‘When introgression takes place between a tetraploid and diploid population, there is a strong tendency for gene flow to proceed in only one direction, from the diploid to the tetraploid. If the hybrids produced in this way, or their backcross progeny, were well adapted to a newly available niche, such rare events could have evolutionary consequences far out of proportion to the rarity of their occurrence.’
G. L. Stebbins (1971, p. 149)
Gene flow from diploid to polyploid species could have significant effects on the morphology and ecology of polyploids. The potential of such introgression for bringing about evolutionary change within polyploids has long been recognized, although there are few examples of the process in the wild. Here, we focus on introgression between the diploid species, Senecio squalidus, and the tetraploid, S. vulgaris, which resulted in the origin of a variant form of S. vulgaris that produces radiate rather than nonradiate flower heads. The radiate variant of S. vulgaris is more attractive to pollinators and has a higher outcrossing rate. We review recent work that has isolated and characterized two regulatory genes, RAY1 and RAY2, that control presence of ray florets in radiate flower heads, and which have been introgressed into S. vulgaris from S. squalidus in the recent past. We identify a copy of RAY2 in S. vulgaris (RAY2b) homeologous to the copy (RAY2a) previously isolated, thus providing further evidence that S. vulgaris is allotetraploid. We also show that the RAY2a-R allele, which is fixed in radiate S. vulgaris, occurs at intermediate frequency in S. squalidus. Thus, based on this result, it is not possible to distinguish whether radiate S. vulgaris originated once or multiple times following hybridization between nonradiate S. vulgaris and S. squalidus.
There has been considerable interest in recent years in the types of genetic and epigenetic changes that take place in allopolyploid species causing the release of novel genetic variation and alterations to gene expression (reviewed in Doyle et al., 2008). Such changes occurring during the origin and early establishment of allopolyploids have received particular attention relative to those generating genetic diversity at later stages. This is particularly true with regard to the possible importance of introgressive hybridization (introgression) occurring at later stages. Stebbins (1971, p. 149) emphasized the potential importance of introgression as a means of generating adaptive genetic variation in polyploids in stating that, ‘Evidence from variation patterns in nature suggests that unilateral introgression has played a significant role in increasing both the morphological range of variation and the ecological range of tolerance of many polyploids. The type of variation pattern which could have been produced most easily by this process is one in which a widespread tetraploid occurs sympatrically in different parts of its range with several different diploids, and in each region tends to possess races which resemble the diploids found in that particular region.’ Although there is much evidence that gene flow can occur across a ploidy barrier, for example between diploid and tetraploid species (see later), there is currently very little detailed evidence that such gene flow may have an important effect on the morphology and ecology of a polyploid as proposed by Stebbins (1971). It is the purpose of this paper to draw attention again to this possibility and, in particular, to recent work in Senecio which has documented the transfer of genes from a diploid to a polyploid species, resulting in significant alterations to the morphology and ecology of the polyploid.
Introgression of genes affecting fitness
Introgressive hybridization between species is well documented in plants and animals (Rieseberg & Wendel, 1993; Arnold, 2006; Baack & Rieseberg, 2007), and results in the transfer of DNA between species following recurrent backcrossing of hybrid to parental individuals. Evidence for introgression occurring in the wild is based largely on studies employing genetic markers thought to be neutral to the effects of selection (Arnold, 2006). By contrast, firm evidence for the introgression of genes that directly affect fitness is surprisingly lacking. Genomic regions that do not contribute to reproductive isolation move relatively freely between hybridizing diploid species of Helianthus (Yatabe et al., 2007) and Silene (Minder & Widmer, 2008), in contrast to regions that contribute to hybrid sterility and ecological divergence between these species. Such species are described as having porous genomes with regard to interspecific gene flow, which fits the ‘genic’ concept of speciation (Wu, 2001) that reproductive isolation is controlled by a moderate number of individual genes rather than the whole genome. However, although genes that contribute to ecological divergence may be exchanged between interfertile species less easily than genes that do not affect fitness, adaptive introgression is expected to proceed if a gene from a donor species has a positive effect on the fitness of the recipient species, even if F1s exhibit very low fertility (Piálek & Barton, 1997).
