Large-scale asymmetric introgression of cytoplasmic DNA reveals Holocene range displacement in a North American boreal pine complex

Jack pine (Pinus banksiana) and lodgepole pine (Pinus contorta var. latifolia) are two North American boreal hard pines that hybridize in their zone of contact in western Canada. The main objective of this study was to characterize their patterns of introgression resulting from past and recent gene flow, using cytoplasmic markers having maternal or paternal inheritance. Mitochondrial DNA (mtDNA) and chloroplast DNA (cpDNA) diversity was assessed in allopatric populations of each species and in stands from the current zone of contact containing morphological hybrids. Cluster analyses were used to identify genetic discontinuities among groups of populations. A canonical analysis was also conducted to detect putative associations among cytoplasmic DNA variation, tree morphology, and site ecological features. MtDNA introgression was extensive and asymmetric: it was detected in P. banksiana populations from the hybrid zone and from allopatric areas, but not in P. contorta populations. Very weak cpDNA introgression was observed, and only in P. banksiana populations. The mtDNA introgression pattern indicated that central Canada was first colonized by migrants from a P. contorta glacial population located west of the Rocky Mountains, before being replaced by P. banksiana migrating westward during the Holocene. In contrast, extensive pollen gene flow would have erased the cpDNA traces of this ancient presence of P. contorta. Additional evidence for this process was provided by the results of canonical analysis, which indicated that the current cpDNA background of trees reflected recent pollen gene flow from the surrounding dominant species rather than historical events that took place during the postglacial colonization.


Introduction
The impact of Quaternary glacial cycles on the distribution of species and their genetic diversity is now well documented. Indeed, the modern population genetic structure of several species currently found at temperate and boreal latitudes still carry the signature of vicariance induced by their retreat in isolated refugia (for reviews, see Hewitt 2004;Soltis et al. 2006;Jaramillo-Correa et al. 2009;Shafer et al. 2010). At the interspecific level, these migrations likely resulted in repeated hybridization events between interfertile species in spatially and temporally dynamic zones of contact. Thus, genetic footprints of historical contact are expected to be found in presently allopatric areas, far from current cross-species gene influence (Bouillé et al. 2011). Anderson and Hubricht (1938) defined introgression as the result of the incorporation of genes from one species into another via hybridization and through repeated backcrossing of hybrids with the parental species. From an evolutionary perspective, the role of introgression is not fully understood. One benefit previously suggested is an increase in intraspecific genetic diversity, and thus, of raw material on which natural selection could possibly act ª 2012 The Authors. Published by Blackwell Publishing Ltd. This is an open access article under the terms of the Creative Commons Attribution Non-Commercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
on (Anderson 1949). However, although introgression can be easily detected using molecular or morphological markers (Klier et al. 1991), the significance of its role in adaptation is much more difficult to demonstrate (Rieseberg and Wendel 1993;Arnold 2004; but see also Anderson et al. 2009;Whitney et al. 2010). Modeling has also shown that more neutral factors such as the direction and amplitude of gene flow can be key factors in shaping introgression patterns (Currat et al. 2008;Meirmans et al. 2009). From a biogeographic perspective, investigating modern geographic patterns of introgression could also provide insight into historical gene exchanges in species complexes and how natural ranges might have evolved and fluctuated in the recent past.
In Pinaceae, mitochondrial and chloroplast genomes are inherited maternally (Dong and Wagner 1993) and paternally (Neale and Sederoff 1988), respectively. Hence, using markers from both genomes allows tracking gene exchange transiting via seeds only (mitochondrial DNA [mtDNA]) or pollen and seeds (chloroplast DNA [cpDNA]). These contrasted modes of dispersion usually lead to differential gene flow (Petit et al. 2005), resulting in different mtDNA and cpDNA patterns potentially revealing cytoplasmic genome capture events among species (e.g., in trees: Tsumura and Suyama 1998;Senjo et al. 1999;Du et al. 2009Du et al. , 2011Bouillé et al. 2011) or among genetically distinct intraspecific lineages (Gérardi et al. 2010;Wei et al. 2011). Moreover, such asymmetric gene dispersion is thought to affect the direction and rate of cytoplasmic introgression (Currat et al. 2008;Du et al. 2009Du et al. , 2011. Jack pine (Pinus banksiana Lamb.) and lodgepole pine (Pinus contorta Dougl. ex. Loud.) belong to subgenus Pinus and sub-section Contortea along with Pinus virginiana and Pinus clausa. They are the only two boreal pines in North America (Critchfield and Little 1966). The combined natural ranges of P. contorta and P. banksiana extend from the Pacific to the Atlantic coasts and cover the entire North American boreal region (Fig. 1). Pinus contorta has a typically latitudinal distribution in the Pacific Northwest, whereas P. banksiana rather stretches longitudinally from western Canada to the Atlantic Maritimes region in eastern Canada (Critchfield and Little 1966). Pinus contorta comprises four subspecies (ssp. latifolia, contorta, murrayana, and bolanderi). Subspecies contorta and bolanderi are found along the Pacific coast and murrayana in the Sierra Nevada. Subspecies latifolia extends eastward from the interior of British Columbia (BC) to Alberta where its distribution overlaps with that of P. banksiana.
