Direction and extent of organelle DNA introgression between two spruce species in the Qinghai-Tibetan Plateau


  • Fang K. Du,

    1. Molecular Ecology Group, Key Laboratory of Arid and Grassland Ecology, Lanzhou University, Lanzhou, Gansu 730000, China
    2. INRA, UMR1202 BIOGECO, 69 route d’Arcachon, F–33610 Cestas, France
    3. Université de Bordeaux, UMR1202 BIOGECO, F–33610 Cestas, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Xiao Li Peng,

    1. Molecular Ecology Group, Key Laboratory of Arid and Grassland Ecology, Lanzhou University, Lanzhou, Gansu 730000, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Jian Quan Liu,

    1. Molecular Ecology Group, Key Laboratory of Arid and Grassland Ecology, Lanzhou University, Lanzhou, Gansu 730000, China
    Search for more papers by this author
  • Martin Lascoux,

    1. Program in Evolutionary Functional Genomics, Evolutionary Biology Centre, Uppsala University, Norbyvägen 18D, NO–75326 Uppsala, Sweden
    2. Laboratory of Evolutionary Genomics, CAS-MPG Partner Institute for Computational Biology, Chinese Academy of Sciences, Shanghai, China
    Search for more papers by this author
  • Feng Sheng Hu,

    1. Departments of Plant Biology and Geology, and Program in Ecology, Evolution and Conservation, University of Illinois, 265 Morrill Hall505 S. Goodwin Avenue, Urbana, IL 61801, USA
    Search for more papers by this author
  • Rémy J. Petit

    1. INRA, UMR1202 BIOGECO, 69 route d’Arcachon, F–33610 Cestas, France
    2. Université de Bordeaux, UMR1202 BIOGECO, F–33610 Cestas, France
    Search for more papers by this author

Authors for correspondence:
Jian Quan Liu
Tel: +86 931 8914305
Rémy J. Petit
Tel: +33 557122837


  • A recent model has shown that, during range expansion of one species in a territory already occupied by a related species, introgression should take place preferentially from the resident species towards the invading species and genome components experiencing low rates of gene flow should introgress more readily than those experiencing high rates of gene flow.
  • Here, we use molecular markers from two organelle genomes with contrasted rates of gene flow to test these predictions by examining genetic exchanges between two morphologically distinct spruce Picea species growing in the Qinghai-Tibetan Plateau.
  • The haplotypes from both mitochondrial (mt) DNA and chloroplast (cp) DNA cluster into two distinct lineages that differentiate allopatric populations of the two species. By contrast, in sympatry, the species share the same haplotypes, suggesting interspecific genetic exchanges. As predicted by the neutral model, all sympatric populations of the expanding species had received their maternally inherited mtDNA from the resident species, whereas for paternally inherited cpDNA introgression is more limited and not strictly unidirectional.
  • Our results underscore cryptic introgressions of organelle DNAs in plants and the importance of considering rates of gene flow and range shifts to predict direction and extent of interspecific genetic exchanges.


Introgression, the transfer of limited amounts of genetic material across species boundaries, can take place without necessarily threatening species persistence (Rieseberg & Wendel, 1993; Arnold, 1997; Rieseberg et al., 2006; Kim et al., 2008). This process is caused by successive back crosses of hybrids with one parental species (Anderson & Hubricht, 1938). It can have profound evolutionary consequences, as realized early on by breeders seeking to incorporate in a domesticated species a given attribute of a wild relative (Bessey, 1906; Gur & Zamir, 2004). Significant progress in our understanding of introgression has been made recently with the development of a neutral model, proposed by Currat et al. (2008). This model offers a number of new insights into the genetic consequences of introgression that can now be tested empirically.

