Self-incompatibility of Prunus tenella and evidence that reproductively isolated species of Prunus have different SFB alleles coupled with an identical S-RNase allele

Authors


For correspondence (fax +44 1732 849 067; e-mail nada.surbanovski@emr.ac.uk).

Summary

Many species of Prunus display an S-RNase-based gametophytic self-incompatibility (SI), controlled by a single highly polymorphic multigene complex termed the S-locus. This comprises tightly linked stylar- and pollen-expressed genes that determine the specificity of the SI response. We investigated SI of Prunus tenella, a wild species found in small, isolated populations on the Balkan peninsula, initially by pollination experiments and identifying stylar-expressed RNase alleles. Nine P. tenella S-RNase alleles (S1S9) were cloned; their sequence analysis showed a very high ratio of non-synonymous to synonymous nucleotide substitutions (Ka/Ks) and revealed that S-RNase alleles of P. tenella, unlike those of Prunus dulcis, show positive selection in all regions except the conserved regions and that between C2 and RHV. Remarkably, S8-RNase, was found to be identical to S1-RNase from Prunus avium, a species that does not interbreed with P. tenella and, except for just one amino acid, to S11 of P. dulcis. However, the corresponding introns and S-RNaseSFB intergenic regions showed considerable differences. Moreover, protein sequences of the pollen-expressed SFB alleles were not identical, harbouring 12 amino-acid replacements between those of P. tenella SFB8 and P. avium SFB1. Implications of this finding for hypotheses about the evolution of new S-specificities are discussed.

Introduction

Self-incompatibility (SI) prevents fertile hermaphrodite plants from producing zygotes after self-pollination. In several families, including Rosaceae (to which the economically important genus Prunus belongs) and Solanaceae, the SI system is gametophytic. The genetic control, attributed to the single multi-allelic S-locus, was first explained in Nicotiana of the Solanaceae (East and Mangelsdorf, 1925) and later demonstrated in many rosaceous species including Prunus avium (sweet cherry; Crane and Lawrence, 1929), Prunus dulcis (almond; Gagnard, 1954), Malus pumila (apple; Kobel et al., 1939) and Pyrus serotina (Japanese pear; Terami et al., 1946). The S-locus is considered to contain two complementary, genetically linked parts, encoding a stylar-specific product and a pollen-specific product.

The molecular basis of gametophytic SI has been studied most thoroughly in Solanaceae, and the achievements have been summarized and reviewed by Franklin-Tong and Franklin (2003); Kao and McCubbin (1996); McClure (2004); McCubbin and Kao (2000); Newbigin et al. (1993); Roalson and McCubbin (2003). Stylar S-proteins were identified as glycoproteins with ribonuclease activity (S-RNases) by McClure et al. (1989) that appear to be essential for the rejection of incompatible pollen (Lee et al., 1994). Solanaceous S-RNases have five short conserved regions (C1–C5) and two short hypervariable regions, named HVa and HVb, located between C2 and C3 and flanking the site of the intron. The S-RNase gene sequences are flanked by diverse highly repetitive sequences thought to play a role in preventing recombination within the S-locus (Coleman and Kao, 1992).

Incompatibility S-proteins have been shown also to have RNase activity in Rosaceae, and procedures have been developed, based on isoelectric focusing (IEF) of stylar protein extracts followed by activity staining (Bošković and Tobutt, 1996; Sassa et al., 1992) and PCR amplification with consensus primers to reveal intron length polymorphism (Broothaerts et al., 1995; Sonneveld et al., 2003; Sutherland et al., 2004; Tao et al., 1997), to detect allelic diversity. Ushijima et al. (1998) pointed out that rosaceous S-RNases have five short conserved regions (C1–C5) that are, with the exception of C4, homologous to conserved regions in solanaceous sequences. The short hypervariable region in rosaceous S-RNases, named RHV, is located between C2 and C3. Rosaceous S-RNase alleles have an intron, in the RHV region, showing considerable length polymorphism and Prunus S-RNases have an extra intron, less variable in length, positioned between the sequence regions encoding the signal peptide and the mature protein (Ma and Oliveira, 2001). The S-RNase sequences of Rosaceae are also flanked by diverse, highly repetitive sequences (Romero et al., 2004; Ushijima et al., 2003).

Recent work in Prunus has identified a gene specifically expressed in the pollen, adjacent to the S-RNase gene and revealing S-specific polymorphism, that is likely to represent the pollen component of the S-locus (Ushijima et al., 2003; Yamane et al., 2003) and is named S-haplotype-specific F-box (SFB). It has a conserved F-box motif at its 5′ end, and recent data show two variable (V1, V2) and two hypervariable (HVa, HVb) regions and an intron positioned in the 5′ untranslated region (Ikeda et al., 2004; Vaughan et al., 2006). Further evidence that the SFB genes are indeed involved in SI recognition in Prunus comes from haplotypes that confer self-compatibility when the SFB allele has been deleted or codes for a truncated protein (Sonneveld et al., 2005; Ushijima et al., 2004). Candidates for the pollen S gene were also identified in Scrophulariaceae and Solanaceae (Qiao et al., 2004; Šijačićet al., 2004).

