Coevolution of the S-Locus Genes SRK, SLG and SP11/SCR in Brassica oleracea and B. rapa

Corresponding author: Takeshi Nishio, Tohoku University, Sendai 981-8555, Japan., nishio{at}bios.tohoku.ac.jp (E-mail)

Communicating editor: M. K. UYENOYAMA

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Brassica self-incompatibility (SI) is controlled by SLG and SRK expressed in the stigma and by SP11/SCR expressed in the anther. We determined the sequences of the S domains of 36 SRK alleles, 13 SLG alleles, and 14 SP11 alleles from Brassica oleracea and B. rapa. We found three S haplotypes lacking SLG genes in B. rapa, confirming that SLG is not essential for the SI recognition system. Together with reported sequences, the nucleotide diversities per synonymous and nonsynonymous site (_S and _N) at the SRK, SLG, and SP11 loci within B. oleracea were computed. The ratios of _N:_S for SP11 and the hypervariable region of SRK were significantly >1, suggesting operation of diversifying selection to maintain the diversity of these regions. In the phylogenetic trees of 12 SP11 sequences and their linked SRK alleles, the tree topology was not significantly different between SP11 and SRK, suggesting a tight linkage of male and female SI determinants during the evolutionary course of these haplotypes. Genetic exchanges between SLG and SRK seem to be frequent; three such recent exchanges were detected. The evolution of S haplotypes and the effect of gene conversion on self-incompatibility are discussed.

SELF-INCOMPATIBILITY (SI) in Brassica is controlled by a set of closely linked genes at the S locus, called the S haplotype. These genes have multiple alleles and are expressed either in the stigma or in the pollen. Stigma cells inhibit pollen tube growth to prevent self-fertilization when the expressed S specificity of the pollen matches that of the stigma. In Brassica, self-recognition specificity of the pollen is determined sporophytically. It depends on the S haplotype of the pollen parent rather than on that of the pollen grain itself. About 50 S haplotypes in Brassica oleracea (OCKENDON 2000 ) and 30 in B. rapa (NOU et al. 1993 ) have been identified so far.

The first S-locus gene to be isolated, SLG (S-locus glycoprotein), encodes a secreted protein, which localizes in the wall of stigma papillar cells (NASRALLAH et al. 1988 ). The SLG alleles are classified into two groups, class I and class II, on the basis of their nucleotide sequences (NASRALLAH et al. 1991 ). Subsequently, the SRK gene (S-locus receptor kinase) was isolated (STEIN et al. 1991 ). SRK is a membrane protein consisting of an extracellular domain (S domain), which is similar in sequence to SLG, a single-pass transmembrane domain, and a cytoplasmic domain with protein kinase activity. The coding region of SRK is 2.6 kb in length and is partitioned by six introns. Loss of the function of SRK was found to result in a breakdown of SI (GORING et al. 1993 ; NASRALLAH et al. 1994 ). S domain sequences of SRK alleles are similar to SLG sequences in the same class (CABRILLAC et al. 1999 ). Class II S haplotypes show a pollen-recessive phenotype. Introduction of the SRK gene alone was found to confer a new S-haplotype specificity in the stigma (TAKASAKI et al. 2000 ). Consequently, SRK is considered to be an indispensable factor in the stigma for both SI recognition and response leading to the rejection of self-pollen. The determinant of S-haplotype specificity in pollen has recently been identified by two groups of researchers. This gene has been called S-locus protein 11 (SP11) by SUZUKI et al. 1999 and S-locus cysteine-rich protein (SCR) by SCHOPFER et al. 1999 . The deduced amino acid sequences of SP11 in B. rapa have been shown to be highly divergent except for the presence of conserved cysteine residues (WATANABE et al. 2000 ). It is considered that recognition of SP11 by SRK results in inhibition of self-pollen germination and pollen tube growth.

KUSABA et al. 1997 sequenced >30 SLG alleles from B. oleracea and B. rapa and subjected them to phylogenetic analysis with other reported sequences. They found a high extent of intraspecific variation and interspecific similarity between SLG alleles and confirmed that the divergence of SLG alleles predates the speciation of B. oleracea and B. rapa. They also demonstrated that recombination or gene conversion plays a major role in the course of the evolution of SLG genes. This observation was confirmed by a further study by AWADALLA and CHARLESWORTH 1999 .

KUSABA et al. 1997 also observed a striking sequence similarity (97.5% identity in their amino acid sequences) between SLG genes from S-8 and S-46 haplotypes in B. rapa. However, the S domains of the SRK genes in these haplotypes are not as similar to each other (85.6% amino acid identity) as are the SLG genes, suggesting that SLG is not essential for self-recognition (KUSABA and NISHIO 1999 ). Further evidence suggests that SLG is not crucial for the SI phenotype: a nonsense mutation and frameshift that eliminated function were found in SLG-18 and SLG-60 in B. oleracea (SUZUKI et al. 2000 ). Furthermore, the SLG gene in B. oleracea S-24 appears to have been deleted, having not been detected in a DNA gel-blot analysis (OKAZAKI et al. 1999 ). On the other hand, transgenic studies have shown that the function of SLG is to intensify the strength of SI (TAKASAKI et al. 2000 ).