First, the model predicts that, when one species invades an area already occupied by a related species, introgression of neutral genes should take place mainly from the local species towards the invader. This is a consequence of the asymmetric demography of the two species, the expanding species being more likely to incorporate genes from the resident species than vice versa. In a review of published studies for animals and for plants (Currat et al., 2008), a majority (82%) of the examples pointed to asymmetric introgression in the expected direction. Since then, new case studies have provided further support for the model’s predictions. In particular, the finding of genetic material of the Neanderthal ancestry in the genome of modern nonAfrican humans has been accounted for by introgression of genes from the resident Neanderthal into the genome of modern humans expanding out of Africa (Green et al., 2010). In such empirical studies, however, obtaining independent evidence (e.g. using fossil data) on the status of the species (resident vs invading) can be difficult. Using genetic data to reconstruct species demography could help identify the resident and invading status of introgressing species, thereby improving the model’s utility.

Second, the model predicts that, following a contact between two hybridizing species, introgression should be particularly high for genome components experiencing little intraspecific gene flow (Currat et al., 2008). The rationale for this prediction, which has received little attention to date (but see Du et al., 2009; Bai et al., 2010; Zhou et al., 2010), is that genes from the resident species, once introgressed in the genome of the invading species, will persist only if intraspecific gene flow is limited. Hence, if one species has expanded into the range of a related species, different geographic patterns of introgression are expected, depending on the type of genetic marker considered. In particular, for genome components experiencing little gene flow (e.g. maternally inherited markers in plant species with limited seed dispersal), introgression by alleles from the local species should be detected in most populations of the invading species that become sympatric with the resident species. By contrast, for genome components experiencing higher gene flow (e.g. paternally or biparentally inherited markers in plant species with high pollen and/or seed dispersal), introgression is expected only in populations located closer to the front of expansion, or not at all (Currat et al., 2008) (Fig. 1).

Figure 1.

Schematic model showing the expected spatial genetic variation in an invading species that becomes sympatric with a resident species. The haplotype of the invading species is represented in black and the haplotype introgressed from the local species in white. Introgression is only observed in the part of the range where the two species are sympatric. The extent of introgression varies according to the rate of intraspecific gene flow: when gene flow is low, introgression is massive; when gene flow is intermediate, introgression is delayed; when gene flow is high, introgression is nearly absent.

The use of genetic markers characterized by different patterns of inheritance would help test simultaneously predictions regarding the direction and extent of introgression. Conifers are ideal model systems for this purpose. They have two uniparentally inherited organelle genomes (mtDNA and cpDNA) with contrasted modes of inheritance (maternal for mtDNA, gene flow mediated by seeds; paternal for cpDNA, gene flow mediated by seeds and pollen) (Neale & Sederoff, 1988; Mogensen, 1996; Petit & Vendramin, 2007), hence providing comparable but genetically unlinked markers to evaluate the directionality of introgression. Moreover, the two genomes typically experience very different rates of gene flow, as gene flow by seeds is typically much more reduced than gene flow by pollen in plants, especially in forest tree species and/or conifers (Petit et al., 2005). They can therefore be used to test the second prediction of Currat et al. (2008) regarding the extent of introgression as a function of gene flow. Finally, incomplete reproductive isolation is a frequent phenomenon in trees (Petit & Hampe, 2006), making them particularly suited for these types of investigations. In line with these characteristics, cryptic organelle DNA introgression has been detected in numerous plant species, including trees, especially those whose ranges have fluctuated in response to past climatic changes (Du et al., 2009; Willyard et al., 2009; Bai et al., 2010; Kikuchi et al., 2010).