Significant advances in understanding the molecular basis of SI, the extraordinarily high level of allelic polymorphism of the gene and the balancing selection that operates and counteracts genetic drift, make the S-locus particularly interesting to population and evolutionary genetic studies (Richman and Kohn, 1996). The S-RNase sequence analysis revealed that some alleles are more similar to alleles from different species than to other alleles from the same species, indicating that the S-alleles are likely to pre-date the divergence of species in Solanaceae (Ioerger et al., 1990). Comparison of rosaceous S-RNase sequences showed that they did not fall into species-specific classes, but formed a Prunus (almond, japanese apricot, japanese plum, sweet cherry and sour cherry) clade and a Maloideae (apple and japanese pear) clade (Igic and Kohn, 2001; Ma and Oliveira, 2002; Sonneveld, 2002; Ushijima et al., 1998). Even though the Prunus S-RNases often show more similarity between than within species, there appear to be no reported cases of two reproductively isolated species sharing the same full-length S-RNase allele. Phylogenetic relationships of Prunus SFB alleles have been studied by Ikeda et al. (2004) and by Vaughan et al. (2006).

The dwarf almond, Prunus tenella (Batsch.), is a wild species growing in small and isolated populations scattered throughout the Balkan Peninsula. The plants have a compact habit, narrow leaves and attractive, dark pink flowers. As well as reproducing sexually, P. tenella also propagates vegetatively by suckering. It is of potential interest for breeding drought-tolerant rootstocks resistant to extremes of temperature. The self-(in)compatibility status of P. tenella has not been established.

We first studied self-(in)compatibility in 18 accessions of P. tenella by pollination tests and by characterizing stylar-expressed RNase alleles using IEF, PCR and DNA sequencing, as well as identifying sequence regions where positive selection may operate. When we found one of the S-RNase alleles cloned corresponded to a protein sequence identical to that of an S-RNase allele previously sequenced from sweet cherry (P. avium), and different by just one amino acid from another from almond (P. dulcis), we investigated the similarity of the SFB sequences corresponding to these three alleles and of the intergenic regions separating the S-RNase and SFB alleles. Interestingly, the SFB sequences of these haplotypes were not identical.

Results

S-RNase alleles in Prunus tenella

When 18 P. tenella accessions (A1–A18) were screened with primers that reveal first- and second-intron length polymorphisms of the S-RNase gene, each accession showed two PCR products for the first intron and two PCR products for the second intron, indicating two putative S-RNase alleles per accession. Analysis of stylar protein extracts by IEF and staining for RNase activity revealed two functional S-RNase alleles for each accession.

Full correspondence was found when the PCR and IEF data for the 18 accessions were compared. Figure 1 shows six of the accessions containing all the S-RNase alleles identified in this study. Distinct alleles were identified as S-RNases that had different sizes of first- and second-intron PCR products and/or different positions on the IEF gel. In all, nine S-RNase alleles of P. tenella were thus characterized, S1S9, and the S-genotypes of the accessions were determined. Of the accessions sampled from Northern Serbia, four (A1, A2, A7, A8) shared the S4S9 genotype; four accessions (A4, A5, A6, A9) were S7S8; four (A12, A13, A15, A16) S6S8; two (A10, A11) S1S5; one (A14) S2S5; and one (A3) S8S9. The two accessions sampled from Southern Serbia (A17, A18) shared the genotype S2S3.

Figure 1.

 Stylar S-RNase alleles in six accessions of Prunus tenella.
(a) PCR products amplified with consensus primers for the first intron of the S-RNase gene separated by capillary electrophoresis.
(b) PCR products amplified with consensus primers for the second intron of the S-RNase gene separated on agarose gels.
(c) Stylar ribonucleases separated by isoelectric focusing and stained for ribonuclease activity. Deduced alleles are annotated. Alleles S4 and S9, were deduced by reference to genotype S8S9 (data not shown). Accessions and genotypes, from top to bottom or left to right, are as follows: A8 (S4S9), A9 (S7S8), A16 (S6S8), A11 (S1S5), A14 (S2S5) and A17 (S2S3).

