Abstract
In Brassica species that exhibit self-incompatibility, two genes, SLG and SRK, at the S locus are involved in the recognition reaction with self and non-self pollen. From a pollen-recessive S29 haplotype of Brassica rapa, both cDNA and genomic DNA clones for these two genes were isolated and characterized. The nucleotide sequence for the S domain of SRK29 showed a high degree of similarity with that of SLG29, and they belong to Class II type. RNA gel blot analysis showed that the transcript of SLG29 consisted of the first and second exons, and no other transcript containing any part of the intron sequence was detected. Because no transmembrane domain was encoded by the second exon of SLG29, SLG29 was designated a secreted type glycoprotein. SLGs of two other pollen-recessive haplotypes, S40 and S44, of B. rapa also had a similar structure to that of SLG29. Previously, SLG2 from a pollen-recessive haplotype, S2, of Brassica oleracea was found to produce two different transcripts, one for the secreted type glycoprotein and the other for a putative membrane-anchored form of SLG. Therefore, the nature of these SLGs from pollen-recessive haplotypes of B. rapa is different from that of SLG2 of B. oleracea.
SELF-INCOMPATIBILITY is a mechanism by which many flowering plants prevent self-fertilization and promote outbreeding. The self-incompatibility system in Brassica is controlled sporophytically by alleles at a single locus called the S locus (Bateman 1955). More than 30 S alleles have been identified in Brassica rapa (Nouet al. 1993) and more than 40 S alleles in Brassica oleracea (Ockendon 1974). Because the activity of S alleles is controlled sporophytically, codominance and dominance relationships influence the ultimate phenotype of stigma and pollen. The following observations have been made about the dominance relationships between S alleles: (1) Codominance is common; (2) dominance/recessiveness is frequent in pollen; (3) dominance relationships are different between stigma and pollen; and (4) dominance relationships are nonlinear (Thompson and Taylor 1966; Ockendon 1975; Visseret al. 1982; Hatakeyamaet al. 1998).
Results from molecular analyses have revealed that the Brassica S locus consists of at least two physically linked genes expressed in stigma papillae (Boyes and Nasrallah 1993). One of the genes is the S locus glycoprotein (SLG) gene, which encodes a secreted glycoprotein (Nasrallahet al. 1987; Takayamaet al. 1987), and the other is the S locus receptor kinase (SRK) gene, which encodes a receptor protein kinase with an extracellular domain (S domain) that shares a high degree of sequence similarity with SLG (Steinet al. 1991; Watanabeet al. 1994). Several lines of evidence suggest that these two genes are involved in the recognition reaction of self-incompatibility (Shibaet al. 1995; Conneret al. 1997). Because the S locus contains multiple genes, S alleles are referred to as S haplotypes (Nasrallah and Nasrallah 1993). In addition, there are three SLG-related genes (SLR1, SLR2, and SLR3), which are unlinked to the S locus but show sequence similarity with SLG (Lalondeet al. 1989; Trick and Flavell 1989; Scuttet al. 1990; Boyeset al. 1991; Isogaiet al. 1991; Watanabeet al. 1992; Yamakawaet al. 1993; Cocket al. 1995; Tantikanjanaet al. 1996; Watanabeet al. 1997). SLR2 shows a high degree of sequence similarly to SLGs isolated from pollen-recessive haplotypes (Tantikanjanaet al. 1996; Watanabeet al. 1997).
Based on the degree of sequence similarity among SLGs and dominance relationships among their corresponding S haplotypes, Nasrallah and Nasrallah (1993) have classified SLGs into two groups, Class I and Class II. The SLGs in Class I all correspond to haplotypes that show dominance in pollen, whereas Class II SLGs correspond to haplotypes that show recessiveness. Molecular investigations on S2 haplotype in Class II have revealed that the SLG2 gene produces two transcripts that differ at their 3′ ends (Tantikanjanaet al. 1993). One transcript encodes the expected secreted glycoprotein, while the other encodes a putative membrane-anchored form that is not found in Class I S6 haplotype. This unusual feature of the SLG of the S2 haplotype is thought to be related to its pollen-recessive nature (Chen and Nasrallah 1990; Steinet al. 1991; Tantikanjanaet al. 1993). So far, SLG2 is the only SLG gene from a Class II haplotype that has been studied in detail, although SLG and SRK genes from several Class I haplotypes of B. rapa have been studied (Watanabeet al. 1994; Yamakawaet al. 1994; Yamakawaet al. 1995; Suzukiet al. 1995; Matsushitaet al. 1996) and of B. oleracea (Steinet al. 1991; Delormeet al. 1995).