In this article, we report the following new sequences: the S domain sequences of 21 BoSRK alleles (SRK in B. oleracea), 15 BrSRK alleles (SRK in B. rapa), 14 BoSP11 alleles (SP11 in B. oleracea), 11 BoSLG alleles (SLG in B. oleracea), and 2 BrSLG alleles (SLG in B. rapa). Together with previously reported sequences of SLG, SRK, and SP11 (NASRALLAH et al. 1988 ; STEIN et al. 1991 ; KUSABA et al. 1997 ; SCHOPFER et al. 1999 ; SUZUKI et al. 1999 ; TAKAYAMA et al. 2000 ; WATANABE et al. 2000 ), a total of 25 BoSRK, 34 BoSLG, 19 BoSP11, 18 BrSRK, 21 BrSLG, and 18 BrSP11 alleles were used for the analysis. This data set includes the majority of S haplotypes known in B. oleracea and B. rapa. We herein discuss the mode of evolution of S haplotypes, focusing in particular on the coevolution of male and female SI determinants and the molecular mechanisms affecting the diversity of SI genes.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

DNA sources and sequencing:
Forty-five S homozygous lines of B. oleracea L. provided by D. Astley (Horticulture Research International, Warwick, UK) and 15 homozygous lines of B. rapa L. maintained at Tohoku University were used as plant materials. SLG and the SRK alleles that have a short first intron were amplified by PCR using genomic DNA as a template. SP11 and those SRK alleles having a long first intron were amplified by RT-PCR from anther and stigma RNA, respectively. Genomic DNA was prepared from young leaves by the method of ROGERS and BENDICH 1985 , and RNA was extracted from stigmas and anthers of the S homozygotes using Isogen (Nippongene) or Micro-Fast track mRNA isolation kit (Invitrogen, San Diego). The primers used for the amplification of SLG were the pair PS5 (5'-ATGAAAGGCGTAAGAAAAACCTA-3') and PS15 (5'-CCGTGTTTTATTTTAAGAGAAAGAGCT-3'; NISHIO et al. 1996 ) or the pair PS22 (5'-ATCGATGGGATGAAAAAGTCATCG-3'; SAKAMOTO et al. 1998 ) and PS15. The primers for amplification of the S domain of SRK were the pair PK7 (5'-ATGCAAGGTGTACGATACATCTATCATCATTCTTAC-3') and PK8 (5'-GATCAGAAGAAGCAGAACAGTAACTCCAACAGTC-3') or the pair of PK7 and PK9 (5'-CCTTGTCCGAGTTTGTTACAGTTGGAGAAATTTTCGG-3'; NISHIO et al. 1997 ). The conditions for PCR and the cloning of the PCR products have been reported previously (NISHIO et al. 1997 ). For RT-PCR of the S domain of SRK, the first strand cDNA was synthesized with SuperScriptII RT (GIBCO BRL, Gaithersburg, MD) using the primer PK8 or PK9, and PCR amplification was carried out under the same conditions as for genomic DNA using the pairs of primers PK7 + PK8 and PK7 + PK9. For the amplification of SP11, the primer pair of pSP11-1 (ATGAAATCTGCTATTTATGCTTTATTATG) and NotI(dT)18 (Amersham Pharmacia Biotec) was used for the first cycle of RT-PCR, and the pair pSP11-2 (TTCATATTCATCGTTTCAAGTC) and RT-1 (ACTGGAAGAATTCGCGGC) was used in the second cycle.

The nucleotide sequences of the PCR products were determined with PRISM377 (Perkin-Elmer ABI). To eliminate errors that may have occurred during the PCR process, three independent clones obtained from the same plant were sequenced. The DNA sequence data were analyzed with the Genetyx version 10 program (Software Development, Tokyo).

DNA blot analysis:
DNA gel blotting was performed as described by OKAZAKI et al. 1999 except for the probe and the washing conditions. Two micrograms of genomic DNA was digested with EcoRI or HindIII, electrophoresed on agarose gel, and transferred to a nylon membrane. The membranes were hybridized with a mixture of S domain probes of BrSRK-32, BrSRK-33, and BrSRK-36, which were amplified from plasmids by PCR. After hybridization, the membrane was washed twice in 0.5x SSC, 0.1% SDS at 65° for 20 min.

Phylogenetic analysis:
Sequences were aligned by using CLUSTAL W (THOMPSON et al. 1994 ) and modified manually. Synonymous and nonsynonymous sites and differences were counted by using an algorithm of the modified Nei and Gojobori method (NEI and KUMAR 2000 ) with R = 1. Phylogenetic analyses were performed by using programs of Neighbor, DNAML (PHYLIP version 3.5; FELSENSTEIN 1993 ), PROTML (MOLPHY version 3.2; ADACHI and HASEGAWA 1994 ), and PAML (YANG 2000 ).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Deletion of the SLG gene in some S haplotypes:
Only one band was detected in the DNA blot analysis of HindIII and EcoRI fragments of BrS-36, BrS-32, and BrS-33 homozygotes with the bulked SRK probe (Fig 1), while two bands, corresponding to SRK and SLG, have been found in most other S haplotypes (OKAZAKI et al. 1999 ). Since these three haplotypes maintain distinct SI specificities, SLG is unlikely to play any essential role in SI recognition (KUSABA and NISHIO 1999 ; OKAZAKI et al. 1999 ; SUZUKI et al. 2000 ).

View larger version (79K): In this window In a new window Download PPT slide   Figure 1. DNA blot analysis of genomic DNA isolated from BrS-9, BrS-32, BrS-33, and BrS-36 homozygotes. E, EcoRI; H, HindIII. DNA sizes are shown on the left.

The BrSRK-36 had extremely high similarity to BoSRK-24 [93.5% identity in amino acids (aa)]. The hypervariable regions (HVRs), which are considered to be important for recognition specificity of SLG and SRK (KUSABA et al. 1997 ; NISHIO and KUSABA 2000 ), were identical except for two amino acid residues, and the SP11 sequences were also similar (96.8% identity in aa). Additionally, the BoS-24 haplotype lacks the SLG gene (OKAZAKI et al. 1999 ). Therefore, BoS-24 and BrS-36 are likely to have been derived from a single ancestral haplotype and the SLG gene in the haplotype was probably deleted in the common ancestor of B. oleracea and B. rapa. The BrS-32 homozygote also showed only one band, suggesting the deletion of SLG. Since BrS-32 has an SRK sequence similar to BrSRK-36 (88.5% identity in aa), BrS-32 most likely originated from the same ancestor as BoS-24 and BrS-36. However, this DNA blot analysis indicated that BrS-33, which is distantly related to BrS-32, BrS-36, and BoS-24 in their SRK sequences, has also lost the SLG gene. This implies that the SLG-deletion event occurred independently at least twice.

