Introduction

One effective way to assure genetic variability in hermaphrodite plants is self-incompatibility (SI), that consists of recognition of self- or self-related pollen by cells of the pistil, followed by rejection of the incompatible pollen through aborted development, either immediately after pollination on the stigma surface, or at a subsequent stage during pollen tube growth into the stigma or style (de Nettancourt, 1977). Recognition is a ‘lock-and-key’ system, and each ‘lock-and-key’ is defined as a specificity (Charlesworth et al., 2005). In gametophytic SI (GSI), pollen specificity is determined by the S-locus genotype of the haploid gametophyte. Therefore, for a given specificity, the rate of successful fertilization is inversely related to the specificity frequency in the population (frequency-dependent selection). The number and frequency of specificities is expected to follow an equilibrium between selection, mutation and drift. In diploid species, specificities are expected to be maintained at equal frequencies—isoplethy, if they are selectively equivalent, and if the population size is infinite (Wright, 1939; see review by Castric and Vekemans, 2004).

Castric and Vekemans (2004) reviewed the empirical studies in natural populations of plant species with GSI (twelve diploid species analyzed, five Solanaceae, three Rosaceae, and one Onagraceae, Papaveraceae and Polemoniaceae species), and found that the expectation of equal frequencies in populations at equilibrium holds in general. Nevertheless, the use of small samples in most studies may be responsible for the failure to reject isoplethy (Stoeckel et al., 2008). Indeed, that the specificity frequencies appear to be stochastically equal in populations is surprising (Lawrence, 2000). The time to steady state after specificity frequencies have been perturbed can take many generations, and this time is greater for populations containing a large number of specificities in comparison with those with a small number of specificities (Brooks et al., 1996, 1997a, 1997b). Unequal frequencies for different specificities have been reported for all Rosaceae species studied so far (Crataegus monogyna (Maloideae), Prunus lannesiana and P. avium (Prunoideae)) but Sorbus aucuparia (Maloideae) (Kato and Mukai, 2004; de Cuyper et al., 2005; Schueler et al., 2006; Raspé and Kohn, 2002, 2007). Two general alternative hypotheses to account for the observed unequal specificity frequencies have been proposed by these authors: some specificities can be subject to natural selection indirectly by close linkage with other genes, or founder effects. Nevertheless, as recently suggested by Stoeckel et al. (2008), the apparent rejection of isoplethy may be due to the noninclusion of genetic drift in the commonly used statistical tests.

There is no theoretical expectation regarding specificity frequencies in polyploid self-incompatible species. Although, polyploid Solanaceae and Scrophulariaceae species usually display a self-compatible phenotype (de Nettancourt, 1977), in Prunus, allopolyploid species can be self-incompatible (Hauck et al., 2006; Nunes et al., 2006). In this genus, self-compatibility is observed only when a minimum of two nonfunctional S-haplotypes are present in the same individual (Hauck et al., 2006). The issue of isoplethy in natural populations of polyploid self-incompatible species was never addressed from the empirical point of view.

P. spinosa is a nondomesticated allopolyploid species, that is self-incompatible (Salesses, 1970; Nunes et al., 2006). This species is thought to be the result of crossbreeding of diploid species P. cerasifera and P. microcarpa (Eryomine, 1990; Salesses and Bonnet, 1993; Reynders-Aloisi and Grellet, 1994), but varying ploidy levels are described (2n varing from 16 to 64; Baiashvili, 1980; Castroviejo- Bolivar et al., 1986).

Here, we study the SFB gene, that is the pollen component of GSI (Entani et al., 2003; Ushijima et al., 2003; Yamane et al., 2003; Romero et al., 2004; Nunes et al., 2006). Using a sample of 26 P. spinosa individuals from a natural population we find 36 SFB alleles. Those 36 alleles are predicted to correspond to 32 different specificities. Each putative specificity differs from all others at amino-acid sites identified as being positively selected. Those sites are likely responsible for determining specificity differences (Nunes et al., 2006). The frequency distribution of putative specificities is determined. This work differs from that of Nunes et al. (2006), that used the same population, as these authors report the frequencies of 19 likely different specificities using 11 individuals only. Moreover, no effort was made by these authors to systematically search for all specificities present in the population, as done here. Isoplethy is not observed in the P. spinosa population studied.