In a previous study, we determined the dominance relationships between 24 S haplotypes of B. rapa by examining pollen tube behavior in diallel crosses. We have classified these S haplotypes into codominant (CD), dominant/recessive (DR), and recessive (R) groups on the pollen side (Hatakeyamaet al. 1998). S haplotypes belonging to the CD group are codominant to each other. The members of the DR group are either codominant or recessive to those of the CD group as well as to others within the DR. The members of the R group are generally recessive to those of the other two groups. Here, we report a study of a set of SLG and SRK genes from one of the pollen-recessive haplotypes (R), S29, of B. rapa to determine whether SLG and SRK genes from a pollen-recessive haplotype exhibit any unique features.
MATERIALS AND METHODS
Plant materials and determination of S genotype: All S homozygous lines of B. rapa L used in this study were described by Hatakeyama et al. (1998). The F2 plants used for RFLP analysis were obtained from an F1 heterozygous individual by bud pollination. The S genotype of the F2 progeny was determined by reciprocal test-pollinations between each F2 plant and the two parental lines (Hatakeyamaet al. 1998).
Reverse-transcriptase PCR (RT-PCR) and cloning of PCR products: Stigmas of S29, S40, and S44 homozygotes were collected from buds at 2–3 days before anthesis. Poly(A)+RNA was isolated with a Fast Track mRNA Isolation Kit (Invitrogen, San Diego, CA). The first strand cDNA was synthesized from 1 μg of poly(A)+RNA using the NotI-(dT)18 adapter primer with the T-Primed First Strand Kit (Pharmacia LKB, Uppsala, Sweden). PCR reactions contained total cDNAs in 100 μl with Class II-specific primer (PS3; Nishioet al. 1996) corresponding to the 5′ terminal region of the SLG2 ORF (Chen and Nasrallah 1990), the adapter primer, and Taq DNA polymerase (TaKaRa Shuzo, Shiga, Japan). PCR was performed as described in Nishio et al. (1996) with minor modifications. The PCR condition was 35 cycles for 1 min at 94°, 2 min at 50°, and 3 min at 72° with a DNA thermal cycler (Perkin Elmer/Cetus Corp., Norwalk, CT). For the 3′ end amplification of SLG29 cDNAs, R-6 primer (Figure 4; 5′-GAGGACGACGCAGATGAGCT-3′) was used as a 5′-primer.
Amplification of the intron sequence of the SLG40 and SLG44 genes was performed with a forward primer (R-7 in Figure 4; 5′-AGTCAGTGAGTTCACACTCG-3′) located 828 bp downstream of the initiation codon of SLG29 and a reverse primer (5′-CGTCTACGTGGCCAATTGA-3′) complementary to the sequence of the C-terminal region of the SLG40 cDNA (Figure 7A). Genomic DNA extracted from young leaves of S40 and S44 haplotypes was used as a template. PCR was carried out as described by Nishio et al. (1996). The amplified products were cloned into the pCRII plasmid vector using the TA cloning kit (Invitrogen).
Isolation of cDNA and genomic clones: An S29 stigma cDNA library was constructed in lambda ZAPII (Stratagene, La Jolla, CA) as described by Matsushita et al. (1996). Approximately 2 × 105 plaques were screened using digoxigenin-labeled (diglabeled) PCR-amplified fragments, pRT26 and pRT37 (see results for details of these two clones), as probes. Prehybridization and hybridization were performed as described by Watanabe et al. (1994). Filters were washed twice with 0.2× SSC, 0.1% sodium dodecyl sulfate (SDS) at 65° for 20 min. In vivo excision of the inserts was performed from positive clones by following the manufacturer's protocol.
