Should I stay or should I go: biogeographic and evolutionary history of a polyploid complex (Chrysanthemum indicum complex) in response to Pleistocene climate change in China



  • Quaternary climatic oscillations greatly influenced the distribution and pattern of biodiversity in the Northern Hemisphere. Here we examine how such oscillations in South East Asia may have affected the demographic and evolutionary history of a polyploid plant complex associated with semi-dry habitats.
  • We analyzed plastid and nuclear ribosomal DNA (rDNA) internal transcribed spacer (ITS) sequence variation within the Chrysanthemum indicum complex (Asteraceae), which comprises diploid and polyploid plants distributed throughout China. In total, 368 individuals from 47 populations across the geographical range of the complex were analyzed.
  • We show that the relatively widespread tetraploid form of C. indicum expanded its range southward in the Pleistocene, possibly during the most recent or previous glacial period when conditions became drier and forests retreated in southern China. In marked contrast, diploid and other polyploid members of the complex failed to expand their ranges at these times or have since undergone range contractions in contrast to tetraploid C. indicum.
  • We conclude that hybridization and gene flow between taxa occurred frequently during the evolutionary history of the complex, causing considerable sharing of chlorotypes and ITS types. Nevertheless, taxa within ploidy levels could be largely distinguished according to chlorotype and/or ITS type.


During the Quaternary, many organisms underwent major changes in geographical distribution in parallel with glacial–interglacial cycles. Thus, species ranges retreated, fragmented or expanded at different times according to climatic conditions, which, in turn, greatly affected lineage divergence within and between species (Comes & Kadereit, 1998; Hewitt, 2000, 2004; Petit et al., 2003; Alsos et al., 2012). Range expansions often led to secondary contact between lineages that had diverged in isolated refugia during periods of range fragmentation, and this in some cases resulted in hybridization and occasionally the origin of hybrid species via homoploid hybrid speciation or allopolyploidy (Stebbins, 1971, 1984; Comes & Kadereit, 1998; Abbott et al., 2003; Brochmann et al., 2004; Paun et al., 2009). Most studies of these biogeographic and evolutionary events in the Northern Hemisphere have been conducted in Europe and North America, while their occurrence in East Asia is less well known (although see Qiu et al., 2011; Liu et al., 2012a).

Unlike in Europe, West Asia and North America, where large continuous ice-sheets covered northern regions during Quaternary glaciations, most areas of East Asia remained ice-free during these periods (Hultén, 1937; Shi et al., 1987; Wu & Wu, 1996; Hewitt, 2000, 2004; Shi, 2002; Abbott & Brochmann, 2003). Consequently, East Asia was an important refugium for many representatives of the Tertiary flora and fauna (‘Tertiary relics’) and is today a species-rich region of the north temperate biota (Shi et al., 1987; Wu & Wu, 1996; Hewitt, 2000; Zhou et al., 2004). Although East Asia was unaffected by extensive ice cover during the Quaternary, changes in climatic variables such as temperature and aridity, as well as in topography and sea level in the region, had a profound effect on structuring genetic diversity. East Asian organisms also underwent cycles of range shifts as forests retreated southward during glaciations and re-colonized northern regions during interglacials (Qian & Ricklefs, 2000; Harrison et al., 2001; Gao et al., 2007; Gong et al., 2008; Tian et al., 2009; Guan et al., 2010; Qiu et al., 2011). In China, aridity increased during glaciations, particularly between the Qinling Mountains–Huai River (c. 34°N) and the tropical south (≤ 22°N), promoting replacement of subtropical forest by steppe and desert vegetation (Sun & Chen, 1991; Zhou et al., 1991). Fossil pollen evidence indicates that during these periods herbaceous plants associated with dry environments, for example, Artemisia, Ajania, Aster and Filifolium, expanded their ranges in southern China, but retreated northward during interglacials (Wang & Xu, 1985; Tong et al., 1992; Yu et al., 2000; Harrison et al., 2001). It remains unknown, however, how such range changes may have affected the genetic structure of these plants and whether they led to divergence, secondary contact events, and on occasion hybrid speciation as occurred in other parts of the Northern Hemisphere. Here we focus on the demographic and evolutionary history of the Chrysanthemum indicum complex, which is generally associated with semi-arid habitats in China.

Chrysanthemum is a temperate Eurasian genus comprising c. 40 species mainly distributed in eastern Asia (Bremer & Humphries, 1993; Oberprieler et al., 2007; Liu et al., 2012b). According to distribution, morphological and molecular phylogenetic data, the Chinese Chrysanthemum species can be divided into two groups, the Chrysanthemum zawadskii group and the Chrysanthemum indicum group (Lin et al., 2011; Liu et al., 2012b). The former occurs in northern China and comprises taxa with erect stems and large capitula containing white-purple ray florets, while the latter group is distributed from north to south China and comprises taxa with creeping stems and capitula with yellow or white ray florets. Here, we focus on the C. indicum complex, which is considered to comprise six species – C. indicum, Chrysanthemum lavandulifolium, Chrysanthemum potentilloides, Chrysanthemum rhombifolium, Chrysanthemum hypargyrum and Chrysanthemum vestitum (Supporting Information Fig. S1) – although clear morphological gaps between these taxa are often obscured in areas of overlap (Lin et al., 2011; Liu et al., 2012b). The species are herbs mainly found in open and relatively dry habitats, for example, on mountain slopes or at the edge of forests. A previous study on the geographic distribution of cytotypes in the C. indicum complex confirmed that C. rhombifolium and C. lavandulifolium were diploid (2= 18), C. hypargyrum and C. potentilloides were tetraploid (2= 36), and C. vestitum was hexaploid (2= 54), while C. indicum comprised two cytotypes, one diploid and the other tetraploid (Li et al., 2013). In contrast to the narrowly distributed C. rhombifolium, C. hypargyrum and C. potentilloides, for which only single populations are known, C. indicum is a widespread species in central and southern China (Fig. S2). However, whereas tetraploid C. indicum is distributed widely, the diploid form and other taxa that comprise the C. indicum complex occur only in central and northern China (Li et al., 2013). Li et al. (2013) suggested that tetraploid C. indicum may have expanded its range southward during a recent Quaternary glacial period when forests retreated in south China as conditions became drier, and then remained there in dry areas when forests expanded their ranges northward during the current post-glacial period. To test this hypothesis and to examine the evolutionary history of the C. indicum complex in more detail, we surveyed both plastid and nuclear ribosomal internal transcribed spacer (nrITS) DNA sequence variation within and among all recognized members of the complex.

