Abstract
Self-incompatibility (SI) is a mechanism for preventing self-fertilization in flowering plants. In Brassica, it is controlled by a single multi-allelic locus, S, and it is believed that two highly polymorphic genes in the S locus, SLG and SRK, play central roles in self-recognition in stigmas. SRK is a putative receptor protein kinase, whose extracellular domain exhibits high similarity to SLG. We analyzed two pairs of lines showing cross-incompatibility (S2 and S2-b; S13 and S13-b). In S2 and S2-b, SRKs were more highly conserved than SLGs. This was also the case with S13 and S13-b. This suggests that the SRKs of different lines must be conserved for the lines to have the same self-recognition specificity. In particular, SLG2-b showed only 88.5% identity to SLG2, which is comparable to that between the SLGs of different S haplotypes, while SRK2-b showed 97.3% identity to SRK2 in the S domain. These findings suggest that the SLGs in these S haplotypes are not important for self-recognition in SI.
MANY flowering plants show self-incompatibility (SI), which prevents self-fertilization. In Brassicaceae, pollen tube development from self-pollen is specifically inhibited on the stigma (Nasrallah and Nasrallah 1993). This inhibition is controlled by a single multi-allelic locus, S. The S locus in fact comprises a number of genes, including SLG (the S-glycoprotein gene) and SRK (the S-locus receptor kinase gene). SRK is a highly polymorphic gene that encodes a putative transmembrane receptor protein kinase. SLG shows high similarity to the putative ligand-binding domain of SRK (the S domain). SLG and the S domain of SRK have basically the same structure, including 12 conserved cysteine residues and three regions that are hypervariable between different S haplotypes (Kusabaet al. 1997).
It is believed that SLG and SRK play central roles in the recognition event in stigmas. Both SLG and SRK are expressed predominantly in stigmas just before flowering, the stage of expression of SI. Matton et al. (1997) suggested that the hypervariable regions of S-RNase, a determinant of the self-recognition specificity in the style of the Solanaceae (Leeet al. 1994; Murfettet al. 1994), are responsible for the determination of the specificity. Similarly, the hypervariable regions of SLG and SRK may be determinants of self-recognition in Brassica. An unknown pollen gene in the S locus, encoding a ligand for SLG and SRK, is thought to function in the recognition event. A physiological experiment has suggested that the pollen product is a small molecule in the pollen coating (Stephensonet al. 1997). In this context, the hypervariable regions of SLG and SRK may be ligand binding sites.
SLG and SRK alleles are assigned, on the basis of sequence similarity, to class I and class II. Class I SLGs and SRKs exhibit ~65% similarity in amino acid sequence to class II SLGs and SRKs (Nasrallahet al. 1991), compared to the ~80% similarity typical of within-class comparisons. Class I S haplotypes, which comprise a class I SLG and a class I SRK, constitute most of the known Brassica S haplotypes (Hatakeyamaet al. 1998; Okazakiet al. 1999). All class II S haplotypes are recessive in pollen to all class I S haplotypes and are thought to express weak SI. These characteristics of class II S haplotypes are considered to be related to a structural feature unique to a class II SLG, the transmembrane domain (Tantikanjanaet al. 1993).
Several observations are consistent with the view that both SLG and SRK function in SI. Transgenic experiments showed that transformation with anti-sense SLG (Shibaet al. 1995) or sense-cosuppression of SLG or SRK (Conneret al. 1997; Stahlet al. 1998) reduced expression of SLG and SRK and caused conversion to the self-compatible phenotype. Nasrallah et al. (1992) demonstrated that the expression of SLG, but not SRK, is reduced in a self-compatible mutant, while another self-compatible line was found to have a nonfunctional SRK and a normal SLG (Nasrallahet al. 1994a). However, some contradictory observations also have been reported. Kusaba et al. (1997) and Kusaba and Nishio (1999) showed that some S haplotypes with very similar SLG alleles do not have the same self-recognition specificity. Gaude et al. (1995) showed that the strength of SI is not related to the expression level of SLG in the class II haplotypes. Furthermore, the S24 haplotype of B. oleracea shows normal SI even though it seems to be lacking SLG (Okazakiet al. 1999). These observations challenge the view that SLG is one of the key molecules in self-recognition (Nasrallahet al. 1994b).