Pollination tests

After the determination of S-genotypes, pollination tests were conducted on several accessions in order to investigate (in)compatibility relationships. In the self-pollinated pistils, of accessions A8 (S4S9) and A11 (S1S5), as well as in selfings of two accessions harbouring allele S8, A9 (S7S8) and A15 (S6S8), the pollen germinated well, small numbers of pollen tubes traversed the upper third of the style and their growth was arrested in the second third of the style (Figure 2a), so that no pollen tubes were observed in the last third or the base of the style (Figure 2b). All these accessions set fruit after open pollination. As accessions A8, A9, A11 and A15 demonstrated the SI reaction, the alleles S1, S4, S5, S6, S7, S8 and S9 do confer SI. Accessions A8 and A11 were reciprocally cross-pollinated: large numbers of pollen tubes passed through the transmitting tissue of the upper and second third of the style (Figure 2c), and pollen tubes were observed reaching the base of the style (Figure 2d). The two genotypes were therefore cross-compatible.

Figure 2.

 Fluorescence microscopy of pollen tube growth in styles of Prunus tenella.
(a, b) Self-pollination of accession A15 (S6S8); (c, d) cross-pollination of accession A8 (S4S9) with pollen from A11 (S1S5).
(a, c) are from the mid-section of the style (as indicated by the location of lateral hairs), showing arrested pollen tubes indicated by arrows (a) and pollen tubes growing uninterrupted (c).
(b, d) are from the base of the style, showing no pollen tube growth (b), and three pollen tubes reaching the base of the style (d). Note that the pollen tubes are thinner than the elements of the transmitting tissue, with callosic plugs giving a ladder-like appearance.

Cloning and sequence analysis of P. tenella S-RNase alleles

S-RNase alleles (S1S9) were successfully cloned and sequenced, and the sequences submitted to Genbank. The alignment of the deduced amino-acid sequences of the eight S-RNase alleles showed that those of P. tenella share features of other Prunus S-RNase alleles, having one hypervariable (RHV) and five conserved (C1–C5) regions as well as conserved cysteine and histidine residues (Figure 3).

Figure 3.

 Alignment of deduced amino acid sequences of nine S-RNase alleles of Prunus tenella.
Hypervariable and conserved regions (Ushijima et al., 1998) are underlined. Shading indicates variant residues. Conserved cysteines and histidines involved in RNase activity are annotated with • and o, respectively. Accession numbers of nucleotide sequences are as follows: DQ983373 (S1), DQ983374 (S2), DQ983375 (S3), DQ983363 (S4), DQ983364 (S5), DQ983365 (S6), DQ983366 (S7), DQ983367 (S8), DQ983370 (S9).

The graph produced from the ‘sliding window’ analysis of the ratio of Ka (non-synonymous substitutions per non-synonymous site) and Ks (synonymous substitution per synonymous site) along the sequence is given in Figure 4. An excess of non-synonymous over synonymous substitutions occurred in the regions between C1 and C2, RHV and C3, C3 and C4, and C4 and C5. The overall mean values for Ka, Ks and Ka/Ks were 0.145, 0.147 and 0.99, respectively.

Figure 4.

 The ratio of non-synonymous to synonymous nucleotide substitutions, Ka/Ks, along the sequence of nine S-RNase alleles cloned from Prunus tenella, based on a sliding windows analysis of 20 codons. The positions of rosaceous conserved regions C1–C5 and the RHV region (Ushijima et al., 1998) and the PS1–PS3 regions in Maloideae (Ishimizu et al., 1998) are indicated.

Amino-acid alignment revealed high levels of diversity among P. tenella S-RNase alleles, and pairwise comparisons showed similarity ranging from 68.3% between S3-RNase and S4-RNase to 83.9% between S4-RNase and S8-RNase. When the deduced amino-acid sequences of P. tenella alleles were compared with those of other Prunus species, the similarity at the amino-acid level to other Prunus alleles was higher, ranging from 71 up to 100%. S8-RNase P. tenella revealed 100% identity with S1-RNase P. avium (Sonneveld et al., 2001; Tao et al., 1999), from within the signal peptide region to the stop codon. In addition, the full sequences of S8-RNase P. tenella and S1-RNase P. avium shared 99% identity (one amino acid difference) with that of the recently sequenced S11-RNase of P. dulcis, S11 (Ortega et al., 2006; Figure 5). Furthermore, S1-RNase P. tenella was found to be 99% identical to S4-RNase Prunus armeniaca (Romero et al., 2004), having only one amino acid difference in the region from C1 to C5.

Figure 5.

 Alignment of genomic DNA sequences of S8-RNase Prunus tenella (Pt), S1-RNase Prunus avium (Pa) and S11-RNase Prunus dulcis (Pd).
The consensus amino-acid sequence is given below. Coding regions are shown in bold. Shading indicates variant nucleotides. (Note that intron sequence similarity between 668 and 981 bp is very low and the alignment should be considered arbitrary). The black-shaded residue H is replaced with N in S11-RNase of P. dulcis. The stop codon is indicated with an asterisk.