A previous survey of molecular variation within the C. indicum complex (Yang et al., 2006) examined variation for nuclear random amplified polymorphic DNA (RAPD) and inter simple sequence repeat (ISSR) markers, and chloroplast SSR markers within and among one population of C. lavandulifolium, two populations of diploid C. indicum and nine populations of tetraploid C. indicum. Although limited in terms of number of taxa and populations studied, this investigation revealed widespread sharing of chloroplast and nuclear markers between members of the complex. It was concluded that the complex probably evolved through cycles of multiple differentiation followed by frequent hybridization and polyploidization (via both autoployploidy and allopolyploidy). In the study reported here, we expanded the analysis of the complex so as to cover all recognized diploid and polyploid members. In addition, we greatly increased the number of populations of C. lavandulifolium and tetraploid C. indicum surveyed to obtain a better understanding of the biogeographic and evolutionary history of the complex.

Materials and Methods

Population sampling

Leaves were collected from 368 individuals from 47 populations across the geographical distribution of the C. indicum complex (Table 1, Fig. 1). Only typical representatives of each species were sampled according to an existing collection of herbarium specimens of the complex at Peking University (Table S1). We did not sample plants with intermediate morphology that might be hybrid. Plants sampled were separated from each other by at least 10 m and approximately half had their ploidy level determined using flow cytometry (see Li et al., 2013). Following collection, leaves were immediately dried with silica gel. The populations sampled comprised three and 32 populations of diploid and tetraploid Chrysanthemum indicum L., respectively, six populations of diploid Chrysanthemum lavandulifolium (Fish. ex Trautv.) Makino, one population each of the rare diploid Chrysanthemum rhombifolium (Ling & Shih) H. Ohashi & Yonekura and the tetraploids Chrysanthemum potentilloides Handel-Mazzetti and Chrysanthemum hypargyrum Diels, and three populations of the hexaploid Chrysanthemum vestitum (Hemsley) Stapf. No other populations of diploid C. indicum, C. rhombifolium and C. hypargyrum are known to exist according to our field and herbarium records.

Table 1. Locations in China, ploidy level and sample sizes (N) of populations of members of the Chrysanthemum indicum complex surveyed for chloroplast DNA and internal transcribed spacer (ITS) sequence variation
TaxonPopulation codeLocationLatitude, longitude (°N, °E)Elevation (m)Ploidy N Chlorotypes (no. of individuals)ITS types (no. of individuals)
  1. The numbers of particular chlorotypes and ITS types detected per population are also indicated.

  2. Voucher specimens are deposited in the Herbarium of Peking University (PKU).

  3. a

    Samples from populations contain diploid (2×) C. indicum.

C. indicum Southern region
1Nanning, Guangxi22.81, 108.30797H17(7)G5(7)
2Guangzhou, Guangdong23.17, 113.29479H25(9)G5(3), G7(6)
3Taibei, Taiwan25.03, 121.532076H39(6)G1(6)
4Kunming, Yunnan25.12, 102.7118065H22(1), H23(1), H37(3)G1(3), G4(2)
5Guilin, Guangxi25.24, 110.2820710H21(10)G5(10)
6Putian, Fujian25.48, 118.989575H37(5)G6(3), G7(2)
7Ganzhou, Jiangxi25.85, 114.917867H11(5), H37(2)G5(4), G7(3)
8Guiyang, Guizhou26.64, 106.6310718H11(1), H28(3), H37(4)G4(8)
9Hengshan Mt, Hunan26.92, 112.5411258H11(5), H37(1), H40(2)G2(2), G4(6)
10Lijiang, Yunnan26.98, 100.23241810H11(10)G1(10)
11Wugong Mt, Jiangxi27.62, 113.8418157H9(2), H11(5)G1(1) G2(2),G4(2),G7(2)
12Minning, Sichuan28.54, 102.2117259H8(2), H11(7)G1(6), G5(1), G7(2)
13Pan'an, Zhejiang29.05, 120.448738H11(8)G4(4), G6(2), G7(2)
14Zhangjiajie, Hunan29.14, 110.473175H18(3), H30(2)G4(5)
15Chongqing29.54, 106.562766H33(6)G1(6)
16Emei Mt, Sichuan29.59, 103.37272810H1(1), H37(9)G1(8), G2(1), G4(1)
Central region
17Lushan Mt, Jiangxi29.66, 115.98121010H32(10)G3(2), G5(6), G7(2)
18Lushan Mt, Jiangxi29.68, 115.95103510H24(5), H32(5)G3(5), G5(5)
19Huangshan Mt, Anhui29.69, 118.3015238H24(3), H34(5)G6(2), G7(6)
20Tianzhu Mt, Anhui30.52, 117.0514209H20(1), H34(8)G4(9)
21Yichang, Hubei30.69, 111.283577H24(5), H26(1), H29(1)G4(6), G7(1)
22Xinyang, Henan31.28, 113.957495H11(1), H13(4)G1(3), G4(2)
23Shennongjia, Hubei31.64, 110.61156710H14(2), H28(8)G1(3), G4(7)
24aShennongjia, Hubei31.73, 110.62285710H24(8), H35(2)G2(4), G4(6)
25aShennongjia, Hubei31.75, 110.6529769H24(9)G2(3), G4(6)
26Shennongjia, Hubei31.76, 110.64253910H14 (5), H24(5)G2(10)
27Baohua Mt, Jiangsu31.92, 119.181456H19(2), H24(2), H35(2)G7(6)
28aBaohua Mt, Jiangsu31.94, 119.161289H24(9)G1(6), G4(3)
29Zijin Mt, Jiangsu32.11, 118.6320210H2(1), H19(7), H26(1), H29(1)G1(4), G2(2), G4(2), G7(2)
30Guangyuan, Sichuan32.49, 105.8613475H37(5)G1(2), G2(1), G4(2)
31Shiyan, Hubei32.65, 110.798247H1(1), H24(5), H27(1)G1(5), G4(2)
32Hanzhong, Shanxi33.06, 107.0119576H11(2), H37(4)G1(4), G4(2)
33Laojun Mt, Henan33.75, 111.64211610H19(5), H24(5)G1(6), G4(4)
34Laojun Mt, Henan33.76, 111.6621189H24(5), H31(1), H34(1), H36(2)G4(9)
35Baoji, Shaanxi34.36, 107.226175H11(2), H12(3)G5(1), G7(4)
C. rhombifolium 36Wuxi, Chongqing31.25, 109.636577H11(1), H16(1), H37(5)G8(4), G9(3)
C. vestitum 37Tianzhu Mt, Anhui30.51, 117.0611796H11(6)G2(3), G4(3)
38Shiyan, Hubei32.64, 110.767796H2(2), H30(4)G1(2), G4(4)
39Laojun Mt, Henan33.74, 111.6519057H11(1), H37(5), H38(1)G1(2), G4(5)
C. hypargyrum 40Taibai Mt, Shaanxi34.54, 107.69370110H15(9), H37(1)G10(10)
C. potentilloides 41Yuanqu, Shanxi35.30, 111.642159H1(2), H24(5), H34(1), H35(1)G1(4), G4(5)
C. lavandulifolium 42Pingliang, Gansu35.53, 106.6711077H4(7)G1(4), G4(1), G7(2)
43Taishan Mt, Shandong36.19, 117.1310287H3(7)G2(2), G4(1), G7(4)
44Yantai, Shandong37.42, 121.36417H10(7)G2(7)
45Kongtongdao, Shandong37.56, 121.51639H6(9)G2(2), G4(7)
46Dongling Mt, Beijing39.98, 115.4312049H1(1), H7(8)G2(9)
47Dongling Mt, Beijing40.20, 116.1310539H5(9)G2(9)
Figure 1.