Examples of adaptive introgression in fungi, plants and animals have recently been reviewed by Arnold et al. (2008). Several studies indicate that introgression has occurred, based on the distributions of anonymous and putatively neutral markers across species, and imply that it is adaptive, resulting in range expansion and niche shifts (Klier et al., 1991; Neuffer et al., 1999; Milne & Abbott, 2000; Rieseberg et al., 2007; and examples therein). However, the link between introgression and adaptive change in these examples is usually only correlative. In plants, some of the best evidence of adaptive introgression comes from studies on sunflowers, Helianthus, (Whitney et al., 2006) and Iris(Anderson, 1949; Arnold et al., 2008). In sunflowers, there is preliminary evidence that herbivore resistance was transferred from Helianthus debilis to H. annuus annuus during the origin of the stabilized introgressant, H. annuus texanus (Whitney et al., 2006). In Iris it has been shown that quantitative trait loci (QTLs) can be transferred by hybridization and backcrossing, allowing backcrossed recipient individuals to survive conditions to which the donor species is adapted (Martin et al., 2006). Although this evidence is important, it has not been demonstrated that the process has occurred in the wild and caused an extension to the ecological range of a particular Iris species. Also, the specific genes within the QTLs shown to affect fitness in Iris have yet to be isolated and characterized.
As well as enabling the transfer of fitness genes between species, introgression has been shown in some instances to cause widespread genomic and epigenomic changes in a recipient species similar to those caused by the merger of divergent genomes during allopolyploid speciation. For example, introgression of a small proportion of the genome of wild rice (Zinzania latifolia Griseb) into cultivated rice (Oryza sativa L.) caused the mobilization of transposable elements (Liu & Wendel, 2000; Shan et al., 2005), changes to DNA methylation patterns (Liu et al., 2004), and extensive sequence modification (Wang et al., 2005) throughout a large part of the genome of the recipient species. By causing such changes, introgression clearly has the potential for bringing about significant evolutionary change in a recipient species.
Introgression across a ploidy barrier
Most documented cases of introgression in the wild involve species that have the same chromosome number. Less common are examples of gene exchange between species of different ploidy level, for example diploid and tetraploid species, because triploid hybrids are either not produced (triploid block, Woodell & Valentine, 1961; Ramsey & Schemske, 1998) or, if produced, generate a high proportion of unbalanced, nonfunctional gametes and consequently are highly sterile (Ramsey & Schemske, 1998). Occasionally, however, functional balanced gametes (haploid, diploid or triploid) might be produced by triploid hybrids, leading to the production of viable and fertile F2 and backcross offspring. Alternatively, fertile tetraploid F1 hybrids might be formed through the fusion of an unreduced gamete produced by the diploid parent and a normal gamete from the tetraploid parent. In these ways, therefore, a ploidy barrier can be breached and result in gene transfer. Stebbins (1971) proposed that such gene transfer is more likely to occur from diploid to tetraploid species, although examples are known of it also proceeding in the reverse direction (Levin, 1978). The transfer of genes across a ploidy barrier has been documented between several crops and their wild relatives that differ in ploidy level (Jorgensen & Andersen, 1994; Guadagnuolo et al., 2001; Hansen et al., 2003; Morrell et al., 2005; Weissmann et al., 2005). However, there are very few cases of the same happening between wild diploid and polyploid species in the wild, (although see Elkington, 1984; Brochmann et al., 1992; and Slotte et al., 2008 for some examples).