We identified a broccoli line that is incompatible with the S2 line in the B. oleracea S tester lines, a standard collection for S haplotypes (Braceet al. 1994). The DNA banding pattern of this line using a class II SLG probe clearly differed from the S2 line and was designated S2-b (Okazakiet al. 1999). S2-b is found to be widely distributed in commercial cultivars of broccoli and cabbage (Sakamotoet al. 1999). Furthermore, an S13 haplotype, which showed a banding pattern different from the S13 haplotype of the S tester lines, was also found in broccoli cultivars and designated S13-b (Okazakiet al. 1999). S2 and S2-b have different PCR-restriction fragment length polymorphisms (RFLPs) in the SLG alleles (Sakamotoet al. 1999), suggesting sequence differences in SLG; S13, and S13-b also have different SLG PCR-RFLPs. In this article, we describe sequence differences of SLG and SRK between S2 and S2-b and between S13 and S13-b and report that among S haplotypes with the same self-recognition specificity, SLG has accumulated more amino acid substitutions than the S domain of SRK. The implications of these results for the recognition mechanism of Brassica SI are discussed.
MATERIALS AND METHODS
Plant materials and pollination analysis: The S2 and S13 S tester lines of B. oleracea were kindly provided by Dr. D. Ockendon. The S2-b and S13-b lines were derived from selfed progeny of cv. “Marimidori.” Pollen tube development in stigmas after pollination was observed as described in Nakanishi and Hinata (1973). To examine the ability to set selfed seeds, seven open flowers from one individual were selfed and the resulting seeds were counted 1 month after pollination. In this experiment, the S2 S tester line and an S2-b line of kale were used (Okazakiet al. 1999).
Isolation of genomic and cDNA clones: Genomic DNA was isolated from young leaves according to Rogers and Bendich (1985). Genomic libraries for S13, S13-b, and S2-b were constructed on λ Dash II (Stratagene, La Jolla, CA) according to Kusaba and Nishio (1999). The genomic clones for SRK13 and SRK13-b were isolated by a PCR-based method (Kusaba and Nishio 1999). All PCR experiments were conducted under the conditions described by Nishio et al. (1996). An SRK-specific primer set, PK1 and PK4 (Nishioet al. 1997), was used for the screening. Poly(A)+ RNA was isolated from stigmas 1 or 2 days before flowering using a Fast-track kit (Promega, Madison, WI), and a stigma cDNA library of S2-b was constructed on λZAP (Stratagene) according to the manufacturer's instructions. The genomic and cDNA clones of SLG2-b and SRK2-b were isolated by a hybridization-based method using a Dig-labeled PCR product amplified from the S2-b haplotype with the PS3 and PS21 primer set (Nishioet al. 1996). This PCR product corresponds to SLG2-b (see text). SLA, an anther-expressed S-locus gene, was amplified, using primers CAAGCGCCCGCAAAGCAGGAAAA and CAGTACATGACCAAATCGACATC, from genomic DNA of the S2 homozygote S tester line and cloned into the pCR2 vector using a TA cloning kit (Invitrogen, Carlsbad, CA).
DNA and protein gel blot analysis: DNA gel blot analyses were carried out as described by Kusaba and Nishio (1999), using the Dig labeling and detection system (Boehringer Mannheim, Indianapolis) using CSPD as a substrate or the ECL system (Amersham Pharmacia Biotech, Piscataway, NJ). Protein gel blot analysis was performed as described by Okazaki et al. (1999). Protein was extracted from 30 stigmas with 0.15 ml of 0.02 m phosphate buffer, pH 7.0, and, after centrifugation, glycerol was added to the supernatant to a final concentration of 5%. An aliquot of the supernatant (15 μl) was subjected to nonequilibrium pH gradient electrophoresis. After transfer to a PVDF membrane, SLG and SLG-like proteins were detected with anti-SLG22 antiserum.