We selected two morphologically distinct but closely related spruce species (Picea likiangensis and Picea purpurea) growing in the eastern declivity of the Qinghai-Tibetan Plateau (QTP) to test the predictions of the model. This area, which has highly diverse reliefs, comprises one of the world’s most important alpine biodiversity hotspots (Wu, 1987). A number of previous studies have inferred allopatric divergence through range fragmentation in this region (Liu et al., 2002, 2006; Yang et al., 2003; Xu et al., 2010). Major range expansions in response to climatic amelioration during interglacial periods should have resulted in frequent secondary contacts, facilitating subsequent introgression between previously isolated species (Opgenoorth et al., 2009; Wang et al., 2009; Sun et al., 2010; Wu et al., 2010). Picea likiangensis is widely distributed in the southern and south-eastern declivity of the QTP, while P. purpurea occurs at the east edge of the plateau where it overlaps with P. likiangensis in a zone at the north-eastern QTP. In this region, P. purpurea is found typically at higher altitudes than P. likiangensis and the two species are easily distinguished by leaf shapes, stomatal lines and seed cones (Fu et al., 1999). Phylogenetic analyses based on mtDNA and cpDNA data have shown that these two species belong to different clades (Ran et al., 2006; Bouilléet al., 2011). Moreover, preliminary nuclear DNA data suggest that the demographic history of these species might differ (Li et al., 2010). Both are dominant species of the local forest ecosystems and none of them has shown to be endangered (Wu, 1987). The present study was therefore designed to better characterize their demographic histories and then test if: shared mtDNA and cpDNA haplotypes are found preferentially in the region where the ranges of P. purpurea and P. likiangensis overlap, as expected if introgression has taken place; introgression is asymmetric, with sympatric populations from the invading species possessing organelle sequences that are otherwise specific to the local species; and introgression is more frequent at mtDNA than at cpDNA markers, owing to the differential gene flow experienced by these markers (Fig. 1).

Materials and Methods

Sampling strategy

We sampled 23 populations of P. likiangensis Franch. E. Pritz in the northern part of the species natural distribution range and 16 populations of P. purpurea Mast. distributed across most of its range (see the Supporting Information, Tables S1, S2). In the region where the species’ ranges overlap, the two species usually occupy different mountains and form monospecific stands. We failed to find obvious hybrids with intermediate morphology between the two species. Needle samples were collected from adult trees for genetic analyses, with individual trees located at least 100 m apart. As seed weight is correlated with dispersal ability in conifers with wind-dispersed seeds and may have implications for range shifts and genetic structures (Grotkopp et al., 2002), we also measured seed weight from a representative sample of each species. Seed cones were therefore collected in a subset of 30 populations (17 of P. likiangensis and 13 of P. purpurea) (Table S3). The latitude, longitude and altitude of each sampling site for both leaf and seeds material were measured by Extrex GIS (Garmin, Olathe, USA). All the needle samples were dried in silica gel immediately and seed cones were stored in envelopes and exposed to air at room temperature in the laboratory after collection. Seeds were individually weighted to the nearest milligram using a digital balance.

DNA extraction, amplification and sequencing

Genomic DNA was isolated from each needle sample using an established procedure (Doyle & Doyle, 1987) or a Plant Mini Kit (Qiagen). To study genetic structure and introgression at organelle markers, we sequenced two mtDNA fragments (nad1 intron b/c and nad5 intron 1 (Meng et al., 2007) and three cpDNA fragments (trnL–trnF) (Taberlet et al., 1991), trnS–trnG (Hamilton, 1999) and ndhK/C (Anderson et al., 2006). Each amplification reaction contained 20–50 ng DNA, 50 mM Tris-HCl, 1.5 mM MgCl2, 250 μg ml−1 BSA, 0.5 mM dNTPs, 0.5 μM of each primers and 0.5 unit of Taq polymerase following the method of Zhang et al. (2005). The PCR products were purified using a CASpure PCR Purification Kit following the recommended protocol (Casarray, Shanghai, China). Sequencing reactions were performed with the PCR primers described earlier. Internal primers were designed for sequencing of the mtDNA nad1 intron b/c (forward: TCT TAT GGA GGG TTG ACT; reverse: CCC CAT ATA TTC CCG GAG C). An ABI Prism Bigdye Terminator Cycle Sequencing Ready Reaction Kit (Life Technologies, Carlsbad, CA, USA) was used for all sequencing reactions. Only sequences with high quality and single peaks were used. Both forward and reverse sequences were checked carefully by eye for all new mutations. Depending on the quality of the sequences and the site of mutation, either the forward or the reverse primers were then used for large-scale sequencing. DNA sequences were aligned with clustal x (Thompson et al., 1997) and MEGA version 3.1 (Kumar et al., 2004) and with manual modifications.