Sequence diversity of SRK, SLG, and SP11 in each species:
The average proportion of identical amino acids per site among all pairwise comparisons of class I S-locus sequences was 80% in 28 BoSRKs, 79% in 21 BrSRKs, 82% in 34 BoSLGs, and 79% in 21 BrSLGs. In SLG sequences, the highest extent of amino acid sequence identity was 99.5% between BoSLG-23 and BoSLG-29 in B. oleracea and 98.2% between BrSLG-43 and BrSLG-46 in B. rapa. Likewise, the highest similarity in the S domain of SRK sequences was 89.9% in B. oleracea (BoSRK-23 and BoSRK-29) and 88.5% in B. rapa (BrSRK-32 and BrSRK-36). KUSABA and NISHIO 1999 showed that the high extent of similarity between some SLG genes was due to homogenization by genetic exchanges among alleles. To investigate whether or not a similar number of exchanges occurred among SRK alleles, we applied the method of KUSABA et al. 1997 —in which topologies of neighbor-joining (NJ) trees using nucleotide sequences of the hypervariable regions I, II, III, and the C-terminal variable region were compared—to all available SRK sequences from B. oleracea and B. rapa. However, no evidence of frequent recombination was detected among SRK alleles. In contrast to the relatively high extent of homology among SRK or SLG amino acid sequences, SP11 sequences, responsible for the S specificity of the pollen, showed extraordinary diversity. As shown in Fig 2, although six cysteine residues are conserved among the sequences, the remaining residues show extensive variation, including frequent insertions and deletions (indels). Since the inclusion of many indels reduces the number of comparable amino acid sites, we selected six sequences (BoSP11-7, -18, -24, -29, -39, and -64) whose alignment requires a relatively small number of indels. The highest amino acid identity among the six sequences was 41.2%, between BoSP11-39 and BoSP11-64.

View larger version (56K): In this window In a new window Download PPT slide   Figure 2. Multiple sequence alignment of SP11 amino acid sequences in B. oleracea. Cysteine residues are in boxes, and conserved cysteine residues are indicated by solid triangles.

The nucleotide diversities per synonymous and nonsynonymous site (_S and _N ) at the SRK, SLG, and SP11 loci within B. oleracea were computed. The values _S and _N are the average numbers of synonymous or nonsynonymous nucleotide differences per site between two randomly chosen sequences. Because the extent of sequence difference of SLG and SRK at both the amino acid and nucleotide levels varies along the coding region (HINATA et al. 1995 ), the sequence was divided into two subregions: the HVR and the remaining conserved region (CR), as designated by KUSABA et al. 1997 . Table 1 shows _S and _N in the HVR, CR, and entire gene (ALL) of the SRK and SLG loci separately. For SP11, since the number of comparable sites is small, the division of the sequence into subregions is not useful. Therefore, we calculated _S and _N for the entire region only. The ratio () of _N:_S for each region is also shown (Table 1).

View this table: In this window In a new window   Table 1. The average number of synonymous (S) and nonsynonymous (N) differences per site of SRK, SLG, and SP11 alleles in B. oleracea

To estimate the -value for each region, we used the modified Nei and Gojobori method (NEI and KUMAR 2000 ). To calculate the parameter R in the modified method, where R is defined as /2ß, we estimated the ratio of transitional () to transversional (ß) substitutions by maximum-likelihood methods by using the PAML (YANG 2000 ). We estimated the ratio, /ß, under the maximum-likelihood topology, using the sites at the third codon positions of the entire gene. The ratio, /ß, in SRK was almost 2 and that in SP11 was 1.5. Because the total number of the third codon positions is small in SP11, we decided to use the ratio 2, setting R equal to 1. We also applied the unbiased Nei and Gojoboris method to our data. This actually reduced the -value, but the tendency did not change.

It is clear that _N in HVR is significantly greater than that in CR for both SRK and SLG, and the -value in HVR of SRK and SLG and in SP11 exceeds unity. Under the neutral theory of molecular evolution (KIMURA 1968 , KIMURA 1983 ), of a particular gene depends on the strength of functional constraints imposed on the product of the gene. However, does not exceed one, unless mutations are selectively advantageous. In other words, if in a gene or a part of a gene is significantly larger than one, this indicates an operation of balancing selection or Darwinian (positive natural) selection in these regions.

To examine whether or not these -values are significantly larger than unity, we calculated , computed the variance of D by bootstrap samplings with 1000 replications, and applied the Z test (NEI and KUMAR 2000 , p. 55). The result showed that D was significantly larger than zero in SP11 and in the HVRs of both SRK and SLG (P < 0.01, Table 1). The observation of > 1 suggests the operation of Darwinian selection or balancing selection at SP11 and at the HVRs of SRK and SLG. A similar result was obtained for B. rapa.

Phylogenetic analysis of SRK and SP11:
To examine whether or not the linkage between SP11 and SRK is tight, we compared phylogenetic relationships of these genes. For both B. oleracea and B. rapa, nucleotide sequences from the two loci in 26 different haplotypes (13 for each species) are available. However, as mentioned, reliable alignment among all available SP11 amino acid or nucleotide sequences is difficult to achieve due to the large number of indels. We therefore used six BoSP11 alleles for further phylogenetic analysis (Table 1, Fig 2). In addition to these six BoSP11 alleles, we chose six other alleles from B. rapa (BrSP11-36, SP11-45, SP11-49, SP11-47, SP11-41, SP11-46), which are seemingly closely related to BoSP11 and therefore can be aligned with each other with a relatively small number of indels (data not shown).

Fig 3A shows the NJ tree (SAITOU and NEI 1987 ) estimated on the basis of both synonymous and nonsynonymous differences among 12 SP11 sequences. It is clear that six major lineages are among them: lineage I (BoSP11-24, BrSP11-36), lineage II (BrSP11-45, BrSP11-49), lineage III (BoSP11-64, BrSP11-41, BoSP11-39), lineage IV (BoSP11-7, BrSP11-46, BrSP11-47), lineage V (BoSP11-29), and lineage VI (BoSP11-18). Lineages I–IV have significant bootstrap support (P > 0.99). This pattern did not change even when we used synonymous or nonsynonymous differences separately (Fig 3B and Fig C), although the bootstrap probabilities for the synonymous tree became a bit smaller. The relationship among the six lineages was not resolved at the root of the tree. This low resolution at the root might be due to either the small number of compared sites or the possible large number of recurrent substitutions or frequent recombination/gene conversions.

