SEARCH

SEARCH BY CITATION

Keywords:

  • Antirrhinum majus;
  • chloroplast DNA;
  • contact zone;
  • genetic introgression;
  • haplotype sharing

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Assessing processes of geographic expansion in contact zones is a crucial step towards an accurate prediction of the evolution of species genetic diversity. The geographic distribution of cytonuclear discordance often reflects genetic introgression patterns across a species geographic range. Antirrhinum majus pseudomajus and A. m. striatum are two interfertile subspecies that occupy nonoverlapping areas but enter in contact in many locations at the margin of their geographic distribution. We found that genetic introgression between both subspecies was asymmetric at the local scale and geographically oriented in opposite directions at both ends of their contact zone perimeter in the Pyrenees. Our results suggest that the geographic expansion of A. majus subspecies was circular around the perimeter of their contact zone and pinpoint the need to integrate different spatial scales to unravel complex patterns of species geographic expansion.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Species range expansion is a key mechanism that shapes the genetic diversity of species (Hewitt, 2000; Excoffier et al., 2009) and modifies their evolutionary potential (Lavergne & Molofsky, 2007; Pujol & Pannell, 2008; Pujol et al., 2009). Range expansion frequently leads to the formation of contact zones between populations of differentiated species (Anderson, 1949; Grant, 1971; Arnold, 1997; Barton, 2001). In such cases, genes from a foreign species might replace genes of, and therefore introgress, the gene pool of the native species (Potts & Reid, 1988, 1990; Schemske & Morgan, 1990). Genetic introgression (i.e. the transfer of genetic material in the genome of another species) is often rendered possible by fertile hybrids that occupy the contact zones and act as ‘bridges to gene flow’, therefore allowing gene exchange between species to occur (Broyles, 2002). Recurrent pollen exchanges between species and recurrent backcrossing between hybrids and parental species are then likely to generate large geographic areas of genetic introgression inside and outside contact zones (Campbell et al., 1998; Leebens-Mack & Milligan, 1998). The characterization of genetic introgression is a key step towards a better understanding of the expansion dynamics of species in contact and of the evolution of their diversity (Currat et al., 2008; Excoffier et al., 2009).

Recent work on Antirrhinum majus showed that floral trait segregation, in combination with pollinator behaviour, can explain, at least partly, the maintenance of flower colour polymorphism in one particularly narrow hybrid zone between A. majus pseudomajus and A. m. striatum subspecies (Whibley et al., 2006; Tastard et al., 2008). It is, however, currently unknown whether contact zones and gene exchanges between A. m. pseudomajus and A. m. striatum are widespread across the geographic range of both these subspecies. The general aim of our study is to understand the biogeography of these two interfertile subspecies. We expect that contact might be frequent across the species range if one subspecies progressively expands its range into the range occupied formerly by the other subspecies because the boundary between A. majus subspecies is not linear. A. m. pseudomajus is distributed around the range of A. m. striatum. Ultimately, moving boundaries sometimes result in the local replacement of the invaded species by the species pushing off the contact zone on its front of colonization (Buggs & Pannell, 2007; Pannell & Pujol, 2009). Alternatively, stable boundaries between both taxa can be maintained at equilibrium between migration and selection (Barton & Hewitt, 1989; Bull, 1991). When the contact zone between two species ranges is not linear, one could expect geographically complex expansion patterns and/or reciprocal gene exchanges to occur and result in multiple sites of genetic introgression. The detection of such widespread pattern is important because it might result in the long term in the genetic admixture of their formerly differentiated genomes. Evidence to determine whether taxa replacement, maintenance of species boundaries or admixture are the most likely evolutionary outcomes can be provided by the analysis of the geographic distribution of relict uniparentally inherited DNA (i.e. mitochondrial or chloroplastic DNA) where the nuclear genome is being replaced (Potts & Reid, 1988, 1990; Schemske & Morgan, 1990).