A case study of introgression of fitness genes across a ploidy barrier in Senecio
In the remainder of this paper, we consider gene transfer via introgressive hybridization across a ploidy barrier between two herbaceous species of Senecio (Asteraceae). The two species are Senecio squalidus L. (2n = 20), a diploid, invasive, self-incompatible, short-lived perennial species in Britain and Ireland, and S. vulgaris L. (2n = 40), a tetraploid, self-compatible, annual species native to Britain and Ireland, and widely distributed throughout temperate regions of the world. Senecio squalidus is a newly originated homoploid hybrid species derived from hybrid material between S. aethnensis Jan. ex DC. and S. chrysanthemifolius Poiret on Mount Etna, Sicily, that was introduced to the UK c. 300 yr ago (James & Abbott, 2005; Abbott et al., 2009). We focus here on the transfer from S. squalidus to S. vulgaris of a particular locus that controls presence of ray florets in flower heads (capitula), which in turn has a marked effect on attractiveness to pollinators, rate of outcrossing and fitness of S. vulgaris plants (Marshall & Abbott, 1982, 1984; Abbott & Irwin, 1988; Abbott et al., 1998). In addition to reviewing the literature that relates to this example of introgression, we present new data on the duplication of this locus in S. vulgaris, and the distribution of allelic variants of one of the introgressed genes in the donor parent species, S. squalidus, and its progenitors that grow on Mount Etna.
In Britain and Ireland S. vulgaris occurs in two forms. One form, S. vulgaris var. vulgaris L., possesses nonradiate capitula containing only hermaphrodite disc florets, while the other, S. vulgaris var. hibernicus Syme, has radiate capitula with an outer whorl of female ray florets in addition to the central disk of hermaphrodite disk florets (Fig. 1). The radiate variant was first recorded in Cork, Ireland, in 1866 (Syme, 1875); however, a herbarium specimen of the same variant collected in Oxford in 1832 was later discovered by Crisp (1972). Trow (1912) showed that crosses between the two variants yielded F1 plants with capitula-bearing short ray florets, and that on selfing these plants radiate, short-rayed and nonradiate plants appeared in the F2 generation in a 1 : 2 : 1 ratio. Thus, it was concluded that the presence of ray florets in S. vulgaris is controlled by alleles at a single locus, and that heterozygotes can be recognized directly through the production of short ray florets. Soon after its discovery, it was speculated that the radiate variant originated as a result of introgressive hybridization between nonradiate S. vulgaris var. vulgaris and radiate S. squalidus. The two taxa occasionally produce a highly sterile triploid hybrid (2n = 30) in the wild (Marshall & Abbott, 1980), and records indicate that radiate S. vulgaris frequently appeared in an area shortly after S. squalidus had invaded that area. Thus, between 1870 and 1930, the two taxa exhibited a parallel geographical spread throughout parts of Britain and Ireland with the spread of radiate S. vulgaris following that of S. squalidus (Crisp, 1972). However, Stace (1977) disputed the introgressive origin of radiate S. vulgaris, and suggested that it was equally likely to have arisen from nonradiate S. vulgaris by single gene mutation.
The hypothesis of an introgressive origin of radiate S. vulgaris was strengthened by studies that successfully synthesized plants resembling the taxon from crosses made between S. squalidus and nonradiate S. vulgaris (Ingram et al., 1980; Lowe & Abbott, 2000), and also by the finding that an allozyme marker present in S. squalidus occurs at relatively high frequency among radiate plants of S. vulgaris, but is absent from nonradiate plants (Abbott et al., 1992). This allozyme marker is encoded by a gene not linked to the locus controlling presence/absence of ray florets, suggesting that different parts of the genome of S. squalidus have been introgressed into S. vulgaris during the production of the radiate variant. Conclusive evidence that radiate S. vulgaris originated as a result of gene transfer from S. squalidus was obtained following the isolation and molecular characterization of two genes controlling presence of ray florets in S. vulgaris (Kim et al., 2008).
