.71–9.81 × 105
	T2 (years)	1.23 × 106	1.23 × 106	1.20 × 106	0.48–1.90 × 106
	T1 (years)	4.77 × 105	4.11 × 105	2.38 × 105	0.15–1.08 × 106
	μ	3.47 × 10−9	2.84 × 10−9	1.93 × 10−9	1.21–7.56 × 10−9
	ra	2.97 × 10−1	2.37 × 10−1	1.49 × 10−1	0.35–7.63 × 10−1

N₁, N₂, and N₃ represent effective population size of Group_1, Group_2, and Mixed populations, respectively. T, divergence time or admixture time of different groups; N_a, ancestral population size; μ, the mutation rate; r_a, admixture rate.

). In addition, these simulated results showed that N₁ and N₂ have experienced slight population expansion, which was supported by the results of Bayesian skyline analysis (Figure 3). Both ABC simulations suggested the average substitution rate of nuclear sequences is close to the general rate of 3.46 × 10⁻⁹ s/s/y for herbaceous angiosperms [46].

The mixed group (N₃) was excluded from isolation with migration (IM) analysis. IM analysis found that migration rate between N₁ and N₂ was asymmetric, with much higher gene flow from N₁ to N₂ (2Nm = 0.122) than the reverse direction (2Nm = 0.059). The divergence time between them was ~1.2 Ma (95% HPD: 0.89–1.51 Ma) (

Table 5. Maximum likelihood estimates and 95% highest posterior density (HPD) intervals of demographic parameters of two nuclear groups of Asteropyrum estimated by IMa2.

Par.	θ1	θ2	θa	m1	m2	t	N1	N2	Na	2N1m1	2N2m2	T
Mean	3.398	1.761	0.615	0.072	0.067	1.349	7.53 × 105	3.90 × 105	1.36 × 105	0.122	0.059	1.20 × 106
HPD95Lo	2.886	1.377	0.068	0.015	0.000	1.011	6.39 × 105	3.05 × 105	1.50 × 104	0.022	0.000	8.96 × 105
HPD95Hi	3.998	2.152	1.327	0.135	0.137	1.704	8.86 × 105	4.77 × 105	2.94 × 105	0.270	0.147	1.51 × 106

θ₁ and N₁, effective population size of Group_1; θ₂ and N₂, effective population size of Group_2. m₁, population migration rate from Group_1 to Group_2; m₂, population migration rate in the opposite direction; 2N₁m₁ and 2N₂m₂, population migration rate estimated by m₁ and m₂; t and T, divergent time between each other. θ, m, and t are scaled by the mutation rate, while N₁, N₂, 2N₁m₁, 2N₂m₂, and T are scaled by individuals or years.

; Figure 4), which is almost identical with that estimated by ABC modeling. Effective population sizes in N₁ and N₂ were larger than ancestral population size (Na), consistent with the results of ABC modelling and Bayesian skyline analysis (Table 5, Figure 4).

3.4. Effects of IBD and IBE on Population Divergence

The regression coefficient of environmental distances (β_E = 0.463, p = 0.013) was about one order of magnitude greater than that of geographic distance (β_D = 0.046, p = 0.766), indicating that IBE has much greater contribution than IBD does. Out of 19 bioclimatic variables, nine variables (bio3: isothermality, bio4: temperature seasonality, bio5: max temperature of warmest month, bio7: temperature annual range, bio10: mean temperature of warmest quarter, bio14: precipitation of driest month, bio15: precipitation seasonality, bio17: precipitation of driest quarter, bio19: precipitation of coldest quarter) showed significant differences between the two genetic groups (Supplementary Table S4). Bio14, bio15, and bio17 exhibited the highest significance level (p < 0.001), followed by bio4 and bio19 (p < 0.01), and the remaining four with p < 0.05. In addition, there was also significant difference in elevation between the two groups, with mean elevation 943 m higher in Group_1 than in Group_2.