For the isolation of SRK29, an S29 stigma cDNA library was constructed in lambda gt10 (Stratagene) as described by Watanabe et al. (1994) because we could not obtain positive clones from the cDNA library constructed in lambda ZAPII. Approximately 4 × 105 plaques were screened using diglabeled SLG29 cDNA and a 0.7-kb EcoRI fragment of the SRK9 cDNA that corresponded to the kinase domain (Watanabeet al. 1994) as a probe. Filters hybridized with both SLG29 cDNA and the 0.7-kb fragment of SRK9 cDNA were washed twice in 0.2× SSC, 0.1% SDS at 65° for 20 min or in 0.5× SSC, 0.1% SDS at 65° for 20 min, respectively. An insert was subcloned from a cross-hybridized phage into the pCRII plasmid vector (Invitrogen).
An S29 genomic library was constructed in the bacteriophage vector lambda GEM11 (Promega, Madison, WI) as described by Suzuki et al. (1995). A total of 7.0 × 105 recombinants were screened with dig-labeled SLG29 cDNA. For isolation of an SRK29 genomic clone, an additional probe, a kinase-encoding domain of the SRK29 cDNA, was also used. Putative SLG29-containing recombinant clones were identified by their intense hybridization signal to the probe. The SLG29-homologous regions from the genomic clones were subcloned into the pCR II plasmid vector.
DNA sequencing, DNA gel blot analysis, and RNA gel blot analysis: DNA sequencing was carried out by the dideoxynucleotide chain termination method (Sangeret al. 1977) using a model 373 DNA sequencer (Applied Biosystems, Foster City, CA). Sequence data were analyzed using DNASIS software (Hitachi Software Engineering, Yokohama, Japan). The homology search was performed using the FASTA program (Pearson and Lipman 1988).
For DNA gel blot analysis, genomic DNA was extracted from 3 g of young leaves by the procedure of Murray and Thompson (1980). The procedures for gel electrophoresis and blot hybridization were performed as described in Watanabe et al. (1994) with minor modifications. A full-length SLG45 cDNA (K. Hatakeyama, T. Takasaki, M. Watanabe and K. Hinata, unpublished results) was used as a probe for DNA gel blot analysis of 24 S haplotypes. Filters were washed twice in 0.2× SSC, 0.1% SDS at 65° for 20 min.
For RFLP linkage analysis, genomic DNA isolated from parental plants, an F1 plant and F2 plants was digested with several restriction enzymes. After hybridization with the full-length pRT37 clone or a 1.0-kb SacI-fragment of the pRT26 clone or 1.7-kb BamHI-XbaI fragment of the SRK29 cDNA which contains 456 bp of the S domain and 692 bp of the kinase catalytic domain, filters were washed twice in 0.1× SSC, 0.1% SDS at 65° for 20 min.
For RNA gel blots, poly(A)+RNA was extracted from stigmas and anthers of flower buds at 1 day before anthesis with the Micro FastTrack mRNA isolation kit (Invitrogen). After denaturation in glyoxal, 1 μg of mRNA was subjected to electrophoresis on 1% (w/v) agarose gel and transferred to a Nytran nylon membrane by blotting with 20× SSC. The blots were prehybridized and hybridized as described in Watanabe et al. (1994) and washed twice in 0.1× SSC, 0.1% SDS at 65° for 20 min. To check the integrity of RNA, the RNA blot was probed with a part of the genomic clone of the beta subunit of the mitochondrial ATP synthase gene of Nicotiana plumbaginifolia (Boutry and Chua 1985). After hybridization at 65°, the filter was washed twice with 0.5× SSC and 0.1% SDS at 65° for 20 min.
DNA gel blot analysis of 16 different S haplotypes of B. rapa. Two micrograms of genomic DNA were digested with EcoRI and separated on 0.8% agarose gels. After transfer to a nylon membrane, the blots were hybridized with the SLG45 cDNA probe. S29, S40, and S44 are pollen-recessive haplotypes. DNA size markers, in kb, are shown on the left.