Alignment of the DNA-coding sequences of these three alleles revealed several positions with single nucleotide polymorphisms: six were observed between P. tenella and P. avium, seven between P. avium and P. dulcis, and eight between P. tenella and P. dulcis (Figure 5). When intron sequences of S8-RNase P. tenella, S11-RNase P. dulcis and S1-RNase P. avium were compared, a number of nucleotide substitutions and in/del events were apparent in both introns (Figure 5). In addition to smaller in/dels, a 246-bp sequence was present in the second intron of S1-RNase P. avium and lacking in S11-RNase P. dulcis and S8-RNase P. tenella.

Cloning and comparison of SFB alleles of S8P. tenella, and S11P. dulcis with S1P. avium

SFB genes of P. tenella were sequenced from two accessions (A9 and A16) with different S-genotypes both having S8-RNase (S7S8 and S6S8, respectively), and the sequence that the two accessions shared was determined to be the SFB8 sequence. The same approach was used for determining the sequence of P. dulcis SFB11 in cultivars Marcona (S11S12) and Bertina (S6S11). Comparison of the sequences determined as SFB8P. tenella and SFB11P. dulcis with other SFB alleles showed they are highly similar to SFB1 from P. avium.

SFB8P. tenella (accession number DQ983369) and SFB11P. dulcis (accession number EF061758) are very similar to SFB1P. avium (Vaughan et al., 2006), but the protein sequences of these three are not identical. The differences scattered throughout the sequence are indicated in Figure 6. Between SFB8P. tenella and SFB1P. avium there are 12 different amino acids out of 364, and seven of these changes are non-conservative, as indicated by Dayhoff et al. (1979); the replacement in the HVb region of M (methionine) with T (threonine) is noteworthy. Between SFB1P. avium and SFB11P. dulcis there are 11 different residues, of which six are non-conservative replacements. Between SFB8P. tenella and SFB11P. dulcis there are three different residues, of which one is a non-conservative replacement. The V1, V2 and HVa regions are identical in all three alleles.

Figure 6.

 Alignment of the DNA coding sequences of SFB8Prunus tenella (Pt), SFB1Prunus avium (Pa) and SFB11Prunus dulcis (Pd).
The consensus amino-acid sequence is given below. Hypervariable (HVa, HVb) and variable (V1, V2) regions and the F-box (Ikeda et al., 2004) are underlined. Grey shading indicates variant nucleotides. Black shading indicates polymorphisms in the amino acid sequence; and the differences from consensus residues are as follows: A, G, V, S, V, A, H, A, V, G, E, N, T.

Comparison of the SFB alleles at the DNA level revealed 17 single nucleotide polymorphisms between P. tenella and P. avium, 18 between P. dulcis and P. avium, and five between P. tenella and P. dulcis (Figure 6).

Intergenic region between the S-RNase and SFB genes in S8P. tenella, S1P. avium and S11P. dulcis

The intergenic regions of the three haplotypes were successfully amplified using a forward primer from the signal-peptide region of the S-RNase and a reverse complement primer specific for the 3′ region of SFB1P. avium, suggesting that a gene highly similar to SFB1P. avium is present adjacent to the S-RNase in P. tenella and P. dulcis, and that the two genes are in opposite transcriptional orientations. Amplified products were cloned and sequenced (Figure 7a). The intergenic region of S8P. tenella (approximately 1.1 kb) is considerably larger than that of the other two (approximately 0.4 kb) because of an in/del of 709 bp. The length of the intergenic region falls within the reported range of 380 bp to 30 kb (Romero et al., 2004).

Figure 7.

 The intergenic region between the S-RNase and SFB genes of S8Prunus tenella (Pt), S1Prunus avium (Pa), and S11Prunus dulcis (Pd).
(a) Alignment of DNA sequences. Grey shading indicates variant nucleotides.Accession numbers of sequences aligned are: DQ983368 (S8P. tenella), DQ983372 (S1P. avium), DQ983371 (S11P. dulcis).
(b) Schematic representation indicating the 709-bp insertion in P. tenella, and microsatellite and direct and inverted repeats.

The in/del detected in the intergenic region of P. tenella appears to be an insertion in P. tenella rather than a deletion in P. dulcis and P. avium. When the sequence was compared with database entries, no significant matches were found. The insertion contained two types of direct repeat (5′-TTTATCACAAATGGTCCTT-3′ repeated twice and 5′-GACCATTTGTGAT-3′ repeated three times) and a microsatellite (ATTT)4 (Figure 7b). In addition, a pair of inverted repeats (5′-TCTCCCTCTGTTTTTTT-3′/5′-AAAAAAACAGAGGGAGA-3′) was identified, suggesting transposition. No substantial repetitive sequences were found in the rest of the intergenic region of any of the three haplotypes, nor were any open reading frames detected.