Geographical distribution of sampled populations and taxa of the Chrysanthemum indicum complex surveyed for chlorotype and internal transcribed spacer (ITS) variation in China. Numbers refer to populations listed in Table 1. Pie diagrams indicate the chlorotype group composition of each population.

DNA extraction, PCR amplification and sequencing

Genomic DNA was extracted from dried leaves using the cetyltrimethyl ammonium bromide (CTAB) method (Doyle & Doyle, 1987). Three intergenic spacers of the plastid (cp) DNA (psbA-trnH, trnY-rpoB and trnL-F) and the nuclear ribosomal ITS regions were sequenced. Amplifications were conducted using primer pairs psbA/trnHGUG, trnYretF/rpoBretR, trnL5′UAA/trnFGAA (Taberlet et al., 1991; Hamilton, 1999; Shaw et al., 2005; Timme et al., 2007) and ITS1/ITS4 (White et al., 1990). All polymerase chain reactions (PCRs) were carried out in a volume of 25 μl containing 25 ng of genomic DNA, 50 mM Tris-HCl, 1.5 mM MgCl2, 0.5 mM dNTPs, 0.6 mM of each primer and 0.75 units of Taq DNA polymerase. Amplifications were conducted using a Peltierthermocycler (Bio-Rad, Hercules, CA, USA) and the following program: initial denaturation at 94°C for 3 min, 30 cycles of denaturation at 94°C for 30 s, 30 s of annealing at 48–55°C (for the plastid regions) or 58°C (for ITS), and 1 min 20 s at 72°C, and a final extension at 72°C for 10 min. The PCR products were purified using a TIANquick Midi Purification Kit following the manufacturer's protocol (Tiangen, Beijing, China). Cycle-sequencing reactions were conducted using the ABI Prism BigDye Terminator (Cycle Sequencing Kit, version 3.1; Applied Biosystems, Life Technologies, Carlsbad, CA, USA) and the same primer pairs used for amplification. The sequencing products were run on an ABI 3130 XL sequencer (Applied Biosystems). Sequences were aligned with clustalx (Thompson et al., 1997). Haplotypes were identified using DnaSP version 4.0 (Rozas et al., 2003). All sequences analyzed have been submitted to GenBank with accession numbers KC694194KC695665.

Cloning and sequencing of ITS repeats

Many individuals sequenced for ITS possessed what appeared to be an additive ITS phenotype (see ITS sequence variation in Results section below). To check this, we cloned and sequenced ITS repeats from individuals exhibiting a particular additive ITS type. Between one and three individuals per additive ITS type per population were examined in this way. ITS PCR products were ligated into the pGEM-T Vector with a Promega Kit (Promega Corporation, Madison, WI, USA) and sequenced using the procedure described above (see DNA extraction, PCR amplification and sequencing). From each diploid, tetraploid and hexaploid individual examined, eight, 12 and 20 ITS repeats were cloned and sequenced, respectively. The number of individuals examined per taxon in this way was as follows: nine of diploid C. indicum (from three populations), 97 of tetraploid C. indicum (from 32 populations), ten of diploid C. lavandulifolium (from three popuations), three of diploid C. rhombifolium (from one population), three of tetraploid C. potentilloides (from one population) and nine of hexaploid C. vestitum (from three populations). No individuals of C. hypargyrum were examined as none possessed an additive ITS type based on direct sequencing.

Analyses of plastid DNA variation

Phylogenetic analysis

Sequences of each plastid DNA region were concatenated into one alignment and each indel was coded as a binary character (e.g. 0/1 = A/C) using GapCoder (Young & Healy, 2003). Phylogenetic relationships between haplotypes (chlorotypes) were reconstructed by maximum likelihood implemented in paup* 4.0b10 (Swofford, 2002). The phylogenetic tree was bootstrapped with 1000 replicates. Phylogenetic relationships were also examined with MrBayes 3.1.2 (Huelsenbeck & Ronquist, 2001; Ronquist & Huelsenbeck, 2003) with the best-fitting GTR + G model selected by Model test version 3.7 (Posada & Crandall, 1998) using the Akaike information criterion (AIC) (Akaike, 1974). Four chains, three heated and one cold, ran simultaneously each with different starting seeds for 2 × 106 generations. Trees were sampled every 100 generations and remained stationary after 400 000 generations. The first 4000 trees were discarded as ‘burn-in’. network (Polzin & Daneshmand, 2003) was also used to generate a median-joining network of chlorotypes (Bandelt et al., 1995, 1999) with weights = 10 and e = 0.

Population diversity analysis

Nucleotide diversity (Pi) and haplotype diversity (Hd) (Nei & Li, 1979; Nei, 1987) were calculated for chlorotypes using the DnaSP 4.0 program (Rozas et al., 2003). In addition, gene diversity within populations (HS) and total gene diversity (HT) were estimated, and two indices of population differentiation, NST (Pons & Petit, 1996; Grivet & Petit, 2002) and GST (Nei, 1987), for chlorotypes were computed using the program permut (available at (Pons & Petit, 1996). Whereas GST only considers chlorotype frequencies, NST takes into account both chlorotype frequencies and their genetic distances. We tested whether the value of NST was significantly larger than that of GST using a permutation test with 1000 random permutations of chlorotypes across populations to infer whether there was significant phylogeographic structure (Pons & Petit, 1996). We also conducted analyses of molecular variance (AMOVA using Arlequin version 3.0; Excoffier et al., 2005) to determine if chlorotype variation among populations was significant. Mantel tests were further used to examine the significance of isolation-by-distance with 1000 random permutations on matrices of pairwise population FST values and the natural logarithm of geographical distances (Rousset, 1997).

Range expansion analysis

The possibility that the C. indicum complex underwent a recent range expansion was tested using mismatch distributions implemented in Arlequin version 3.0 (Excoffier et al., 2005). Mismatch distributions were constructed by plotting the number of nucleotide differences between all pairs of chlorotypes (Rogers & Harpending, 1992). Multimodal or random and rough distributions of pairwise differences are assumed to characterize populations that have been stable for a long time, whereas populations that have experienced a recent demographic expansion should display a unimodal and smooth distribution (Rogers & Harpending, 1992). The sum of squared deviations (SSD) was used to test the validity of the expansion model. P values were calculated as the proportion of simulations producing a larger SSD than the observed SSD. The raggedness index (r) of the observed distribution was used to quantify the smoothness of mismatch distribution (Harpending, 1994). Fu's FS value (Fu, 1997) and Tajima's D (Tajima, 1989) were also calculated to test the equilibrium of the populations. These statistics are particularly sensitive to past population expansions and indicate demographic expansion when large, negative values are obtained. Both Fu's FS and Tajima's D were tested for significance in Arlequin version 3.0 using 10 000 bootstrap replicates.