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. Neighbor-joining tree of selected SP11 sequences. One thousand bootstrap trials were performed and the bootstrap values are shown in ovals. The scale shows the number of nucleotide differences per site. The reason for the selection of SP11 sequences is shown in the text. (A) NJ tree based on all nucleotide changes. (B) NJ tree based on synonymous nucleotide changes. (C) NJ tree based on nonsynonymous nucleotide changes.

Phylogenetic trees of the 12 SP11 alleles and their linked SRK alleles were constructed separately by using the deduced amino acid sequences. First, maximum-likelihood analysis (PROTML in the MOLPHY version 2.3; ADACHI and HASEGAWA 1994 ) was applied (Fig 4). The topologies of these SP11 and SRK trees were a little different from each other. In the following, we call the tree topology of SRK genes Tree 1 and that of SP11 genes Tree 2 (Fig 4). To test the significance of topological differences between the two trees, likelihood values were calculated for both genes. Because the topology of the best tree, which shows the maximum-likelihood value among different topologies, was different between SRK and SP11, we also calculated the likelihood value of the SRK gene tree under the assumption of the Tree 2 topology and vice versa. For each tree, the difference of the likelihood values between the best and alternative trees and its standard error were calculated (KISHINO and HASEGAWA 1989 ; Table 2). There were no significant differences between the maximum-likelihood values of the two topologies for both the SP11 and SRK sequences, and therefore the hypothesis that the topologies of phylogenetic trees supported by these two loci were the same was not rejected. To address whether heterogeneity in amino acid substitution rate affects the conclusion, we carried out a similar analysis using PAML (YANG 2000 ). A gamma distribution was used for the rate heterogeneity model and the shape parameter of the distribution was determined to fit the data. Even after taking heterogeneity of substitution rate into consideration, the hypothesis of the same topology of the two gene trees (SP11 and SRK) was not rejected.

View larger version (15K): In this window In a new window Download PPT slide   Figure 4. The phylogenetic relationship of SRK and SP11 genes on a single haplotype. Number shown at each node indicates the bootstrap value of the OTU cluster connecting at the node. The scale of each tree is indicated by a thick bar. Note that these are unrooted trees.

If the topology is the same, one may consider whether or not the divergence time of each operational taxonomic unit (OTU) is the same. Indeed, in the study of parasites and host coevolution, there are several such kinds of discussions (HUELSENBECK et al. 1997 , HUELSENBECK et al. 2000 ). In these studies, the nucleotide or amino acid sequences of the same genes or proteins are available from both hosts and parasites (e.g., mitochondrially encoded COI genes in HUELSENBECK et al. 1997 ), permitting direct comparison of maximum-likelihood estimates of external branch length between host and parasites. However, in our case, divergence time of different genes is compared. It is obvious that the amino acid substitution rate is different between two genes (Table 1, Fig 4) and therefore we cannot compare the maximum-likelihood estimate of external branches, as in the case of host-parasite coevolution, to examine whether the divergence time of genes is the same or not. Furthermore, synonymous substitutions were not used because of large standard errors due to the small number of compared sites in SP11. Therefore to evaluate whether the divergence time of each gene at SRK and SP11 on a single haplotype is similar, we examined the correlation of external branch lengths. If the two genes on each pair of haplotypes diverged at the same time, there would be a correlation between SRK and SP11 external branch lengths. Since we do not know the true tree, we estimated branch lengths under both topologies (Tree 1 and Tree 2) separately by using PROTML (ADACHI and HASEGAWA 1994 ). As we expected, the correlation of branch lengths between the two genes was quite high ( for Tree 1 and for Tree 2). This observation shows that the SP11 and SRK genes on a single haplotype seem to have diverged at the same time.

Phylogenetic relationship and tracing gene conversion between SRK and SLG:
Regarding the generation of diversified haplotypes, the involvement of the frequent duplication of the S domain of SRK and gene conversion between SRK and SLG has been pointed out (GORING et al. 1993 ; WATANABE et al. 1994 ; TANTIKANJANA et al. 1996 ). However, these conclusions are based on a limited number of samples. Thus, here we reexamined the phylogenetic relationship between SLG and SRK on the basis of an extensive number of samples (49 SRK and 55 SLG sequences).

Since amino acids are likely to be a target of diversifying selection, only synonymous changes were used for the phylogenetic analysis (Fig 5). Among 43 haplotypes for which both SLG and SRK were sequenced, 18 cases show that SLG and its linked SRK are more closely related to each other than to their alleles from different haplotypes. Furthermore, of these, 10 (BrS-26, BoS-25, BrS-37, BoS-1, BrS-28, BrS-30, BrS-45, BrS-49, BoS-33, and BoS-35) showed close relationships between SLG and SRK that were significantly supported by high bootstrap probability (95% bootstrap support, Fig 5). The number of synonymous changes per site between a pair of SRK and SLG genes on the same haplotype ranges from 0.004 ± 0.004 (BrSRK-45:BrSLG-45) to 0.088 ± 0.020 (BoSRK-1:BoSLG-1) with an average of 0.045 ± 0.014. Compared with the minimum divergence (0.022 ± 0.010) observed in interspecific comparisons between B. oleracea and B. rapa (BoSRK-32 and BrSRK-43), the relatively small synonymous changes between SRK and SLG suggest relatively recent conversion, including conversion after the species divergence, of SLG by SRK or vice versa.

