Molecular evidence for multiple polyploidization and lineage recombination in the Chrysanthemum indicum polyploid complex (Asteraceae)


Author for correspondence: Ji Yang Tel: +86 021 65643494 Fax: +86 021 65643494 Email:


  • • The Chrysanthemum indicum polyploid complex comprises morphologically differentiated diploids, tetraploids and hybrids between C. indicum and C. lavandulifolium. The relationships between species and cytotypes within this complex remain poorly understood.
  • • Random amplified polymorphic DNAs (RAPDs), intersimple sequence repeats (ISSRs) and chloroplast SSR markers were used to elucidate the genetic diversity and relationships of the C. indicum polyploid complex.
  • • Molecular analysis of three diploid and nine tetraploid populations provided strong evidence for recurrent origins and lineage recombination in the C. indicum polyploid complex. The high similarity in molecular marker profiles and cpDNA haplotypes between the diploids and tetraploids distributed in the Shen-Nong-Jia Mountain area of China suggested an autopolyploid origin of the tetraploids, while the tetraploids from other populations may have originated via allopolyploidization. Lineage recombination was revealed by the extensive sharing of chloroplast haplotypes and genetic markers among the tetraploid populations with different origins.
  • • Multiple differentiation and hybridization/polyploidization cycles have led to an evolutionary reticulation in the C. indicum polyploid complex, and resulted in the difficulties in systematic classification.


Polyploidy is a widespread phenomenon in flowering plants, with perhaps as many as 70% of species having experienced at least one recognizable genome duplication (Masterson, 1994; Soltis & Soltis, 1999). A growing body of evidence indicates that polyploid evolution is an ongoing, dynamic process in plant speciation, and many taxonomically recognized polyploid species have originated multiple times (Soltis & Soltis, 1993, 1999). Recurrent polyploidization involving genetically different diploids can create a series of genetically distinct polyploid populations. Populations of independent origin can subsequently come into contact and hybridize, generating new genotypes by recombination (Doyle et al., 1999; Soltis & Soltis, 1999). Recurrent polyploid formation and the hybrid nature of many polyploids thus provide multiple avenues for the introduction of genetic variation into populations, which might contribute to the success and persistence of new polyploid species (Soltis & Soltis, 2000). Early cytological and morphological studies have suggested that both polyploid formation and hybridization are prevalent in the genus Chrysanthemum (Asteraceae) (according to a recent decision of the International Commission for Botanical Nomenclature the generic name Chrysanthemum has now been reserved for all taxa of the former Dendranthema) (Dowrick, 1952; Shimotomai et al., 1956; Tanaka, 1959, 1960; Nakata & Tanaka, 1987; Tsukaya, 2002; Kim et al., 2003). Several species are known to contain both polyploids and hybrids, forming species complexes (Lee, 1975; Nakata et al., 1987; Hotta et al., 1996). However, very little is known about the genetic diversity and relationships within and among the closely related taxa in each complex. The processes that have given rise to polyploid hybrid taxa in Chrysanthemum remain poorly understood.

Chrysanthemum indicum L., along with C. sinense, C. erubescens, C. ornatum, C. japonense, C. makinoi, C. chanetii, C. zawadskii and C. vestitum (Stapf, 1933; Kitamura, 1948; Bailey, 1949; Dowrick, 1953; Chen, 1985; Nakao, 1986; Dai & Chen, 1997; Fukai, 2003), is widely believed to be a wild progenitor of the cultivated chrysanthemum, and has long been used in the development of new forms of garden chrysanthemum. The C. indicum polyploid complex comprises both the diploid (2n = 2x = 18) and tetraploid (2n = 4x = 36) cytotypes of C. indicum and the closely related diploid species C. lavandulifolium (Lin & Shi, 1983; Du et al., 1989; Li et al., 1991; Wang et al., 1993; Zhou & Dai, 2002). Two types of diploid have been found in C. indicum that differ in morphology and karyotype (Wang et al., 1993). One is confined to Nanjing, in Jiangsu Province of China (referred to as the NJ diploid); the other is endemic to a limited area of Shen-Nong-Jia Mountain, in Hubei Province of China (referred to as the SNJ diploid). In contrast to the fragmented distribution of diploid populations, the tetraploids of C. indicum are widespread and continuously distributed throughout southern China. In some systems, morphological characters such as leaf morphology and the number and size of heads are used as diagnostic traits to distinguish diploids from tetraploids in C. indicum, but in many cases they are indistinguishable because the tetraploids display a great diversity of morphology. Meanwhile, morphologically intermediate individuals have been found between C. indicum and C. lavandulifolium (Lin & Shi, 1983), suggesting the occurrence of interspecific hybridization between these two species. Nakata et al. (1987) proposed that the diploid C. lavandulifolium is one of the putative ancestors of allopolyploid forms of C. indicum, based on conventional karyotypic parameters.