To establish whether parapatric boundaries between A. majus subspecies are moving following a complex geographic expansion pattern, we studied the geographic distribution of the association between chloroplast haplotypes (maternally inherited) and a nuclear gene (biparentally inherited) regulating the main taxonomic criterion, which is the magenta flower colour for A. m. pseudomajus and the yellow flower colour for A. m. striatum (see the study system section for details on the taxonomy of the species; Rothmaler, 1956 and Sutton, 1988) over the range of the species. We then searched for evidence of genetic introgression between A. m. pseudomajus and A. m. striatum. Although chloroplast haplotype sharing across taxa boundaries is often the outcome of genetic introgression (Rieseberg & Soltis, 1991; Wendel & Doyle, 1998; Linder & Rieseberg, 2004), caution must be taken when interpreting patterns of haplotype sharing because convergence or incomplete sorting of ancestral polymorphism might generate similar patterns (Muir & Schlotterer, 2005; Lexer et al., 2006). In cases of retention of ancestral polymorphism or convergence, we would expect cytonuclear associations to be randomly distributed in a mosaic pattern over the species geographic range (Fig. 1a). In contrast, if chloroplast sharing between both subspecies is the result of introgression, we would expect discordant cytonuclear associations to be located close to the perimeter zone formed by the contact between subspecies (Fig. 1b). Geographic sectors characterized by the high frequency of one cytonuclear association are also expected if heterogeneous selection spatially structured A. majus ancestral polymorphism. In this article, we confront those hypotheses to establish the most likely scenario of evolutionary history that can explain the observed geographic distribution of cytonuclear associations in A. majus. Our investigation of those scenarios was rendered possible by the broad scale at which our study was conducted, i.e. the species geographic range, which allowed us to uncover the geographic direction of genetic introgression between both subspecies around their geographic boundaries.

image

Figure 1.  Expected patterns of chloroplast sharing between species. Black and white colours represent Species 1 and Species 2, respectively. The contact zone perimeter between both species is symbolized by a dotted line. Squares represent the chloroplast haplotype 1 most often associated with Species 1, and diamonds represent the chloroplast haplotype 2 most often associated with Species 2. Under the hypothesis that black squares and white diamonds reflect the ancestral state, then the two less frequent associations (i.e. black diamonds and white squares) are called ‘discordant’. (a) Random geographic distribution of discordant associations within each species range expected that results from the retention of ancestral polymorphism or convergence. (b) Geographic structure of discordant associations found only around the contact zone perimeter that results from local introgression between both species where genes can be exchanged.

Download figure to PowerPoint

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Study system

Antirrhinum majus (Scrophulariaceae) is a herbaceous short-lived perennial plant characterized by a patchy distribution in southern Europe. Its geographic distribution is centred over the Pyrenees, between north-eastern Spain and south-western France. The two subspecies A. m. striatum and A. m. pseudomajus occupy largely parapatric geographic regions. The geographic area occupied by A. m. striatum is surrounded by the geographic area occupied by A. m. pseudomajus (Fig. 2). Taxonomic determination of A. majus subspecies is mostly based on the colour of flower corolla. A. m. pseudomajus is characterized by magenta flowers. It is referred to interchangeably in the literature as A. m. ssp. majus and A. m. ssp. linkianum. Some authors include the ssp. cirrhigerum as a variety of A. m. ssp. linkianum. A. m. striatum is characterized by yellow flowers. It is referred to interchangeably in the literature as A. latifolium ssp. striatum, A. huetii and A. braun-blanquetii (Rothmaler, 1956; Sutton, 1988). It is important to note that A. m. pseudomajus and A. m. striatum are interfertile and share pollinators (Whibley et al., 2006; Andalo et al., 2010).

image

Figure 2.  Geographic distribution of A. majus cytonuclear associations. Magenta and yellow layers represent, respectively, Antirrhinum majus pseudomajus and A. m. striatum geographic ranges. Within each subspecies, black symbols represent populations sampled for this study. Squares and diamonds represent populations characterized by Haplotype I and Haplotype II, respectively. Pie charts are only presented for populations that are polymorphic at the ROS1 locus. The magenta and the yellow proportion of the pie charts represent the respective frequencies of ROS1-M and ROS1-Y alleles in the population. Magenta and yellow arrows indicate the hypothetical scenario of range expansion followed by A. m. pseudomajus and A. m. striatum that is supported by our data. Brown lines represent elevation isoclines above 1800 m.