Materials and methods

Plant material and DNA extraction

P. spinosa is a common shrub in all Portugal (Franco, 1984). Leaves from individuals of one wild Portuguese P. spinosa population located in Rabal-Bragança (41°52′0N, 6°45′0W, assigned as B) were collected individually and frozen. A total of 26 individuals were sampled along the Sabor river margins (the river is depicted by a continuous black line in Figure 1).

Figure 1
figure 1

Distribution of the individuals studied of the wild P. spinosa Rabal-Bragança population. The zigzag line represents the margin of the Sabor river where all individuals were sampled. For each individual, SFB alleles are presented. ∨—individuals having the same SFB alleles.

Full size image

In P. spinosa populations, individuals with varying ploidy levels have been described (Baiashvili, 1980). Directly determining ploidy levels in a large number of individuals, in large shrubs such as P. spinosa, is difficult, as chromosome counts are usually performed from the root tip meristems (Salesses, 1970). Moreover, in natural populations, chromosome numbers are not always a multiple of the chromosome base number (Baiashvili, 1980). As it is not known in which chromosome the S-locus region is located, determining the chromosome number does not necessarily indicate how many S-alleles are present in a given individual. Different approaches aimed at fully characterizing the allele set present in the 26 individuals from the Bragança population were thus used (see below). One of the approaches involved, looking for the presence of SFB alleles known to be present in a small sample (N=7) from the unrelated Santo António- Pêso da Régua population, here characterized (41°10′0N, 7°46′60W, assigned as Re; see Figure 2). As described below, three specificities were found in the Bragança population, only when specific primers based on sequences obtained from Re population were used.

Figure 2
figure 2

Maximum likelihood phylogenetic tree of P. spinosa SFB sequences. Individuals that present identical sequences are indicated in the brackets. B codes for Bragança and Re for Pêso da Régua populations, respectively. Using Modeltest (Posada and Crandall, 1998) the model that best describes the P. spinosa SFB data set is determined to be the TVM+G model, with nucleotide frequencies A=0.27940; C=0.19720; G=0.20950; T=0.31390, the substitution rate matrix A↔C=1.3819; A↔G=3.1999; A↔T=0.9383; C↔G=1.8718; C↔T=3.1999 and G↔T=1.0000, a proportion of invariable sites of 0.0803 and a γ-distribution for the rates at variable sites with shape parameter 0.9095. A molecular clock was not enforced.

Full size image

Genomic DNA was extracted from leaves of individual plants using the method of Ingram et al. (1997).

SFB PCR amplifications

Primers 62F and 1010R (Supplementary Table 1; Nunes et al., 2006) were used to obtain an amplification product of 930 bp. As in the coding region of the SFB gene there are no introns, the different alleles of an individual, amplified using this general set of primers, cannot be distinguished on the basis of length differences on an agarose gel. This is in contrast with what is usually done when studying the S-RNase gene, where significant intron size variation is often observed between alleles (Vieira et al., 2003). Thus, the amplification product was cloned using the TA cloning kit (Invitrogen, Carlsbad, CA, USA). The DdeI restriction pattern was analyzed on average for 60–80 colonies. For each restriction pattern, three colonies of the same type were sequenced to correct for possible nucleotide miss-incorporations that may have occurred during the PCR reaction. Sequencing was performed by STABVIDA (Lisbon, Portugal), using BigDye cycle-sequencing kit (PerkinElmer), and specific primers or the primers for the M13 forward and reverse priming sites of the pCR2.1 vector.

Due to chance, SFB alleles could have been missed during the screening of clones. Therefore, specific primers were designed for each putative SFB specificity (including those found in the Re population; Supplementary Table 1) except for SFB3−1/SFB3−2. The latter two putative specificities were distinguished using primers 1624F (TCTAACCTTCTTCTTTGA) and 2108R (GGGGGTTTTGTTTTTGTG) that amplify the S-RNase and SFB intergenic region, and digestion with Sau3AI. For the 26 individuals of the Bragança population, PCR amplifications were performed using genomic DNA, and each primer combination separately. To confirm the specificity of the primer pair, amplification products not expected in a given individual according to the above-described cloning experiment, were cloned and sequenced. We also used all possible combinations of each specific primer and the appropriate general primer (either the 62F or the 1010R primer; Supplementary Table 1), to complete the SFB characterization of the individuals of the Bragança population. Unexpected amplification products, according to the characterization thus far performed were cloned and sequenced.