To assess the goodness of fit of our data to the exponential growth model characterized by the equation θt θinitial–gt, and to compute maximum likelihood-based estimators of exponential growth rate (g), we used the Markov chain Monte Carlo approach implemented in lamarc 2.1.3 (Kuhner, 2006), which takes into account phylogenetic relationships between sequences. When consistent estimates were obtained between runs with different numbers of steps in chains and different initial values of g, final estimates were obtained by running the program with 10 short (10 000 steps) and two long (200 000 steps) chains (sampling increment 20). The 95% confidence intervals of g were also assessed by lamarc. The run was conducted twice with random seeds and results were consistent. Additionally, a maximum-likelihood approach of Galtier et al. (2000) implemented in sweep-bottleneck was used to test expansions by calculating the log likelihood values of the no-founder model versus the bottleneck model. If the hypothesis of rapid expansion is not rejected, then s (s = 2ut) can be converted to time since expansion (τ) using the formula t = 0.5τu − 1, where τ is the mode of the mismatch distribution expressed in units of evolutionary time (years before present (ybp)) and u is the mutation rate per generation for the whole sequence (Rogers & Harpending, 1992). As plastid (cp) DNA is maternally inherited, u = μkg, where μ is the substitution rate per nucleotide site per year (s s−1 yr−1), k is the average sequence length used for analysis and g is the generation time in years. Because cpDNA mutation rates remain unknown in the Asteraceae, we used in our calculations both a high (1.0 × 10−8 s s−1 yr−1) and a low (1.0 × 10−9 s s−1 yr−1) substitution rate according to estimates reported for plants in general (Wolfe et al., 1987; Muse, 2000; GuhaMajumbdar & Sears, 2005). Generation time (g) was taken to be 1 because most of the Chrysanthemum species examined reproduce annually (Lin et al., 2011).

Analysis of ITS diversity

Examination of ITS variation revealed the occurrence of ten different ITS repeat sequences in the complex, with many individuals containing additive sequences, that is, more than one type of ITS sequence in the same individual (see ITS sequence variation in Results section). We treated ITS variants among individuals, whether of additive or nonadditive type, as different ITS types. To quantify levels of ITS type variation, we used the Shannon diversity index (Shannon, 1948). Thus, we estimated total Shannon diversity (Htotal), within-population diversity (Hwithin-population), and between-population diversity (Hbetween-population) for diploid and tetraploid C. indicum, and C. lavandulifolium.


Phylogenetic relationships and distribution of chlorotypes

The lengths of aligned sequences and percentages of substitution sites across individuals were 442 bp and 4.07% for trnH-psbA, 371 bp and 2.43% for trnL-F, and 674 bp and 2.07% for trnY-rpoB, respectively. The combined aligned sequence contained 43 substitution sites, which enabled identification of 40 different haplotypes (chlorotypes) (Tables 1, S2, Fig. 1). Maximum likelihood and Bayesian phylogenetic analyses provided strong support for the distinction of two main chlorotype clades (only the maximum likelihood tree is shown; Fig. 2a). network analysis similarly resolved two major chlorotype groups (Hg1 and Hg2) with four subgroups (Hg 2a–d) detected in the Hg2 group (Fig. 2b). Relationships between ancestral chlorotypes (interior position and frequent) versus derived chlorotypes (tip position and relatively rare) could be inferred from the network.

Figure 2.

Phylogenetic relationships of chlorotypes. (a) Maximum likelihood tree with bootstrap percentages (BP)/posterior probabilities from the Bayesian inference (PP) values indicated. (b) Network of chlorotypes. The relative sizes of the circles in the network are proportional to haplotype frequency (the smallest circles representing n = 1 and the largest representing n = 66). The small black bars represent mutation steps and the black dots represent missing chlorotypes. After each chlorotype (H1–H40) in the key are listed for each taxon the population and number of individuals per population that possessed the chlorotype. ind, Chrysanthemum indicum; lav, C. lavandulifolium; rho, C. rhombifolium; pot, C. potentilloides; hyp, C. hypargyrum; ves, C. vestitum.

All six populations of diploid C. lavandulifolium contained only Hg1 chlorotypes (Figs 1, 2, Table 1) with each population fixed or nearly fixed for a different chlorotype. By contrast, diploid C. indicum contained only Hg2 chlorotypes with populations fixed or nearly fixed for chlorotype H24, the most common and interior haplotype in subgroup Hg2b. In the very rare diploid species, C. rhombifolium, five of seven individuals surveyed possessed the Hg2d chlorotype (H37), while the two other individuals contained chlorotypes in the Hg2a subgroup.

In contrast to diploid C. indicum, tetraploid C. indicum was highly polymorphic. Although largely comprising a wide range of different Hg2 chlorotypes that included representatives of all four Hg2 subgroups, it also contained a few individuals (five of 247 examined) possessing an Hg1 chlorotype. The two rare tetraploids, C. potentilloides and C. hypargyrum, for which only one population of each is known, were also polymorphic for chlorotypes from different subgroups or groups, as were two of the three populations examined in the hexaploid C. vestitum.

Genetic structure of chlorotype variation in the C. indicum complex

Chrysanthemum lavandulifolium and tetraploid C. indicum contained similar levels of chlorotype diversity, based on estimates of Pi, Hd and HT, that were much higher than those estimated for diploid C. indicum (Table 2). However, C. lavandulifolium contained almost no chlorotype diversity within populations (HS = 0.037), in contrast to diploid and tetraploid C. indicum (HS = 0.119 and HS = 0.360, respectively).

Table 2. Estimates of nucleotide diversity (Pi), haplotype diversity (Hd), total gene diversity (HT), average gene diversity within populations (HS), interpopulation differentiation (GST), and the number of substitution types (NST) (mean + SE in parentheses) as indicators of chlorotype diversity in members of the Chrysanthemum indicum complex distributed in China
Populations Pi Hd H T H S G ST N ST
  1. Total includes all members of the C. indicum complex; **NST is significantly different from GST (< 0.01); ns, not significant.