Hybrid individuals presenting intermediate morphological features between P. contorta ssp. latifolia and P. banksiana have been frequently reported in the zone of contact found in western and northern Canada (Rudolph and Laidly 1990;Fig. 1). The zone located in western Alberta has been extensively studied using morphological (Moss 1949;Critchfield 1957), biochemical (Zavarin et al. 1969;Pollack and Dancik 1985), and genetic markers (Wagner et al. 1987;Wheeler and Guries 1987;Wagner et al. 1991;Dong and Wagner 1993;Ye et al. 2002). However, little is known about the hybridization dynamics and the role of Holocene postglacial migrations in shaping introgression patterns between these species. Indeed, although these studies contributed to gain an understanding of gene exchange in this species complex, their geographically limited coverage provided an incomplete picture of introgression patterns at both temporal and spatially large scales.
The geographic distribution of mtDNA and cpDNA diversity was previously analyzed in P. banksiana and P. contorta (including all of its four subspecies) to explore their postglacial histories. No clear geographic structure was detected using cpDNA markers in either species (Marshall et al. 2002;Godbout et al. 2010). In contrast, mtDNA data showed that both pines had colonized their current natural ranges from multiple and genetically differentiated glacial populations (Godbout et al. 2005(Godbout et al. , 2008. At least two distinct mtDNA lineages were detected in P. contorta ssp. latifolia (out of four to five lineages identified in P. contorta [Godbout et al. 2008]), whereas at least three mtDNA lineages were identified in P. banksiana (Godbout et al. 2005(Godbout et al. , 2010. A typical mitotypes were also found in one population of P. banksiana in eastern Alberta (Godbout et al. 2005), suggesting possible introgression at the far east of the known zone of interspecific hybridization with P. contorta ssp. latifolia.
The first objective of this study was to assess the extent of cytoplasmic introgression between P. contorta ssp. latifolia and P. banksiana in their current zone of contact in western Alberta and in adjacent parts of their natural ranges (central BC through central Canada) using markers from the differentially dispersed mtDNA and cpDNA genomes. In doing so, we wanted to test if long-range introgression indicative of historical species displacement was present and if it was asymmetric. These results were examined in the light of previously inferred species postglacial histories (Godbout et al. 2005(Godbout et al. , 2008, and in relation to predictions regarding the direction and intensity of introgression using markers experiencing different rates of gene flow (Currat et al. 2008;Du et al. 2011). The decoupling observed between mtDNA and cpDNA introgression patterns was further assessed from an ecological perspective and in relation to tree morphological classification, with the aim to determine whether introgression was driven by determinants at the local microgeographic scale including the local abundance of the different species.

Materials and Methods
Sampling and polymerase chain reaction (PCR) amplification A total of 925 trees were sampled from 50 populations identified as P. banksiana, P. contorta var. latifolia (hereafter designated more simply as P. contorta), or hybrid (see below for criteria used to designate hybrid populations). Nineteen of these populations (P. banksiana = 14 [n = 226]; P. contorta = 5 [n = 50]) originated from two provenance tests managed by the Ministère des Ressources naturelles et de la Faune of Québec, where each provenance was established from seeds harvested from 8 to 100 trees per stand, as previously described (Godbout et al. 2005(Godbout et al. , 2008. Additionally, 17 populations were obtained from the National Tree Seed Centre (Fredericton, New Brunswick, Canada; P. banksiana = 5 [n = 77]; P. con-torta = 12 [n = 182]). These populations were sampled in provenance trials where each provenance was established from seed harvested from a minimum of 20 trees per stand. Furthermore, 14 populations (hereafter named "in situ sampling") were sampled and analyzed in the zone of sympatry and adjacent areas in western Alberta. These in situ populations were selected to be representative of the various morphological population types encountered in the zone of contact, with six populations (n = 169) labeled as hybrid (each with at least 90% of their trees classified as morphological hybrids), four populations (n = 109) labeled as P. banksiana, and four populations (n = 122) labeled as P. contorta. Both P. banksiana and P. contorta in situ populations presented a maximum of 5% of morphological hybrids. Species or hybrid identification was based on morphological characters (growth form, cone shape and orientation, foliage, and branching habits) according to criteria proposed by Rudolph and Yeatman   1982). Ecological classification (ecoregion type and ecosite type, as described in Beckingham et al. 1996) was also determined for each of these 14 in situ populations. DNA was extracted from frozen needles, embryos, or young seedlings. DNA extraction was performed with the Dneasy Plant Mini Kit (Qiagen; Mississauga, Ontario, Canada) following the manufacturer's instructions.