Download figure to PowerPoint

Plant material sampling strategy

A total of 685 plants were sampled from 2002 to 2007 in 55 allopatric or parapatric populations distributed over the geographic range of the species. Geographic coordinates of populations were recorded by using a GPS device (Garmin, Olathe, KS, USA). A numerical scoring system was used to rank magenta and yellow flower colour phenotypes visually, following methods developed by Whibley et al. (2006). Obviously, plants that displayed yellow flowers were classified as A. m. striatum whereas plants that displayed magenta flowers were classified as A. m. pseudomajus. Population characteristics are summarized in Table S1a and S1b. For each individual, young leaves and shoot tips were collected and stored at −20 °C until DNA was extracted by using the DNeasy Plant Mini kit (Qiagen, Hilden, Germany).

Molecular analyses

ROSEA genotyping

The ROSEA locus is made of 2 MYB – myeloblastosis – regulatory genes controlling floral pigmentation intensity, out of which ROS1 has the main role in flower colour variation (Schwinn et al., 2006). ROS1 sequences can be grouped in three main haplotypes ROS1-Ma, ROS1-Mb and ROS1-Y (Whibley, 2004). ROS1-Ma and ROS1-Mb haplotypes are diagnostic of A. m. pseudomajus and are grouped under the name of ROS1-M whereas the ROS1-Y haplotype is diagnostic of A. m. striatum (Whibley, 2004). ROS1 genotypic data were available for the 14 populations (n = 166 plants) that were previously examined by Whibley et al. (2006). We obtained ROS1 genotypic data for the remaining 41 populations (n = 519 plants) using the RG4/RR21, RG6/RR21 and RG1/RR21 primers in a single PCR, following the protocol established by Whibley (2004).

PCR–RFLP analysis of chloroplast DNA

Maternal lineages were determined in the 55 populations (n = 685 plants) by genotyping the 1.6-kb psbC [psII 44-kDa protein] – trnS [tRNA-Ser(UGA)] intergenic region, using the CS universal primers (Demesure et al., 1995). Sequencing of this chloroplast region revealed two haplotypes that differed at two SNP loci, one of which was included in a MseI restriction site. We therefore obtained two different haplotypes after digestion of the psbC-trnS fragment by the Mse I enzyme. Haplotype I was characterized by eight Mse I restriction sites that generated a nine-band profile on agarose gel. Haplotype II was characterized by a 10-band profile. The PCR amplification protocol is presented in the supplementary online material.

Data analyses

To examine cytonuclear associations, we calculated ROS1 allelic frequencies and chloroplast haplotype frequencies within each population and mapped these frequencies using ArcGis (ESRI, Redlands, CA, USA) software. To determine whether subspecific patterns of chloroplast haplotype sharing were the result of evolutionary convergence, incomplete sorting of ancestral polymorphism or introgression, we assessed the role of the geographic distance between populations of different subspecies on the geographic distribution of chloroplast haplotypes. To do so, we calculated the Euclidian geographic distance to the closest population of the other subspecies for every population within each subspecies. We then tested whether this geographic distance differed between populations that share chloroplast haplotypes with the other subspecies and populations that do not share chloroplast haplotypes with the other subspecies. We performed a two-sample t-test built on the basis of 2000 permutations (Good, 2000) in R (R Development Core Team, 2007, Vienna, Austria).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Relationship between flower colour and ROS1 genotype

At the population level, 67% of the 55 populations were assigned to A. m. pseudomajus and 33% to A. m. striatum on the basis of their flower colour phenotype (= 37 A. m. pseudomajus populations and = 18 A. m. striatum populations). Forty-six populations out of 55 were monomorphic at the ROS1 locus. All plants in those populations presented the same homozygote genotype at the ROS1 locus, being either ROS1-M/ROS1-M in 34 A. m. pseudomajus populations or ROS1-Y/ROS1-Y in 12 A. m. striatum populations (Table 1). Six A. m. striatum populations (Els, Fab, Pal, Pom, Thu and Tri) and three A. m. pseudomajus populations (Hor, Lag and Lou) were polymorphic at the ROS1 locus (allelic frequencies are presented on Fig. 2).

Table 1.   Cytonuclear associations.
Chloroplast haplotypePopulation subspeciesROS-1 genotypes
  1. *Most frequent genotype in bold. The number of populations is indicated between parentheses.