Because of the great variation in morphology, the occurrence of interspecific hybridization, and the absence of an unambiguous set of criteria for designating a true species, considerable differences exist among systematists with regard to classification of the C. indicum complex. For example, in some systems the NJ diploids are classified as an independent species, named C. nankingense (Bremer & Humphries, 1993; Dai & Chen, 1997). However, this name is treated as a synonym rather than a distinct species in the Flora Republicae Popularis Sinicae (Lin & Shi, 1983). Molecular markers developed during the past two decades have overcome many of the problems associated with phenotype-based classification, and have been widely used to elucidate the genetic relationships within species complexes (Soltis & Soltis, 1995; Lowe & Abbott, 1996; Brochmann et al., 1998; Cook et al., 1998; Doyle et al., 1999; Hedren et al., 2001; Wallace, 2003; Awasthi et al., 2004; Budak et al., 2004; Souframanien & Gopalakrishna, 2004; Guo et al., 2005). In this study, PCR-based marker assays including random amplified polymorphic DNAs (RAPDs) and intersimple sequence repeats (ISSRs) were employed to assess the genetic diversity and relationships of the C. indicum complex. RAPD provides a simple, PCR-based molecular tool for the evaluation of genetic variation, while ISSR generates a large number of markers by using SSR repeat-anchored primers that target multiple microsatellite loci distributed throughout the genome (Awasthi et al., 2004; Budak et al., 2004).

To determine the origin of the tetraploid C. indicum and the relatedness of different diploids, genetic variation at chloroplast microsatellites (chloroplast simple sequence repeats, cpSSRs) was also surveyed. Chloroplast SSRs are a new class of cytoplasmic markers, which feature variable numbers of mononucleotide repeats and allow the high-resolution analysis of chloroplast genomes (Powell et al., 1995; Provan et al., 2001). They are useful for population genetic and phylogenetic analysis between closely related taxa, especially when conventional chloroplast sequences are not variable enough to resolve relationships (Powell et al., 1996; Provan et al., 1997; Bucci et al., 1998; Ishii et al., 2001; Provan et al., 2001; Hashimoto et al., 2004; Provan et al., 2004; Molina-Cano et al., 2005). As microsatellites are generally assayed as differences in PCR product lengths using gel electrophoresis, the true nature of mutation in microsatellite analysis is usually unclear. It is difficult to determine whether the size differences are caused by the length of the SSR itself or that of its flanking sequence. We therefore performed DNA sequence analysis of the SSR–PCR products to exclude ambiguity resulting from size homoplasy (Doyle et al., 1998), where alleles of identical size contain different internal mutations.

The objectives of this study were to characterize the distribution of genetic diversity within and among various cytotypes and elucidate the species relationships within the C. indicum complex; to identify the original diploid donor of the C. indicum tetraploids and document the actual extent of recurrent formation of polyploids; and to assess the species delimitations and establish a more precise systematic treatment of the group. This investigation has provided us with a basis for better understanding of polyploid evolution and species differentiation in the C. indicum complex.

Materials and Methods

Plant materials

A total of 110 individuals were sampled from nine tetraploid and two diploid populations of Chrysanthemum indicum L. (10 individuals for each population). The tetraploid populations were selected from across the distribution range of C. indicum (Fig. 1). Both NJ and SNJ diploids have a very limited range and thus form single natural populations, NJ(2x) and SNJ(2x), respectively. C. lavandulifolium is a diploid species recorded in Sichuan, Hubei, Jiangxi, Guangdong and other provinces of southern China (Lin & Shi, 1983). However, we failed to collect individuals with characters typical of C. lavandulifolium in these provinces. It is unclear whether C. lavandulifolium has diverged greatly in southern China, or its distribution has contracted to northern China. We thus included a population from Beijing (LAV) for comparative analysis. As the taxa may spread clonally via rhizomes, we collected individuals that were at least 5 m apart to avoid collecting multiple individuals of the same genotype.

Figure 1.

Distribution within China of the 12 populations analysed in this study. NJ(2x), 2x population from Nanjing, Jiangsu Province; SNJ(2x), 2x population from Shen-Nong-Jia Mountain, Hubei Province; LAV(2x), 2x population of C. lavandulifolium; NJ(4x), 4x population from Nanjing, Jiangsu Province; HZ(4x), 4x population from Hangzhou, Zhejiang Province; GZ(4x), 4x population from Guangzhou, Guangdong Province; JJ(4x), 4x population from Jiujiang, Jiangxi Province; LS(4x), 4x population from Lushan Mountain, Jiangxi Province; JFS(4x), 4x population from Jin-Fu-Shan Mountain, Chongqing; SNJ(4x), 4x population from Shen-Nong-Jia Mountain, Hubei Province; WHA(4x), 4x population from Wuhan, Hubei Province; WHB(4x), 4x population from Wuhan, Hubei Province.