The screening of mtDNA and cpDNA polymorphisms was conducted on an exploratory panel combining 12 individuals of each species. These trees were selected from populations located far from the reported hybrid zone, so as to avoid hybrid or introgressed individuals. The main goal was to identify polymorphisms that were species-specific for each cytoplasmic genome. In addition, a hyper-variable mtDNA marker was sought to identify the geographic origin of mtDNA introgression (see Results). In total, 29 mtDNA regions (Demesure et al. 1995;Jeandroz et al. 2002;Jaramillo-Correa et al. 2003;Godbout et al. 2005) and 19 cpDNA regions (Taberlet et al. 1991;Wang et al. 1999) were amplified following published PCR protocols in these studies. All loci yielding a single amplification product were then sequenced for both strands with a Sequenase GC-rich kit (Applied Biosystems, Cleveland, Ohio) using the dideoxynucleotide chain termination procedure on an ABI-3130xl (Applied Biosystems) to detect polymorphism among the two pine species. For mtDNA genes, the intron b/c of nad1 (Demesure et al. 1995) showed an interspecific polymorphism and the intron 1 of nad7 (Godbout et al. 2005) showed a hyper-variable intraspecific polymorphism. Internal primers were developed for the nad1 intron b/c region (internal-nad1(b/c)_F: 5′GATCGATCCATAGTGCC GTAA3′ and internal-nad1(b/c)_R: 5′GGGTGCCGCAAGG TTATAC3′; annealing temperature = 60°C). The nad7 intron 1, whose fragment length polymorphism is caused by a variation in the number of a minisatellite-like repeat, had already been used in previous studies of P. banksiana and P. contorta populations (Godbout et al. 2005(Godbout et al. , 2008(Godbout et al. , 2010. For cpDNA, the matK1, rbcL2, rbcL3 (Wang et al. 1999), and trnL-trnF (Taberlet et al. 1991) regions harbored polymorphisms that were species-specific. The internal nad1 (intron b/c), matK1, rbcL2, rbcL3, and trnL-trnF PCR products were, respectively, used in combination with RsaI, HhaI, SspI, HinfI, and MfeI restriction endonucleases (New England BioLabs, Pickering, Ontario, Canada) following the manufacturer's instructions, and electrophoresed on 2% agarose gels to visualize DNA polymorphisms. Thus, except for the fragment length polymorphism of nad7 intron 1 directly detectable on 2% agarose gel, all other polymorphisms were revealed using a PCR-RFLP (restricted fragment length polymorphism) assay. Each variant observed on agarose gel was further sequenced to confirm the nature of polymorphism and detect possible cases of fragment length homoplasy.

Diversity indices and analyses
The two detected polymorphic mtDNA loci on one hand, and the four polymorphic cpDNA loci on the other hand, were considered simultaneously to define multi-locus mitotypes and chlorotypes, which resulted in 22 mitotypes and two chlorotypes (see Results). The number of mitotypes or chlorotypes (nh), the total unbiased diversity index (H T ), the mean unbiased diversity index (H M ; both equivalent to the expected heterozygosity, H E , for diploid data; [Weir 1996]), and the among-population differentiation F ST index were estimated for the whole dataset, for each morphological group (populations of P. banksiana, P. contorta, and hybrid populations), and for each mtDNA Bayesian group defined below.
For the more complex mtDNA dataset, a mitotype network was drawn using the software TCS (Clement et al. 2000), setting the maximum number of steps between two connected mitotypes to seven. This network aimed at revealing related mtDNA variants possibly associated to each pine species. A neighbor-joining (NJ; Saitou and Nei 1987) population dendrogram was built using mtDNA data to infer the population structure of each species, as well as to identify those populations presenting intermediate genetic content, which could be interpreted as introgressed. A matrix of population pairwise F ST was first calculated using the pairwise difference distance method implemented in the software Arlequin v.3.1 (Excoffier et al. 2005) and loaded in the software MEGA v.4.0 (Tamura et al. 2007) to construct the dendrogram. The tree was anchored using mid-point rooting. A spatial Bayesian analysis of population structure (BAPS) aiming to delineate genetically differentiated and spatially coherent groups was carried out on mtDNA data using the "spatial clustering of groups" option implemented in BAPS v.5.4 (Corander et al. 2008). K-values ranging from 2 to 10 were separately tested and the value providing the highest F CT value (corresponding to the genetic differentiation among clusters) without yielding a singleton group was considered optimal. This method was preferred over the original log-likelihood method because BAPS delineated over 10 distinct mtDNA groups of little historical significance according to previous studies (see Godbout et al. 2005Godbout et al. , 2008. A similar population structure analysis was conducted on cpDNA data to compare the patterns of interspecific structuring between cpDNA and mtDNA. The original log-likelihood method was used on cpDNA data to determine the optimal k-value. A canonical redundancy analysis (RDA) was conducted on the 14 natural populations sampled in situ in the zone of contact to test for possible associations among mtDNA or cpDNA genotypes and the morphological classification of trees (P. contorta, P. banksiana, or hybrid) or site ecological classification following ecoregions (boreal mixed woods, upper foothills, lower foothills) or ecosites (corresponding to species association identified as a1, b1, c1, d1, e1, and f1; see Beckingham et al. 1996 for a description). These three categorical factors were considered as three independent explanatory matrices. The tested null hypothesis assumed the independence of genotypic information and each of the three factors. Main effects of explanatory matrices were tested via Monte Carlo permutation tests. The canonical analyses and statistical tests were performed using the Vegan Community Ecology Package ver. 1.17-2 for R software v. 2.11.0 (Oksanen et al. 2010).