Haplotype I (35)Antirrhinum majus pseudomajus (31)ROS1-M/ROS1-M (29)
ROS1-M/ROS1-M; ROS1-M/ROS1-Y; ROS1-Y/ROS1-Y (2)
A. m. striatum (4)ROS1-Y/ROS1-Y (3)
ROS1-M/ROS1-Y; ROS1-Y/ROS1-Y (1)
Haplotype II (20)A. m. pseudomajus (6)ROS1-M/ROS1-M (5)
ROS1-M/ROS1M; ROS1-M/ROS1-Y (1)
A. m. striatum (14)ROS1-Y/ROS1-Y (9)
ROS1-M/ROS1-M; ROS1-M/ROS1-Y; ROS1-Y/ROS1-Y (5)

Over all individuals, plants that displayed yellow flowers were characterized by the genotypes ROS1-Y/ROS1-Y, ROS1-Y/ROS1-M or ROS1-M/ROS1-M at the respective frequencies of 94%, 4.5% and 1.5%. Plants that displayed magenta flowers were characterized by the genotypes ROS1-Y/ROS1-Y, ROS1-Y/ROS1-M or ROS1-M/ROS1-M at respective frequencies of 0.8%, 2.2% and 97%. Such correlation between ROS1-Y and the yellow colour and between ROS1-M and the magenta colour is in agreement with the previous study conducted by Whibley et al. (2006).

Distribution of chloroplast DNA genotypes

Each particular population was characterized by a unique psbC-trnS chloroplast haplotype. The geographic distribution of population haplotypes was nonoverlapping across the geographic range of the species (Fig. 2). Haplotype I was found in 79% of A. m. pseudomajus populations and in 20% of A. m. striatum populations. Haplotype II was found in the remaining 21% of A. m. pseudomajus populations and 80% of A. m. striatum populations (Table 1).

Among A. m. pseudomajus populations, the chloroplast haplotype depended significantly on whether populations were located closely to A. m. striatum populations. Most of the A. m pseudomajus populations that were characterized by Haplotype I were distant from A. m. striatum populations (Fig. 2). In contrast, A. m. pseudomajus characterized by Haplotype II could only be found in populations located closely to the contact zone perimeter. The mean distance between A. m. pseudomajus populations characterized by Haplotype I and the closest A. m. striatum population (mean distance ± SD = 40.3 ± 30.1 km) was significantly larger (using a permutation t-test < 0.05) than the mean distance between A. m. pseudomajus populations characterized by Haplotype II and the closest A. m. striatum population (mean distance ± SD = 12.9 ± 5.7 km). Within the A. m. striatum geographic range, no correlation between the occurrence of a chloroplast haplotype and the distance to the nearest A. m. pseudomajus population was detected. The mean distance between A. m. striatum populations characterized by Haplotype II and the closest A. m. pseudomajus population (mean distance ± SD = 16.6 ± 7.0 km) was not significantly different (using a permutation t-test > 0.05) than the mean distance between A. m. striatum populations characterized by Haplotype I and the closest A. m. pseudomajus population (mean distance ± SD = 18.3 ± 11.1 km). It is important to note that the few A. m. striatum populations characterized by chloroplast Haplotype I (n = 4) were all grouped on the west border of A. m. striatum geographic distribution (see Fig. 2). It is also important to note that such analysis in A. m. striatum was limited by the small number of A. m. striatum populations that are located far from the contact zone perimeter, which is a direct consequence of the narrower geographic area occupied by A. m. striatum. Furthermore, A. m. pseudomajus populations characterized by Haplotype II were located in the east of the contact zone perimeter whereas A. m. striatum populations characterized by Haplotype I were located in the west of the contact zone perimeter (Fig. 2). The direction of the geographic gradient formed by ROS1 allele frequencies in the east was different from the one in the west of the contact zone.

Cytonuclear association

Because of the correlation between ROS1 and flower colour, the overall pattern of cytonuclear association was very similar to the pattern presented above. Most of the A. majus populations were characterized either by the association of chloroplast Haplotype I and the ROS1-M allele or by the association of chloroplast Haplotype II and the ROS1-Y allele. Among populations characterized by Haplotype I, most of them (83%) were also characterized by the fixation of the ROS1-M allele whereas the remaining populations were characterized either by polymorphism at the ROS1 locus (8.5%) or by the fixation of the ROS1-Y (8.5%). Among populations characterized by Haplotype II, only 45% were characterized by the fixation of the ROS1-Y allele whereas the other populations were characterized either by polymorphism at the ROS1 locus (30%) or by the fixation of the ROS1-M allele (25%) (Table 1).