Young leaves of an average of 10 individuals from each population were collected and dried with silica gel. Total DNA was extracted from 50–100 mg dried leaf material using the method of Doyle & Doyle (1990). Voucher specimens were deposited at the herbarium of Peking University (PEY).

RAPD and ISSR amplification and detection

RAPD  10 individual plants were analysed from each population. A prescreening of 100 RAPD primers (Operon Technologies, Almeda, CA, USA) was performed, and nine (Table 1) were selected for further analysis based on the number of reproducible polymorphic bands generated. PCR amplifications were performed in a reaction volume of 20 µl containing 20 mm Tris–HCl, 50 mm KCl, 2 mm MgCl2, 8 pmol primer, 0.2 mm of each dNTP, 20 ng of template DNA and 1 U Taq polymerase (Huabei BioEngineering Company, Beijing, China). PCR reactions were carried out in a PTC-100 Programmable Thermal Controller (MJ Research, Waltham, MA, USA) with the following cycle profile: one cycle of 1 min denaturing at 94°C, and 40 cycles of 45 s at 94°C, 45 s at 37°C, 2 min at 72°C, followed by a 10-min final extension at 72°C. Negative controls were used in all experiments with sterile H2O to replace the DNA template. Amplification products were separated by gel electrophoresis on 1.5% agarose gels in 1 × TAE buffer, then stained in ethidium bromide (2 µg ml−1) for 15 min and destained for 30 min. Gels were visualized and photographed under UV light.

Table 1.  Random amplified polymorphic DNAs (RAPD) and inter-simple sequence repeat (ISSR) primers used in this study
RAPD primerISSR primer

To ensure reproducibility and consistent results, all working solutions were prepared in advance using the same reagents and high-quality ddH2O. The optimized PCR conditions were strictly followed in all reactions. Each PCR amplification and gel electrophoresis was repeated twice to ensure repeatability. Only sharp, reproducible bands were scored and used in genetic analysis.

ISSR  100 different ISSR primers (University of British Columbia, Vancouver, Canada) were tested, nine of which (Table 1) were used for screening the full set of accessions included in this study. Amplification reactions were carried out in 20 µl containing 10 mm Tris–HCl, 50 mm KCl, 2.5 mm MgCl2, 10 pmol primer, 0.2 mm of each dNTP, 20 ng of template DNA and 1 U Taq polymerase (Huamei BioEngineering Company, Beijing, China). PCR reactions were performed using the following amplification conditions: initial denaturation at 94°C for 2 min, followed by 40 cycles of 94°C for 30 s, annealing at 51.5–53.5°C (optimized for each individual primer) for 45 s, extension at 72°C for 2 min, and a final extension at 72°C for 7 min. Amplification products were separated on 2% agarose gels and visualized under UV after staining with ethidium bromide.

Compared with RAPD primers, the ISSR primer sequence is larger, allowing for a higher primer-annealing temperature that may result in greater band reproducibility than RAPD markers. However, we found that the ISSR fragments were sensitive to small changes in annealing temperature. We thus optimized the annealing temperature for each primer, and performed duplicate PCR reactions for each sample to ensure that we generated repeatable profiles.

Chloroplast SSR analysis

All 10 pairs of conserved chloroplast microsatellite primers (CCMP 1–10) proposed by Weising & Gardner (1999) and three individuals from three populations were used for an initial screening to investigate variation. Two pairs of primers (ccmp2 and ccmp5) were then used to screen all individuals sampled, as they detected a relatively high level of polymorphism. PCR was performed in 20 µl reaction mixture containing 10 mm Tris–HCl, 50 mm KCl, 2.5 mm MgCl2, 0.5 pmol each of forward and reverse primer, 0.2 mm each of dNTP, 20 ng of template DNA and 1 U Pfu DNA polymerase (TIANWEI Co., Beijing, China). After an initial denaturation at 94°C for 5 min, PCR was conducted for 30 cycles, each consisting of 94°C for 1 min, 50°C for 1 min, and 72°C for 1 min. Final elongation was at 72°C for 8 min. All PCR products were cloned into a plasmid vector (pGEM-1, Promega) and sequenced in both directions using an ABI 377 automatic sequencer. Five individual plants were analysed from each population. To reduce the effect of ‘stuttering’ in PCR amplification, we used the high-fidelity proofreading polymerase Pfu in place of Taq in all PCR reactions. Meanwhile, we repeated the PCR reactions and sequencing of the samples from Jin-Fu-Shan (JFS) and Guangzhou (GZ), which showed more variation in haplotype than other populations, and found that the results were reproducible.