Analyses of sequences

The DNA sequences were deposited in GenBank (accession numbers EU760896–EU760574 for sequences obtained with 62F and 1010R; for SFB sequences using specific primers, EU760575–EU760619). Here, we use two data sets, namely, P. spinosa SFB alleles only, and all Prunus SFB alleles available. The nucleotide sequences were aligned using ClustalX v. 1.64b (Thompson et al. 1997), and minor manual adjustments were performed using Proseq version 2.43 (http://helios.bto.ed.ac.uk/evolgen/filatov/proseq.html). Analyses of DNA polymorphism were performed using DnaSP 4.1 (Rozas et al. 2003) and ProSeq version 2.43 software. Maximum likelihood trees were obtained using PAUP (Swofford, 2002), after running Modeltest (Posada and Crandall, 1998) to find the simplest model of nucleotide sequence evolution that best fits the data (using the Akaike information criterion (AIC)).

Nunes et al. (2006) used a set of 37 Prunus SFB sequences to identify amino-acid sites under positive selection. Nevertheless, a set of 70 Prunus SFB sequences (including those here reported) is now available in GenBank. Therefore, in principle, there is more power to identify amino-acid sites under positive selection. The analyses performed by Nunes et al. (2006) were thus repeated using the larger data set now available (Supplementary Table 2). Such information is needed to identify all putative P. spinosa specificities (sets of amino-acid sequences that differ from all other amino-acid sequences at least in one site identified as positively selected). The codeml software, as implemented in PAML 3.13 (Yang, 1997), and the method of Wilson and McVean (2006), as implemented in the omegaMap v 0.5 software (www.danielwilson.me.uk), that uses a population genetics approximation to the coalescent with recombination, were used. The model that best describes this data set is TVM+I+G model, with nucleotide frequencies A=0.2842; C=0.2215; G=0.2047 and T=0.2896, the substitution rate matrix A↔C=1.1161; A↔G=2.8711; A↔T=0.8283; C↔G=1.4636; C↔T=2.8711 and G↔T=1.0000, a proportion of invariable sites of 0.0803 and a γ-distribution for the rates at variable sites with shape parameter 0.9095. A molecular clock was not enforced. When using the Yang (1997) method, of all models tested (M0, M1, M2, M3, M7 and M8) the simplest model that best fits the data is the discrete model with 3 categories (M3). In the method of Wilson and McVean (2006) two 250 000 iterations were performed. A burn-in of 25 000 was also used and the results checked for convergence. All codons were assumed to be at equal frequencies. Ten random sequence orders were used. The parameters to be estimated are the selection parameter (ω=Ka/Ks), the population recombination rate (ρ), the rate of synonymous transversion (μ), the transition–transversion ratio (κ) and the insertion/deletion rate (ϕ) (Wilson and McVean, 2006). The first two parameters may vary along the sequence. A block of 30 codons (approximately 10% of the sequence size) is used when estimating both ω and ρ. For the Wilson and McVean (2006) method both an objective and a subjective approach to prior specification were used. Inverse distributions were used as priors for ω and ρ, and improper inverse distributions were used for the other parameters (μ, κ and ϕ). The bounds for ω are 0.01–1000 and for ρ are 0.00000001–1000. Thus the posterior density outside this range should be about 0. In the second approach, exponential distributions were used for all parameters (μ=1, κ=3, ω=1, ρ=0.0001, ϕ=0.2). The values chosen reflect the prior belief that there is little recombination and that only a few sites along the sequence are positively selected.

To test whether all putative specificities are present at equal frequency in the Bragança population, we used a resampling method, excluding individuals suspected to be the result of vegetative propagation. A total of 50 000 samples were generated with the same number of individuals and ploidy levels as observed in the Bragança sample, under the assumption that the population had a given total number of specificities (varying from 32 to 42) and that all specificities were present at the same frequency in the population. When 33 specificities are assumed, about 33% of the times, the simulated samples show the same number of putative specificities as in the Bragança population sample. If a different number of specificities is assumed, 32 specificities are not observed so often in a sample with the same configuration as that from Bragança population. We then took the 16 387 simulations that, under the assumption of 33 specificities, showed 32 specificities in the simulated sample. On the basis of these simulated samples we calculated the average number of times the commonest, the second commonest and so forth, specificity occurs. The fit of the expected and observed distribution was tested using a χ2-test after merging the appropriate classes in such a way that for each cell the expected number of events is higher than five.