Data analysis

We used the newly developed Isolation-with-Migration (IM) model (Hey, 2010) implemented in the program IMa2 to test if demographic expansion of one or both species has occurred. For this purpose, we relied on a previously published nuclear data set made up of 13 loci (Li et al., 2010). The parameters used in the IM model can be summarized as follows: multiple preliminary runs were conducted to adjust prior ranges and assess convergence. Markov Chain Monte Carlo (MCMC) runs were carried out with a burn-in of 5 000 000 steps for 50 000 000 iterations. The inheritance scalars were set as 1 for the nuclear markers. The final prior parameters (–q1 4 –q2 10 -qa 40 –m1 10 –m2 3 –t 0.5 –hfg –hn 150 –ha 0.95 –hb 0.8) were used. To ensure adequate mixing, runs were continued until all effective sample size values exceeded 200 and the trend-line plots showed no obvious trends, as suggested in the program documentation. Under an expansion scenario, effective population size should be larger than that of the ancestral population, whereas under demographic stability, the estimate of current population size should be similar to that of the ancestral population (Hey, 2010).

Phylogenetic relationships among mtDNA and cpDNA haplotypes were reconstructed using median-joining networks with network version (Bandelt et al., 1999), available at The average gene diversity within populations (HS), total gene diversity (HT) and the coefficients of differentiation GST and NST were estimated for each species or species groups for both mtDNA and cpDNA markers, using permut (available at The difference between NST and GST is that only the former takes into account the similarities between haplotypes (Pons & Petit, 1996). Comparisons between GST and NST (NST > GST) were tested with 1000 random permutations of haplotype identity. We examined hierarchical partitioning of genetic variations among species and populations by AMOVA (Excoffier et al., 1992) using the program arlequin version 3.0 (Excoffier et al., 2005). Significance was tested based on 1000 permutations.


Seed weight was measured in a representative sample (c. 3000 seeds) of the two species (Table S3). We found that P. purpurea seeds are approx. half the weight of those from P. likiangensis, with little overlap in the two distributions (Fig. 2).

Figure 2.

Seed mass variation in Picea purpurea and P. likiangensis. The lower and upper margins of each box correspond to quartiles 25% and 75%, the line in the middle of the box corresponds to the median and the dotted lines correspond to 1.5 times the interquartile range.

We then reanalysed a previously published nuclear data set (13 loci; Li et al., 2010) using the IM model. The estimates of the marginal distribution of the posterior probability of the demographic population sizes (θ) are: for P. likiangensis, θ = 1.2 with the 95% highest posterior density (HPD) ranging from 0.7 and 2.1; for P. purpurea, θ = 4.5 with HPD ranging from 2.6 to 8.5; and for the inferred ancestral population, θ = 1.3 with HPD ranging from 0.6 and 2.1. The threefold increase in the effective population size of P. purpurea suggests demographic expansion, whereas there is no evidence of demographic change in P. likiangensis (Fig. 3).

Figure 3.

Marginal distribution of the posterior probability of the demographic population sizes (θ) estimated by the Isolation-with-Migration model. The estimate of population size for Picea purpurea (red line) is three times larger than that of the ancestral population (green line), indicating population expansion. Conversely, there is no obvious change for the effective population size of Picea likiangensis (blue line).

A total of 1581 bp were sequenced from two mtDNA fragments and 1946 bp from three cpDNA fragments in 330 individuals distributed in 31 localities of the two species (Tables S1, S2). A matrix of combined sequences was constructed for the 330 individuals examined in which 10 different mtDNA sequences (mtDNA haplotype, Table S4) and seven different cpDNA sequences (cpDNA haplotype, Table S5) were identified. These mitochondrial and chloroplast fragment sequences have been deposited in EMBL under accession numbers HM561831HM561859.