Species distribution of genetic diversity
The nad7 intron 1 locus exhibited a minisatellite-motif of 32 nucleotides that yielded a total of 15 haplotypes (corresponding to the numeral code in the mitotype name, see the legend in Fig. 1a). Six indels and one single-nucleotide polymorphism (SNP) were found in the nad1 intron b/c region, which resulted in three haplotypes among the total population sampling (corresponding to the letter code B, C, or C2 in the mitotype name, see Fig. 1a). Two of these were associated to P. contorta (C and C2), whereas the last one was specific to P. banksiana (B). The combination of both mtDNA loci generated a total of 22 distinct mitotypes (Table 1 and Fig. 1a). From the 14 mitotypes detected among P. contorta populations, five were speciesspecific, nine occurred in both species, and one was found in hybrid populations (Fig. 1). Similarly, of the 15 mitotypes observed among P. banksiana populations, six were exclusive to P. banksiana, whereas eight and two mitotypes were shared with P. contorta and hybrid populations, respectively. Trees from hybrid populations carried 10 mitotypes, from which one was only found in a single hybrid population, and seven and eight were shared with P. contorta and P. banksiana populations, respectively. The highest mtDNA diversity was observed in P. contorta populations (H T = 0.744; H S = 0.484), followed by hybrid populations (H T = 0.636; H S = 0.390), and P. banksiana populations (H T = 0.591; H S = 0.346; Table 1). Hybrid populations (F ST = 0.395) were also more differentiated than P. banksiana and P. contorta populations (F ST = 0.349 and 0.282, respectively). Details about diversity parameters for each population can be found in Supplementary Table S1.
The mitotype network revealed two distinct clusters that were separated by an average of six mutational steps (Fig. 2), which corresponded to the divergence observed between two nad1 (intron b/c) major variants and their associated low-frequency variants. The first group was predominantly associated to P. banksiana (banksiana morphological-type trees), and thus contained so-called banskiana-type variants (mitotype numbers followed by the letter B in Fig. 2). The second one was mostly associated to P. contorta (contorta morphological-type trees) and contained contorta-type variants (mitotype numbers followed by the letter C or C2 in Fig. 2). Banksiana-type variants were found in both P. banksiana and hybrid populations, whereas contorta-type mitotypes were found in all three population types (P. banksiana, P. contorta, and hybrid populations) (Fig. 2).
All four polymorphic cpDNA regions harbored one SNP each. Polymorphisms were linked, resulting in the delimitation of two distinct multi-locus chlorotypes (Table 1 and Fig. 1b). Chlorotype B was mostly associated to P. banksiana and almost fixed in this species (H T = 0.005; H S = 0.003; F ST = 0.009). Both chlorotypes (B and C) were observed in hybrid populations, whereas P. contorta populations were all fixed for chlorotype C (Fig. 1b). Therefore, chlorotype B was considered to be of

Geographic distribution of genetic diversity
The geographic distribution of mtDNA variation revealed the presence of mitotypes typically found in P. contorta populations as far east as Saskatchewan in central Canada, well within the natural range of P. banksiana (Fig. 1a). Globally, P. banksiana populations from Alberta and Saskatchewan in the near of the hybrid zone exhibited higher mtDNA diversity than those located further east. These populations from Manitoba and western Ontario were characterized by lower genetic diversity and a high abundance of mitotype 9B (Fig. 1a).