RESULTS
Isolation of SLG-like sequences from S29 haplotype: We first performed DNA gel blot analysis of 24 S haplotypes of B. rapa to ascertain whether S haplotypes in the CD and DR groups had Class I SLG. To detect Class I sequences, we probed with an SLG45 cDNA, the predicted amino acid sequence of which is 78.7% identical to the polypeptide encoded by Class I SLG9 cDNA (Watanabeet al. 1994; K. Hatakeyama, T. Takasaki, M. Watanabe and K. Hinata, unpublished results). Figure 1 shows a representative result obtained from 16 of the 24 haplotypes. The SLG45 cDNA probe hybridized very strongly to genomic DNA fragments from 20 of the 24 S haplotypes. The hybridization pattern was unique for each S haplotype, as expected. However, we were unable to detect a strong hybridization signal from genomic DNA of the 4 other S haplotypes (i.e., S29, S31, S40, and S44), all of which are pollen-recessive. Similar results were obtained when SLG8 cDNA (Yamakawaet al. 1994) and SLG9 cDNA (Watanabeet al. 1994) from pollen-dominant S haplotypes were used as probes (Watanabe, Okazaki, Suzuki and Hinata, unpublished results).
To isolate the SLG-homologous sequences from the S29 haplotype, RT-PCR was conducted using the Class II-specific primer; two clones, pRT26 and pRT37, each containing a DNA insert with the expected size of ca. 1.6 kb, were obtained. The sequences of the two fragments were different (88.1% identity). Database searches revealed that the nucleotide sequence of the pRT37 clone showed the highest identity, 95.3%, with that of SLG5 isolated from the pollen-recessive S5 haplotype (Scutt and Croy 1992). On the other hand, the sequence of the pRT26 showed the highest identity, 94.4% with those of SLR2 genes (Boyeset al. 1991).
RFLP linkage analysis of an F2 population segregating for S45 and S29 self-incompatibility haplotypes. DNA isolated from parental (P) plants homozygous for either the S45 or S29 haplotype, their F1 heterozygotes and 13 F2 progeny was digested with both EcoRI and EcoRV (A) or digested with SacI (B), or BamHI (C), and analyzed by blot hybridization. Segregation patterns observed after hybridization with probes for pRT37 clone (A), pRT26 clone (B), and SRK29 cDNA (C) are shown. The incompatibility phenotype of each plant was determined by pollination tests (D). DNA size markers, in kb, are shown on the left.
Alignment of predicted amino acid sequences of three Class II SLGs and SLR2-S29 from B. rapa. Colons and dashes indicate identical amino acid residues and gaps introduced to optimize the alignment, respectively. Boxed residues represent twelve conserved cystein residues among Class II SLGs. Asterisks below the sequence indicate twenty amino acid residues that differ between the Class II SLG and SLR2 families. Seven residues corresponding to the primer sequence at the N-terminal end of SLG40 and SLG44 are omitted.
Cloning and sequence analysis of the SLG29 gene of S29 haplotype: Genetic linkage to the S locus of the gene corresponding to one of the two clones, pRT37, was examined by RFLP analysis of 13 plants from an F2 progeny segregating for S45 and S29 haplotypes. When genomic DNA was hybridized with the pRT37 clone, two bands of 6.0 and 6.8 kb were detected only in plants carrying the S29 haplotype (Figure 2A; plants 1, 2, 3, 4, 5, 6, 7, 22, 23, and 25). The intensities of the 6.0- and 6.8-kb bands observed in plants 3 and 4 were much weaker than observed in the other F2 plants due to lower amounts of DNA loaded. The genomic DNA fragments detected by the pRT37 clone correlated perfectly with S29 haplotype. Therefore, we concluded that pRT37 corresponds to the SLG29 gene. Genetic linkage analysis for the other clone, pRT26, is described below.
A full-length SLG29 cDNA clone was obtained from an S29 stigma cDNA library by using the pRT37 clone as a probe; this clone was completely sequenced. The SLG29 cDNA encodes a polypeptide of 449 amino acids that begins, as in other SLGs, with a signal peptide sequence of 31 residues (Figure 3). There are six potential sites of N-glycosylation (N-X-S or N-X-T) distributed throughout the protein. Twelve conserved cysteine residues present in the C-terminal region of all SLGs were also found in this protein. An additional cysteine residue was found in the N-terminal region of the protein. The deduced amino acid sequence of SLG29 shows a higher degree of similarity with Class II SLGs than with Class I SLGs. For example, there is 93% identity with the Class II SLG2 protein (Chen and Nasrallah 1990), but only 66% identity with the Class I SLG9 (Watanabeet al. 1994).