Results

P. spinosa SFB sequences

Three to six different amplification products per individual (16 Bragança and seven Pêso da Régua) were revealed after the restriction digestion of 60–80 colonies resulting from the cloning of the 930 bp amplification product obtained from genomic DNA (with primers 62F and 1010R). Using tblastx, it can be seen that the 79 sequences obtained share more than 80% of amino-acid identity with described SFB alleles from Prunus species.

These steps had been already performed for other 11 individuals from the Bragança population analyzed by Nunes et al. (2006), and 10 were here used. Therefore 106 nucleotide sequences (the 79 here reported plus 27 reported by Nunes et al., 2006), are available for P. spinosa SFB gene. The 106 nucleotide sequences define 37 different nucleotide sequences. The phylogenetic relationship of these 37 nucleotide sequences is presented in Figure 2.

There are five phylogenetic groups that show two closely related sequences each (Figure 2). To determine whether they represent the same specificity, controlled crosses must be performed. This is, however, not an easy task, as this species is polyploid, thus multiple controlled crosses and analyses of large numbers of progeny are required. Nevertheless, each of the five pairwise comparisons differs in one (SFB3−1 and SFB3−2, SFB4−1–SFB4−2 and SFB20−1–SFB20−2) or two (SFB7−1–SFB7−2 and SFB14−1–SFB14−2) amino-acid sites. In the pairwise comparisons, none of the differing amino-acid sites is identified as positively selected (see below). In the absence of data on controlled crosses, we thus assume that the amino-acid differences found in the SFB4−1–SFB4−2, SFB20−1–SFB20−2, SFB7−1–SFB7−2 and SFB14−1–SFB14−2 comparisons represent amino-acid polymorphism within specificity classes. In agreement with this assumption, for SFB7−1–SFB7−2 the corresponding S-RNase gene is identical at the amino-acid level in the region analyzed (DQ677586 for SFB7−1 and EU833958 for SFB7−2). For SFB3−1 and SFB3−2 there is, however, evidence to suggest that they do represent two different specificities, as both alleles are amplified from genomic DNA of the same individual (Re1, Re2 and B21; Figure 2). Furthermore, the corresponding S-RNase sequences show two amino-acid differences (Nunes et al., 2006), and one of these has been identified as a positively selected amino-acid site (Vieira et al., 2007). Although, it is unclear, how many amino-acid changes at sites under diversifying selection are required to create a new specificity, we assume that they correspond to two different specificities. The remaining 27 phylogenetic groups (Figure 2) likely represent different specificities, as the number of amino-acid differences between these phylogenetic groups varies from 6 (SFB17 and SFB26, Ka=0.011) to 85 (SFB13 and SFB22, Ka=0.1557). In all cases, at least one of the amino-acid differences is a site identified as under positive selection (see below). P. spinosa thus likely harbors at least 33 different specificities.

To fully characterize the 26 Bragança individuals shown in Figure 1, specific primers were designed for 32 out of 33 putative SFB specificities (as described in ‘Materials and methods’). PCR amplifications using each of the 32 primer sets were then performed using genomic DNA. All unexpected amplification products were cloned and sequenced (see ‘Materials and methods’). The full characterization of the SFB alleles in Bragança is presented in Figure 1. Only one primer set did not support amplification in any of the 26 individuals from the Bragança population. Therefore, the P. spinosa Bragança population likely harbors at least 32 different specificities.

None of the 32 primer sets tested supports amplification in all individuals, as expected for GSI alleles. In the Bragança population, individuals B2–B3, B15–B28, B17–B18–B26, B19–B20–B24 have the same set of SFB alleles.

The number of SFB alleles characterized per individual using the 32 sets of primers varies from three to six (Figure 1). P. spinosa is described as being an allotetraploid species thus four S-alleles were expected per individual (Halliday and Beadle, 1983). Variations in ploidy levels are however described in this species (2n can vary from 16, 24, 32, 40, 43, 44, 48, 53, 56, 59, 64; n=16; Flora Iberica, http://www.rjb.csic.es/floraiberica/PHP/cientificos.php) even in the same population (Baiashvili, 1980).

The combination of 46 specific SFB primers and the appropriate general primer (either 72F or 1010R) revealed no further alleles in the individuals studied (data not shown). We take this as suggestive evidence that the inferred number of specificities per individual is correct.