Disk florets of S. vulgaris are radially symmetrical (actinomorphic) in contrast to ray florets, which are bilaterally symmetrical and dorsoventrally asymmetrical (zygomorphic). Using degenerate primers based on conserved regions of the CYCLOIDEA gene in Antirrhinum major, which encodes a putative transcription factor known to have a major controlling effect on dorsoventral asymmetry in Antirrhinum flowers (Luo et al., 1996), Kim et al. (2008) isolated two CYCLOIDEA-like genes (RAY1 and RAY2) in S. vulgaris that were expressed only in the outer floret primordia of capitula produced by radiate and nonradiate plants. The two variants were shown to be homozygous for different alleles of these two genes, although nucleotide substitutions within the coding regions of both genes could not account for functional differences that might give rise to a difference in floret shape. Rather, nucleotide substitutions in untranscribed flanking regions were thought to be responsible for causing this difference. Linkage analysis showed that RAY1 and RAY2 were tightly linked and that alleles of both genes cosegregated with capitulum type (radiate vs nonradiate) in a very large F2 family of a cross between a radiate and a nonradiate plant. Confirmation that both genes control ray floret development in S. vulgaris was obtained from transformation studies in which the alleles of both genes from the nonradiate variant (NN) were independently overexpressed in plants of the radiate variant. Overexpression of the N allele of RAY1 in radiate plants resulted in the production of either shorter ray florets than normal or disk florets in place of ray florets. Overexpression of the N allele of RAY2 in radiate plants resulted in the production of ‘tubular ray’ florets, that is, florets with ventralized petals. It was concluded that the locus in S. vulgaris which Trow (1912) showed controls presence of ray florets in capitula comprises two tightly linked genes, RAY1 and RAY2.
A survey of sequence variation across both loci among radiate and nonradiate plants of S. vulgaris revealed that all sequences were identical among radiate plants (the R haplotype), whereas two variants (the N and N1 haplotypes) were present among nonradiate plants. When the survey of sequence variation was extended to a small sample of S. squalidus plants, two haplotypes (R and R1) were recorded, with the R haplotype identical in sequence to that present in radiate S. vulgaris. Kim et al. (2008) concluded that ‘These results provide molecular proof that the radiate form of S. vulgaris arose through hybridization with S. squalidus plants and show that the R-haplotype was introgressed through this process.’
Because S. vulgaris is believed to be an allotetraploid (Ashton & Abbott, 1992; Chapman, 2004), both RAY1 and RAY2 loci are expected to be duplicated in this species, with the original homeologues inherited from two divergent diploid parent species. In their analysis, Kim et al. (2008) focused only on copies of RAY1 and RAY2 in S. vulgaris that were orthologous to RAY1 and RAY2 in S. squalidus. Sequence information for the other homeologous copies of RAY1 and RAY2 in S. vulgaris was not provided and no analysis was conducted on relationships between the copies of either gene in this species. In the following section we present a survey of sequence variation for RAY2 in S. vulgaris, S. squalidus and its progenitors S. aethnensis and S. chrysanthemifolius, to determine if both homeologues of this gene are present in S. vulgaris, and, if so, how they are related to each other and to sequences found in the diploid species.
A further point of interest concerns the frequency of the R haplotype in S. squalidus. If the R haplotype were present in this species at very low frequency, then this might indicate that radiate S. vulgaris, which is fixed for the R haplotype, originated only once following hybridization with S. squalidus. If, however, the R haplotype were common in S. squalidus, we could not rule out that radiate S. vulgaris originated on more than one occasion following hybridization with S. squalidus. To examine these alternatives, we surveyed allele frequencies at the RAY2 locus in populations of S. squalidus, and also across the hybrid zone between its progenitors, S. aethnensis and S. chrysanthemifolius, on Mount Etna (Brennan et al., 2009). The results of these surveys are also reported here.