Data analysis

Bands were scored as present (1) or absent (0) for each primer pair. Monomorphic and minor polymorphic bands were excluded from the data sets, because monomorphic bands do not contribute to the genetic distance calculation and minor polymorphic bands can arise artefactually from differences in genomic DNA quality and other factors. Only distinct, reproducible polymorphic bands were considered in the genetic analysis.

Nei's (1973) gene diversity (H) within populations and Nei's (1978) unbiased measure of genetic distance between populations were calculated with popgene ( Phenograms based on Nei's genetic distances were constructed using the unweighted pair group method with arithmetic mean (UPGMA) method implemented in popgene, to demonstrate the relationships among populations. Correlation between the RAPD and ISSR data matrices was estimated by means of the Mantel matrix-correspondence test (Mantel, 1967), using the program zt (Bonnet & Van de Peer, 2002), which yields a product moment correlation (r) to measure the relatedness between two matrices. Principal component analysis based on genetic similarity matrices was performed using the ntsyspc software package (Rohlf, 1998) to obtain a graphical representation of the relationship structure of all accessions.

To elucidate hybrid and polyploid reticulation in the C. indicum complex, we determined the exclusive or ±stabilized RAPD and ISSR bands of each diploid, then traced these bands to different tetraploid populations. The joint presence of specific bands from the suspected parental diploids in different tetraploids reveals phylogenetic reticulations in this complex (Guo et al., 2005). The population-specific bands are those fixed in all the individuals of a population, or at least with ≥ 70% occurrence. Rare or partly stabilized (< 70%) bands were excluded.

As the chloroplast genome is haploid and does not undergo recombination, it can be viewed as a single locus (Molina-Cano et al., 2005). The alleles detected at each locus were thus combined to generate a multilocus haplotype of each individual. Relatedness between haplotypes was represented by a statistical parsimony network, generated using the program tcs (Clement et al., 2000). Mean genetic diversity within populations (Hs), total genetic diversity (Ht) and genetic differentiation among populations (Gst), were estimated following Pons & Petit (1995, 1996) using the program cpssr ( Patterns of genetic relatedness between populations were examined using cluster analysis in the ntsyspc software package, based on the proportion of shared alleles measure generated with microsat (


RAPD and ISSR markers

A total of 127 polymorphic bands were generated from the nine RAPD primers used. Of these, 81 were shared among diploids and tetraploids; 39 were uniquely present in tetraploids; and seven were uniquely present in diploids. The percentage of polymorphic bands for a single population varied greatly among populations, with the tetraploid population from Hangzhou (HZ) showing the highest intrapopulation variation (H = 0.139 ± 0.193) and the population from Guangzhou (GZ) the lowest (H = 0.045 ± 0.134) (Table 2). The genetic distances between populations ranged from 0.082 (between populations WHA and WHB) to 0.334 (between populations NJ(2x) and NJ(4x); Table 3). Cluster analysis of the RAPD data using UPGMA (not shown) revealed a close relationship between the diploid population SNJ(2x) and the tetraploid populations SNJ(4x) and JFS, whereas the other two diploid populations, NJ(2x) and LAV, were genetically distant from the nearest tetraploid populations. plants from the tetraploid population NJ(4x) showed a significant separation from other tetraploid populations, and formed a single clade in the phenogram. The same general pattern expressed in the phenogram was also revealed by principal component analysis (PCA) (Fig. 2a), with the individuals from the population NJ(4x) being clearly separated from others, and the plants from SNJ(2x), SNJ(4x) and JFS forming a close group. The individuals from other tetraploid populations appear to be interspersed.

Table 2.  Nei's genetic diversity within populations (Nei, 1973)
  1. RAPD, random amplified polymorphic DNAs; ISSR, intersimple sequence repeats.