Discussion

Self-incompatibility of Prunus tenella

We investigated SI in P. tenella, a wild species, of which the SI status had not been established previously. All of the accessions under investigation were heterozygous for the S-RNase gene, as expected for an SI species. The nine S-RNase alleles revealed in the 18 accessions characterized were found to correspond to active enzymes capable of degrading RNA in vitro, and had the sequence characteristics of S-RNases. Epifluorescent microscopy of pollinated pistils showed that the self-pollinated styles arrested and the cross-pollinated styles allowed pollen-tube growth. Based on this evidence, we conclude that SI operates in P. tenella.

Sequence regions exposed to positive selection

Our calculation of Ka/Ks, based on a sample of nine S-RNases from P. tenella, gave a value (0.99) higher than those calculated previously for S-RNases of cherry (0.50), almond (0.59) and Maloideae (0.83) (Ma and Oliveira, 2002), and of Prunus lannesiana (0.79; Kato and Mukai, 2004). The high Ka/Ks ratio in P. tenella is due to the very low rate of synonymous substitutions (0.147) compared with the 0.290 and 0.240 calculated for almond and cherry, respectively (Ma and Oliveira, 2002), and is even lower than the 0.17 and 0.183, respectively, calculated for S-RNases of P. armeniaca (Romero et al., 2004) and P. lannesiana (Kato and Mukai, 2004). Despite the likely age of these S-RNase alleles of P. tenella, they have accumulated remarkably few synonymous substitutions over time, unlike the old solanaceous S-RNase alleles described by Richman and Kohn (1996). Kato and Mukai (2004) suggested that the low rate of synonymous substitutions in P. lannesiana S-RNase alleles is due to genetic drift.

Ishimizu et al. (1998) analysed 11 S-RNase sequences from Maloideae and identified four sequence regions, PS1–PS4, with an excess of non-synonymous substitutions where positive selection for specificity may operate. In P. tenella the excess of non-synonymous substitutions is apparent in PS1, PS2 and PS3 and in an additional region between C1 and C2; the length of our sequences did not allow investigation of PS4 in P. tenella. It is interesting that, unlike the situation in almond (Ortega et al., 2006), positive selection operates in S-RNase alleles of P. tenella in the sequence regions previously identified in Maloideae.

An excess of non-synonymous substitutions was observed in antigen-recognition sites of class I MHC antigens, but not in the remaining codons (Hughes and Nei, 1988). By the same token, it is likely that in S-RNase alleles of P. tenella, the majority of codons (apart from those in the conserved regions and that between C2 and RHV) play a role in defining allelic specificity. Very high peaks indicate the particular importance of the region between RC4 and C5. On the basis of sequence comparison of phylogenetically related alleles in P. dulcis, Ortega et al. (2006) highlighted the significance of the region between RC4 and C5 in defining allelic specificity in that species.

Identity and near identity of S-RNase alleles in P. tenella, P. avium and P. dulcis

Surprisingly, one S-RNase allele of P. tenella, S8, was found to have 100% identity at the protein level to an allele from another species, S1 in P. avium. In addition, the allele S11 in almond had only one amino acid different from S1P. avium and S8P. tenella. It has been shown previously that S-alleles from different species may be similar and group together in gene genealogies in Solanaceae and Rosaceae (Ioerger et al., 1990; Ma and Oliveira, 2002; Richman and Kohn, 1996; Ushijima et al., 1998). Here, though, we report the occurrence not of highly similar, but of identical S-RNases in different Prunus species. Although crosses between P. tenella and P. dulcis have been attempted with varying success (Knight, 1969), neither P. dulcis nor P. tenella (sub-genus Amygdalus) hybridizes with P. avium (sub-genus Cerasus), so the occurrence of identical S-alleles in P. avium and P. tenella cannot be explained by introgression. Therefore the allele must have existed before the separation of these two sub-genera – an early event in Prunus speciation (Bortiri et al., 2002) – and remained with the same protein sequence in the two lineages ever since.

In contrast to the coding regions, the introns of the S-RNase alleles have accumulated various mutations and show significant differences in the three haplotypes – independent evidence against introgression. Intron lengths, which are considered (with the minor exception of S1 in almond; Ma and Oliveira, 2001) to be allele-specific, and are often used as markers in determining S-genotypes of Prunus cultivars (Sonneveld et al., 2001, 2003; Sutherland et al., 2004; Tao et al., 1997), differ considerably in the three species. Therefore intron length cannot be used for tracking S-alleles across the genus.