View larger version (40K): In this window In a new window Download PPT slide   Figure 5. Phylogenetic relationships of SRK and SLG sequences in B. oleracea and B. rapa. The NJ tree based on synonymous substitutions was constructed. Internal branches that are highly supported are in red (P > 0.99) and in light blue (0.99 > P > 0.95). External branches and OTUs in blue indicate haplotypes in which SRK and SLG are more closely related to each other than to SRK and SLG alleles in the other haplotypes. External branches and OTUs in green indicate that SRK or SLG alleles are more closely related to SRK or SLG alleles in an interspecific pair of haplotypes than to those in the other haplotypes.

Five pairs of S haplotypes (BoS-45:BrS-22, BoS-51:BrS-24, BoS-7:BrS-46, BoS-12:BrS-47, and BoS-64:BrS-41) show that haplotypes from different species, B. oleracea and B. rapa, are closely related to each other: BoSRK genes are closely related to BrSRK genes and BoSLG genes are closely related to BrSLG genes (Fig 5, Table 3). Among these five, three pairs (BoS-45:BrS-22, BoS-7:BrS-46, and BoS-64:BrS-41) showed comparable levels of nucleotide divergence at SRK and SLG (P > 0.05). Taking the average of SRK and SLG divergence for each pair, we compared these averages with the minimum synonymous divergence (0.022 ± 0.010) of the two species. The divergences of two pairs, 0.047 ± 0.010 between BoS-45:BrS-22 and 0.034 ± 0.009 between BoS-7:BrS-46, were not significantly different from the minimum (P > 0.05). These observations indicate that each of these haplotype pairs is likely to have diverged from a common ancestor at the time of species divergence.

The relationship between BoS-64 and BrS-41 was somewhat different from those of other pairs. In both nucleotide and amino acid sequences, BrSRK-41 is quite similar to BoSRK-64 over the entire coding region (95.1% identity in aa and 96.8% identity in DNA). This close relationship was supported by phylogenetic analysis (P = 0.99, Fig 5). The nucleotide sequence from position 669 to 1115 (from the ATG initiation codon) in BrSLG-41 is highly similar to those of BrSRK-41 and BoSRK-64 (0 and 2.4% difference in DNA, respectively), but not to BoSLG-64 (9.2% difference in DNA). The remaining region of BrSLG-41 is highly similar to that of BoSLG-64 (2.5% difference, Fig 6), but distantly related to that of BrSRK-41 (13.9% difference). This was also reflected in a relatively low bootstrap probability of the branch of BoSLG-64 and BrSLG-41 in Fig 5 (P = 0.868). Partial but high identity observed between BrSLG-41 and BrSRK-41 might be caused by convergent evolution with some natural selection. However, because the highly homologous segment in BrSLG-41 involves synonymous sites as well as nonsynonymous sites, gene conversion is more likely than convergence due to natural selection.

View larger version (111K): In this window In a new window Download PPT slide   Figure 6. Comparison of nucleotide sequences between the S domain of SRK and SLG of BoS-64 and BrS-41. Nucleotides different from those in BoSRK-64 are shown in red boxes.

BoS-51:BrS-24 shows large divergence in SLG compared with a close relationship in SRK, and BoS-12:BrS-47 shows the opposite pattern, namely, large divergence in SRK compared to SLG (Table 3). In the former case, BoSRK-51 and BrSRK-24 are closely related through the entire coding region. Nucleotide differences were detected at only 18 among 1152 sites (1.6%). However, the relationship between BoSLG-51 and BrSLG-24 is complicated. In some regions, BoSLG-51 is almost identical to BoSRK-51 or BrSRK-24, but in others, BrSLG-24 is almost identical to BoSRK-51 or BrSRK-24. This suggests that segmental transfer between SRK and SLG has occurred not only once but several times. In the case of BoS-12 and BrS-47, BrSRK-47 and BoSRK-12 are relatively distantly related (Table 3). In fact, 33 and 25 unique nucleotides are not shared with the other three sequences in BrSRK-47 and BoSRK-12, respectively. In nucleotide position 457–729, the BrSRK-47 sequence is quite similar to BrSLG-47 and BoSLG-12 (0 and 2.2% difference, respectively) but not to BoSRK-12 (4.4% difference), while BrSRK-47 is similar to BoSRK-12 from position 809 to 1014 (0.5% difference) but not to BrSLG-47 (10.7% difference). This fact suggests that segmental transfer between BrSLG-47 and BrSRK-47 has occurred.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Diversity of S-locus genes:
Knowledge of the synonymous nucleotide diversity (_S) at other neutral loci may help to distinguish among the possible processes that may have enhanced the rate of nonsynonymous substitution in the S-locus genes. Balancing selection, including diversifying selection, makes _S at SRK or SLG much larger than that at the neutral loci due to relatively longer persistence time of alleles at these loci, whereas positive selection makes _S rather small due to selective sweep. If the _S at SRK and SLG is significantly larger than _S at other unlinked loci, balancing selection is plausible. Although there has been no systematic analysis of nucleotide diversity at loci unlinked to the S locus in Brassica species, a _S value >10% seems unusually large (Table 1). In a relatively closely related species, Arabidopsis thaliana, nucleotide diversity in different genomic regions ranges from 0.5 to 1.8% (AGUADE 2001 ). To understand the reason for the relatively large synonymous nucleotide diversity in Brassica, we need nucleotide sequence information for other neutral loci in the species.

In the present study, SRK, SLG, and SP11 alleles in class I S haplotypes were compared. Although the frequency of the class II S haplotypes is high in Brassica vegetables, the number of functionally distinct haplotypes is few—three in B. oleracea (CABRILLAC et al. 1999 ). There are a limited number of nucleotide sequence data of the class II S haplotypes (CABRILLAC et al. 1999 ). Through comparison of SLG sequences, an ancient divergence between class I and class II has been inferred (KUSABA et al. 1997 ). SCHIERUP et al. (2001) found greater sequence diversity in A. lyrata SRK alleles than in Brassica SRK alleles, suggesting more ancient diversification of A. lyrata alleles. The class I SRK sequences newly determined in the present study did not enlarge remarkably the sequence diversity of the class I alleles, supporting the finding of SCHIERUP et al. (2001).