RAY2 sequence variation
Initially, PCRs were carried out using primers designed from the RAY2 sequences reported by Kim et al. (2008). Each PCR contained 10 ng of template DNA, 30 mM tricine pH 8.4-KOH, 50 mM KCl, 2 mM MgCl2, 100 μM each dNTP, 0.1 μM each primer (RAY2f – CAAGGAATCAAGAAAAACTCTTCG; RAY2r – TTAGTCCTTTCTCTAGCTCTTGCTC), and one unit of Taq DNA polymerase. Cycling conditions followed a ‘touchdown’ protocol as follows: initial denaturation at 95°C for 3 min; 10 cycles of (30 s at 94°C, 30 s at 65°C (annealing temperature was reduced by 1°C per cycle), and 45 s at 72°C); 30 cycles of (30 s at 94°C, 30 s at 55°C, and 45 s at 72°C); and a final extension time of 20 min at 72°C. PCR amplification and sequencing were carried out on four individuals of each variety of S. vulgaris (radiate and nonradiate), six of S. squalidus and six of each of S. aethnensis and S. chrysanthemifolius (Table 1). Sequencing revealed that only one homeologue was amplified from S. vulgaris and thus a second primer (RAY2f2 – CCTTTTTCACATCTTRCTTCATC), designed from RAY2 sequences from a range of Senecio species (Chapman, 2004), was used with primer RAY2r. This primer pair amplified both homeologues of RAY2 from S. vulgaris.
|S. vulgaris var. vulgaris||Edinburgh (2)||na||na||na||na||na|
|S. vulgaris var. hibernicus||Bradford||4|
|Millers Dale, Derbyshire||4|
|S. aethnensis2||ET3 [2287 m] (3)||1||2||8|
|PRO2 [2061 m] (3)||1||2||4||4|
|Hybrids||SAP0 [1915 m]||1||5||5|
|SAP2 [1613 m]||7||3||1|
|SAP4 [1364 m]||2||4||3|
|MON1 [1045 m]||10||2|
|S. chrysanthemifolius||RAN1 [755 m] (3)||4||6|
|NIC1 [755 m] (3)||5||3||2|
Before sequencing, PCR products from S. vulgaris were cloned into pGEM-T vectors following the manufacturer’s recommendations (Promega, Madison, WI, USA) and six clones were sequenced per individual. Similarly, if the sequencing electropherograms of RAY2 from the diploid species indicated heterozygosity for indels, PCR products were cloned and four clones sequenced per individual. Sequence alignments were constructed in Genedoc (Nicholas & Nicholas, 1997) and sequence relationships resolved using PAUP* (Swofford, 2003) and PHYML (Guindon & Gascuel, 2003).
Ten different RAY2 sequences were detected among the PCR products analysed. These were resolved into two major clades by maximum likelihood (Fig. 2) and neighbour-joining analyses (data not shown). Clade 1 comprised three sequences that were present in both radiate and nonradiate S. vulgaris, but were absent from S. squalidus and its diploid progenitors, S. chrysanthemifolius and S. aethnensis. We conclude that these sequences are variants of the second homeologue of RAY2 in S. vulgaris, which was not analysed or described by Kim et al. (2008). From here onwards we refer to this gene as RAY2b and denote the other homeologue described previously by Kim et al. (2008) as RAY2a. It would be of interest to screen other diploid species of Mediterranean Senecio for the presence of a gene orthologous to RAY2b to identify, if possible, the species (or relative of it) that acted as the other parent of allotetraploid S. vulgaris. Because identical sequences of RAY2b were shared between radiate and nonradiate plants of S. vulgaris, this gene does not appear to have an effect on controlling presence/absence of ray florets in this species.
The second clade (clade 2) contained: (i) a subclade comprising the RAY2a-R sequence previously recorded in both radiate S. vulgaris and S. squalidus; (ii) another subclade comprising two related RAY2a-N sequences that are present in nonradiate S. vulgaris and which are allelic to RAY2a-R (Kim et al., 2008); (iii) a third subclade containing sequences most closely related to the RAY2a-R1 sequence previously isolated from S. squalidus by Kim et al. (2008); and (iv) a fourth subclade comprising two variants of a sequence denoted as RAY2a-R2. Relationships between these subclades lack strong support, although it appears that the RAY2a-R allele is most distantly related to the other RAY2a alleles present in clade 2. Thus, somewhat unexpectedly, the RAY2a-R1 and -R2 alleles seem to be more closely related to the RAY2a-N alleles than to the RAY2a-R allele.