Corresponding SFB alleles are not identical but SI function is not affected

The widely accepted inhibitor model of SI response (Kao and McCubbin, 1996; Luu et al., 2001) proposes that the SFB protein and the S-RNase interact directly in specific recognition. In the light of this prediction, it is interesting that the protein sequences of SFB alleles of haplotypes S1P. avium (Vaughan et al., 2006) and S8P. tenella, which share the same S-RNase, were not identical: 12 residues differed between them. In addition, SFB11P. dulcis is different from both SFB8P. tenella and SFB1P. avium. This raises the question as to whether all three haplotypes are functionally self-incompatible. There is plenty of evidence that S1 in cherry and S11 in almond are functional and cause SI, as these are present in such well studied self-incompatible cultivars as Early Rivers, Roundel and Hertford (cherry) and Marcona, Rumbeta and Bertina (almond). In addition, our pollination tests showed that S8P. tenella confers SI. Thus, despite their variations in sequence, all three SFB alleles are operating in specific recognition, and the changes appear not to have disturbed their SI function.

This suggests some flexibility in the specific recognition – if the variant amino acids are involved in the physical interaction of the two proteins. Alternatively, it may be that the differing residues do not play a role in defining current allelic specificities of the haplotypes. Seven of the 12 residues that differ between SFB1P. avium and SFB8P. tenella are non-conservative replacements, one of which is in the HVb region, which is under positive selection and is likely to be involved in discrimination between self and non-self RNases (Ikeda et al., 2004; Ushijima et al., 2004). Thus it may be that a relatively high number of substitutions is required to change specificity of SFB alleles, rather than that these differing residues do not play a role in defining specificity.

Evolution of new S-specificities

From an evolutionary point of view, and regardless of whether or not the differing amino acids are involved in specific interaction, our data show that the SI function can tolerate a substantial number of mutations in SFB protein sequence, without breakdown. This is interesting information regarding the two models proposed so far to explain the evolution of new S-specificities. Uyenoyama et al. (2001) proposed that the evolution of S-specificities allows self-compatible intermediates, while the model proposed by Matton et al. (1999) predicted dual specificity intermediates that are self-incompatible. Recently, Chookajorn et al. (2004) investigated sporophytic SI in crucifers and determined that the pollen specificity (S-locus cysteine-rich protein) can tolerate a variety of experimentally induced amino acid changes and domain swapping, without compromising its ability to participate in the SI response. Their hypothesis of evolution of S-specificities predicts variability within the pollen and stylar components of each functional allelic class, and proposes each S-specificity to be a cluster of slightly varying sequences or sub-alleles. Recently, Newbigin and Uyenoyama (2005) speculated that a comparable situation may occur in the gametophytic system, and proposed that selection within S-allele specificity classes could accelerate the rate of non-synonymous substitutions, and that periodic selective sweeps would remove variation within classes.

Our data, derived from related Prunus species, show that some degree of variability is observed in both genes of the three S-haplotypes investigated, and that variants of the SFB matching the same S-RNase occur naturally in gametophytic SI. The pollen specificity seems more tolerant of non-conservative replacements. If balancing selection is the major driving force in the evolution of new specificities, it seems logical that evolution of new specificity starts with mutation of the pollen part. Such mutations do not necessarily cause breakdown of SI, and an identical S-RNase can recognize SFB alleles differing in 12 residues. Taking into account that a small number of substitutions is required in the S-RNase to change its specificity, at least in some cases (Matton et al., 1999; Saba-El-Leil et al., 1994), it could be that the plasticity of the SI response to pollen-part mutations allows the pistil component to change in time without breakdown of SI. This would accord with the dual-specificity model proposed by Matton et al. (1999). Whether sub-alleles exist within species, and how they might perform within a population, remains to be seen. The functional significance of differences between the S11 haplotype of P. dulcis and the S8 haplotype of P. tenella, differing in one residue in the S-RNase sequence and three residues in the SFB sequence, could be tested by reciprocal crosses of an appropriate accession of P. tenella with cultivar Marcona (S11S12) of P. dulcis to check for S-haplotype specific rejection. Prunus tenella and P. avium do not interbreed, and the interaction of S8P. tenella and S1P. avium cannot be investigated by crossing.

Even though the S-RNases of P. tenella are exposed to positive selection in most regions other than the conserved, evolution preserved the amino acid sequence of S8-RNase P. tenella and S1-RNase P. avium, as it existed prior to the separation of sub-genera in Prunus, while the cognate SFB underwent changes. This may indicate that the S-RNase is less tolerant of mutations. From a molecular point of view, this may be due to stricter functional constraints associated with the wider range of functions assigned to this gene compared with the SFB. The SFB is a gene for which two roles have been proposed: specific recognition and inactivation of non-cognate S-RNases (Ushijima et al., 2004). However, in the genus Prunus SFB appears to function only in specific recognition (Sonneveld et al., 2005). On the other hand S-RNase has been shown to operate in specific recognition (Kao and McCubbin, 1996; Zurek et al., 1997) and also to have a role in RNA degradation (McClure et al., 1989). In addition, the currently accepted ‘inhibitor model’ for S-RNase mediation of SI response proposes inhibition of the cytotoxic non-self S-RNases during compatible pollinations (Kao and Tsukamoto, 2004; Luu et al., 2001; McClure, 2004); this could be a third molecular process in which the S-RNase takes part.