We used this SLG29 cDNA clone to detect Class II sequences in a DNA gel blot analysis of the 24 B. rapa S haplotypes. This probe showed a strong hybridization signal with the genomic DNA from only the four pollen-recessive S haplotypes (data not shown).
A genomic clone corresponding to SLG29was obtained from an S29 genomic library by using the SLG29 cDNA as a probe. Alignment of the genomic sequence and the cDNA sequence revealed the presence of a 1640-bp intron in the region encoding the C-terminal part of the S domain (Figure 4). This intron interrupts the 1321-bp open-reading frame of 440 amino acids, and the last eight amino acids are encoded by the second exon (one amino acid is encoded by the first and second exon). As has been found for SLG2 (Tantikanjanaet al. 1993), the only intron of the SLG29 gene contains an in-frame stop codon (TAG) that could be used to produce an alternative transcript. Unlike the SLG2, however, the second exon of SLG29 does not encode a transmembrane domain, and the only protein product predicted from the SLG29 sequence is a secreted type glycoprotein.
Restriction map of the subcloned regions of the SLG29 and SRK29 genomic clones. The exon of each gene is indicated by filled black boxes. The initiation codon of each gene (ATG) and stop codon of the SLG29 gene (TAG) are shown. Arrowheads represent an in-frame TAG codon of each gene. The position and orientation of the PCR primers are indicated by horizontal arrows. The striped boxes above the SLG29 indicate the region used as probes in Figure 6. B, BamHI; H, HindIII; E, EcoRI; S, SacI. The GenBank data library accession numbers for the SLG29 genomic clone and the SRK29 cDNA sequences are AB008190 and AB008191, respectively.
Cloning and sequence analysis of the SLR2-S29 cDNA: We used the pRT26 clone to isolate cDNA from what appears to be the SLR2 gene from the S29 line. We designate this gene as SLR2-S29. Upon digestion of genomic DNA with SacI and hybridization with a 1.0-kb SacI-fragment of pRT26 (Figure 2B), a 1.0-kb fragment was detected in S45 and S29 homozygous parents and in all of the F2 progeny. This suggested that the region that hybridized to the pRT26 probe did not cosegregate with the S29 haplotype.
We then used the pRT26 clone to probe the aforementioned cDNA library. Two positive clones containing the ATG initiation codon were obtained, and the longer insert was selected for sequence analysis. This cDNA clone encodes a polypeptide of 439 amino acids that begins with a signal peptide sequence of 31 residues (Figure 3). There are six potential sites of N-glycosylation distributed throughout the protein. The deduced amino acid sequence shows the high degree of similarity (99.8% identity) with that of the SLR2-C636 gene isolated from B. rapa (Watanabeet al. 1997). However, SLR2-S29 did not contain a 4-bp frame-shifting insertion at its C-terminal region, as had been found for SLR2-S8 (Tantikanjanaet al. 1996) and SLR2-C636 (Watanabeet al. 1997) isolated from B. rapa. Eleven of the 12 conserved cysteine residues are present in the C-terminal region of the cDNA clone, as in other SLR2. The seventh conserved cysteine (counted from the N-terminal end) is changed to glycine. These results suggest that pRT26 corresponds to the SLR2 gene.
Cloning and sequence analysis of the SRK29 gene of S29 haplotype: A full-length SRK29 cDNA clone was isolated by using as probes the SLG29 cDNA and the 0.7-kb EcoRI-fragment of the SRK9 cDNA that encodes the kinase domain (Watanabeet al. 1994). Using methods similar to those described for the isolation of SLG29, we identified a genomic clone corresponding to SRK29. The SRK29 genomic clone was subcloned into a plasmid vector, and the DNA sequences were determined. Comparison of the cDNA and the genomic DNA sequences of SRK29 revealed that the subcloned region of the genomic clone lacked the seventh exon (Figure 4). As in other SRKs, the first intron of the SRK29 contains an inframe stop codon (TAG). Hanks and Quinn (1991) have shown that protein kinases have 11 conserved subdomains in which 15 invariant or nearly invariant amino acid residues are located. These subdomains and all the 15 conserved residues are present in SRK29. The sequences in subdomains VI (HRDLKASN) and VIII (GTYGYMSPE) suggest that SRK29 is likely to have serine/threonine kinase activity.