In most of the A. m. pseudomajus populations (= 29 out of 37), all individuals were characterized by the cytonuclear association of chloroplast Haplotype I and ROS1-M. This includes all the A. m. pseudomajus populations that were distant from the contact zone perimeter (Fig. 2). The most frequent cytonuclear association that characterized A. m. striatum populations was found in 50% of A. m. striatum populations (= 9). In those populations, all individuals were characterized by the same cytonuclear association (chloroplast Haplotype II and ROS1-Y). Around the contact zone perimeter (see Fig. 2), we found five A. m. pseudomajus populations (Arl, Div, Per, Pra and Sal) where all individuals were characterized by the cytonuclear association of chloroplast Haplotype II and ROS1-M. Those populations were located at the eastern side of the contact zone perimeter (Fig. 2). Around the contact zone perimeter, we also found three A. m. striatum populations (And, For and Val) where all individuals were characterized by the association of the chloroplast Haplotype I and ROS1-Y. Those populations were located at the western side of the contact zone perimeter (Fig. 2).

In two populations of the three A. m. pseudomajus populations that were polymorphic at the ROS1 locus, Haplotype I was associated with a high frequency of ROS1-M. Similarly, in five populations of the six A. m. striatum populations that were polymorphic at the ROS1 locus, Haplotype II was associated with a high frequency of ROS1-Y alleles (Table 1). Interestingly, such populations at an intermediary stage of genetic introgression were always very close to the contact zone perimeter (Fig. 2).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Heterogeneous selection of ancestral polymorphism vs. genetic introgression

One hypothesis explaining that chloroplast haplotypes are shared between A. m. pseudomajus and A. m. striatum is that such pattern results from the retention of ancestral polymorphism without selection being involved. Under such scenario, we would expect cytonuclear associations to be widespread across the entire range of A. majus (Fig. 1). This was however not the case. We found them to be grouped in four discrete geographic areas. We therefore discarded this hypothesis (Fig. 2). Another hypothesis that can be invoked is that local heterogeneous selection is responsible for the geographic distribution of the four cytonuclear associations in four discrete geographic sectors in the absence of interspecific introgression. Under such scenario, natural selection would have differently advantaged four ancestral cytonuclear associations between the chloroplast Haplotypes I and II and ROS1 alleles in four regions. In populations located between those four regions where cytonuclear associations were fixed, we found populations that were polymorphic for ROS1 alleles but not for chloroplast haplotypes (Fig. 2). These polymorphic populations formed geographically orientated gradients in ROS1 allele frequencies that were all located onto the contact zone perimeter between A. m. pseudomajus and A. m. striatum. Gradients were found on the east side and on the west side of the contact zone perimeter. Because the contact zone perimeter is where we expect gene exchanges between both subspecies to occur, the geographic distribution of chloroplast haplotypes between subspecies and the gradients of ROS1 allele frequencies that we found in the contact zone perimeter are more likely reflecting genetic introgression between subspecies than geographically heterogeneous selection on cytonuclear ancestral polymorphism.

Local patterns of genetic introgression reflect a circular range expansion scenario