Experimental procedures

Plant materials and DNA extraction

This study used 18 accessions of P. tenella: 16 (A1–A16) were sampled from the Deliblato sandstone area in Vojvodina, northern Serbia (44°50′Ν; 21°03′ E), and two (A17, A18) were collected from the Niš area in southern Serbia (43°19′ N; 21°54′ E). Prunus avium Roundel (S1S2) from East Malling Research’s Prunus collection, and P. dulcis Marcona (S11S12) Rumbeta (S11S23) and Bertina (S6S11) from CEBAS-CSIC (Murcia, Spain), were also used. Leaves and styles were collected, frozen in liquid nitrogen and stored at –80°C until used.

Genomic DNA was extracted according to the cetyltrimethylammonium bromide (CTAB) protocol of Doyle and Doyle (1987), adjusted for small quantities of leaf tissue. Polyvinylpyrolidone (PVP 40) in 2% (w/v) concentration was added to the extraction buffer, as well as 1%β-mercaptoethanol. Extracted DNA was diluted to 20 ng μl−1.

PCR analysis of first and second intron of S-RNase alleles:

The forward primer of Sonneveld et al. (2003) (PaConsI-F) designed from sweet cherry sequences, and the reverse primer of Ortega et al. (2005) (EM-PC1consRD) designed from published sequences of five Prunus species, were used to amplify the first introns of putative S-RNase alleles in P. tenella, the forward primer being labelled with VIC dye (Applied Biosystems, http://www.appliedbiosystems.com). The PCR reactions were performed in accordance with Sonneveld et al. (2006). PCR cycling conditions for amplification of the first intron were those used by Ortega et al. (2005).

The amplification products were fractionated by capillary electrophoresis through a 3100 genetic analyser (Applied Biosystems). Data generated by capillary electrophoresis were collected and analysed using the genescan and genotyper (Applied Biosystems) software.

The degenerate consensus primers of Sutherland et al. (2004) (EM-PC2consFD and EM-PC3consRD), designed from published sequences of five Prunus species, were used to amplify the second introns of the putative S-RNase alleles. PCR amplification of the second intron was performed using the same reaction mix and PCR conditions as described by Sutherland et al. (2004) for almond, except that the annealing temperature was 59°C. The PCR products of the second intron were electrophoresed through a 2% agarose gel for 2 h at 85 V, stained with ethidium bromide and visualized over UV light.

Isoelectric focusing of stylar protein extracts and RNase activity staining

Stylar proteins were extracted, separated by their charge on IEF gels, and stained to reveal RNase activity, essentially using the procedures described by Bošković and Tobutt (1996). The samples were electrophoresed for 1 h at 150 V, 2 h at 300 V and 2 h at 500 V.

Fluorescent microscopy of pollinated pistils

Controlled pollinations were performed in the field on plants with previously determined genotypes. Accessions A9 (S7S8) and A15 (S6S8) were selfed and accessions A8 (S4S9) and A11 (S1S5) were crossed and selfed. The flowers were emasculated at the balloon stage, 24 h before pollination, and protected by paper bags. Anthers were collected when plants were emasculated and kept for 24 h at room temperature to release pollen. The flowers were pollinated and re-bagged and the pistils were harvested 96 h later. Pistils were fixed, stained, squashed and examined under an epifluorescence microscope (DMLS, Leica Microsystems, http://www.leica-microsystems.com) using an I3 Blue filter, in general accordance with Cerović (1994).

Cloning and sequencing genomic PCR products of S-RNase alleles of P. tenella

Primers used for PCR amplification of nine S-RNase alleles (S1S9) of P. tenella, from the signal peptide to the C5 region, prior to cloning, were PaConsI-F (Sonneveld et al., 2003) and EM-PC5consRD (Sutherland et al., 2004). PCR amplification was performed with KOD Hot Start DNA Polymerase (Novagen, http://www.merckbiosciences.co.uk) as described by Ortega et al. (2006). Amplification products were cloned into pCR Blunt-II-Topo vector and transformed into TOP10 One-shot chemically competent Escherichia coli (Zero Blunt TOPO Cloning Kit, Invitrogen, http://www.invitrogen.com), according to the manufacturer’s protocol.

Colony PCR for the S-RNase alleles was performed using EM-PC2consFD and EM-PC3consRD primers in a PCR reaction, as described for analysis of the second intron. Plasmid DNA from at least two bacterial colonies per allele was sequenced using M13 forward and reverse primers. The exact sequence of the C5 region, as well as that corresponding to the last 26 amino acids of the protein, was obtained for allele S8P. tenella and also for S11P. dulcis when the intergenic region was sequenced, as explained below.