Total0.002640.9220.934 (± 0.0150)0.310 (± 0.0402)0.668 (± 0.0447)0.760 (± 0.0532)**
C. indicum (2× and 4×)0.001720.9030.893 (± 0.0223)0.339 (± 0.0465)0.620 (± 0.0532)0.584 (± 0.0613)ns
C. indicum-2×0.000190.1380.133 (± 0.1156)0.119 (± 0.1185)0.111 (−)0.111 (−)
C. indicum-4×0.001800.9120.927(± 0.0186)0.360 (± 0.0484)0.612 (± 0.0541)0.597 (± 0.0625)ns
Southern populations0.001580.8270.876 (± 0.0489)0.258 (± 0.0724)0.706 (± 0.0891)0.639 (± 0.1079)ns
Central populations0.001800.8760.911 (± 0.0264)0.462 (± 0.0554)0.493 (± 0.0712)0.533 (± 0.0817)ns
C. lavandulifolium 0.002040.8551.000 (± 0.0186)0.037 (± 0.0370)0.963 (± 0.0370)0.968 (± 0.0302)ns

AMOVA showed that chlorotype divergence was highly significant between C. indicum (diploid and tetraploid forms combined) and C. lavandulifolium (FCT = 0.69; < 0.001; Table 3), but was not significant between the two cytotypes of C. indicum (FCT = 0.05). Divergence among populations within all three taxa (FSC) was significant, although it was greatest for C. lavandulifolium. AMOVA conducted on only tetraploid populations of C. indicum showed that chlorotype divergence between centrally located and southern populations (Table 4) was not significant (FCT = 0.08); however, overall diversity, and particularly within-population diversity (HS), tended to be greater in central tetraploid populations (populations 17–23, 26, 27 and 29–35) than in southern populations (populations 1–16) (Table 2). Phylogeographic structure was evident across all populations of the complex (NST > GST; < 0.001; Table 2), largely because populations of C. lavandulifolium (occurring only in northern China) were composed of Hg1 haplotypes, whereas populations of all other taxa (distributed in central and southern China) were composed mainly of Hg2 haplotypes. No significant phylogeographic structure was detected within diploid and tetraploid C. indicum nor within C. lavandulifolium treated separately (i.e. NST was not significantly greater than GST; Table 2). A Mantel test revealed a significant and positive correlation between population differentiation (FST) and geographical distance when conducted across C. indicum (both ploidy levels) and C. lavandulifolium populations (= 0.307; < 0.05), whereas no significant correlations were detected by Mantel tests conducted within diploid C. indicum (= 0.397), tetraploid C. indicum (= 0.254) or C. lavandulifolium (= 0.342).

Table 3. Analyses of molecular variance (AMOVA) conducted on chlorotype variation within and among diploid (2×) and tetraploid (4×) Chrysanthemum indicum, and C. lavandulifolium
Source of variationdfSSVCPV (%)Fixation index
  1. df, degrees of freedom; SS, sum of squares; VC, variance components; PV, percentage of variation; FSC, correlation within populations relative to group; FST, correlation within populations relative to total; FCT, correlation within groups relative to total. **, < 0.001, 1000 permutations. ns, not significant.

Groups defined by species (C. lavandulifolium versus 2× and 4× C. indicum)
Among species1245.452.91Va68.87FSC = 0.67**
Among populations within species39287.290.88 Vb20.88FST = 0.90**
Within populations282122.130.43 Vc10.25FCT = 0.69**
C. indicum(2× and 4×)
Among populations34219.680.76 Va60.43FST = 0.60**
Within populations240119.460.50 Vb39.57 
C. lavandulifolium
Among populations567.621.69 Va96.37FST = 0.96**
Within populations422.670.069 Vb3.63 
Groups defined by geography in C. indicum-4× (central versus southern populations)
Among regions120.710.115 Va8.21FSC = 0.58**
Among populations within region30187.070.741 Vb53.11FST = 0.61**
Within populations215116.020.540 Vc38.69FCT = 0.08ns
Groups defined by ploidy level in C. indicum (2× versus 4×)
Among ploidy levels111.030.079 Va5.49FSC = 0.60**
Among populations within ploidy levels33208.650.75 Vb56.71FST = 0.62**
Within populations240119.460.50 Vc37.80FCT = 0.05ns
Table 4. Values for the time in number of generations elapsed since a sudden expansion episode (τ), the sum of squared deviations (SSD), and Harpending's raggedness index (RAG) (with probability P values) from mismatch analysis of chlorotype variation, values for Tajima's D and Fu's FS (with P values), and significance of the estimate of population exponential growth rate (g) and τ based on lamarc and sweep-bottleneck (sweep-bott) tests
GroupsMismatch distributionNeutrality tests lamarc sweep-bott
τ (t)SSD (P value)RAG (P value)Tajima's D (P value)Fu's FS (P value) g τ (t1t2)
  1. t1, time in thousand years derived from 1 × 10−8 s s−1 yr−1; t2, time in thousand years derived from 1 × 10−9 s s−1 yr−1; **, < 0.01; ns, not significant.

  2. Diploid, Chrysanthemum indicum and C. rhombifolium in Hg2 chlorotype clade; tetraploid, C. indicum -4×, C. hypargyrum and C. potentilloides in Hg2; hexaploid, C. vestitum in Hg2; polyploid, all tetraploid and hexaploid samples in Hg2.

Total1.020 (34–340)0.148 (0.000)0.022 (1.000)−1.204 (0.076)−15.232 (0.004)1778.414**1.374 ** (46–460)
Hg1 chlorotype clade3.639 (−)0.028 (0.025)0.079 (0.010)0.241 (0.559)0.259 (0.628)291.707 ns4.563 ns
Hg2 chlorotype clade2.225 (75–750)0.001 (0.303)0.039 (0.337)−1.541 (0.023)−16.488 (0.000)1923.370**2.657** (89–890)
Diploid (2×)0 (−)0.333 (0.000)0.502 (0.955)−0.367 (0.335)−0.970 (0.262)4071.343 ns
Tetraploid (4×)2.426 (81–810)0.001 (0.133)0.038 (0.324)−1.407 (0.043)−13.514 (0.002)2015.904**2.478** (83–830)
Hexaploid (6×)1.146 (−)0.023(0.137)0.172 (0.108)0.313(0.563)−0.223 (0.417)223.541 ns1.322 ns
Polyploid (4× + 6×)2.348 (78–780)0.001 (0.152)0.039 (0.295)−1.515 (0.045)−16.159 (0.000)1556.63**2.643**(89–890)
C. lavandulifolium 3.834 (−)0.070 (0.000)0.171(0.000)0.937 (0.741)1.764 (0.880)237.53 ns4.768 ns
C. indicum 2.145 (71–710)0.001 (0.511)0.030 (0.647)−1.558 (0.019)−15.551 (0.001)2369.192**2.349** (75–750)
C. indicum-4×2.137 (72–720)0.001 (0.507)0.028 (0.575)−1.569 (0.013)−15.598 (0.001)1978.743**2.339** (78–780)

Demographic expansion based on chlorotype variation

Mismatch distributions based on chlorotype diversity in the Hg2 clade, and also for tetraploid C. indicum treated separately, exhibited unimodal curves (Fig. 3b,d). Both SSD and raggedness index tests of these curves were nonsignificant, while values for Tajima's D and Fu's Fs were significantly negative (Table 4). Thus, there is a strong indication that the Hg2 clade of chlorotypes underwent a recent demographic expansion and that this was largely a result of a range expansion of tetraploid C. indicum. This was supported by lamarc and sweep-bottleneck tests which showed, respectively, that values of g (exponential growth rate) and τ (time since expansion) for both the Hg2 clade of chlorotypes and tetraploid C. indicum were highly significant (Table 4). By contrast, there was no evidence for a demographic expansion of diploid C. indicum in the recent past (Fig. 3c, Table 4), and similarly there was no indication of diploid C. lavandulifolium having undergone a recent expansion based on mismatch distributions and accompanying tests of Hg1 chlorotype diversity (Fig. 3a, Table 4). Using the fast cpDNA mutation rate (1.0 × 10−8 s s−1 yr−1), the demographic expansion of the Hg2 chlorotype clade was estimated to have occurred 750 kya (based on mismatch analysis) or 890 kya (based on the sweep-bottleneck analysis), while that for tetraploid C. indicum was estimated to have occurred almost simultaneously (720 and 780 kya, respectively) (Table 4). These values were reduced by an order of magnitude using the slow cpDNA mutation rate, that is, to 75 and 89 kya, respectively, for the Hg2 clade, and to 72 and 78 kya, respectively, for tetraploid C. indicum.