Positively selected sites in the P. spinosa SFB sequences

The 70 Prunus SFB sequences have been used to identify positively selected sites (Supplementary Table 2), those likely responsible for defining specificity differences. Two different approaches were used, namely, a phylogenetic one (Yang, 1997) and a population genetics method that uses an approximation to the coalescent with recombination (Wilson and McVean, 2006). The two methods were used previously, but with 37 Prunus sequences only (Nunes et al., 2006). Therefore, it was unclear whether more sites under positive selection would be identified when using many more sequences. Thirty-one amino-acid sites were identified by at least one of the methods used with a posterior probability of selection higher than 95%, and these are shown in Table 1 (see Supplementary Figure 2). Twenty-two of these amino-acid sites were previously identified as being under positive selection by Nunes et al. (2006). It should be noted that we cannot be sure that all positively selected amino-acid sites have been identified. Indeed, the power to detect sites under positive selection may be low when sequences are highly diverged (Castric and Vekemans, 2007).

Table 1 Amino-acid sites under positive selection
Full size table

Both the phylogenetic and population genetics approach present potential problems that can affect the identification of sites under positive selection (see Vieira et al., 2007). Nevertheless, 18 out of 25 amino-acid sites without alignment gaps were identified by both methods as being positively selected with a posterior probability of selection higher than 50%. It is possible that the six amino-acid sites that are in class 5 (Phy>95% and Pop<50%; Table 1) and the one that is in class 6 (Pop>95% and Phy<50%; Table 1) are false positives. Sites in class 5 could be explained if the phylogenetic relationship between the sequences used was inaccurately derived, or if there is recombination meaning that a single tree cannot accurately describe the relationship of the sequences used (Wilson and McVean, 2006). The Bruen et al. (2006) Phi test (as implemented in SplitsTree4; Huson and Bryant, 2006), looks for an excess of homoplasies in a data set relative to those expected given the observed divergence. This test performs well in diverse situations such as exponential growth (star-like topologies) and patterns of substitution rate correlation (Bruen et al., 2006). The Phi test can also be efficiently used to detect recombination between closely, as well as distantly related samples, regardless of the suspected rate of recombination (Bruen et al., 2006). When this test is applied to the 70 SFB sequences here analyzed, a statistically highly significant excess is observed (P<0.00005). A similar result is obtained when a set of 88 Prunus S-RNases compiled by Vieira et al. (2007) is used (P<0.05). For Prunus S-RNases, other features compatible with recombination have been presented, although the evidence is still not unequivocal (Vieira et al., 2003; Nunes et al., 2006; Ortega et al., 2006). It is tempting to take such an excess of homoplasies as evidence for recombination. Nevertheless, in GSI, many old specificities are maintained by strong frequency-dependent selection. It remains to be determined how many times such an excess is observed when comparable simulated data (without recombination) is used, but this is beyond the scope of this article.

Specificity numbers and frequencies

As P. spinosa individuals are polyploid, the total number of specificities present in the population cannot be estimated as done by other authors (Castric and Vekemans, 2004). The 127 alleles sampled in the Bragança population revealed 32 different putative specificities (Figure 1). The other Prunus species exhaustively sampled (P. dulcis, P. armeniaca, P. salicina and P. avium; see the GenBank data compiled by Vieira et al., 2008a) revealed in between 21 and 31 specificities. Furthermore, in two Maloideae species, namely S. acuparia and C. monogyna, 32 and 17 alleles were found, respectively (Raspé and Kohn, 2002, 2007). It is thus likely that most P. spinosa specificities were sampled. Figure 3 shows for the Bragança population, the observed and expected distribution under the assumption of equal frequencies. The two distributions are significantly different (P<0.05). There is an excess of low and high frequency specificities in the Bragança sample (Figure 3). It should be noted, that there is no theoretical expectation regarding specificity frequencies in polyploid self-incompatible species. 

Figure 3
figure 3

Observed and expected distribution under the assumption of equal specificity frequencies for the Bragança population.

Full size image

Discussion

The thorough search performed in a large sample from the Bragança population revealed 32 putative specificities. Although P. spinosa is a polyploid species with varying chromosome numbers, the number of putative specificities found is similar to that found for diploid Prunus species (Kato and Mukai, 2004; de Cuyper et al., 2005; Schueler et al., 2006).

Only 12% of the P. spinosa putative specificities show amino-acid polymorphism, even when sequences from different populations are compared (data not shown). This finding is in agreement with that of Nunes et al. (2006), that suggested little amino-acid polymorphism within specificities. Such a result is expected, because of the small effective population size of each specificity class (Charlesworth et al., 2005). As pointed out by Nunes et al. (2006), it is possible that the amino-acid variation could have been inherited from P. spinosa's two diploid ancestors. In addition, the variation observed within specificities could have been obtained by gene flow from P. insititia that may hybridize with P. spinosa (Nunes et al., 2006).