DNA sequencing and sequence analysis: Sequencing was carried out by the dye-terminator method using PRISM 377 (Perkin-Elmer, Norwalk, CT) as described by Kusaba and Nishio (1999). SLG13 and SLG13-b were amplified with the class I-SLG-specific primer set, PS22 and PS15 (Sakamotoet al. 1998). The PCR products were purified with a QIAquick PCR purification kit (QIAGEN) and their DNA sequences were directly determined. Sequence analyses were performed using Genetyx v. 10.0 (Software Kaihatsu, Tokyo). Sequences reported in this article correspond to DDBJ accession nos. AB024415–AB024422.
RESULTS
Characterization of the S2-b and S13-b haplotypes by test crossing: Class I S haplotype S13 is dominant in pollen and codominant in stigmas to a class II S haplotype S2 (Thompson and Taylor 1966). S13-b was incompatible with S13 in reciprocal pollinations, dominant to S2 in pollen, and codominant with S2 in stigmas (Figure 1, A and B), suggesting that S13-b has the same dominance-recessiveness relationship as S13 of the S tester lines. Like S2 of the S tester lines, S2-b was incompatible with S2 in reciprocal pollinations and was found to be recessive to S13-bin pollen and codominant with S13-b in stigmas (Figure 1, C and D). To compare the strength of SI of the S2-b haplotype to that of the S2 haplotype, the ability of the plants to set selfed seeds was examined. S2 homozygotes set 0.71 seeds/pod by selfing and S2-b homozygotes set 1.0 seeds/pod, suggesting that the strengths of their SIs are comparable.
Isolation of SLG and SRK clones from the S13 and S13-b haplotypes: An SRK13 genomic clone was isolated from a genomic library constructed from the leaves of S13 homozygotes derived from the S tester lines (Figure 2A). An SRK13-b genomic clone was isolated from a genomic library constructed from the leaves of S13-b homozygotes derived from selfed progeny of the broccoli F1 hybrid cultivar Marimidori (Figure 2A). Both clones contained the entire coding region of SRK gene and showed similar restriction maps. SLG13 and SLG13-b were amplified by genomic PCR using a class I SLG-specific primer set, PS22 and PS15 (Sakamotoet al. 1998). Both SLGs have the 12 conserved cysteine residues (Figure 2C). Our SLG13 sequence is slightly different from the published SLG13 sequence (Dwyeret al. 1991): there is one amino acid substitution in the signal sequence and one in hypervariable region I. These differences probably reflect genetic differences in the lines studied. A DNA gel blot analysis using the SLG13 probe of an F2 population segregating for S13-b and S2-b revealed two HindIII fragments, which were determined by test crossing to be perfectly linked to the S13-b genotype (Figure 3A). The two bands were thought to correspond to SLG13-b and SRK13-b. The PCR-amplified product of SLG13-b was also perfectly linked to the S13-b genotype in 18 segregating plants (Figure 3B), confirming that the PCR products were derived from SLG13-b. In a previous study (Okazakiet al. 1999), a DNA gel blot analysis of S13 showed only two bands detected after HindIII digestion; those bands apparently correspond to SLG13 and SRK13.
Dominance-recessiveness relationship between S2, S2-b, and S13-b. Test crossing between (A) the S13 and S13-b, (B) the S13-b and S2, (C) the S2 and S2-b, and (D) the S2-b and S13-b haplotypes. In experiments A and C, the S25 (class I) haplotype was used as a control. −, incompatible; +, compatible.
Sequence comparison of SLG and SRK of the S13 and S13-b haplotype: Both SRK13 and SRK13-b showed high similarity to the class I SRKs reported so far, including the 12 conserved cysteine residues in the S domain and the conserved amino acid residues in their kinase domain, which are important for kinase activity (Figure 2B). These characteristics suggest that both SRK13 and SRK13-b are functional alleles. The S domains of SRK13 and SRK13-b exhibited a very high similarity in their amino acid sequence (99.8%). There was only one amino acid difference in hypervariable region II. On the other hand, the signal sequences and the transmembrane-kinase domains showed lower amino acid identities (96.8 and 97.6%, respectively). The much higher similarity in the S domain suggests that maintenance of the same self-recognition specificity requires conservation of the S domain.