4. Discussion

4.1. Subtle East–West Differentiation in Asteropyrum

Due to different geological history, the development of contrast monsoon systems, or the cold and dry climate during the glacial periods, distinct east–west phylogeographic breaks (EWPBs) have been observed in an array of plant species [2,8,23,24,25,26,27]. In this study, we also detected two distinct clades using chloroplast DNA and two genetic groups using DNA sequences of multiple nuclear loci. Populations of Clade A are distributed in the east and southeast, while most populations of Clade B in the west and southwest. The two nuclear groups have similar pattern, most populations of Group_1 are located in the west and southwest while those of Group_2 in the east and southeast.

However, it is clear that the two clades/groups are not strictly associated with geography. For example, in the center of the range, several populations (MB, SJ, LB, YL, and ZX) belonging to different chloroplast and nuclear clades/groups intertwine together. Four populations (PL, SNJ, XE, and SZ) of Clade B extend to the northeast. In addition, three populations (XE, SZ, and WS) display genetic admixture between the two nuclear groups. This subtle east–west differentiation pattern, on the one hand, is largely consistent with an emerging east–west differentiation pattern that has been observed in many plant species in subtropical China. On the other hand, this pattern reflects the fact that different plant species may have responded to historical events idiosyncratically [52] and a diverse of plant species with various evolutionary trajectories should be investigated to fully understand the evolutionary history of plant species in subtropical China.

4.2. Historical Factors Responsible for Population Differentiation

In this study, we found a subtle EWPB in Asteropyrum that happened during the Mid-Pleistocene (MPT; 0.8–1.2 Ma) [53]. The time estimates (ABC: 1.23 Ma; IMa: 1.2 Ma) are very close to the divergence time of two chloroplast clades in Sophora davidii (1.28 Ma). However, we believe that our estimate could be more robust to stochastic errors because this study used two coalescent-based approaches that account for stochasticity of gene trees [12] and the results of two approaches are almost identical, although they rely on different philosophies (a mutation rate for sequences is needed for parameter estimation in IMa analysis, but not in ABC modelling [14,44,54]).

Pleistocene glaciations have been the most frequently quoted factors responsible for demographic dynamics and genetic differentiation in subtropical China [1]. For example, Qiu et al. (2009) detected an east–west differentiation (likely incipient speciation) in Dysosma versipellis based on cpDNA sequences along the line of 105° E since ~0.48 Ma, which was interpreted as a consequence of cold and dry climate in central China (glaciation-driven hypothesis) [23]. On the contrary, Fan et al. (2013) ascribed an EWPB (~1.28 Ma) observed in Sophora davidii to the establishment of presently differing monsoon regimes on either side of Tanaka-Kaiyong Line (extending from c. 33° N/102° E to 19° N/108° E, see Figure 1 for Kaiyong Line) during the (late) Pleistocene (monsoon-driven hypothesis) [2]. They also proposed that the Tanaka-Kaiyong Line is a climatically-driven barrier to present-day plant dispersal. The third hypothesis is related to the uplift of Yungui Plateau and Qianghai-Tibetan Plateau. In Cephalotaxus oliveri, Wang et al. (2014) also found a EWPB in cpDNA; however, the two cpDNA groups (Yungui and Hunan-Hubei groups) diverged at ~9.15 Ma, which is associated with the rapid uplift of Yungui Plateau during 8–10 Ma (plateau uplift-driven hypothesis) [25].