To confirm the linkage between the SRK29 gene and the S locus, RFLP analysis was performed on the same F2 progeny used in the linkage analysis of the SLG gene. When the genomic DNA was digested with BamHI and hybridized with the BamHI-XbaI fragment of the SRK29 cDNA, an intense band of 12 kb was identified in the S29 but not in the S45 homozygous parent (Figure 2C). A perfect correlation was observed between the presence of the SRK29 band and the S29 haplotype in 13 plants of the F2 family segregating for S45 and S29. These results indicate that SLG29 and SRK29 were linked to the S locus.
Genomic structural similarities between SLG29 and SRK29: Comparison of the genomic sequences of the SLG29 and SRK29 genes revealed a region of sequence similarity, which extends from 370 bp upstream of the ATG codon of the S domain to 4 bp downstream of the in-frame stop codon (at position 1330) in the first intron (Figure 5A). In the S domain, the sequence identity is 84%, and in the 5′ flanking region, the sequence identity is 71%. The five conserved elements (box I to V) previously identified in the 5′ flanking region by Dzelzkalns et al. (1993) were also found in these two genes (Figure 5B). This finding suggests that the SLG29 and SRK29 genes may have an expression pattern similar to other SLGs and SRKs. The promoter regions of the SLG29 and SRK29 genes are more similar to those of Class II SLG2 and SRK2 genes (79.8 and 83.7% identity, respectively) than to those of Class I SLG9 and SRK9 genes (53.9 and 54.0% identity, respectively).
Comparison of the genomic sequences of the SLG29 and SRK29 genes. (A) Schematic representation of the sequenced regions of the SLG29 and SRK29 genomic clones. The shaded boxes represent the protein coding regions in both SLG29 and SRK29. The striped box corresponds to the untranslated region of the SLG29 gene. The initiation and stop codons of the SLG29 gene are shown. The transmembrane domain (TM) of the SRK29 is indicated under the diagram. The stippled areas between the two diagrams indicate the region showing high similarity. The sequences are numbered relative to the ATG initiation codon of each gene. (B) Nucleotide sequences of the 5′ flanking region of the SLG29 and SRK29 genes aligned with those of SLG2 and SRK2. Dashes represent gaps introduced to optimize the alignment. Asterisks indicate nucleotides that are conserved in all four sequences. The boxed sequences show regions corresponding to the five boxes that were identified previously by Dzelzkalns et al. (1993). The sequences are numbered from the translation initiation codon of each gene. (C) Nucleotide sequence comparison of the first intron of the SLG29 and SRK29 genes. Colons represent nucleotide identities and dashes represent gaps introduced to optimize the alignment. Exon sequences are shown by uppercase letters and intron sequences by lowercase letters. Boxed sequence indicates the inframe stop codon. The SRK29 intron sequences (1415–2758) that do not show sequence similarity to the SLG29 sequences are omitted. The sequences are numbered relative to the ATG initiation codon of each gene.
Comparison of the amino acid sequences of different domains of SRK29 and some other Brassica SRKs
The alignment of the sequences at the 3′ end of the S domain is shown in Figure 5C, with the nucleotides numbered from the ATG initiation codon in each gene. When the first introns of the SRK29 and SLG29 genes were compared, only a 384-bp region (from position 1413 to 1797 in SLG29) located 87 bp downstream of the in-frame stop codon showed 84% sequence identity, albeit several small deletions/insertions were observed in this region (Figure 5, A and C). The sequence similarity for the rest of the intron (from position 1798 to 2960 in SLG29) was less than 50%.