SLG13 and SLG13-b also showed high similarity to class I SLGs that have been reported so far, including the 12 cysteine residues (Figure 2C). Protein gel blot analysis of soluble stigma proteins using an anti-SLG22 antibody revealed that the positions of S-haplotype-specific bands for S13-b are the same as those for S13 (Figure 4). Consistent with this observation, the pIs estimated from the deduced amino acid sequences in the mature protein region (pI 8.12) and potential N-linked glycosylation sites are identical in SLG13 and SLG13-b. However, direct sequencing of SLG13 and SLG13-b revealed a number of nonsynonymous differences (98.3% identity in amino acid sequence in the mature protein region), particularly in hypervariable region I, which is thought to be involved in the determination of self-recognition specificity: 6 of 11 amino acid residues in hypervariable region I are substituted.
The base substitutions observed between SLG13 and SLG13-b were clustered in hypervariable region I (data not shown). Interestingly, on the nucleotide sequence level, SLG13-b in this region is identical to SRK13-b and very similar to SRK13 (data not shown), suggesting that the evolution of S13 and S13-b has involved gene conversion between SLG and SRK or recombination between different SLG alleles. This is consistent with previous evidence that gene conversion between and recombination in SLG and SRK alleles has occurred in the evolution of S haplotypes (Goringet al. 1993; Kusabaet al. 1997; Suzukiet al. 1997; Kusaba and Nishio 1999).
Isolation of SLG and SRK clones of the S2-b haplotype: A genomic library was constructed from leaves of S2-b homozygotes derived from the selfed progeny of broccoli cv. Marimidori. The PCR product amplified from S2-b with class II-specific primer set PS3 and PS21 was used as a probe to screen the library. Three classes of clones with sequences highly similar to the class II SLG alleles were isolated. On the basis of nucleotide sequence similarity (97.8%; data not shown), the first class is thought to be SLR2, an SLG-like gene unlinked to the S locus (Boyeset al. 1991). However, clones within this class presumably encode only a truncated protein, due to a frame-shifting 14-bp deletion in the coding region. These observations suggest that the clones represent a nonfunctional allele of SLR2. The second class is identical in nucleotide sequence to the PCR product amplified with PS3 and PS21 (data not shown) and was designated as GS2b-1. The third class was designated as GS2b-2.
A DNA gel blot analysis of the selfed progeny of broccoli cv. Marimidori using the GS2b-1 clone as a probe revealed two HindIII bands (5.8 and 6.8 kb) that were determined by test crossing to be completely linked to the S2-b genotype (Figure 3C). The GS2b-1 clones contain a 5.8-kb HindIII fragment and GS2b-2 clones contain a 6.8-kb HindIII fragment, both of which hybridized to the class II SLG probe (Figure 5A). This result indicates that the two bands in the DNA gel blot correspond to GS2b-1 and GS2b-2. The PCR product amplified by the primer set of PS3 and PS21, corresponding to GS2b-1, was also perfectly linked to the S2-b genotype (Figure 3D). The 8.5-kb band, which is common to all segregating plants, was thought to correspond to SLR2. cDNA clones corresponding to the GS2b-1 and GS2b-2 genomic clones were isolated from an S2-b stigma library using the GS2b-1 probe. Sequence analysis of the clone corresponding to GS2b-2 revealed that it has a kinase domain, suggesting that GS2b-2 is SRK. This gene was designated as SRK2-b. The cDNA clone corresponding to GS2b-1 did not have a kinase domain, suggesting that it encodes SLG of the S2-b haplotype (SLG2-b).
Sequence diversity of SLG and SRK between the S13 and S13-b haplotypes. (A) Restriction maps of the SRK13 and SRK13-b genomic clones. Thick lines indicate exons. H, HindIII; X, XhoI. (B) Comparison of the amino acid sequences of SRK13 and SRK13-b. (C) Comparison of the amino acid sequences of SLG13 and SLG13-b. Boxed N residues show the positions of potential N-linked glycosylation sites. Solid circles show the positions of conserved cysteine residues. The three larger boxes represent the three hypervariable regions (I, II, and III): hypervariable regions I and III are separated into two parts by conserved amino acid residues. The signal sequence and the transmembrane domain are underlined. Open circles show the positions of the conserved amino acid residues among the protein kinases. Closed circles indicate identical amino acid residues.