Based on our time estimates, we can reject the plateau uplift-driven hypothesis directly because even the most recent rapid uplift of Qinghai-Tibetan Plateau (between the late Miocene and late Pliocene [55,56] greatly predates the east–west divergence of Asteropyrum. For the remaining two hypotheses, we favor the glaciation-driven hypothesis over the monsoon-driven hypothesis for two reasons. First, although changes in the strength of monsoon during the Pleistocene have been reported in many studies [57], the initiation and development of monsoon system in East Asia predate substantially the onset of the Pleistocene [58,59,60]. Therefore, the population divergence of Asteropyrum is unlikely to be the result of the establishment of presently differing monsoon regimes (i.e., Pacific monsoon and Indian Monsoon) during the Pleistocene as suggested by Fan et al. (2013). Second, the Mid-Pleistocene climate transition (MPT; 0.8–1.2 Ma) [53] is an important period marked by an increase in the severity of glaciations and the emergence of the ~100-kyr glacial cycles [55]. In coincidence, the earliest moraine was reported on the QTP at the beginning of MPT [61], which was assumed to be the earliest glaciation in China [62]. Palaeoecological studies suggested that central China and adjacent regions to the southwest (between 20° to 30° N and around 105° to 120° E), which today harbor subtropical evergreen forests interspersed with warm temperate deciduous forests, had been displaced by temperate deciduous to mixed temperate and boreal forest vegetation or no forest vegetation at all during the Pleistocene glacial periods [63,64]. Therefore, it is very likely that cold and dry climate during the earliest glacial period could have expelled the populations of Asteropyrum and other floristic components in subtropical China to east and west refugia and then initiated genetic differentiation among refugia. However, we should bear in mind that the monsoons (particularly winter monsoon) and the Pleistocene glaciations have interacted with each other during the glacial periods [65], and they could have shaped the biome of subtropical China in combination. The categorization of glaciation-driven and monsoon-driven hypotheses might be arbitrary.

4.3. Demographic History and Population Genetic Admixture

In this study, we found that both groups of Asteropyrum had experienced gradual demographic expansions, with their effective population sizes (Ne) much larger than that of their ancestral populations. In addition, an admixture event was dated to 0.46 Ma, which led to several admixed populations (WS, XE, and SZ). Thus, our results contradict the notion of limited demographic expansion and population genetic admixture of temperate plant species due to both glacial and interglacial isolation in subtropical China in old literature [23,66,67,68,69]. In contrast, our results support the conclusion that the early generalization is premature possibly because early works have biased toward endangered species with narrow distribution [3]. To our knowledge, distinct east–west divergence in subtropical has been mostly observed in endangered plants with fragmented populations, such as Dysosma versipellis [23], Davidia involucrata [24], and Liriodendron chinense [8,26]. In such plant species, genetic differentiation may be strong due to genetic drift in fragmented and isolated populations. In contrast, for widespread species such as Asteropyrum, range expansion could have enhanced genetic admixture among populations as observed in ABC modeling, resulting in much less distinct east–west divergence.

Note that the time estimate of the genetic admixture event in Asteropyrum corresponds to an interglacial between 500 and 460 ka (the Large Interglacial [62]). It is most likely that populations of the west group might have experienced demographic expansion and migrated along the Daba Mountains during the Large Interglacial, and then met populations of the east group in the Three-Gorge Area. This dispersal event can be evidenced by the genetic components of XE and SZ, both populations possessing the chlorotype (H6) of Clade B. Interestingly, two adjacent populations (PL and SNJ) have both the west chloroplast and nuclear types, clearly indicating a dispersal route along the Daba Mountains. In contrast to the eastward dispersal of PL and SNJ, WS might be the result of westward dispersal and genetic admixture, because it has a Clade A haplotype (H4).

4.4. The Relative Contribution of Geographical Isolation and Environmental Heterogeneity to Population Differentiation

Geography and environment represent two key landscape components that can potentially influence gene flow and population differentiation [28]. In this study, MMRR analysis found significant IBE but non-significant effect of IBD on genetic divergence. Significant differences in nine bioclimatic variables and elevation between two genetic groups further suggest that environmental heterogeneity have substantial contribution to population differentiation of Asteropyrum. In Emmenopterys henryi, a temperate tree species with similar distribution to Asteropyrum, significant contribution of environment to genetic divergence was also revealed [70]. However, the study found that IBE is more important in Asteropyrum (β_E = 0.463, β_D = 0.046) than in Emmenopterys henryi (β_E = 0.181, β_D = 0.360), suggesting environmental heterogeneity is stronger in the former than in the later. This is reasonable because a few populations of Asteropyrum extend to the Hengduan Mountains, a region with extreme topographic relief and climatic diversity [71]. Recently, the effects of ecology in population divergence in subtropical China have received increasing attention because of high environmental heterogeneity in this region [7,70,72]. Nonetheless, more attempts are needed to fully disentangle the relative roles of geography and ecology in driving population divergence in subtropical China.