NJ(4x)0.083¡ ± 0.1730.119¡ ± 0.164
GZ0.045¡ ± 0.1340.054¡ ± 0.146
HZ0.139¡ ± 0.1930.068¡ ± 0.141
JJ0.126¡ ± 0.2070.039¡ ± 0.122
LS0.104¡ ± 0.1890.108¡ ± 0.177
WHA0.081¡ ± 0.1690.055¡ ± 0.146
WHB0.107¡ ± 0.1860.098¡ ± 0.176
NJ(2x)0.084¡ ± 0.1770.043¡ ± 0.122
LAV0.112¡ ± 0.1860.099¡ ± 0.175
JFS0.126¡ ± 0.1790.100¡ ± 0.175
SNJ(4x)0.112¡ ± 0.1740.056¡ ± 0.138
SNJ(2x)0.079¡ ± 0.1550.033¡ ± 0.116
Table 3.  Genetic distance between populations based on random amplified polymorphic DNAs (RAPDs) (above diagonal) and intersimple sequence repeat (ISSR) data (below diagonal)
NJ(4x) 0.3050.2850.2980.3000.2420.2380.3340.2920.2850.3030.313
GZ0.143 0.1460.2010.2460.2510.2160.3180.3120.2450.2420.243
HZ0.1430.143 0.1080.1800.1960.1310.2100.2190.1510.1640.214
JJ0.2160.1660.172 0.2600.2380.1870.2630.2340.2100.2350.262
LS0.1010.1890.1500.186 0.2800.1850.250.3050.1670.1860.216
WHA0.1630.2010.1770.2180.143 0.0820.2890.2930.2190.2600.252
WHB0.1280.1680.1240.1410.1440.123 0.2200.2010.1460.1700.198
NJ(2x)0.2250.2900.2110.3040.2680.2640.186 0.2260.2520.2150.271
LAV0.2230.2470.2130.2690.2040.2130.1910.268 0.2230.2290.242
JFS0.1630.2120.1890.1960.1610.1840.1450.2490.204 0.0820.099
SNJ(4x)0.1380.1710.1660.220.1610.1750.1410.2690.2280.167 0.125
Figure 2.

Three-dimensional plot of principal component analysis based on (a) random amplified polymorphic DNAs (RAPD); (b) intersimple sequence repeat (ISSR) banding patterns, showing relationships among the 12 populations included in this study. Closed circles, 2x individuals; open circles, 4x individuals.

Analysis of all accessions with nine ISSR primers identified a total of 163 reproducible polymorphic fragments. Among these, 74 were shared among diploids and tetraploids, but were not present in all accessions. Seventy-nine fragments were uniquely present in the tetraploid genomes, and 10 fragments were unique to the diploid genomes. The tetraploid population from Nanjing exhibited the highest intrapopulation variation (H = 0.119 ± 0.164) and the diploid population SNJ(2x) the lowest (H = 0.033 ± 0.116) (Table 2). The highest value of genetic distance among populations was 0.304, between the tetraploid population JJ and the diploid population NJ(2x). The lowest value of 0.101 was between the tetraploid populations NJ(4x) and LS (Table 3). Both UPGMA (not shown) and PCA (Fig. 2b) of the ISSR data showed that the diploid population SNJ(2x) was closely related to the tetraploid population SNJ(4x), while the individuals from the other two diploid populations, NJ(2x) and LAV, formed two dense clusters and were clearly separated from other populations. The results also showed that the tetraploid population NJ(4x) had a close relationship with other tetraploid populations, but the tetraploid population JJ formed a distinct group. This pattern is partially inconsistent with the results obtained using RAPDs. To obtain a more robust comparison, the Mantel matrix-correspondence test was used to evaluate the correlation between the RAPD and ISSR distance matrices. The correlation coefficient was 0.392 (P = 0.042), indicating that there is a faint concordance between RAPDs and ISSRs.

The genetic links between diploid and tetraploid populations were assessed further with the joint presence of diploid-specific bands in different tetraploid populations. As shown in Fig. 3, the SNJ diploid-specific bands mainly appeared in SNJ(4x) and JFS, but were also present in LS, JJ, NJ(4x) and WHA with lower frequencies. The C. lavandulifolium-specific bands were widespread in tetraploid populations, except the populations SNJ(4x) and JFS for RAPDs, but only SNJ for ISSRs (Fig. 3). Compared with the SNJ diploid and C. lavandulifolium, the NJ diploid-specific bands were shared by fewer tetraploid populations, and most appeared in WHA (RAPDs) and WHB (ISSRs) (Fig. 3). Tetraploids NJ (RAPDs, ISSRs), JJ (RAPDs), and LS (ISSRs) had only one NJ diploid band in common (Fig. 3). It is notable that the tetraploid population SNJ(4x) shared diploid-specific bands exclusively with SNJ diploids for both RAPD and ISSR markers, and only the C. lavandulifolium-specific bands were fixed in the tetraploid population from Guangzhou (GZ) for RAPD and ISSR markers. However, populations HZ and WHB had C. lavandulifolium-specific bands fixed in the RAPD markers only (Fig. 3).

Figure 3.

Bands, specific to individual diploid populations, linking 2x (quadrangles) and 4x (ellipses) taxa in the Chrysanthemum indicum polyploid complex, demonstrating the reticulate relationships among all taxa. Numbers on lines indicate the number of shared bands.

We also examined the relevant bands shared by different pairs of diploids. It was revealed that five RAPD and 10 ISSR bands were shared between the NJ diploid and C. lavandulifolium, while the SNJ diploid shared one RAPD and four ISSR bands with the NJ diploid, and five RAPD and five ISSR bands with C. lavandulifolium.