Linkage analysis of the SLG and SRK alleles of the S2-b and S13-b haplotypes. (A) DNA gel blot analysis of selfed progeny of cv. Marimidori segregating for S2-b and S13-b using the SLG13-b PCR product as a probe. Genomic DNA (5 μg/lane) was digested by HindIII and separated on a 0.8% agarose gel. The DNA was detected by the Dig detection system. The hybridization and wash were carried out at 68°. The S genotype determined by pollination analysis is shown above each lane. (B) PCR analysis of the same plants using the class I-SLG-specific primer set, PS22 and PS15. (C) The same blot used in A was reprobed with the SLG2-b PCR product under the same conditions as in A. (D) PCR analysis of the same plants using the class II-specific primer set, PS3 and PS21.
Sequence and structural diversity of SLG and SRK in S2 and S2-b: While class I SLG alleles have no intron and produce only soluble proteins, SLG2 has an intron and a second exon, which encodes a transmembrane domain (Tantikanjanaet al. 1993). SLG2 produces alternative mRNA molecules: the spliced mRNA produces a membrane-anchored protein and the unspliced mRNA a soluble protein. Sequence analyses of the genomic and cDNA clones of SLG2-b revealed that like SLG2, the SLG2-b gene also contains an intron and a second exon; however, the second exon cannot encode a transmembrane domain as it encodes only four amino acid residues (Figure 5, A and B).
Protein gel blot analysis of water-soluble stigma proteins from the S2, S2-b, S13, and S13-b haplotypes. Stigma proteins were separated by nonequilibrium pH gradient electrophoresis and detected with an anti-SLG22 antibody. The non-S-haplotype specific bands are thought to be derived from SLG-like genes unlinked to the S locus, such as SLR1 (Lalondeet al. 1989). +, anode; −, cathode.
In addition to these structural differences, SLG2 and SLG2-b showed significant sequence differences in the first exon. Amino acid sequence identity between SLG2-b and SLG2 was only 88.5% in the mature protein region and a number of differences were observed in the hypervariable regions (Figure 5B). The divergence between SLG2 and SLG2-b is comparable to that between SLGs of different S haplotypes. For example, SLG2 and SLG2-b exhibit 88.7 and 92.9% identity to SLG5 (Scutt and Croy 1992), respectively. Protein gel blot analysis revealed multiple S-haplotype-specific bands in the S2-b haplotype, unlike S2, which shows a single band (Figure 4). All of these bands are thought to be SLG. SLG2-b was produced at a level much higher than that of SLG2. Low production of SLG by the S2 haplotype was also reported by Gaude et al. (1995).
SRK2-b has retained the 12 conserved cysteine residues in the S domain and the conserved amino acid residues important for protein kinase activity in the kinase domain (Figure 5C). In the S domain, SRK2 and SRK2-b showed 97.3% identity. Hypervariable regions I and II were identical and hypervariable region III had one amino acid substitution at the end of the region. In the transmembrane-kinase domain, SRK2 and SRK2-b showed a similarity slightly lower (96.1%) than that in the S domain. The much higher similarity in the SRKs than in the SLGs indicates that SI recognition specificity is less sensitive to amino acid substitutions in SLG than in SRK. This suggests that SLG is not as important as SRK for the determination of self-recognition specificity.
Absence of SLA in the S2-b haplotype: SLA is an anther-expressed S-locus gene specific to the S2 haplotype (Boyes and Nasrallah 1995) that was once thought to be a candidate for the pollen ligand gene. A DNA gel blot analysis was performed using SLA as a probe (Figure 6). No signal was detected in the class I haplotypes (S12, S13-b, S18) or the S5 (class II) haplotype. Unexpectedly, no band was detected in the S2-b haplotype. It is considered that S2-b haplotype does not harbor a sequence that is highly similar to that of SLA. The absence of SLA and the transmembrane-anchored SLG in the S2-b haplotype indicates that the S2-b haplotype has features that are quite different from the S2 haplotype in the S tester lines.