Within the natural range of P. contorta, the highest mtDNA diversity was observed in populations from central BC and western Alberta where mitotypes 5C and 11C were the most frequent. As mentioned above, diversity at the investigated cpDNA loci was markedly split between the two species (Fig. 1b), given our specific interest to identify species-specific cytoplasmic markers to monitor interspecific gene flow. Except for one P. banksiana and one hybrid population from Alberta, all other stands were fixed for a single chlorotype (Fig. 1b). In P. banksiana, only one tree (1/412), belonging to population BANK19 from Alberta, presented the chlorotype typical of P. contorta. All trees from P. contorta populations possessed typical contorta mitotype/chlorotype assemblages (Fig. 3). Trees showing composite (or hybrid) mitotype/chlorotype assemblages were found in 3 hybrid populations and 11 P. banksiana populations. The easternmost tree presenting such a mixed cytoplasmic genome assemblage was found in a population from Ontario where a cpDNA typical of P. banksiana was found in association with an mtDNA typical of P. contorta. Of the 106 trees carrying a mixed cytoplasmic genome assemblage, a single case of banksiana mitotype/contorta chlorotype was observed in population BANK19 from Alberta (see above), whereas all other individuals carried contorta mitotype/banksiana chlorotype combinations.

Cluster and canonical analyses
The most basal branching of the NJ tree based on pairwise F ST distances derived from mitotype frequencies separated well the two species (Fig. 4). However, five P. banksiana populations clustered between the two major groups. These populations were characterized by the cooccurrence of banksiana and contorta mitotypes, as previously defined from the mitotype network (Fig. 2). The P. contorta group was separated in two subgroups of 6 and 21 populations. These two subgroups were geographically coherent and included populations from the northwestern and south-central parts of the P. contorta sampling area, respectively. Five of the six hybrid populations were found in this last group. Additionally, two populations from southern BC fixed for mitotype 5C, CON4, and CON5 were found separated from the rest of the large subgroup of 21 populations.
The pattern of population grouping that resulted from the spatial Bayesian analysis (BAPS) of mitotype frequency data was very similar to the NJ dendrogram (Fig. 5). The optimal number of groups (k-value) was five (F CT = 0.399). The BAPS also separated well the two species, with the exception of the "Light Blue" group ( Fig. 5) that was composed by four P. contorta, six P. banksiana, and three hybrids populations located from central BC through the Saskatchewan province (H S = 0.499, H T = 0.579). The "Purple" group ( Fig. 5) was characterized by P. contorta populations presenting a high diversity of mitotypes (especially 7C, 11C, 14C, and 16C; H S = 0.685, H T = 0.820).
The "Dark Blue" group of populations (Fig. 5) contained mostly a mix of 5C and 11C mitotypes (H S = 0.513, H T = 0.642). Two groups of P. banksiana populations were also delineated: the "Orange" group (   noted between the Bayesian analysis and the NJ dendrogram based on F ST distances could be attributed to differences in the distribution of low-frequency mitotypes. Globally, population groupings obtained by both nonspatial and spatial analyses were coherent with the glacial lineages previously inferred for P. banksiana and P. contorta using the intron 1 region of the mtDNA gene nad7 (Godbout et al. 2005(Godbout et al. , 2008; see Discussion).
As expected, the BAPS analysis of cpDNA data showed a clear division between populations from the two species (Fig. 5). The two hybrid populations located in the P. contorta natural range were associated to the P. contorta group while the four remaining hybrid populations were clustered with P. banksiana.
The canonical redundant analysis conducted on the 14 natural populations sampled in situ in the current zone of contact between the two species identified significant associations (P ≤ 0.05) between cpDNA variation and the tree morphological characters defining species, as well as the ecoregions (Table 2). Such a significant relationship was not detected between cpDNA variation and ecosites. No significant relationships could be found between mtDNA variation and any of the three candidate explanatory factors.

Discussion
Postglacial mtDNA capture by P. banksiana Although the natural range of P. contorta does not currently extend further eastward than western Alberta, the present results show evidence of past mtDNA introgression of P. contorta into P. banksiana as far east as Saskatchewan in central Canada, which represents a distance in excess of 10 degrees of longitude or nearly 1000 km (Fig. 1). Indeed, while P. contorta trees carried exclusively contorta-associated mitotypes, almost half (44.3%) of P. banksiana trees from the provinces of Saskatchewan and Alberta also possessed such mtDNA background (Fig. 1). The absence of contorta-associated mitotypes in P. banksiana trees from eastern Canada (data not shown) points toward past introgression rather than ancestral polymorphism to explain the observed pattern in P. banksiana from Alberta and Saskatchewan. Indeed, shared ancestry would have translated into a more even geographic distribution of these mitotypes across the natural range of P. banksiana. This mostly unidirectional introgression pattern indicates that P. banksiana captured P. contorta mtDNA in the western part of its range. Such mtDNA capture likely occurred following the early dispersal of P. contorta through part of central Canada after the retreat of the continental ice sheet. The possibility that the natural range of P. contorta once extended further eastward than its current range was previously proposed to explain the finding of disconnected P. contorta stands in Alberta (Critchfield and Little 1966) and the distribution patterns of mtDNA-RFLP in Saskatchewan (Dong and Wagner 1993). Although the presence of a contorta-type mitotype in a population from Ontario may point to a postglacial expansion of P. contorta in this region, the absence of such mitotypes in Manitoba rather indicates the effect of a possible long-range dispersal of a limited number of individuals in the area (Fig. 1). The fossil record also supports this hypothesis, as P. contorta expanded into southern Alberta around 12 ky BP (Mac-Donald and Cwynar 1991), whereas the first expansion of P. banksiana into south Saskatchewan (~500 km eastward) was estimated at a much later period, around 7.5 ky BP (McLeod and MacDonald 1997).  