Chloroplast SSRs

Sequencing of the microsatellite-containing PCR products revealed five polymorphic microsatellite loci in the material studied. A total of 12 alleles were detected, and the number of alleles per locus ranged from two to three. When all alleles were combined, 10 multilocus haplotypes were identified. Of these, four (A, D, E and J) were common to both diploid and tetraploid populations (A in populations SNJ(2x), SNJ(4x), NJ(4x), LS and JFS; D in populations SNJ(2x), NJ(2x), LAV, SNJ(4x), LS, JFS, JJ and GZ; E in populations SNJ(2x) and JFS; J in populations LAV and GZ); and six (B, C, F, G, H and I) were exclusively present in the tetraploid populations. Of the six haplotypes found exclusively in tetraploids, haplotypes B, H and I were observed in one or a few individuals in single populations, while the remainder were shared among different populations. The most common haplotype, D, was detected in 19 individuals from nine populations. The second most frequent haplotype, G, was found in only five tetraploid populations.

The haplotype network in Fig. 4 indicates the minimum numbers of evolutionary events separating the haplotypes. Most haplotypes were related to the two most common haplotypes, D and G, by a single microsatellite length mutation. This pattern is consistent with proposal of Crandal & Templeton, 1993) that the common haplotypes are at interior nodes and rare haplotypes at the tip. The haplotypes that occurred at low frequency are probably derived from the major haplotypes, and represent more recent mutations.

Figure 4.

Relatedness between chloroplast haplotypes detected in the Chrysanthemum indicum polyploid complex, represented in a statistical parsimony network. The sizes of circles are roughly proportional to haplotype frequencies (%). The numbers on lines indicate the minimum number of evolutionary events separating haplotypes.

A high level of diversity in haplotype composition was detected in the populations analysed (ht = 0.821, SE = 0.048). The cpDNA diversity appears to be distributed less between populations (Gst = 0.243, SE = 0.078) than within populations (hs = 0.621, SE = 0.079). It was noted that, with the exception of NJ(2x), all populations have multiple haplotypes, and most of the haplotypes were not unique to a particular population. In the tetraploid population JFS, we found four different cpDNA haplotypes (2G, 1A, 1D and 1E) in a sample size of five. Each of these haplotypes was also shared by other populations.

Using cluster analysis of cpssr data, based on the proportion of shared alleles, we were able to differentiate three main population groups (Fig. 5): group 1, comprising three diploid populations (NJ2x, SNJ(2x) and LAV) and two tetraploid populations (SNJ(4x) and LS); group 2, comprising five tetraploid populations (GZ, HZ, WHB, JJ and JFS); and group 3, comprising a single tetraploid population, NJ(4x). The most common haplotype, D, was dominant in group 1, whereas the second most frequent haplotype, G, was prevalent in group 2. The population NJ(4x) formed a distinct group possessing the population-specific haplotype B.

Figure 5.

Dendrogram derived from UPGMA analysis of the chloroplast simple sequence repeat (cpSSR) data, showing genetic relatedness between populations.


PCR-based marker analyses used in this study revealed a high level of genetic diversity in the C. indicum polyploid complex. Both RAPD and ISSR analyses showed a close relationship between the diploid population SNJ(2x) and the tetraploid population SNJ(4x). Among all the RAPD and ISSR bands observed in SNJ(2x), 69.5% were also present in SNJ(4x). Furthermore, the individuals from SNJ(4x) shared bands specific to single diploid populations exclusively with the diploids from SNJ(2x) (Fig. 3). There is thus strong evidence that the SNJ diploids were involved in the origin of the tetraploids distributed in the Shen-Nong-Jia Mountain area. Additional evidence comes from the plastotype analysis, which showed that the haplotypes possessed by the SNJ diploids and tetraploids overlapped completely. All haplotypes identified in the tetraploid population (4D and 1A) were also found in the diploid population (3D, 1A and 1E). This strong genetic similarity, coupled with the distribution pattern of tetraploids relative to diploids, appears to suggest an autopolyploid origin of the tetraploids distributed in this area.