DISCUSSION
Uniqueness of the S2 haplotype of B. oleracea among the class II S haplotypes: Among Brassica S haplotypes, the S2 haplotype of B. oleracea has been intensively investigated and has been regarded as an exemplar of the class II S haplotypes. It has an SLG allele with a second exon encoding a transmembrane domain and SLA, an anther-expressed S locus gene (Tantikanjanaet al. 1993; Boyes and Nasrallah 1995). Certain genetic and physiological features of the S2 haplotype, including its recessiveness and relative weakness of SI, have been attributed to the existence of the membrane-anchored SLG. In this study, we have shown that the S2-b haplotype, which is incompatible in reciprocal crosses with the S2 haplotype, does not encode an SLG having a transmembrane domain, although its genetic and physiological properties are similar to those of the S2 haplotype: both are recessive to the class I haplotype S13-b in pollen and codominant with it in stigmas, and the difference between S2-b and S2 in the rate of seed set by selfing in homozygotes was not significant. These results suggest that the membrane-anchored SLG neither influences the dominance/recessiveness relationship nor has a remarkable effect on the strength of SI. This is consistent with a recent report that S5, which is recessive and shows relatively weak SI, also lacks a membrane-anchored SLG (Gaudeet al. 1995; Cabrillacet al. 1999).
Because S2 and S2-b have the same self-recognition specificity, their self-recognition genes in pollen would be expected to show high similarity to each other. DNA gel blot analysis using SLA as a probe revealed that the S2-b haplotype does not possess any gene showing high homology to SLA, suggesting that SLA is not involved in SI recognition. Pastuglia et al. (1997) showed that self-incompatible lines of B. oleracea var. botrytis as well as the self-compatible p57Sc (a pollen-part mutant of the S15 haplotype; Cabrillacet al. 1999) have a nonfunctional SLA allele (interrupted by a retrotransposon) and claimed that SLA is not involved in SI. Our observation supports their claim.
Sequence diversity of SLG and SRK between the S2 and S2-b haplotypes. (A) Restriction maps of GS2b-1 (SLG2-b) and GS2b-2 (SRK2-b). Thick lines indicate exons. H, HindIII; B, BamHI. (B) Comparison of the amino acid sequences of SLG2 and SLG2-b. (C) Comparison of the amino acid sequences of SRK2 and SRK2-b. For explanation of symbols, see Figure 3. Arrowheads in B show the positions of introns.
With respect to structural features, SLG2-b is less similar to SLG2 than it is to SLG5 of B. oleracea (Scutt and Croy 1992) and SLG40 and SLG44 of B. rapa (Hatakeyamaet al. 1998): those genes also lack a transmembrane domain, and their C-terminal amino acid sequences encoded in the second exons are identical. Other than S2, only the S15 haplotype and its pollen-part mutant P57Sc possibly produce a membrane-anchored SLG, which was designated SLGA15 (Pastugliaet al. 1997; Cabrillacet al. 1999). However, the very high sequence similarity of SLGA15 to SLG2 (99.0% in nucleic acid sequence) and the existence of an SLA-like sequence may indicate that SLGA15 and the SLA-like sequence were recently transmitted from the S2 haplotype as a segment. Furthermore, the S15 haplotype has another SLG gene (SLGB15) (Cabrillacet al. 1999), suggesting that SLGB15 could represent the original SLG. Interestingly, SLGB15 also has the same four amino acid residues as SLG2-b in its C-terminal end. The unique features of S2, particularly the existence of a transmembrane domain and SLA, indicate that the S2 haplotype is not a typical class II S haplotype.