Figure 3.

Mismatch distributions. The distribution of the number of pairwise nucleotide differences between chlorotype sequences in (a) the northern clade (Hg1 chlorotypes), (b) the central and southern clade (Hg2 chlorotypes), (c) diploid Chrysanthemum indicum and (d) tetraploid C. indicum is shown. The solid lines (circles) are drawn from observed differences between chlorotype sequences, whereas dashed lines (squares) represent simulated differences under a model of sudden (stepwise) population expansion (Rogers & Harpending, 1992).

ITS sequence variation

Aligned ITS sequences of all individuals surveyed were 703 bp in length. Three unambiguous polymorphic sites were detected, one in ITS1 (site 1 at 158 bp, A or G) and two in ITS2 (site 2 at 476 bp, A, G or T; and site 3 at 592 bp, T or C; Fig. 4). A total of 10 different ITS types were identified from variation across these three sites (Table 1, Fig. 4). Five of these were of the nonadditive type (AAT, GAT, GGT, GAC and AAT) and five were of the additive polymorphic type at site 1 or 2, or at both of these sites ((A/G)AT, A(A/G)T, (A/G)(A/G)T, (A/G)AC and G(A/G)T). Additive types were recognized by double peaks occurring at polymorphic sites 1 and 2 in electropherograms, while nonadditive types had single peaks at these sites. Proof that these types were additive came from the analysis of cloned ITS repeats (Table S3). In all but three of the 131 individuals tested, the expected ITS repeats comprising an additive ITS type were present in an individual. Thus, individuals possessing the (A/G)AT additive type were shown to contain both the AAT and GAT repeats, while those possessing the A(A/G)T contained both the AAT and AGT repeats, etc. The three individuals that did not contain all expected repeats possessed the (A/G)(A/G)T ITS additive type (Table S4). In each case, one of the four repeats expected to be present was not found among the sample of repeats sequenced, probably as a result of a sampling effect.

Figure 4.

Distribution of different internal transcribed spacer (ITS) types (G1–G10) detected among populations and taxa of the Chrysanthemum indicum complex. Solid-line circles represent ITS types detected; broken-line circles represent missing types (e.g. AGT and GAC). Additive ITS types are indicated by overlapping circles. Besides each ITS type are listed for each taxon the population and number of individuals per population that possessed the ITS type. ind, C. indicum; lav, C. lavandulifolium; rho, C. rhombifolium; pot, C. potentilloides; hyp, C. hypargyrum; ves, C. vestitum.

Of the 47 populations analyzed, 39 contained at least some individuals possessing an additive ITS type (Fig. 4, Table S3). Such individuals were present in all species except C. hypargyrum, which was monomorphic for the ATT type not found in other members of the complex (Table S4). Diploid C. rhombifolium was also unique in possessing a C nucleotide rather than a T at the third polymorphic site (site 3), giving rise to two ITS types, GAC and (A/G)AC, not found in other members of the complex. The two other diploids, diploid C. indicum and C. lavandulifolium, and also the hexaploid, C. vestitum, had three ITS types in common, AAT, GAT and (A/G)AT. Chrysanthemum lavandulifolium also possessed the (A/G)(A/G)T type (Table S4).

As was the case for chlorotype variation, tetraploid C. indicum was most polymorphic for ITS type, containing seven of the 10 types detected. Of these, AAT and (A/G)AT (both of which were shared with diploid C. indicum, C. lavandulifolium, C. potentilloides and C. vestitum) were most common in the taxon, being found among 67 and 73 individuals, respectively, and also 14 and 17 populations, respectively. The (A/G)(A/G)T type (shared with C. lavandulifolium) was also relatively common in tetraploid C. indicum (present in 38 individuals across 12 populations), as was G(A/G)T (present in 37 individuals across eight populations). G(A/G)T and the relatively rare A(A/G)T type (present in seven individuals across three populations) were unique to tetraploid C. indicum, as was another rare type, GGT (present in seven individuals across two populations). The remaining ITS type found in tetraploid C. indicum, GAT, was shared with diploid C. indicum, C. lavandulifolium and C. vestitum.

From these results, it is clear that both sharing of ITS types and the presence of additive ITS polymorphic sites within individuals commonly occur among diploid and polyploid members of the C. indicum complex. Nonetheless, the rare diploid C. rhombifolium and also the rare tetraploid C. hypargyrum could be distinguished from each other and all other members of the complex by their possession of unique ITS types. Furthermore, tetraploid C. indicum contains some individuals possessing either unique additive (G(A/G)T and A(A/G)T) or unique nonadditive (GGT) ITS types, although both A(A/G)T and GGT were rare. Overall, the most commonly encountered ITS types in the complex are AAT, GAT and (A/G)AT.

Estimates of ITS diversity based on the Shannon index confirmed that total diversity (Htotal) was greatest in tetraploid C. indicum relative to diploid C. indicum and C. lavandulifolium (Table 5). This was largely a result of a greater level of between-population diversity (Hbetween-population) in tetraploid C. indicum, as within-population diversity (Hwithin-population) was higher in diploid C. indicum. Interestingly, central populations of tetraploid C. indicum contained greater levels of within-population diversity than southern populations, but there was no difference between the two groups for between-population diversity.

Table 5. Shannon diversity values for the Chrysanthemum indicum complex based on internal transcribed spacer (ITS) type variation
  H total H within-population H between-population
  1. Htotal, total diversity; Hwithin-population, within-population diversity; Hbetween-population, between-population diversity.

  2. Total includes all members of the C. indicum complex.

C. lavandulifolium 1.0850.4070.678
C. indicum (2× and 4×)1.6650.5071.158
C. indicum (2×)1.0110.6490.362
C. indicum (4×)1.6790.4941.185
Southern populations1.5740.4441.130
Central populations1.6760.5431.133


We undertook the study reported in this paper to clarify the pattern of divergence within the C. indicum polyploid complex in China, and to test the hypothesis that it underwent a southward expansion during the Pleistocene (Li et al., 2013). In this way we aimed to provide information on the broader issue of how Quaternary climatic oscillations affected the demographic and evolutionary history of plants adapted to semi-arid conditions in South East Asia. Our analysis of chlorotype diversity was successful in resolving two main clades within the C. indicum complex that largely distinguished the northern diploid taxon C. lavandulifolium from all other taxa. Most importantly, our analysis provided strong support for the hypothesis that the C. indicum complex underwent a massive range expansion in southern China during the Pleistocene and that this was attributable to an expansion in the range of the tetraploid form of C. indicum. Our results provided no evidence for diploid and other polyploid members of the complex having undergone similar range expansions during the same period.