For mtDNA, we found seven substitutions and five indels, which combine into 10 haplotypes (M1–M10) (Fig. 4a). The most common haplotype (M10) is fixed in the northern populations of P. purpurea. Haplotypes M1 and M2 are only found in P. likiangensis and in populations of P. purpurea sympatric with P. likiangensis. All remaining haplotypes are restricted to P. likiangensis (Fig. 4a). These haplotypes cluster in two distinct groups separated by at least four mutations (Fig. 4b). One group includes the M10 haplotype found only in allopatric populations of P. purpurea while the other group includes all remaining haplotypes found in allopatric P. likiangensis populations and in the zone of sympatry with P. purpurea. In the area of sympatry, the most commonly distributed haplotypes in P. likiangensis (M1 and M2, Fig. 4d) are also found in P. purpurea (Fig. 4c).

Figure 4.

Mitochondrial (mt) DNA variation in Picea purpurea (area indicated with dotted line) and P. likiangensis (area indicated with solid line). (a) Geographic variation of mtDNA in both species. Populations 1–8 correspond to allopatric populations of P. purpurea, populations 9–16 indicate samples from both P. purpurea and P. likiangensis, and populations 17–31 correspond to pure populations of P. likiangensis. The different colors correspond to haplotype with the same colors in (b). (b) Network of the 10 mtDNA haplotypes. For each haplotype, the circle size is proportional to its frequency over all populations. (c) mtDNA variation in P. purpurea in the region of sympatry. (d) mtDNA variation in P. likiangensis in the region of sympatry.

For cpDNA, we detected 11 substitutions and one indel that combine into seven haplotypes (A-G) (Fig. 5a). The most common haplotype (C) is nearly fixed in the northern populations of P. purpurea, predominant in the southern populations of P. purpurea and found at low frequency in P. likiangensis, where it is restricted to populations in sympatry with P. purpurea. In P. likiangensis, the most frequent haplotypes (A and B) are also found in some populations of P. purpurea sympatric with P. likiangensis. As for mtDNA, these cpDNA haplotypes cluster in two distinct groups separated by at least five mutational steps (Fig. 5b): one comprising haplotypes G, C and F, which are mainly found in P. purpurea, and the other including the remaining haplotypes, which are mainly found in P. likiangensis. In the area of sympatry of the two species, the most common haplotype in P. likiangensis, haplotype A, is also found in P. purpurea (Fig. 5c). Similarly, the dominant haplotype in P. purpurea, haplotype C, is also found in P. likiangensis (Fig. 5d).

Figure 5.

Chloroplast (cp) DNA variation in Picea purpurea (area indicated with dotted line) and P. likiangensis (area indicated with solid line). (a) Geographic variation of cpDNA in both species. Populations 1–8 correspond to allopatric populations of P. purpurea, populations 9–16 indicate samples from both P. purpurea and P. likiangensis, and populations 17–31 correspond to pure populations of P. likiangensis. The different colors correspond to haplotype with the same colors in (b). (b) Network of the seven mtDNA haplotypes. For each haplotype, the circle size is proportional to its frequency over all populations. (c) cpDNA variation in P. purpurea in the region of sympatry. (d) cpDNA variation in P. likiangensis in the region of sympatry.

Genetic structure, as measured by the coefficient of differentiation (GST), was high in the two species for mtDNA as well as for cpDNA (Table 1). Genetic structure was greater at mtDNA than at cpDNA in P. purpurea and similar for both markers in P. likiangensis. When populations from the sympatric zone were excluded, genetic differentiation (GST) and diversity (HS and HT) decreased greatly for both markers in P. purpurea (Table 1). Analysis of molecular variance (AMOVA) of all samples showed that the genetic variation partitioned between species is higher at cpDNA markers (50%) than at mtDNA markers (26%) (Table 2). Moreover, interspecific differentiation is much larger when sympatric populations are excluded, especially for mtDNA.