The sequence encoding the receptor (S domain), juxtamembrane, transmembrane, kinase and C-terminal domains of SRK29 were compared with the corresponding domains of other SRKs, and the results are shown in Table 1. Very low similarity was observed for the juxtamembrane and transmembrane domains between Class I and Class II types of SRKs.
Expression of SLG and SRK: Because the SLG29 gene had an in-frame stop codon in the intron, it could potentially produce two transcripts that differed at their 3′ ends. One transcript (type I) would contain only the first exon and the other (type II) would contain both the first and second exons (as is the case for the SLG29 cDNA). To examine this possibility, we hybridized stigma and anther poly(A)+RNA to probes expected to be specific for each transcript (see Figure 4). Using the full-length SLG29 cDNA as a probe, a strong band of ca. 1.6 kb was observed only in the stigma (Figure 6A). A 1.2-kb HindIII fragment (probe a in Figure 4) from the SLG29 genomic clone was used to detect type II expression. This probe detected a ca. 1.6-kb band in stigmas and none in anthers (Figure 6B). A ca. 1.0-kb fragment corresponding to the 5′ end of the first intron of the SLG29 genomic clone (probe b in Figure 4) was amplified by PCR and used as a probe to detect type I expression. This probe detected only a very weak signal in stigmas and none in anthers, even after overexposure (Figure 6C). Furthermore, RT-PCR was performed to look for transcripts that contained the first intron of the SLG29 gene. Poly(A)+RNA isolated from S29 stigmas was reverse transcribed and amplified with a 20-bp oligonucleotide primer (R-6 in Figure 4) located 1050 bp downstream of the translation initiation codon of SLG29 and a 3′ oligo (dT) primer. Thirty positive clones were isolated by hybridization with an SLG29 cDNA probe. PCR analysis was performed on these clones by using the SLG29 forward primer (R-6) and a type II-specific primer (R-10 in Figure 4) complementary to the sequence of the second exon of the SLG29 gene. We found that all positive clones corresponded to the type II SLG29 transcripts (data not shown). These results suggested that only the type II transcript, consisting of the first and second exons, was produced from the SLG29 gene, and that the SLG29 gene was expressed mainly in the stigma.
RNA gel blot analysis of transcripts of the SLG29 and SRK29 genes. One microgram of poly(A)+RNA isolated from stigma (St) and anther (A) was loaded in each lane. RNA blots were hybridized with probes of the following construction: (A) a full-length SLG29 cDNA; (B) the clone containing a part of the intron and the second exon as illustrated in Figure 4 (probe a); (C) the clone containing the 3′ part of the first exon and a part of the intron as illustrated in Figure 4 (probe b); (D) a gene encoding beta subunit of the mitochondrial ATP synthase. RNA size markers, in kb, are shown on the left.
Comparison of the 3′ terminal regions between pollen-recessive SLG genes. (A) Alignment of nucleotide and deduced amino acid sequences at the 3′ terminus between cDNA of SLG29, SLG40, and SLG44. The stop codon of each SLG cDNA clone is indicated by bold italics. Sequence encoded by the second exon of the SLG gene is underlined. Gaps indicated by dashes are introduced to optimize nucleotide sequence alignment. Sequences are numbered relative to ATG initiation codon of each cDNA. (B) DNA sequence around the first exon/intron junction in the SLG29, SLG40, and SLG44 and the SLG2 of B. oleracea (Tantikanjanaet al. 1993). Exon sequences are shown by uppercase letters and intron sequences by lowercase letters. Gaps indicated by dashes are introduced to optimize nucleotide sequence alignment. The exon-intron junctions are indicated by an arrowhead. The in-frame stop codons as indicated by bold type letters are present in all genes.
In addition to the SLG29 transcript, a band of ca. 3.0 kb was observed in the stigma after long exposure, when the full-length SLG29 cDNA was used as a probe (Figure 6A). On the basis of the length of the transcript and the intensity of the band, this band was ascribed to the SRK29 transcript.