Traditionally, autopolyploidy was considered extremely rare in natural populations because autopolyploids would have reduced fertility caused by irregular distribution of chromosomes caused by unequal separation of multivalents (Stebbins, 1947). However, in recent decades a growing number of autopolyploids have been documented in the plant kingdom (Lumaret et al., 1989; Soltis & Soltis, 1993; Dijk & Bakx-Schotman, 1997; Ramsey & Schemske, 1998). Molecular data have also provided compelling genetic evidence that autopolyploids, like allopolyploids, have a fitness advantage over their diploid progenitors because of their increased genetic diversity (Soltis & Soltis, 2000). Further studies of chromosome association at meiosis and comparative mapping are needed to ascertain whether the tetraploids of SNJ(4x) originated via autopolyploidization or allopolyploidization. However, genome restructuring following polyploidization may have jumbled chromosome configurations (bivalent vs multivalent) during meiosis and collinearity between genomes, because karyotype analysis showed that all tetraploids distributed in this area had a diploidized karyotype (data not shown). Genomic diploidization is thought to be an important step in adaptation of autopolyploids (Stebbins, 1971), although only a few cytogenetic or inheritance studies have documented this process directly (Sybenga, 1996; Qu et al., 1998; Weiss & Maluszynska, 2000). The 2x and 4x cytotypes of Shen-Nong-Jia Mountain area are ecologically differentiated. While the tetraploids are widely distributed in medium- and low-altitude regions of Shen-Nong-Jia Mountain, the diploids are strictly limited to the summit area. This adaptive differentiation is probably caused by the enhanced genetic divergence between the diploid and diploidized tetraploid genomes, which enables the tetraploids to occupy a new environment relative to their diploid parents.

Both RAPD and ISSR analyses demonstrated that the individuals from two other diploid populations, NJ(2x) and LAV, were well separated from each other and from tetraploid populations, suggesting high levels of genetic differentiation between them. However, striking signs of hybrid introgression become apparent when tracing the diploid-specific bands linking 2x and 4x taxa included in this complex. Figure 3 shows that the C. lavandulifolium-specific RAPD and ISSR bands were widely shared among tetraploid populations. Similarly, the NJ diploid-specific bands also appeared in the tetraploid populations NJ(4x), JJ, LS, WHA and WHB. It is thus reasonable to postulate that the NJ diploids and the plants of C. lavandulifolium have also contributed to the formation of the widespread tetraploids of the C. indicum complex.

An important evolutionary aspect of the RAPD and ISSR profiles from the 4x taxa is the appearance of bands that were obviously lacking among the 2x taxa, but were shared among tetraploid populations. These bands may result from genome reorganization following polyploidization, but we cannot exclude the possibility that they came from other diploid donors of these tetraploids that are different from the present-day representatives of the putative parental groups. The clear separation between the tetraploids included in this study and the diploids from Nanjing and C. lavandulifolium could thus be an indication of the rapid divergence of the tetraploids from their diploid progenitors following polyploidization, but may also indicate the hybrid nature of these tetraploids: these tetraploids originated from allopolyploidization following hybridization between diploids with different genetic make-ups. Some of the diploid progenitors may not have been sampled in the present study, or may have become extinct.

cpDNA analysis confirmed the existence of other diploid progenitors, which must have participated in the formation of the widespread tetraploids in the C. indicum polyploid complex. A total of 10 different cpDNA haplotypes were found in this complex. Four (A, D, E and J) were found in the present 2x taxa, but the remaining six haplotypes occurred only in plants from tetraploid populations, including the second most frequent haplotype, G. Based on cpDNA diversity, the populations clustered into three main groups (Fig. 5). All three 2x taxa included in this study were in the group with the dominant haplotype D. The other two groups possessed the dominant haplotype G, and the population-specific haplotype B, respectively. If we assume that tetraploids arose from diploids with the same cpDNA haplotype, then there are at least three independent origins of the tetraploids included in the C. indicum complex, and the diploid progenitors with haplotypes G and B may have been overlooked or have become extinct.

In the context of independent origin, the C. indicum complex comprises at least three polyploid lineages of independent formation. The present data, however, showed that these lineages were not clearly separated. All tetraploid populations in this study were a mixture of a dominating haplotype and one to three less frequent haplotypes, which were often shared by populations from different lineages. PCR-based molecular marker analysis also showed that 53% of the RAPD and 44% of the ISSR bands observed in tetraploids were shared by populations from different lineages. The presence of identical fixed haplotypes or genetic markers in widely separated populations can be indicative of (i) convergence; (ii) common ancestry; or (iii) hybridization/introgression. Convergence appears to be the least likely explanation in this case as identical mutations are rare events, and cpDNA and genetic marker sharing is extensive in this study. Common ancestry is possible, but if each molecular marker profile or each tetraploid haplotype is considered to represent the product of an independent allopolyploidization/autopolyploidization event, many more incidences of independent origin need to be assumed to take place in disjunct areas. This may lead to an overestimate of the number of times a polyploid has evolved de novo because other factors, such as gene flow between polyploid populations of independent origin, could also lead to wide sharing of a specific haplotype or genetic marker among different polyploid populations. The distribution pattern of the SNJ diploid-specific RAPD and ISSR bands appears to be a consequence of lineage recombination occurring among polyploids with different origins. In addition to the tetraploid population SNJ(4x), the SNJ diploid-specific bands were also found in populations JFS, LS, JJ, NJ(4x), WHA and WHB. It is unlikely that the SNJ diploids were directly involved in the formation of the tetraploids from those populations because they are ecogeographically isolated. The autotetraploids are, however, widespread in the medium- and low-altitude region of Shen-Nong-Jia Mountain, and thus have the chance to come into contact with other tetraploid populations with different origins. The wide sharing of the SNJ diploid-specific bands among various tetraploid populations probably resulted from the secondary contact and hybridization between SNJ(4x) and other tetraploid populations with independent origins. Otherwise, a very large geographical distribution of the ancestral diploid, or extensive past migration events among polyploid populations of separate origin, has to be postulated.