Our analyses of chlorotype diversity combined with ITS variation were less successful in terms of clarifying the parentage of polyploid taxa and their origins via autopolyploidy and/or allopolyploidy. However, they confirmed that, although hybridization and interspecific gene flow were widespread in the complex, the different taxa within each ploidy level could be largely distinguished from each other according to chlorotype and/or ITS genotype, thus reinforcing their distinctiveness based previously on differences in morphology (Fig. S1) and genome size (Li et al., 2013).

Ancestral diploid lineages within the C. indicum complex

Our survey of plastid DNA variation revealed the presence of two major chlorotype groups, Hg1 and Hg2, within the C. indicum complex. Whereas diploid C. lavandulifolium, which is distributed in the northern part of the distribution of the complex, contains only Hg1 chlorotypes, all other members of the complex, which are distributed more centrally or in the south of China, contain mainly Hg2 chlorotypes. Thus, Hg1 chlorotypes were not found in the two other diploid members of the complex (diploid C. indicum and C. rhombifolium) nor in the tetraploid C. hypargyrum, and were present only rarely in tetraploid C. potentilloides, hexaploid C. vestitum and tetraploid C. indicum (Table S2). Based on chlorotype evidence and the current geographical distribution of taxa, it is feasible that an ancestral form of the complex diverged early in its evolution to give rise to C. lavandulifolium in northern China, and diploid C. indicum and C. rhombifolium in central China. Because the latter two taxa possess chlorotypes from different Hg2 subgroups, it is feasible that they also diverged from each other early in the evolution of the complex.

Interestingly, although C. lavandulifolium and diploid C. indicum are easily distinguished according to chlorotype, this is not so for ITS type. Our results showed that these two taxa share two nonadditive ITS types (AAT and GAT) and one additive ITS type ((A/G)AT) composed of the AAT and GAT types (Table S4). Only a few individuals of C. lavandulifolium possessed an ITS type ((A/G)(A/G)T) not found in diploid C. indicum. Sharing of ITS types between taxa and the joint possession of an additive ITS type may stem from incomplete lineage sorting and/or hybridization between taxa. Here it is feasible that AAT and GAT were present early in the evolution of the complex, with one possibly originating from the other in isolation by point mutation followed by homogenization of repeats as a result of concerted evolution. This may have occurred during divergence of C. lavandulifolium from diploid C. indicum. The finding that each of these genotypes is now present in both taxa may indicate that on occasion the two taxa have come into contact, hybridized, exchanged ITS types and formed additive ITS types.

An additive ITS type in diploid taxa may be produced by hybridization following a cross between individuals possessing different ITS types, in which case both parental types should be present among segregating offspring. Alternatively, it may be present in hybrid derivatives as a result of recombination between ribosomal DNA (rDNA) repeats containing different ITS sequences, in which case the additive sequence is inherited as a single unit and will not segregate (Fuertes Aguilar & Nieto Feliner, 2003). In both diploid C. indicum and C. lavandulifolium, the additive (A/G)AT type was never found in populations in combination with both putative parental types (AAT and GAT), indicating that it does not segregate and instead contains an intragenomic additive ITS sequence generated by recombination. Another way in which an additive ITS type might be generated in a diploid is through incomplete concerted evolution, resulting in partial homogenization of the parental copies. Although this cannot be ruled out as a cause of the additive ITS types detected in these two diploid species, hybridization is considered to be the more likely cause, certainly of the (A/G)AT additive type, given the sharing of AAT and GAT types by both species.

Whereas C. lavandulifolium and diploid C. indicum are very similar in regard to ITS type, the rare diploid taxon C. rhombifolium contains two ITS types unique to the taxon (GAC and (A/G)AC). Thus, the ITS evidence for this taxon agrees with the chlorotype evidence in indicating that C. rhombifolium diverged in isolation from the other two diploids early in the evolution of the complex and has remained genetically isolated thereafter.

Polyploid lineages within the C. indicum complex

All three tetraploid taxa, that is, the widespread tetraploid C. indicum and the two rare taxa, C. potentilloides and C. hypargyrum, were shown to be polymorphic for chlorotype, while C. indicum and C. potentilloides were also polymorphic for ITS type (Tables S1, S3). The different chlorotypes and ITS types recorded in C. potentilloides were shared with tetraploid C. indicum, indicating a close relationship and possible genetic exchange between these two taxa via introgressive hybridization. By contrast, C. hypargyrum was shown to be fixed for an ITS type (ATT) and almost fixed for a chlorotype (H15) not found in other members of the complex. Hence, this species may have diverged in isolation from other diploid and polyploid taxa and remained largely reproductively isolated from them thereafter.

Tetraploid C. indicum contains 31 of the 40 different chlorotypes detected in the complex and seven of 10 ITS types. The most common chlorotypes in this taxon were H11 (present in 46 individuals), H24 (in 35 individuals) and H37 (in 33 individuals) (Table S2), which represent the ancestral chlorotypes of subgroups Hg2a, Hg2b and Hg2d, respectively (Fig. 2b). The ancestral chlorotype (H34) of subgroup Hg2c was also relatively common in tetraploid C. indicum (present in 14 individuals). Only two of these ancestral chlorotypes were present in diploid members of the complex: H24 was very common in diploid C. indicum, while H37 was common in C. rhombifolium. The fact that the other two ancestral chlorotypes, H11 and H34, commonly found in tetraploid C. indicum, along with the majority of Hg2 tip chlorotypes resolved in this taxon, were not detected in any diploid taxon surveyed raises questions about the origins of these chlorotypes. It is feasible that at an earlier stage in the evolution of the complex they were present in diploid representatives, which have since become extinct, possibly as a result of competitive exclusion by tetraploid C. indicum (see Yang et al., 2006). Alternatively, it is possible that the chlorotypes unique to tetraploid C. indicum originated in this taxon during periods of population isolation and range fragmentation. The chlorotype network (Fig. 2b) shows that many chlorotypes differ from their most closely related chlorotypes by single mutations. This is true even for pairs of ancestral chlorotypes, and indicates that such divergence could have been easily generated during periods of population isolation in tetraploid C. indicum. Either way, the fact that the majority of tetraploid C. indicum populations examined are now polymorphic for chlorotypes shared with other populations of the taxon suggests that during the evolution and spread of tetraploid C. indicum there were frequent opportunities for gene flow to occur between populations.