Table 1.   Genetic diversity estimates for chloroplast (cp) DNA and chloroplast (cp) DNA variations in Picea likiangensis and P. purpurea
  1. HS, Average gene diversity within populations; HT, total gene diversity; GST and NST, coefficients of differentiation.

  2. 1indicates that NST is significantly different from GST (< 0.05).

mtDNA variation
P. purpurea0.120.700.830.941
Allopatric populations of P. purpurea
P. likiangensis0.370.790.530.60
Allopatric populations of P. likiangensis0.390.820.530.53
cpDNA variation
P. purpurea0.120.240.490.52
Allopatric populations of P. purpurea0.
P. likiangensis0.260.710.640.80
Allopatric populations of P. likiangensis0.270.550.520.36
Table 2.   Analysis of molecular variance (AMOVA) of mitochondrial (mt) DNA and chloroplast (cp) DNA variation between and within Picea purpurea and P. likiangensis
Source of variationdf1SS2VC(%)3VariationF-statistics4
  1. 1df, degrees of freedom; 2SS, sum of squares;3 VC, variance component;4FCT,correlation of species relative to total; FSC, correlation within populations relative to species; FST, correlation within populations relative to total.

All samples
 Between P. likiangensis and P. purpurea121.60.1326.2FCT = 0.27
 Among populations within species3671.50.2245.7FSC = 0.62
 Within populations28138.50.1428.1FST = 0.72
 Allopatric samples
 Between P. likiangensis and P. purpurea132.30.2850.8FCT = 0.51
 Among populations within species2034.70.1526.8FSC = 0.55
 Within populations21626.90.1322.4FST = 0.78
All samples
 Between P. likiangensis and P. purpurea136.00.2350.3FCT = 0.50
 Among populations within species3644.30.1430.2FSC = 0.61
 Within populations28125.00.0919.5FST = 0.81
 Allopatric samples
 Between P. likiangensis and P. purpurea138.60.3564.6FCT = 0.65
 Among populations within species2024.10.1019.5FSC = 0.55
 Within populations21618.50.0916.0FST = 0.84


Shared haplotypes and hybridization history

We recovered two groups of haplotypes for cpDNA and for mtDNA. Both markers differentiate the species well, except in the zone of sympatry. Nevertheless, maternally inherited mtDNA haplotypes are more systematically shared in sympatry than paternally inherited cpDNA, resulting in a stronger interspecific partitioning of genetic variation at cpDNA marker than at mtDNA marker. These shared haplotypes may have originated from introgression between species or by retention of ancestral polymorphisms because of incomplete lineage sorting (Gay et al., 2007; Willyard et al., 2009). We believe that introgression is a more likely interpretation for two reasons. First, the shared cpDNA and mtDNA haplotypes are concentrated geographically in sympatry, a pattern expected if introgression is involved but highly unlikely in the case of incomplete lineage sorting (Palméet al., 2004; McGuire et al., 2007; Zhou et al., 2010). Second, at both sets of markers, the two distinct lineages differ by at least four mutational steps, that is, much more than between haplotypes within each lineage. This suggests comparatively ancient interspecific divergence. In fact, the main cpDNA and mtDNA lineages belong to well-supported clades involving other spruce species occurring in the QTP, further suggesting that P. likiangensis and P. purpurea diverged early (Ran et al., 2006; Li et al., 2010; Bouilléet al., 2011; data presented here). Such ancient divergence reduces the likelihood that incomplete lineage sorting is involved (Comes & Kadereit, 1998; Avise, 2004), especially considering that more than one haplotype is shared for each genome (i.e. two mtDNA and three cpDNA haplotypes are shared between species).