The extensive mtDNA introgression detected in the western part of the natural range of P. banksiana range may be explained by the demographic disequilibrium occurring when a colonizing species spreads into an area already occupied by a related species. Indeed, an early arrival of P. contorta before P. banksiana is in concordance with observations and simulations that showed a bias in the direction of introgression between two species in a zone of contact (Currat et al. 2008;Du et al. 2011). Because of the difference in species relative abundances at the time of invasion, the transfer of genes between the two species is more likely to occur from the locally abundant species (here P. contorta) toward the invasive one (P. banksiana). This trend is also most likely for the genome with least gene flow, such as for the seed-dispersed mitochondrial genome in the Pinaceae (Petit et al. 2005;Currat et al. 2008;Du et al. 2009Du et al. , 2011. Assuming that P. contorta was well established in central Canada when P. banksiana first migrants reached the region, P. banksiana from the margin of the migration wave would have first been heavily introgressed by abundant P. contorta, and trees carrying this new genetic background would have then proliferated while P. contorta gradually declined. The decline of P. contorta in western and central Canada presumably occurred between 9 and 6 ky BP during the Hypsithermal (i.e., Holocene climatic optimum; Anderson et al. 1989). During this warmer period, the haploxylon/dyploxylon pollen ratio reversed in the subalpine region of southern Alberta, as a haploxylon pine (probably Pinus flexilis) became more abundant than a diploxylon pine (probably P. contorta; Anderson et al. 1989). The warmer and drier climate may have promoted the northward and eastward expansion of prairie species east of the Rocky Mountains (Williams et al. 2009). Remnants of this ecosystem can still be found in this region, as several grassland species currently present disjunct distributions (Strong and Hills 2003). Thus, P. contorta may have retreated to higher elevations in northern Alberta during this warmer period before it almost completely disappeared.
Putative introgression of P. banksiana into P. contorta populations has been reported using several types of characters including wood resin (Mirov 1956;Smith 1983;Pollack and Dancik 1985), cortex properties (Forrest 1980(Forrest , 1981, morphological traits (Critchfield 1957;Pauly and Rudloff 1971;von Rudloff 1975;von Rudloff and Nyland 1979;Critchfield 1980;Wheeler and Guries 1987), or allozyme data (Dancik and Yeh 1983;Wheeler and Guries 1987). As these occurrences of P. banksiana-type characters were found to be largely scattered and randomly distributed at a large scale and low frequency among P. contorta populations, Critchfield (1985) proposed that the putative introgression observed in P. contorta populations predated the last glaciation. This interpretation is in line with our findings that P. banksiana did not spread west of the Rockies during the Holocene. Furthermore, although P. banksiana has been less studied than P. contorta, variable introgression has been reported among P. banksiana populations using nuclear markers, such as allozymes (Wheeler and Guries 1987) and random amplified polymorphic DNA (RAPD) markers (Ye et al. 2002).
Introgression of P. banksiana in relation to the phylogeography of P. contorta The genotyping of the highly diverse nad7 intron 1 allowed us to determine the intraspecific lineage responsible for the early colonization of part of central Canada by P. contorta following the retreat of the continental ice sheet, as seen today from the introgressed mtDNA genome of jack pine populations from eastern Alberta and western Saskatchewan ("Light Blue" BAPS group, Fig 5). This group is also represented in populations of P. contorta located in BC, and corresponding to a glacial population that was located south of the ice sheet and west of the Rocky Mountains (Godbout et al. 2008). Indeed, based on mitotype variation, the "Light Blue" and "Dark Blue" BAPS mtDNA groups delimited in this study correspond roughly to two previously inferred P. contorta lineages originating from distinct glacial populations of P. contorta subspecies latifolia located south of the ice sheet on the western and eastern sides of the Rockies, respectively (Godbout et al. 2008). However, the large zone of secondary contact between these two intraspecific lineages previously observed in eastern BC and western Alberta was not clearly detected by the Bayesian analysis, possibly because the present sampling did not cover the entire range of P. contorta in this study or because it included populations of P. banksiana. Nevertheless, with the exception of two populations found at the end of two long branches in the NJ dendrogram (CON4 and CON5; Fig. 4), populations from the "Dark Blue" BAPS group were part of this large zone of secondary contact (Fig. 5). Indeed, most of these populations contained an admixture of both mitotypes (5C and 11C) typically representative of each of the two previously mentioned P. contorta glacial lineages. Populations from the "Purple" BAPS group of north-central BC were also part of this large zone of secondary contact, but mostly represented a mix of trees from different lineages including an additional lineage specific to P. contorta subspecies contorta typically found in western BC (Godbout et al. 2008).