Is SLG important for self-recognition specificity? It has generally been believed that SLG and SRK play central roles in self-recognition in stigmas. However, accumulating data question the role of SLG in SI. Gaude et al. (1995) showed that the amount of SLG was not correlated with seed set by self-pollination of class II homozygotes, a measure of the strength of SI. On the basis of protein and DNA gel blot analyses, Okazaki et al. (1999) suggested that SLG was deleted in the S24 haplotype of B. oleracea in spite of its normal expression of SI. Furthermore, some different S haplotypes have very similar SLGs while their SRKs show much lower similarity (Kusabaet al. 1997; Kusaba and Nishio 1999). For example, SLG23 and SLG29 of B. oleracea show 99.5% identity and have identical hypervariable regions, but the S domains of SRK23 and SRK29 exhibit only 87.9% identity in amino acid sequence. In this article, we demonstrated that SLG varies much more than SRK between different lines that share the same self-recognition specificity. While the amino acid sequences responsible for self-recognition specificity are expected to be conserved among such lines, the SLGs of S2/S2-b and S13/S13-b have accumulated a number of amino acid substitutions in the hypervariable regions. In particular, SLG2-b showed only 88.5% amino acid identity to SLG2, which is comparable to that between SLGs of different S haplotypes. These results suggest that SLG is not important for recognition of SI specificity.
Absence of SLA in the S2-b haplotype. DNA gel blot analysis using SLA as a probe. Genomic DNA (5 μg/lane) was digested by HindIII and separated on a 0.8% agarose gel. The blot was detected by the ECL detection system. The hybridization and wash were carried out at 42° in hybridization buffer containing 6 m urea.
On the other hand, Nasrallah et al. (1992) demonstrated that the trans-acting scf mutation, which causes self-compatibility, reduces the expression of SLG, but not SRK. However, expression of SLR1 and SLR2 was also reduced in the scf mutant. This suggests that the self-compatible phenotype could have been caused by the reduced expression of other genes regulated by SCF. Other evidence for the function of SLG in SI is given by transgenic experiments. A sense cosuppression experiment (Conneret al. 1997) and an anti-sense transgenic experiment (Shibaet al. 1995) using SLG seemed to indicate that reduced expression of SLG causes the self-compatible phenotype. However, the expression of SRK might also have been reduced in the transgenic plants because of the high homology between SLG and SRK. Thus, while these transgenic experiments do not unequivocally demonstrate that SLG is essential for SI, they do demonstrate that one or more genes highly similar to SLG are involved in SI. Involvement of SRK in SI is consistent with other observations as well: a self-compatible mutant was shown to have a nonfunctional SRK (Nasrallahet al. 1994a), and introduction of a nonfunctional SRK transgene caused partial self-compatibility without reducing SLG or SRK expression (Stahlet al. 1998). Our observations that the amino acid sequence of the S domain of SRK is highly conserved between S haplotypes with the same recognition specificity suggest that the S domain is important for recognition. Consistent with this view, we have not found within the same species any distinct SRK alleles that show high similarity in the S domain (24 S haplotypes in B. oleracea; T. Nishio, T. Suzuki and M. Kusaba, unpublished results), unlike the case for SLG alleles or the kinase domains of SRK alleles (Kusaba and Nishio 1999).
In the present investigation, it was suggested that the S domain of SRK, but not SLG, is important for recognition in SI. From the observation that the S15 haplotype has two distinct SLG genes, Cabrillac et al. (1999) suggested that the two SLG genes are redundant or that they are not required for recognition in SI. Our observation favors the latter interpretation. Luu et al. (1999) recently suggested that SLG and SLR1 are involved in pollen adhesion to the surface of the stigma. This means that SLG might have a function more general than determination of self-recognition specificity. In any case, further evidence is required to demonstrate that SLG is not essential to SI.
Acknowledgments
We thank D. Ockendon and D. Astley for providing plant materials, M. E. Nasrallah and J. B. Nasrallah for providing anti-SLG antiserum, and M. Uyenoyama for her suggestions for improvement of our manuscript. This work was supported by a grant from the Science and Technology Agency of Japan and in part by a Grant-in-Aid (Bio Design Program) from the Ministry of Agriculture, Forestry and Fisheries.
Footnotes
-
Communicating editor: M. K. Uyenoyama
- Received May 24, 1999.
- Accepted September 10, 1999.
- Copyright © 2000 by the Genetics Society of America