Setting aside these shared haplotypes, P. purpurea was nearly fixed range-wide for a single cpDNA and mtDNA haplotype. This contrasts with the situation in P. likiangensis, where we found six mtDNA haplotypes and three cpDNA haplotypes. The near-absence of genetic variation in P. purpurea at both cpDNA and mtDNA markers suggests that this species has experienced a demographic founder effect followed by an expansion, which resulted in the fixation of a common haplotype in all colonized regions, as found in other species with such expansion histories (Hewitt, 1996; Avise, 2004). Furthermore, P. purpurea is characterized by lighter seeds than P. likiangensis (Fig. 2). In conifers, reduced seed weight characterizes species that are successful invaders in disturbed habitats (Rejmánek & Richardson, 1996; Grotkopp et al., 2002). Finally, the IMa2 analysis of nuclear data also lends support to a scenario of population expansion in P. purpurea.

However, a couple of caveats need to be pointed out for such an implication. First, strictly speaking, the IM analysis only indicates that P. purpurea has a larger effective population size than P. likiangensis and their common ancestor, which is not equivalent to concluding that P. purpurea experienced population expansion. We believe, however, that population expansion is the most likely explanation. Indeed, Li et al. (2010) also obtained a signal, albeit weak, of population growth when considering P. purpurea individually. The weakness of the signal could simply reflect the lack of statistical power of the approximate Bayesian analysis used by Li et al. (2010). Second, P. purpurea has been shown to have previous admixture history with another species located further north, P. wilsonii (Li et al., 2010). Population admixture is expected to lead to a structured ancestral population size. In such a scenario, a signal of population decline rather than of population expansion is expected (Becquet & Przeworski, 2009; Peter et al., 2010). Hence, our interpretation that P. purpurea experienced population expansion should not have been confounded by admixture. Finally, population expansion does not necessarily imply range expansion (Excoffier et al., 2009). However, we believe that the conclusion that P. purpurea experienced range expansion, which is supported by several other lines of evidence, as pointed out before (the distribution patterns of both organelle markers and the lighter seed weight), is robust. By contrast, no such a species-level expansion is apparent in P. likiangensis. In this species, multiple haplotypes are fixed in different regions. Moreover, approximate Bayesian computation of nuclear polymorphism in P. likangensis (Li et al., 2010) failed to detect evidence of departure from the standard neutral model, further supporting this conclusion.

While dating this demographic expansion of P. purpurea would require other types of evidence, these analyses do support the idea that P. purpurea has expanded geographically, thereby invading areas already colonized by P. likiangensis, resulting in introgression. Introgression has significantly affected the genetic structure of the two species, especially of the expanding one, to the point that failure to detect it would have resulted in serious misinterpretations. For example, introgression of mtDNA markers in sympatry not only compromises the specificity of the markers but also results in a misleading phylogeographic signal in this expanding species (e.g. NST > GST,Table 1). Likely errors caused by failure to detect introgression include overestimating historical species effective population sizes and underestimating actual species divergence times.

Direction and extent of mtDNA and cpDNA introgression

Previous studies (Du et al., 2009; Bai et al., 2010; Zhou et al., 2010 and references therein) could not conclusively test all predictions of Currat et al. (2008). Indeed, in these studies, incomplete lineage sorting could not be totally ruled out. Moreover, some of the DNA markers used differed not only in rates of gene flow but also in ploidy levels and hence in effective population sizes, therefore complicating interpretations. Here we relied on markers from two oppositely uniparentally inherited genomes and therefore differing only in rates of gene flow (through seed and pollen for cpDNA and through seed only for mtDNA). Given that P. purpurea was shown to be the expanding species, the direction of introgression of the maternally inherited mtDNA markers is in agreement with the first model prediction: it is the expanding species (P. purpurea) that is introgressed by genetic material from the other species (P. likiangensis). While mtDNA haplotypes found in the allopatric range of P. likiangensis were detected in P. purpurea in the sympatric zone, the opposite is not true: no ‘allopatric’P. purpurea haplotype was detected in P. likiangensis from the sympatric zone, as expected if introgression had taken place only towards the expanding species. By contrast, the prediction regarding the directionality of introgression does not hold well for the paternally inherited cpDNA, whose introgression is bidirectional. While cpDNA haplotypes found in the allopatric range of P. likiangensis were detected in some sympatric P. purpurea populations, cpDNA haplotypes found in the allopatric range of P. purpurea were also detected in some sympatric P. likiangensis populations.