Average sequence divergence within the RAY2a and RAY2b clades was 0.017 and 0.002, respectively (uncorrected P-distance calculated in PAUP*), whereas between the two clades, sequence divergence averaged 0.025. Greater divergence within RAY2a than RAY2b is expected because RAY2a sequences were resolved from several species instead of just S. vulgaris. A formal test for selection was conducted using fitmodel (Guindon et al., 2004) which tests for variation in dN/dS ratio (the ratio of nonsynonymous to synonymous substitution). Comparing the null (M0) model, which assumes the same dN/dS ratio across sequences, with the M3 model, which allows for variation in selective constraint, revealed marginal significance for heterogeneity in selective constraint across sites (P = 0.09). However, a test for positive selection (i.e. comparing the M2a model (dN/dS < 1, dN/dS = 1 or dN/dS > 1) with the null M1a model (dN/dS < 1 or dN/dS = 1)) revealed no evidence for this form of selection as a cause of sequence divergence (P = 0.15).
In the material sequenced, the RAY2a-R sequence was present among individuals of S. aethnensis and S. chrysanthemifolius along with the RAY2a-R1 and -R2 sequences; however, neither RAY2a-R2 sequence was recorded in the eight individuals of S. squalidus surveyed. The 10 different RAY2a and RAY2b sequences identified have been lodged in GenBank with reference numbers as follows: GU065705, (RAY2b-A); GU065706, (RAY2b-B); GU065707, (RAY2b-C); GU065708, (RAY2a-R); GU065709, (RAY2a-Na); GU065710, (RAY2a-N); GU065711, (RAY2a-R1); GU065712, (RAY2a-R1a); GU065713, (RAY2a-R2a); GU065714, (RAY2a-R2).
Distribution of RAY2a allelic variation in S. squalidus and its diploid progenitors
After identifying restriction sites distinguishing the RAY2a-R, -R1 and -R2 alleles from each other (Table 2), PCR-restriction fragment length polymorphism (PCR-RFLP) was used to survey the frequencies of these alleles among 24 individuals of radiate S. vulgaris sampled from six different UK populations, 48 S. squalidus individuals sampled from nine populations across the species range in the UK, and 85 individuals sampled from across the hybrid zone on Mount Etna (Table 1). The two variant sequences of RAY2a-R1 (i.e. -R1 and -R1a) could not be distinguished by PCR-RFLP, nor could variants of RAY2a-R2 (i.e. -R2 and -R2a), and therefore for each of these genes each pair of sequences was treated as the same allele. The primers RAY2f and RAY2r were employed, and PCR conditions for this analysis were the same as those described in the previous section. Eight microlitres of the resulting PCR were then mixed with a 7 μl cocktail containing 2 U of HinfI restriction enzyme, 1.5 μl enzyme buffer and ddH2O. Reactions were incubated at 37°C for 3 h, followed by enzyme inactivation at 80°C for 15 min. RFLPs were resolved on 1.5% agarose gels stained with ethidium bromide.
|Allele||Fragment sizes (bp)|
|R||427 + 421 + 5|
|R1||427 + 334 + 90 + 5|
|R2||791 + 37 + 5|
As expected, all 24 individuals of radiate S. vulgaris were homozygous for the RAY2a-R allele. In S. squalidus, this allele was present at intermediate frequency (48.96%) along with the RAY2a-R1 allele (51.04%), and all nine populations examined were polymorphic for both alleles (Fig. 3) and contained heterozygotes (Table 1). The frequencies of the two alleles appeared to vary among populations of S. squalidus, but low sample sizes meant that the significance of difference could not be tested reliably. Because the RAY2a-R allele is present in S. squalidus at intermediate frequency and was recorded in all populations examined, it is feasible that there may have been several independent origins of radiate S. vulgaris involving the transfer from S. squalidus of the same allele at the RAY2a locus. Thus, based on the present results, we are not able to distinguish between the possibilities of a single origin or multiple origins of radiate groundsel following hybridization between S. squalidus and nonradiate S. vulgaris. More extensive analysis employing many additional genetic markers might help to distinguish between these possibilities in the future.