Coevolution of SP11, SRK, and SLG:
In the phylogenetic analysis of SP11 and the S domain of SRK, the hypothesis that the topology is the same between the SP11 tree and the SRK tree was not rejected. A positive correlation in divergence time between the SRK and SP11 alleles was suggested by comparison of branch lengths in the two trees. This phylogenetic relationship between the SRK and SP11 alleles likely suggests strong linkage disequilibrium of these two genes in the S locus. Recent studies on S-locus structure have demonstrated that the distance between SP11 and SRK and the orientation of these genes are highly variable among different S haplotypes (TAKAYAMA et al. 2000 ). The structural diversity of S locus likely discourages recombination between these genes. Functional interactions between SRK and SP11 may also have contributed to the coevolution of these genes.

Since SRK and SLG genes do not fall into separate clusters in the gene genealogy, genetic exchange between the two loci seems to play a significant role in the diversification of S haplotypes. This pattern of molecular evolution contrasts with the pattern observed in human MHC (HLA) class I genes. In both cases, diversified alleles are favored and selection operates to maintain extensive polymorphism in a population. However, in HLA, a reciprocally monophyletic relationship between different loci is observed (GU and NEI 1999 ) and this suggests infrequent exchanges between different loci. Further information is necessary to understand the molecular mechanism facilitating such frequent exchanges between different loci in the S-locus complex.

The role of gene conversion in SI gene diversity:
DIXIT et al. 2000 showed that a self-compatible mutant line in B. oleracea that lacked SLG synthesized a wild-type level of SRK transcripts but failed to produce the SRK protein, suggesting that SLG plays some role in stabilizing the SRK protein. KUSABA et al. 2001 found that self-incompatible A. lyrata has SRK but lacks SLG, suggesting the dispensability of SLG in SI. Our finding of three distinct SI haplotypes lacking SLG supports the latter view. On the other hand, it has been verified that SRK plays an essential role in self-recognition and SLG may enhance the process (TAKASAKI et al. 2000 ). The role of SRK and SLG in the SI recognition system must be different (CHARLESWORTH 2000 ; DICKINSON 2000 ). Therefore, it is likely that the evolutionary forces operating and the resulting patterns of nucleotide substitutions in the HVRs of SRK are different from those of SLG. However, in practice, a similar diversification pattern between SRK and SLG (Table 1) was observed. We suggest that this similar pattern is mainly due to gene conversion between SRK and SLG. Gene conversion might have occurred so frequently that it masked the natural evolutionary forces acting on SLG.

In disease resistance genes, gene conversion plays a role in maintaining paralogs and in generating new specificities (MICHELMORE and MEYERS 1998 ). In this study, we showed three examples of apparent gene conversion detected because of their high similarity in long stretches in the genes. Also, we observed 18 haplotypes in which SRK and SLG sequences are more closely related to each other than to alleles in different haplotypes (Fig 5). It has been suggested that the presence of SLG highly similar to SRK promotes strong SI (TAKASAKI et al. 2000 ). Gene conversion from SRK to SLG may help to maintain a strong SI phenotype.

In the analysis of BrS-47, gene conversion from SLG to SRK can be speculated. Replacement of SRK sequence with SLG sequence may change the recognition specificity in stigma and result in self-compatibility. The region from 457 to 729 in BrSRK-47, which is the putative converted region, contains HVR1. However, the HVR1 sequence in BrSRK-47 has only one synonymous nucleotide difference from BoSRK-12. The recognition specificity of BrSRK-47 was found to be the same as that of BoSRK-12 in our investigation (Y. SATO, R. FUJIMOTO, K. TORIYAMA and T. NISHIO, unpublished data), as shown between SRK-46 in B. rapa and SRK-7 in B. oleracea (KIMURA et al. 2002 ). These observations suggest that the gene conversion from BrSLG-47 to Br-SRK-47 may have happened but did not influence the recognition specificity of SRK. Alternatively, some mutation in SRK might have been repaired by the SLG sequence. Gene conversion may have played a role in resetting the variation between SRK and SLG.

	FOOTNOTES

Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AB054691–054751.
¹ Present address: Shirane High School, Kamiimasuwa 1180, Shirane, Yamanashi 400-0211, Japan.
³ Present address: Sakata Seed Co., Uchikoshi 358, Sodegaura, Chiba 299-0217, Japan.
⁴ Present address: National Institute of Vegetable and Tea Science, Kusawa 360, Ano, Age-gun, Mie 514-2392, Japan.
⁵ Present address: 7 Talbot Rd., Stratford-on-Avon, Warwick CV37 6SU, United Kingdom.

	ACKNOWLEDGMENTS

We thank Dr. D. Astley, HRI, United Kingdom, for providing plant materials. This work was supported in part by Grant-in-Aid for Special Research on Priority Areas (B)(11238202).

Manuscript received October 1, 2001; Accepted for publication July 26, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ADACHI, J., and M. HASEGAWA, 1994 Molphy Version 2.3: Programs for Molecular Phylogenetics Based on Maximum Likelihood.

AGUADE, M., 2001  Nucleotide sequence variation at two genes of the phenylpropanoid pathway, the FAH1 and F3H genes, in Arabidopsis thaliana.. Mol. Biol. Evol. 18:1-9.[Abstract/Free Full Text]

AWADALLA, P. and D. CHARLESWORTH, 1999  Recombination and selection at Brassica self-incompatibility loci. Genetics 152:413-425.[Abstract/Free Full Text]

CABRILLAC, D., V. DELOME, J. GARIN, V. RUFFIO-CHABLE, and J.-L. GIRANTON et al., 1999  The S₁₅ self-incompatibility haplotype in Brassica oleracea includes three S gene family members expressed in stigmas. Plant Cell 11:971-986.[Abstract/Free Full Text]

CHARLESWORTH, D., 2000  Unlocking the secrets of self-incompatibility. Curr. Biol. 10:184-186.