In the P. spinosa Bragança population isoplethy is not observed. P. spinosa is a polyploid species, and no theoretical expectation regarding specificity frequencies in polyploid self-incompatible species is available. Nevertheless, as this is the first report on specificity frequencies distribution in polyploidy species, it seems useful to discuss possible deviations from population equilibrium.

Identical individuals were found in this sample. Considering the total number of alleles found in the population, and the ploidy level of the individuals, it seems unlikely that two identical individuals would be found in the sample by chance alone. As the maximum physical distance between the identical individuals is less than 11 m (Figure 1), vegetative propagation could be a likely reason for this observation. It should be noted that P. spinosa does not show apomixis (Guitian et al., 1992), and that although natural vegetative multiplication is rare in the genus Prunus, clones of P. spinosa can be maintained by tillering (OECD Environment, Health and Safety Publications, 2002). Therefore, individuals suspected to be the result of vegetative propagation were removed from the analyses related to specificity frequency distributions. It should, be noted, nevertheless, that old clonal individuals may have contributed to a putative initial deviation from expected equilibrium frequencies.

In the Bragança population there is an excess of low and high frequency specificities (Figure 3). This type of deviation is similar to that observed in wild cherry populations, where a significant departure from the isoplethic distribution is also observed (Stoeckel et al., 2008). These authors have shown that the observed allele frequency distribution is compatible with genetic drift and a model of subdivided populations with demes linked by moderate migration rate. The way we tested for the expectation of equal frequencies does not include the effect of drift, but this is also a caveat of other approaches commonly used for diploid species (see for instance Castric and Vekemans, 2004).

For polyploid species such as P. spinosa it is possible that not all S-allele combinations are equally present in pollen grains, as this would imply that every possible chromosome pairing is equally likely during meiosis. In the polyploid P. cerasus this, however, seems to be the case (Hauck et al., 2006). Moreover, we did not find any obvious association between pairs of alleles.

Furthermore, in Prunus, heteroallelic pollen retains its SI phenotype, and two nonfunctional S-haplotypes are needed to be present in the same individual for self-compatibility to arise (Hauck et al., 2006). Thus, in principle, nonfunctional haplotypes may reach significant frequencies in the population. At the SFB gene, no evidence has been found suggestive of nonfunctionality. So far only one nonfunctional haplotype (the S-RNase6-SFB6 haplotype that shows a stop codon in the S-RNase gene; data not shown) has been found. It should be noted that this haplotype seems to be at low frequency in the Bragança sample. Nunes et al. (2006) studied a sample of 10 randomly chosen individuals from Bragança population and found them to be self-incompatible.

The pollen contribution of a large tree may be very different from that of a young small tree. If for any reason, in a local population and for a given age cohort, the frequency of a given specificity increases beyond its predicted equilibrium frequency, then the local population will be out of the equilibrium for many years as large trees and shrubs live for many years. Nevertheless, Schueler et al. (2006) studied a large population of German diploid P. avium trees with different age cohorts and found that the frequency distribution did not change in the different age cohorts. This can be taken either as evidence that the population is at equilibrium or that, once disturbed, it takes many years for the frequencies to go back to equilibrium. The two main explanations (founder effects and selection at linked loci) that have been put forward for the observed unequal frequencies in diploid Rosaceae species may also, in principle apply. P. spinosa is a long-lived perennial species. Because of recent dispersal and establishment of secondary populations, local populations of long-lived species are not expected to be at evolutionary equilibrium (Raspé and Kohn, 2007).

S-alleles with different sets of linked recessive deleterious alleles cannot coexist in the same diploid individual (Uyenoyama, 1997). Therefore, isoplethy may not be observed if S-alleles are commonly linked with recessive deleterious alleles. This effect is expected to be more pronounced in species where recombination is suppressed in a large region around the S-locus, as thought for Maloideae (Sassa et al., 2007). Nevertheless, in Prunus the S-locus seems to encompass at most three genes (Entani et al., 2003; Ushijima et al., 2003; Vieira et al., 2008b).

More data are needed from natural populations of other polyploid self-incompatible species to establish the generality of the observation made for P. spinosa, regarding specificities frequencies distribution.