It is of interest to note that, although the present data indicated the existence of gene flow between SNJ(4x) and other tetraploid populations, the plants from SNJ(4x) were completely free of the NJ diploid- and C. lavandulifolium-specific bands, as well as the less frequent haplotypes appearing in other tetraploid populations. In contrast, the plants from JFS possess some of the C. lavandulifolium-specific ISSR bands, although they are even further away than SNJ(4x) from C. lavandulifolium and other tetraploid populations. This pattern seems to suggest an asymmetrical hybridization with unidirectional gene flow from SNJ(4x) to other tetraploid populations. It is not clear what factors could be responsible for such an unusual pattern of evolution. The asymmetrical hybridization appears not to exist between JFS and other tetraploid populations, thus the plants from JFS might acquire the C. lavandulifolium-specific bands via hybridization with those from other tetraploid populations.

In conclusion, the C. indicum complex consists of diploids and tetraploids with polyphyletic origins. Recurrent polyploidization involving genetically different diploids has created a series of genetically distinct tetraploid populations. Subsequent contacts between polyploid populations with independent origins generated even more genetic and morphological complexity. Such differentiation–hybridization cycles (Ehrendorfer, 1959) have led to difficulties in delimitation of the diploid and tetraploid taxa included in the C. indicum complex. The diploids distributed in Nanjing have sometimes been recognized as an independent species, named C. nankingense (Bremer & Humphries, 1993; Dai & Chen, 1997), but this name is treated as a synonym of C. indicum in the Flora Republicae Popularis Sinicae (Lin & Shi, 1983). Some recent studies (Dai et al., 1995; Dai & Chen, 1997; Wang et al., 2002; Zhou & Dai, 2002) have supported recognition of C. nankingense at the specific level based on morphological and cytological analyses. Molecular data from this study demonstrated a significant genetic distinction between the NJ diploids and other diploids or tetraploids included in this study. Although the plants from NJ(2x) and NJ(4x) grow in close proximity, they did not have identical haplotypes, and showed great difference in the RAPD and ISSR profiles, suggesting that there is little or no gene flow between them. NJ(2x) is the only population possessing a single cpDNA haplotype in this study. The results of this study thus also support recognition of C. nankingense at the specific level, as it not only comprises a morphologically recognizable entity, but also has diverged genetically from the closely related species C. indicum.

With regard to the relationships between C. lavandulifolium and C. indicum, the present study provides strong evidence that C. lavandulifolium is a potential diploid progenitor of the tetraploids of C. indicum. Chrysanthemum lavandulifolium and C. indicum are very similar in morphological features, but plants of C. lavandulifolium can be distinguished from those of C. indicum by their bipinnate leaves and relatively small flower heads. However, plenty of morphologically intermediate individuals were found between C. lavandulifolium and C. indicum, which blurred species boundaries. Chrysanthemum lavandulifolium has been recorded in both southern and northern China. The present study also shows that C. lavandulifolium was involved in the formation of the tetraploids of the C. indicum complex. We found, however, that currently plants with typical characters of C. lavandulifolium are mainly distributed in northern China, while the tetraploid populations are located mostly in southern China. This is contrary to the common prediction that existing tetraploid populations originated from the geographically closest diploids. We cannot exclude the possibility that the C. lavandulifolium diploids involved in formation of the tetraploids in this complex have been overlooked, or have become extinct. But the lack of congruence between the geographical distribution of tetraploid origins and their diploid parents is more likely to result from differences in the ecological attributes of diploids and polyploids, because tetraploids of C. indicum appear to be preadapted to invade diploid populations. The diploids of C. indicum still coexist with the tetraploid populations, but are restricted to very narrow regions. The evolutionary significance of polyploids has often been attributed to their increased ability to invade new habitats and fill new niches, allowing for expansion of the geographical range of a species (Roose & Gottlieb, 1976; Brammall & Semple, 1990; Segraves et al., 1999). The C. indicum polyploid complex, however, provides an example of tetraploid range expansion via exclusion of diploids through competition.


This work was supported by the National Key Project for Basic Research (973) (G2000046804) and a National Natural Science Foundation of China grant (39970051) to J.Y.