The geographic distribution of cytonuclear associations suggests that chloroplast Haplotype I was historically associated with A. m. pseudomajus. This is because chloroplast Haplotype I was more frequent in A. m. pseudomajus populations, especially those that were geographically isolated from A. m. striatum populations (i.e. allopatric populations) whereas Haplotype II was only found in A. m. pseudomajus populations located in the contact zone perimeter (i.e. parapatric populations). The distribution of chloroplast Haplotypes was less strikingly structured among A. m. striatum populations. It would seem nevertheless logical, in regard of Haplotypes I and II distribution in A. m. pseudomajus, that Haplotype II was historically associated with A. m. striatum. Under the assumption that Haplotype I and Haplotype II were originally associated specifically with A. m. pseudomajus and A. m. striatum, respectively, the geographic distribution of subspecies, chloroplast haplotypes and nuclear ROS1 alleles revealed areas of cytonuclear discordance. In those areas, chloroplast haplotypes were not associated with the expected subspecies. This was the case on the east side of the contact zone perimeter for six A. m. pseudomajus populations characterized by chloroplast Haplotype II and a high frequency of ROS1-M alleles that had often reached fixation. This is probably because A. m. striatum plants were previously occupying the sites where those A. m. pseudomajus populations are nowadays found. Historically, those six populations were probably displaying yellow flowers and were characterized by matching chloroplast nuclear genotypes, i.e. Haplotype II and ROS1-Y. It is plausible that cytonuclear discordance emerged because nuclear genes of foreign populations were dispersed and introgressed the gene pool of local populations. The exact inverse scenario can be observed on the west side of the contact zone perimeter in four populations of A. m. striatum, which habitat was probably occupied previously by A. m. pseudomajus populations. Such geographic distribution of cytonuclear associations could be interpreted as reflecting asymmetric introgression between subspecies at a local scale, i.e. unidirectional introgression of ROS1 alleles of one subspecies into the gene pool of the second subspecies. At the broad scale of the species geographic distribution, such directional genetic introgression, however, appeared to be inverted between the east side and the west side of the contact zone perimeter. Because the genotype at the ROS1 locus determines whether a plant belongs to A. m. pseudomajus or to A. m. striatum, the spread of ROS1 alleles reflects the spread of the corresponding subspecies. Our results therefore reflect a progressive shift in the geographic range of both A. m. pseudomajus and A. m. striatum. Under such scenario, both subspecies expanded and/or still expand their ranges in opposite directions on the east and the west side of the contact zone perimeter, which ultimately results in their global range expansion being articulated around each other into a circular pattern (Fig. 2).

The relative role of selection and dispersal in the spread of ROS1 alleles

Either selection or dispersal can generate and maintain genetic introgression patterns, such as those detected in our study (Currat et al., 2008). Local selection might explain the local asymmetry in the introgression pattern, even in the presence of bidirectional gene flow. In such case, we would expect cytonuclear discordant associations ‘Haplotype II/ROS1-M’ and ‘Haplotype I/ROS1-Y’ to provide a selective advantage, respectively, on the east side and on the west side of the contact zone perimeter. When patterns of introgression are asymmetric, they might result from intrinsic attributes of species, such as prezygotic asymmetric barriers [e.g. asymmetric pollen-style incompatibilities (Cruzan & Arnold, 1994)], sex-biased dispersal (Petit et al., 2003) or post-zygotic asymmetric barriers [e.g. partial hybrid sterility (Shuker et al., 2005), which are commonly attributed to cytonuclear interactions (Levin, 1971; Tiffin et al., 2001)]. The hypothesis of one subspecies having an intrinsic advantage over the other subspecies when introgressing a foreign gene pool can be discarded because reciprocal patterns of introgressive hybridization between subspecies were detected on the west and the east side of the contact zone perimeter. Our results bring evidence that genes of each subspecies have the potential to introgress the other subspecies. They therefore corroborate the absence of intrinsic post-pollination barriers to reproduction between both subspecies previously found in an experimental study by Andalo et al. (2010). Local selection might also be driven by extrinsic factors. Environmental conditions might exert selective pressures on the ROS1 locus that vary between regions where genetic introgression was found. Such selective pressures might also target nuclear genes that are linked with ROSEA. We acknowledge the limits of our genetic assay based only on the single-locus ROSEA, which is responsible for the taxonomic criterion determining to which subspecies a plant belongs. Investigating more markers would bring a more complete picture about the extent of genetic introgression between both subspecies and would be informative on the role played by local selection on the spread of ROS1 alleles. Local asymmetric introgression patterns might also be explained by recurrent unidirectional gene exchanges. Because asymmetric introgression was restricted to specific geographic areas, local environmental barriers to gene flow (valleys, mountains, etc.) might be responsible for local unidirectional gene flow. Our study therefore calls for testing whether specific environmental or physical conditions on each side of the contact zone might exert directional constraints to gene flow. Finally, biotic interactions might also be involved in the spread of ROS1 alleles at the local scale. Experimental pollination studies brought evidence of a constancy phenomenon in the pollinating behaviour of bumblebees that was driven by A. majus flower colour, i.e. pollinators visited preferentially the same morph during a foraging sequence (Jones & Reithel, 2001; Tastard, 2009). Such pollinator behaviour was already shown to affect the evolution of a floral trait coded by a single locus (Jones & Reithel, 2001). In our case, such behaviour might result in positive frequency-dependent selection on flower colour that would ultimately reinforce or accelerate the spread of ROS1 alleles. Such process would counteract the spread of rare variants in a population and is therefore not expected to be at the origin of the asymmetric introgression of ROS1 alleles. It might, however, participate to the fixation of a new variant in a population that is submitted to massive unidirectional gene flow from the other subspecies.