What was evident for chlorotype variation in tetraploid C. indicum was also apparent, though to a lesser extent, for ITS variation. Thus, some ITS types found in this taxon (GGT and G(A/G)T) were not detected in diploid members of the complex and again sharing of different ITS types was common among populations of the tetraploid (Table 1, Table S4). Thus, our findings for this taxon confirm in broad terms those reported previously by Yang et al. (2006), albeit based on surveys of a different set of molecular markers, and fit a model of multiple differentiation and hybridization/polyploidization cycles that Yang et al. (2006) proposed for the evolutionary history of tetraploid C. indicum.

Based on their results, Yang et al. (2006) further proposed that some populations of tetraploid C. indicum originated by autoploidy while others were more likely to have originated by allopolyploidy. Thus, Yang et al. (2006) considered that a tetraploid population of C. indicum located at Shennongjia Mountain, Hubei Province, probably originated by autopolyploidy from a diploid population of C. indicum occurring nearby, because both populations shared many of the same nuclear and chloroplast markers. By contrast, there was no obvious association between markers present in a different diploid population of C. indicum, from Baohua Mountain, Jiangsu Province, and its nearest tetraploid population of C. indicum. Yang et al. (2006) proposed that this tetraploid population may have originated through allopolyploidy. However, our results failed to confirm an autopolyploid origin of the tetraploid population of C. indicum from Shennongjia Mountain (population 26; Table 1). Instead, they showed this population to be polymorphic for chlorotype, with some individuals possessing a chlorotype not detected in the closest diploid population of C. indicum (population 25). Consequently, at least some individuals in this tetraploid population could not have originated via autopolyploidy from diploid individuals nearby or, if they did, the genetic evidence for such an origin has been erased through subsequent hybridization and capture of another chlorotype.

Interestingly, hexaploid C. vestitum was also highly polymorphic for both chlorotype and ITS type. In this taxon, each of the three populations examined was dominated by a different high-frequency chlorotype, representing a different chlorotype subgroup. This might indicate that each population represents an independent origin of the taxon. Because two of the populations are polymorphic for chlorotype, it seems they would have been subject to interpopulation or interspecies gene flow following their origin (populations 38 and 39). For none of the three populations is it possible to determine accurately the original parentage of plants based on chlorotype and ITS data.

Pleistocene southward range expansion of tetraploid C. indicum

A southward range expansion of tetraploid C. indicum during the Pleistocene was strongly indicated by its unimodal mismatch distribution of chlorotype diversity and significantly negative values obtained for Fu's Fs and Tajima's D. By contrast, no such demographic expansions were indicated for either diploid C. indicum or C. lavandulifolium, while other diploid and tetraploid taxa of the complex currently exist only as single populations. Using a fast cpDNA mutation rate, the range expansion of tetraploid C. indicum was dated to have occurred between 720 and 780 kya ago, which is within the period from 600 to 800 kya when South East Asia was subjected to the effects of the longest and most severe Pleistocene glaciation recorded (Zhou & Li, 1998; Shi, 2002; Zheng et al., 2002). However, using a slow mutation rate the expansion was dated to have occurred between 72 and 78 kya, which is at the time of the last glaciation in the region (Shi, 2002). The climate of South East Asia was further affected by several more glaciations between the largest and the most recent glaciations (Zheng et al., 2002). Overall, our analysis clearly shows that the expansion occurred during the Pleistocene, and possibly during the most recent or the previous large glacial period when a marked increase in environmental aridity occurred and subtropical forests were replaced by steppe and semi-desert vegetation in large parts of southern China (Sun & Chen, 1991; Zhou et al., 1991).

There are many examples of polyploids being geographically more widespread than their diploid ancestors, and it has been argued that this is partly a result of a greater ability to colonize open habitats generated by changes in climate or other factors (Stebbins, 1984; Brochmann et al., 2004). The greater colonizing ability of polyploids has been attributed to several causes. First, allopolyploids contain two or more divergent genomes, which may provide adaptation to different environmental conditions. Secondly, levels of heterozygosity in both allo- and autopolyploids are likely to be greater than in their diploid ancestors (Brochmann et al., 2004), and this might further enable adaptation to a wider range of conditions and promote colonization of new open areas (Wu et al., 2010). Thirdly, polyploids can potentially generate considerable genetic and phenotypic novelty immediately following their origin through a range of diverse genetic and epigenetic mechanisms (Doyle et al., 2008; Soltis & Soltis, 2009; Abbott et al., 2013). Given the presumed greater ability of polyploids to colonize available habitats, it is somewhat surprising that the other polyploid taxa in the complex (C. hypargyrum, C. potentilloides and C. vestitum) were not as successful as tetraploid C. indicum in becoming widespread in central and southern China. A possible reason for this is that these polyploids originated later than tetraploid C. indicum and that suitable habitat for colonization had already been occupied by tetraploid C. indicum before they had a chance to spread. Alternatively, they may have become widespread at one stage, but then experienced a range contraction, in contrast to tetraploid C. indicum.

Comparisons of chlorotype diversity and ITS diversity within tetraploid C. indicum indicated that both types of diversity were greater in a group of centrally located populations (populations 17–23, 26, 27 and 29–35) than in southern populations (populations 1–16). This was particularly true for within-population chlorotype and ITS diversity. The latter finding does not fit the ‘stable rear edge’ model of Hampe & Petit (2005) for a species that historically has experienced cycles of range expansion and contraction in response to Pleistocene cycles of climate change. This model predicts that populations occurring at the rear edge of a species distribution will be of small size and geographically isolated from each other and hence should contain low levels of within-population genetic diversity, but high levels of between-population diversity. Rather, our results suggest that populations of large size were retained in central China during the southern expansion of tetraploid C. indicum, enabling maintenance of high levels of within-population diversity during this period. The fact that central China is also home to a rare diploid species, two rare tetraploid species, and a hexaploid species of the C. indicum complex emphasizes that it is an important region for the long-term conservation of genetic diversity within the complex.

Another question of interest concerns how tetraploid C. indicum has remained in southern China during the current post-glacial period when environmental aridity has decreased and subtropical forest is once again favored by prevailing conditions. It is feasible that tetraploid C. indicum has a particularly broad range of habitat tolerance as a result of its polyploid nature and/or has undergone rapid adaptation to changed conditions in southern China during post-glacial times and that this is partly responsible for it remaining widespread at the present time. It is of interest that plants comprising southern populations of tetraploid C. indicum tend to have more entire leaves than those comprising central populations (Fig. S1). It will be of interest to investigate whether these and other possible differences between centrally located and southern populations may be of adaptive significance. Certainly it would be of value to conduct transplant studies to determine if central and southern plants of tetraploid C. indicum are differentially adapted to present-day, local conditions in central and southern China.


The authors are grateful to Professor J. Q. Liu for his constructive comments and valuable discussion of the manuscript. We also would like to thank Z. C. Liang for collecting plant material, J. Q. Zhang for technical assistance and Y. S. Sun for help with data analysis. This research was supported by the National Natural Science Foundation of China (grants 30770142 and 30970207) and a China Scholarship Council award to J.L.