According to the present results, P. contorta trees from the glacial lineage located east of the Rocky Mountains (Godbout et al. 2008; this study) did not participate significantly to the eastward colonization of central Canada, despite their geographic proximity to the region. Indeed, only three P. banksiana individuals exhibited introgression from mitotype 5C, the most typical variant of this lineage (Fig. 2). However, this lineage was presumably of smaller size than those located west of the Rockies prior to the Holocene. This hypothesis is supported by the lack of mitotype diversity observed in the two P. contorta populations located southwest of the zone of contact with P. banksiana (CON4 and CON5; Fig. 1a), confirming a previous observation based on a larger sampling of P. contorta in this region (Godbout et al. 2008). This hypothesis is also supported by the low Pinus pollen frequencies at the last glacial maximum in this region (Williams et al. 2004). Thus, at the onset of the colonization east of the Rockies, after the continental ice sheet had retreated eastward, the presumably larger size of the P. contorta population located west of the Rocky Mountains might have compensated for the migration delay imposed by the crossing of the mountain range when racing to colonize Alberta and Saskatchewan against the presumably smaller P. contorta glacial population located on the southeastern side of the range.
The eastward expansion of P. contorta into central Canada did not result in a decrease of mtDNA diversity, although the colonization was proposed to have occurred via a series of founding events (Cwynar and MacDonald 1987). Indeed, the introgressed P. banksiana trees presented seven distinct P. contorta mitotypes (n = 105; H T = 0.569; Fig. 1). When considering the "Light Blue" BAPS group only, all mitotypes detected in P. contorta ssp. latifolia were also found in P. banksiana (Figs. 1 and  5). Therefore, only a few rare and infrequent mitotypes may have been lost during the species eastward migration (Fig. 2). This diversity trend suggests that colonization may have occurred through frequent long-dispersal events that would have favored a reshuffling of genes at the migration front, and thus maintained the genetic diversity (Bialozyt et al. 2006;Fayard et al. 2009).
Asymmetry in mtDNA and cpDNA introgression: the effect of differential gene flow The quasi-absence of cpDNA introgression between the two boreal pines contrasted with the long lasting traces of introgression observed in the mitochondrial genome (Fig. 5). This difference may be explained by the contrasted mode of dispersal of the two cytoplasmic genomes in the Pinaceae, including pines: through pollen and seeds for the paternally inherited cpDNA (Neale and Sederoff 1988) and through seeds only for the maternally inherited mtDNA (Dong and Wagner 1993). Indeed, pollen flow was shown to be theoretically several orders of magnitude larger than seed flow in pines (Ennos 1994). Several comparative studies of population differentiation between seed-only dis-persed mitochondrial genomes and seed-and-pollen dispersed chloroplast, and nuclear genomes in the Pinaceae have also shown much higher population differentiation and much reduced levels of gene flow for the mitochondrial seed-dispersed genome (e.g., Burban and Petit 2003;Gamache et al. 2003;Petit et al. 2005;Gérardi et al. 2010;Godbout et al. 2010;Wei et al. 2011). Consequently, such high pollen gene flow may have rapidly resulted in the replacement of the introgressed cpDNA background by that predominantly found in the local surrounding pollen clouds. Similar asymmetric introgression of cytoplasmic genomes, though spatially more restricted, has been recently observed between coastal and interior varieties of Douglas fir (Pseudotsuga menziesii) in the northern part of their distributions in BC, where many trees with mixed mtDNA and cpDNA backgrounds indicated the likely capture of coastal mtDNA by an invading cpDNA lineage from the interior of the Rocky Mountains (Wei et al. 2011). A similar pattern of genome capture has also been reported between western and more eastern glacial lineages of Picea mariana in an area of Holocene contact in Alberta (Gérardi et al. 2010).