Cytonuclear discordance as a result of pollen flow

Genetic introgression patterns such as those detected between A. m. pseudomajus and A. m. striatum are a common outcome when invading populations can spread their nuclear genes at a long distance by means of pollen flow and seed dispersal is limited (Petit et al., 2003). Such hypothesis is not exclusive because the geographic distribution of cytonuclear associations that we observed could also be explained by demographic expansion through seed dispersal. In such case, the demographic imbalance between invaders and residents would result in the asymmetric introgression of genes from the resident species genome into the invader genome (Currat et al., 2008). Dispersal characteristics of A. majus, however, bring support to the first hypothesis, i.e. genetic introgression by pollen flow. Indeed, A. majus seeds are very small and light [<15 mg (Andalo et al., 2010)] and can mostly be dispersed at a short distance of the maternal plant by gravity. In contrast, A. majus pollen is transported by bumblebees (several Bombus species) and carpenter bees (Xylocopa sp.) and is therefore likely to migrate across long distances (Whibley, 2004). Indeed, distance covered by carpenter bees of the species Xylocopa violacea can reach 1.2 km (Molitor, 1937) whereas bumblebees of the species Bombus terrestris can cover up to 2.8 km (Chapman et al., 2003; Darvill et al., 2004). In the light of such dispersal characteristics, the geographic scale at which we observed the signature of genetic introgression reinforces our view that the spread of nuclear genes across subspecies boundaries in A. majus was/is progressive. Such progressive spread certainly involved populations, either disappeared or still present, that were separated by close distances suitable for pollinator browsing. Such populations would then play the role of a relay for pollinators and act as directional bridges to gene flow.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Documented examples of species geographic expansion in a contact zone generally imply a unique geographic direction at the scale of the species (Martinsen et al., 2001; Rohwer et al., 2001; Melo-Ferreira et al., 2005). Here, we found that A. m. pseudomajus invaded what was previously the habitat of A. m. striatum by expanding its range northward on the east side of the contact zone perimeter whereas A. m. striatum expanded its range southward within the initial habitat of A. m. pseudomajus on the west side of the contact zone perimeter. Both subspecies appear thus to replace each other in a rotation movement at the scale of the species geographic range. Ultimately, this circular mode of geographic expansion might result in the global admixture of both subspecies nuclear genomes. Evolutionary consequences of genetic admixture in A. majus might therefore be expected to influence the evolutionary dynamics of the species at a global scale. This system, because it integrates reciprocal gradients of range expansion and genetic admixture in the two subspecies, constitutes a unique opportunity to evaluate their relative impact on the evolutionary potential of a species. It was possible to detect this surprising geographic pattern because we evaluated the geographic distribution of few but spatially structured chloroplastic and nuclear loci in multiple populations from geographically distinct sectors of the whole contact zone perimeter between A. m. pseudomajus and A. m. striatum. Our study therefore reinforces the current view that direction and speed of hybrid zone displacement can vary across replicates (Hairston et al., 1992; Britch et al., 2001; Buggs & Pannell, 2007). It also pinpoints the need to take into account multiple sites when studying contact zones between species because a broad geographic scope might reveal different patterns than those observed at a local scale. Indeed, focusing on a restricted area of the contact zone might shed light on species geographic range expansion patterns that are not representative of the whole species expansion dynamics.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

All the work presented in this article was supported by core funding from the French CNRS and the University of Toulouse Paul Sabatier, France. AK is supported by a PhD grant from the French ministry of research. We thank Nick Barton, Pauline Garnier-Géré, Ferran Palero and the anonymous reviewers for their useful comments on the manuscript and the Coen group from the JIC Norwich for useful discussions in the early stage of the study.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. Supporting Information

Table S1aAntirrhinum majus pseudomajus population characteristics.

Table S1bAntirrhinum majus striatum population characteristics.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JEB_2276_sm_TableS1.doc96KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.