Across the hybrid zone on Mount Etna, populations NIC1 and RAN1 comprised plants that were morphologically classified as S. chrysanthemifolius. Samples MON1, SAP4, SAP2 and SAP0 were from populations located at intermediate altitudes and comprised plants of intermediate hybrid morphology, while samples PRO2 and ET1 were from populations at higher altitudes and comprised plants that, morphologically, appeared to be S. aethnensis partly introgressed with genetic material from S. chrysanthemifolius. Plants resembling pure S. aethnensis are not normally found below 2400 m on Mount Etna (James & Abbott, 2005). Allele frequencies showed a clinal association with altitude across the hybrid zone, with RAY2a-R1 being common at lower altitudes and RAY2a-R2 at higher altitudes (Fig. 4). Whereas all populations examined were polymorphic for the RAY2a-R1 allele (overall frequency = 53.53%) and -R2 allele (overall frequency = 44.12%), the RAY2a-R allele was present only in samples from the highest altitudes and at very low overall frequency (2.35%, Fig. 4).
A surprising result to emerge from the analysis, therefore, is that the RAY2a-R allele which has been introgressed into S. vulgaris from S. squalidus, although present at intermediate frequency in S. squalidus, is very rare in hybrid material between S. chrysanthemifolius and S. aethnensis on Mount Etna that gave rise to S. squalidus. This allele was found only at higher altitudes in the hybrid zone and is therefore likely to be descended from S. aethnensis rather than S. chrysanthemifolius. It is possible that chance alone has been responsible for this rare allele in Mount Etna material becoming relatively common in S. squalidus. A genetic bottleneck experienced during the introduction of hybrid material from Mount Etna to the UK, and the subsequent origin and establishment of S. squalidus, could have resulted in the exclusion of the RAY2a-R2 allele from S. squalidus.
Although Stebbins (1971) emphasized almost 40 yr ago the potential importance of introgression across ploidy barriers to the subsequent evolution of polyploid species, there are few known examples of this phenomenon occurring in the wild leading to significant alterations to morphology and ecology of a polyploid. The only example of this process occurring between a wild diploid and tetraploid species, where the introgressed genes affecting fitness have been isolated and characterized, involves introgression between the diploid species S. squalidus and the tetraploid S. vulgaris. In this case, introgression of two tightly linked genes from S. squalidus to S. vulgaris resulted in the origin of a radiate variant of S. vulgaris that is more attractive to pollinators (Abbott & Irwin, 1988) and exhibits a higher outcrossing rate than the nonintrogressed nonradiate variant of S. vulgaris (Marshall & Abbott, 1982). New data presented here show that the allele of one of the two genes (RAY2a-R) introgressed into S. vulgaris and known to affect ray floret development in flower heads is present at intermediate frequency in the British population of S. squalidus, but surprisingly is very rare in hybrid material on Mount Etna from which this diploid hybrid species is derived. Our survey of RAY2 sequence variation in both radiate and nonradiate variants of S. vulgaris also recovered a homeologue of the RAY2 gene, denoted as RAY2b to distinguish it from the other RAY2a homeologue. The discovery of these two divergent homeologues provides further confirmation that S. vulgaris is allotetraploid, although the diploid species that donated the RAY2b copy during the origin of S. vulgaris remains unknown.
Part of the work was conducted while MAC was in receipt of a PhD studentship from NERC and while RJA was supported by BBSRC grant G10929. We are grateful to David Forbes for technical assistance and to Adrian Brennan for providing seed collected from plants growing on Mount Etna. We thank Enrico Coen, Pilar Cubas, Minsung Kim and Amanda Gillies for their contributions and help in the past.