The canonical analysis revealed that, contrary to mtDNA, cpDNA variation within and near the zone of contact was significantly explained by the morphological classification of local trees and the ecoregion classification (Table 2), thus being well in line with species morphological differences and ecological preferences. For instance, P. contorta chlorotypes were predominant in the upper foothills, where P. contorta pollen is expected to be the most abundant, owing to the occurrence of habitats particularly favorable to this species and high local presence of this species under such conditions (Lotan and Critchfield 1990). Also, hybrid populations (except HYB3) were also fixed or nearly fixed for a single chlorotype, whether it was typical of P. banksiana in some populations or typical of P. contorta in other populations, and it matched the most abundant species in the local landscape. These various observations indicate that the differences observed among hybrid populations in their cpDNA content reflect a decisive effect of the local pollen cloud on determining the cpDNA background of trees in hybrid populations. These results are also well in line with observations on spontaneous hybridization rates between exotic and native species of poplars where the genetic origin of the local pollen cloud has been found as the most important factor to explain variation in spontaneous hybridization (Meirmans et al. 2010). These results are concordant with recent observations suggesting that genomes experiencing higher gene flow, such as the pollen and seed-dispersed chloroplast genome in the Pinaceae, are less sensitive to introgression (Currat et al. 2008;Du et al. 2009), and thus may be better suited for species delimitation (Petit and Excoffier 2009) than the seed-dispersed mitochon- drial genome. In the present case, cpDNA markers were better than mtDNA markers at delimitating current species ranges and identifying the local presence of one or the other species. This trend also reflects similar conclusions at the genus level in Picea, another Pinaceae with different transmission and dispersal of its cytoplasmic genomes. In this genus, cpDNA and mtDNA phylogenies were not congruent and the mtDNA phylogeny appeared as a better footprint of phylogeographic history while cpDNA phylogenies would more reflect the result of repeated reticulate evolution (Bouillé et al. 2011).
Although the present set of DNA polymorphisms did not allow to separate F 1 from advanced-generation hybrids, it is likely that most trees in hybrid populations were of the second type, because flowering times have been shown to differ between the two species, and true hybrid trees have been reported to produce a higher proportion of aborted pollen than pure P. banksiana or P. contorta trees (Critchfield 1980). A high prevalence of introgressed trees over true hybrids is frequent in tree species, as reported for the zone of contact between Picea mariana and P. rubens in eastern Canada (Perron and Bousquet 1997) and also between Populus alba and P. tremula in Europe (Lexer et al. 2005).
Hybrid populations HYB3 and BANK19 were the only ones to carry both species-specific chlorotypes together, where the typical contorta chlorotype was observed in 5 of 30 and 1 of 25 trees, respectively. Such a likely recent and limited incursion of contorta cpDNA alleles in banksianaidentified populations may have been caused by the prevalence of westerly winds in the region (Bartlein et al. 1998). This weak introgression trend contrasts with that observed in the same region at the intraspecific level in Picea mariana, where extensive capture of a central Canada mtDNA lineage by a BC cpDNA lineage was observed, a phenomenon presumably driven largely by westerly dominant winds carrying pollen over large distances (Gérardi et al. 2010). These different trends follow the expectation that scale of differences in phenology timing or variations in hybrid fitness are likely more important driving factors of gene exchange at the interspecific level than wind intensity and direction, such as between P. contorta and P. banksiana.
The present cpDNA results are to a large extent concordant with those obtained using cpDNA RFLPs (Wagner et al. 1987). However, while similarly reporting unambiguous species delimitation as in this study, Wagner et al. (1987) and Govindaraju et al. (1989) also detected new cpDNA variants in the region of contact. These new variants were proposed to be the result of either new mutations or recombination (Govindaraju et al. 1989). No such new chlorotypes were detected in the present sampling, and we did not notice recombination between the four cpDNA loci sampled, which may indicate that the new variants discov-ered by these authors were the result of mutation. However, evidence for recombination among cpSSR has been reported in P. contorta (Marshall et al. 2001) and mtDNA recombinants of recent origin have been observed in the zone of contact between Picea mariana and P. rubens (Jaramillo-Correa and Bousquet 2005).

Conclusions
The mtDNA and cpDNA data showed marked asymmetric introgression patterns between P. contorta and P. banksiana. The noticeable presence of P. contorta mitotypes well within the allopatric range of P. banksiana suggests an ancient mtDNA capture by P. banksiana several thousand years ago following an early postglacial expansion of P. contorta in central Canada. Contrastingly, introgression at the cpDNA level was limited because of the extensive pollen gene flow from the surrounding local species over generations. Within the zone of contact between the two species where interspecific gene flow still occurs, the cpDNA background of a given population could be well predicted from the species that was locally most abundant or from the ecological characteristics of the local environment that determine the presence of either of the two species. This trend indicates that in absence of tree morphological characterization, the most dispersed cpDNA genome should be best suited to delineate boundaries between these two species, as indicated for other Pinaceae by Du et al. (2009). Some P. banksiana and hybrid trees from Alberta and Saskatchewan differed from allopatric P. contorta and P. banksiana populations in terms of cytoplasmic genome assemblage (Fig. 3). Except for one tree, these individuals now carry a composite organelle background where cpDNA and mtDNA were inherited from P. banksiana and P. contorta, respectively. These de novo cytoplasmic assemblages may represent a new "nested" form of genomic diversity on which natural selection may act (Gérardi et al. 2010). It indicates convincingly that even far away from glacial refugia, genetic diversity can be high and bear distinctive features not seen in ancestral lineages. access to provenance tests. This research was made possible by funding from the National Sciences and Engineering Research Council of Canada grants to J. B. (strategic grant #321969 and discovery grant #46273).