DICKINSON, H. G., 2000  Pollen stigma interactions: so near yet so far. Trends Genet. 16:373-376.[Medline]

DIXIT, R., M. N. NASRALLAH, and J. B. NASRALLAH, 2000  Post-transcriptional maturation of the S receptor kinase of Brassica correlates with co-expression of the S-locus glycoprotein in the stigmas of two Brassica strains and in transgenic tobacco plants. Plant Physiol. 124:297-311.[Abstract/Free Full Text]

FELSENSTEIN, J., 1993 PHYLIP (Phylogeny Inference Package), version 3.5c. Department of Genetics, University of Washington, Seattle.

GORING, D. R., T. L. GLAVIN, U. SCHAFER, and S. J. ROTHSTEIN, 1993  An S receptor kinase gene in self-compatible Brassica napus has a 1-bp deletion. Plant Cell 5:531-539.[Abstract]

GU, X. and M. NEI, 1999  Locus specificity of polymorphic alleles and evolution by a birth-and-death process in mammalian MHC genes. Mol. Biol. Evol. 16:147-156.[Abstract]

HINATA, K., M. WATANABE, S. YAMAKAWA, Y. SATTA, and A. ISOGAI, 1995  Evolutionary aspects of the S-related genes of the Brassica self-incompatibility system: synonymous and nonsynonymous base substitutions. Genetics 140:1099-1104.[Abstract]

HUELSENBECK, J. P., B. RANNALA, and Z. YANG, 1997  Statistical tests of host-parasite cospeciation. Evolution 51:410-419.

HUELSENBECK, J. P., B. RANNALA, and B. LARGET, 2000  A Baysian framework for the analysis of cospeciation. Evolution 54:352-364.[Medline]

KIMURA, M., 1968  Evolutionary rate at the molecular level. Nature 217:624-626.[Medline]

KIMURA, M., 1983 The Neutral Theory of Molecular Evolution. Cambridge University Press, Cambridge, UK.

KIMURA, R., K. SATO, R. FUJIMOTO, and T. NISHIO, 2002  Recognition specificity of self-incompatibility maintained after the divergence of Brassica oleracea and Brassica rapa. Plant J. 29:215-223.[Medline]

KISHINO, H. and M. HASEGAWA, 1989  Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in Hominoidea. J. Mol. Evol. 29:170-179.[Medline]

KUSABA, M. and T. NISHIO, 1999  Comparative analysis of S haplotypes with very similar SLG alleles in Brassica rapa and B. oleracea.. Plant J. 17:101-109.

KUSABA, M., T. NISHIO, Y. SATTA, K. HINATA, and D. OCKENDON, 1997  Striking sequence similarity in inter- and intra-specific comparisons of class I SLG alleles from Brassica oleracea and Brassica campestris: implications for the evolution and recognition mechanism. Proc. Natl. Acad. Sci. USA 94:7673-7678.[Abstract/Free Full Text]

KUSABA, M., K. DWYER, J. HENDERSHOT, J. VREBALOV, and J. B. NASRALLAH et al., 2001  Self-incompatibility in the genus Arabidopsis: characterization of the S locus in the outcrossing A. lyrata and its autogamous relative A. thaliana.. Plant Cell 13:627-643.[Abstract/Free Full Text]

MICHELMORE, R. W. and B. C. MEYERS, 1998  Clusters of resistance genes in plants evolve by divergent selection and a birth-and-death process. Genome Res. 8:1113-1130.[Abstract/Free Full Text]

NASRALLAH, J. B., S. M. YU, and M. E. NASRALLAH, 1988  Self-incompatibility genes of Brassica oleracea: expression, isolation, and structure. Proc. Natl. Acad. Sci. USA 85:5551-5555.[Abstract/Free Full Text]

NASRALLAH, J. B., T. NISHIO, and M. E. NASRALLAH, 1991  The self-incompatibility genes of Brassica: expression and use in genetic ablation of floral tissues. Annu. Rev. Plant Physiol. Plant Mol. Biol. 42:393-422.

NASRALLAH, J. B., S. J. RUNDLE, and M. E. NASRALLAH, 1994  Genetic evidence for the requirement of the Brassica S-locus receptor kinase gene in the self-incompatibility response. Plant J. 5:373-384.

NEI, M., and S. KUMAR, 2000 Molecular Evolution and Phylogenetics. Oxford University Press, New York.

NISHIO, T. and M. KUSABA, 2000  Sequence diversity of SLG and SRK in Brassica oleracea L. Ann. Bot. 85(Suppl. A):141-146.[Abstract/Free Full Text]

NISHIO, T., M. KUSABA, M. WATANABE, and K. HINATA, 1996  Registration of S alleles in Brassica campestris L by the restriction fragment sizes of SLGs. Theor. Appl. Genet. 92:388-394.

NISHIO, T., M. KUSABA, K. SAKAMOTO, and D. J. OCKENDON, 1997  Polymorphism of the kinase domain of the S-locus receptor kinase gene (SRK) in Brassica oleracea L. Theor. Appl. Genet. 95:335-342.

NOU, I. S., M. WATANABE, A. ISOGAI, and K. HINATA, 1993  Comparison of S-alleles and S-glycoproteins between two wild populations of Brassica campestris in Turkey and Japan. Sex. Plant Reprod. 6:79-86.

OCKENDON, D. J., 2000  The S-allele collection of Brassica oleracea.. Acta Hort. 539:25-30.

OKAZAKI, K., M. KUSABA, D. J. OCKENDON, and T. NISHIO, 1999  Characterization of S tester lines in Brassica oleracea: polymorphism of restriction fragment length of SLG homologues and isoelectric points of S-locus glycoproteins. Theor. Appl. Genet. 98:1329-1334.

ROGERS, S. O. and A. J. BENDICH, 1985  Extraction of DNA from milligram amounts of fresh, herbarium and mummified plant tissues. Plant Mol. Biol. 5:69-76.