Cloning and sequencing genomic PCR products of SFB8 allele of P. tenella and SFB11 allele of P. dulcis

Cloning and sequencing of SFB8 from P. tenella accessions A9 (S7S8) and A16 (S6S8) and SFB11 from P. dulcis cultivars Marcona (S11S12) and Bertina (S6S11) was done using the degenerate primers designed for amplifying Prunus SFB alleles, F-BOX 5′A and F-BOX 3′A (Vaughan et al., 2006). PCR reactions were performed as described by Vaughan et al. (2006) with minor alterations in PCR cycling conditions: 3 min at 95°C, with 10 cycles of 45 sec at 94°C, 1 min at 65°C decreasing the temperature 1°C per cycle and 2 min at 68°C, followed by 30 cycles of 45 sec at 94°C, 1 min at 50°C and 2 min at 68°C, with 10 min final extension at 68°C. Amplification products were cloned using the Zero Blunt TOPO Cloning Kit (Invitrogen).

Colony PCR was performed with the same primers as those used before cloning. Three clones from each accession were sequenced using M13 forward and reverse primers, and the sequence of the SFB8 allele was identified as the SFB sequence that the two accessions had in common. The sequences were compared with that of SFB1P. avium (Vaughan et al., 2006). We agree with Vaughan et al. (2006) in regarding the SFB1P. avium sequence published by Ikeda et al. (2004) as erroneous.

Cloning and sequencing S-RNase–SFB intergenic region of haplotypes S8P. tenella, S1P. avium and S11P. dulcis

The SFB gene in Prunus is positioned downstream and adjacent to the S-RNase. Primers used for amplifying the intergenic region in all three haplotypes were PaConsI-F, matching the signal peptide of the S-RNase gene, and a newly designed primer SFBa1-3RC (5′-TTG CAA TTA CTT ACA TAG AAA GTT CTG G-3′) that fully matched the 3′ region of the P. avium SFB1 gene and none of the other SFB sequences available in the database at the time. PCR reactions were performed in a final volume of 20 μl containing 50 ng genomic DNA, 1 × PCR buffer (Qiagen, http://www.qiagen.com), 2.5 mm MgCl2, 0.2 mm dNTPs, 0.25 μm of each primer, 0.1× Q solution (Qiagen) and 0.5 U Taq polymerase (Qiagen). Cycling conditions were as described for the second intron of cherry S-RNases (Sonneveld et al., 2003). PCR products were purified (QIAquick, Qiagen) and cloned into pGEM-T vector (Promega, http://www.promega.com). Electrocompetent E. coli were used for transformation. Colonies were screened using the same primers and PCR protocol as prior to cloning, and inserts were sequenced using M13 primers, the reverse complement of EM-PC3consRD from the C3 region of the S-RNase (Sutherland et al., 2004) and the newly designed primer, SFBa1-3RC.

Sequence analysis

Contigs were assembled from raw sequence data using SeqMan 4.06 (DNAStar, http://www.dnastar.com). Sequences were translated using EditSeq 4.06 (DNAStar). DNA and protein alignments of the nine P. tenella S-RNase alleles were carried out using the clustal V method in MegAlign 4.06 software (DNAStar). Database searches were performed using the National Center for Biotechnology Information (NCBI)’s basic local alignment search tool (blast). To identify the sequence regions in which positive selection may operate, the mean ratio of the number of non-synonymous substitutions (nucleotide changes resulting in a different amino acid) per non-synonymous site (Ka) to the number of synonymous substitutions per synonymous site (Ks) was calculated for ‘sliding windows’ of 20 codons for all 36 pairs using the program DnaSP (ver. 4.10.8) (Rozas and Rozas, 1999). To perform pairwise comparisons of P. tenella alleles, lalign software (W. Pearson, http://www.ch.embnet.org/software/LALIGN_form.html), implementing the algorithm of Huang and Miller (1991), was used. Repetitive sequences were identified using repfind (Betley et al., 2002) and emboss palindrome (M. Faller, http://bioweb.pasteur.fr/docs/EMBOSS/palindrome.html) software (Rice et al., 2000).

Acknowledgements

This work was supported in part by the Ministry of Science and Environment Protection of the Republic of Serbia, Grant 143017. Stone-fruit genetics at EMR is financed by the Department for Environment, Food and Rural Affairs, UK. Nada Šurbanovski thanks the British Council in Belgrade for a short-term travel grant, and Radovan Bošković acknowledges a grant from the Mount Trust. We thank Dr Simon Vaughan for valuable advice and information.

Accession numbers:DQ983363; DQ983364; DQ983365; DQ983366; DQ983367; DQ983368; DQ983369; DQ983370; DQ983371; DQ983372; DQ983373; DQ983374; DQ983375; EF061758.

Ancillary