In Currat et al.’s (2008) model, an analogy is made between introgression and the surfing of genes on the front of expansion (Edmonds et al., 2004; Klopfstein et al., 2006). Surfing is caused by high levels of genetic drift occurring at the front of an expanding species (Hallatschek & Nelson, 2008), thereby resulting in a strong stochastic component. This could explain cases where the model does not hold (i.e. 16% of the cases reviewed by Currat et al., 2008). Other exceptions from the expected trend might be caused by local population size advantage for the invader, once it has firmly established itself in a colonized area (e.g. Scascitelli et al., 2010), or extremely low abundance of the resident species (rare or narrowly endemic), which could then become endangered as a consequence of introgression with the invading species, but only in cases of extreme differences of abundance (Rieseberg & Gerber, 1995; Levin et al., 1996; Petit & Excoffier, 2009).

The second model prediction states that introgression extent should vary according to the type of marker considered. All populations of P. purpurea sympatric with P. likiangensis are fixed for mtDNA haplotypes originating from P. likiangensis. Thus, as expected for markers experiencing reduced gene flow, massive introgression has taken place at mtDNA markers. By contrast, at cpDNA markers, only those populations of P. purpurea located further away along the inferred axis of expansion (i.e. farther south and east) are introgressed by P. likiangensis. Hence, the second model’s prediction is well supported. The reduced introgression of paternally inherited genes (cpDNA) can be accounted for by the increased gene flow (through pollen and seed) that they experience compared with the maternally inherited (and only seed-dispersed) mtDNA (Fig. 1).


Our empirical study indicates that information on past range shifts and rates of gene flow can help predict the direction and magnitude of introgression, which represents an important step forward for evolutionary studies (Du et al., 2009; Petit & Excoffier, 2009). The consistency of our data with the model’s predictions suggests that the order of colonization of closely related species could be inferred using genetic markers experiencing low rates of gene flow. Thus maternally inherited markers may help decipher the long-term histories of closely related species that do not have good fossil records or whose pollen cannot be identified on the basis of morphological features (Hu et al., 2009), such as the spruce species in the QTP. Furthermore, our data suggest large-scale cryptic introgression of organelle DNA following historical expansions in response to climatic changes. Thus caution should be exerted when using organelle DNA markers for species identification (Chase et al., 2005; Rubinoff, 2006; CBOL Plant Working Group, 2009; Kress et al., 2009).


We thank Georg Miehe, Outi Savolainen, Richard Abbott and Rongling Wu for comments on a previous version of the manuscript and thank Spencer Barrett for discussions on this topic. Seed mass measurements were carried out by Long Li and MingFei Ji. We thank Dr Zhonghu Li for help with the analysis for the nuclear data using IMa2. This research was supported by grants from the National Natural Science Foundation of China (30930072 and 30725004), the Key Project of International Collaboration Program, the Ministry of Science and Technology of China (2010DFB63500) and the International Collaboration ‘111’ Project to J.Q.L. The stay of F.K.D. in France was supported by the Chinese Scholarship Council and the University of Bordeaux I. R.J.P. and M.L. are supported by the EU sponsored network of excellence EVOLTREE.

Author contributions

J.Q.L., F.K.D., and R.J.P. designed research; J.Q.L. collected materials; X.L.P. performed experiments; F.K.D. analyzed data; and F.K.D., J.Q.L. M.L., F.S.H. and R.J.P. wrote the paper.