SAITOU, N. and M. NEI, 1987  The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406-425.[Abstract]

SAKAMOTO, K., M. KUSABA, and T. NISHIO, 1998  Polymorphism of the S-locus glycoprotein gene (SLG) and the S-locus related gene (SLR1) in Raphanus sativus L. and self-incompatible ornamental plants in the Brassicaceae. Mol. Gen. Genet. 258:397-403.[Medline]

SCHOPFER, C. R., M. E. NASRALLAH, and J. B. NASRALLAH, 1999  The male determinant of self-incompatibility in Brassica.. Science 286:1697-1700.[Abstract/Free Full Text]

SHIERUP, M. H., B. K. MABLE, P. AWADALLA, and D. CHARLESWORTH, 2001  Identification and characterization of a polymorphic receptor kinase gene linked to the self-incompatibility locus of Arabidopsis lyrata.. Genetics 158:387-399.[Abstract/Free Full Text]

STEIN, J. C., B. HOWLETT, D. C. BOYES, M. E. NASRALLAH, and J. B. NASRALLAH, 1991  Molecular cloning of a putative receptor protein kinase gene encoded at the self incompatibility locus of Brassica oleracea. Proc. Natl. Acad. Sci. USA 88:8816-8820.[Abstract/Free Full Text]

SUZUKI, G., N. KAI, T. HIROSE, K. FUKUI, and T. NISHIO et al., 1999  Genomic organization of the S locus: identification and characterization of genes in SLG/SRK region of S⁹ haplotype of Brassica campestris (syn. rapa). Genetics 153:391-400.[Abstract/Free Full Text]

SUZUKI, T., M. KUSABA, M. MATSUSHITA, K. OKAZAKI, and T. NISHIO, 2000  Characterization of Brassica S-haplotypes lacking S-locus glycoprotein. FEBS Lett. 482:102-108.[Medline]

TAKASAKI, T., K. HATAKEYAMA, G. SUZUKI, M. WATANABE, and A. ISOGAI et al., 2000  The S receptor kinase determines self-incompatibility in Brassica stigma. Nature 403:913-916.[Medline]

TAKAYAMA, S., H. SHIBA, M. IWANO, H. SHIMOSATO, and F.-S. CHE et al., 2000  The pollen determinant of self-incompatibility in Brassica campestris. Proc. Natl. Acad. Sci. USA 97:1920-1925.[Abstract/Free Full Text]

TANTIKANJANA, T., M. E. NASRALLAH, and J. B. NASRALLAH, 1996  The Brassica S gene family: molecular characterization of the SLR2 gene. Sex. Plant Reprod. 9:107-116.

THOMPSON, J., D. G. HIGGINS, and T. GIBSON, 1994  CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673-4680.[Abstract/Free Full Text]

WATANABE, M., T. TAKASAKI, K. TORIYAMA, S. YAMAKAWA, and A. ISOGAI et al., 1994  A high degree of homology exists between the protein encoded by SLG and the S receptor domain encoded by SRK in self-incompatible Brassica campestris L. Plant Cell Physiol. 35:1221-1229.[Abstract/Free Full Text]

WATANABE, M., A. ITO, Y. TAKEDA, C. NINIMIYA, and T. KAKIZAKI et al., 2000  Highly divergent sequences of the pollen self-incompatibility (S) gene in class-I S haplotypes of Brassica campestris (syn. rapa) L. FEBS Lett. 473:139-144.[Medline]

YANG, Y. W., 2000 Phylogenetic Analysis by Maximum Likelihood (PAML), version 3.0. University College, London.

This article has been cited by other articles:

				S. Takuno, R. Fujimoto, T. Sugimura, K. Sato, S. Okamoto, S.-L. Zhang, and T. Nishio Effects of Recombination on Hitchhiking Diversity in the Brassica Self-incompatibility Locus Complex Genetics, October 1, 2007; 177(2): 949 - 958. [Abstract] [Full Text] [PDF]

				J. S. Bechsgaard, V. Castric, D. Charlesworth, X. Vekemans, and M. H. Schierup The Transition to Self-Compatibility in Arabidopsis thaliana and Evolution within S-Haplotypes over 10 Myr Mol. Biol. Evol., September 1, 2006; 23(9): 1741 - 1750. [Abstract] [Full Text] [PDF]

				K. L. Bubb, D. Bovee, D. Buckley, E. Haugen, M. Kibukawa, M. Paddock, A. Palmieri, S. Subramanian, Y. Zhou, R. Kaul, et al. Scan of Human Genome Reveals No New Loci Under Ancient Balancing Selection Genetics, August 1, 2006; 173(4): 2165 - 2177. [Abstract] [Full Text] [PDF]

				R. Fujimoto, K. Okazaki, E. Fukai, M. Kusaba, and T. Nishio Comparison of the Genome Structure of the Self-Incompatibility (S) Locus in Interspecific Pairs of S Haplotypes Genetics, June 1, 2006; 173(2): 1157 - 1167. [Abstract] [Full Text] [PDF]

				D. Charlesworth, E. Kamau, J. Hagenblad, and C. Tang Trans-specificity at Loci Near the Self-Incompatibility Loci in Arabidopsis Genetics, April 1, 2006; 172(4): 2699 - 2704. [Abstract] [Full Text] [PDF]

				Y. Sato, K. Sato, and T. Nishio Interspecific Pairs of Class II S Haplotypes Having Different Recognition Specificities between Brassica oleracea and Brassica rapa Plant Cell Physiol., March 1, 2006; 47(3): 340 - 345. [Abstract] [Full Text] [PDF]

				N. L Clark, J. E Aagaard, and W. J Swanson Evolution of reproductive proteins from animals and plants Reproduction, January 1, 2006; 131(1): 11 - 22. [Abstract] [Full Text] [PDF]

				S. Glemin, T. Gaude, M.-L. Guillemin, M. Lourmas, I. Olivieri, and A. Mignot Balancing Selection in the Wild: Testing Population Genetics Theory of Self-Incompatibility in the Rare Species Brassica insularis Genetics, September 1, 2005; 171(1): 279 - 289. [Abstract] [Full Text] [PDF]