Gene structure of other Class II SLGs of B. rapa: SLG cDNAs were also amplified from stigma poly(A)+ RNA of two other pollen-recessive haplotypes, S40 and S44, by using the PS3 primer and an oligo (dT) primer. Alignment of SLG29, SLG40 and SLG44 cDNA sequences of B. rapa revealed that the sequences of the 3′ terminal regions of SLG40 and SLG44 cDNA were very similar to those of the second exon of SLG29, as shown in Figure 7A. On the basis of the results of the cDNA sequence analysis, the SLG40 and SLG44 genes were predicted to contain an intron that interrupted the 1321-bp ORF, as did SLG29. To confirm this prediction, amplification of the first intron sequence from S40 and S44 haplotype genomic DNA was performed (see materials and methods). DNA sequences of approximately 2.0-kb amplified fragments were determined. As found for SLG29, the amplified products from both S40 and S44 haplotypes contained an in-frame stop codon following the GT motif at the 5′ end of the amplified fragment (Figure 7B), indicating that the SLG genes of pollen-recessive haplotypes of Brassica have in common an intron at their C terminus.
DNA sequence analysis of cDNA clones showed that the predicted amino acid sequences of SLG40 and SLG44 had strong similarity with that of SLG29 (96.3% and 95.9% identity, respectively) and contained all of the 12 conserved cysteine residues (Figure 3). However, the amino acid residues (TCTG) encoded in the second exon of SLG40 and SLG44 were different from those (TIR-KRHKI) of SLG29, because of a ca. 20-bp deletion or insertion in the sequences of the second exon (Figure 7A).
DISCUSSION
We have characterized three SLG genes, SLG29, SLG40, and SLG44, from pollen-recessive haplotypes of B. rapa in this experiment. These three SLG genes, belonging to the Class II SLG, all contain an intron at their C terminus. In contrast, none of the Class I SLG genes so far reported contains an intron. The nucleotide sequences of the second exon of the three SLG genes are highly conserved, except for a ca. 20-bp deletion/insertion (Figure 7A). This deletion/insertion provided different amino acid sequences at the C terminus of SLG29 relative to the other two. In SLG40 and SLG44, a specific amino acid sequence, TCTG, was found in the C-terminal region (Figures 3 and 7A). This sequence was also found in the SLG from another pollen-recessive haplotype, S5 in B. oleracea and self-compatible Brassica napus, although it was not determined whether or not these four amino acids are encoded by the second exon (Scutt and Croy 1992; Robertet al. 1994). The structure of the SLG genes observed here seems to be typical of Class II haplotypes of Brassica species. However, it is different from that of the SLG2 gene isolated from a pollen-recessive haplotype of B. oleracea, in which the second exon encodes the transmembrane and a part of the cytoplasmic domain (Tantikanjanaet al. 1993). Tantikanjana et al. (1993) suggested that the existence of a membrane-anchored form of SLG might be involved in a leaky self-incompatibility phenotype or the pollen-recessive nature of this haplotype. Our data demonstrated that an unusual structure of SLG2 was not the sole determinant of its pollen-recessive nature.
In a previous article, we showed, based on pollination results, that 24 S haplotypes in B. rapa could be classified into three groups: codominant (CD), dominant/recessive (DR), and recessive (R) (Hatakeyamaet al. 1998). DNA gel blot analysis demonstrates, without exception, that the SLG and SRK genes isolated from the R group belong to Class II and those from the CD and DR groups belong to Class I (Figure 1). In the S29 haplotype of the R group also, the deduced amino acid sequence of the S domain of SRK29 is highly similar to that of SLG29 (93%), as is true in many SLG/SRK gene pairs. Both SLG29 and the S domain of SRK29 show a high degree of sequence similarity (89.8% to 95.7% identity) to the SLGs of S2 and S5, which have been classified as Class II, whereas they show less than 70% sequence identity to Class I SLGs. Furthermore, the transmembrane, juxtamembrane, kinase, and C-terminal domains of SRK29 are also divergent from those of Class I SRK. In particular, the juxtamembrane and transmembrane domains show the lowest similarity (37.7% and 45.5% identity, respectively) to those of Class I SRK9 (Table 1). Similar trends have been also observed in the detailed analysis of SLG2 and SRK2 isolated from the pollen-recessive S2 haplotype in B. oleracea (Chen and Nasrallah 1990; Steinet al. 1991). The feature that pollen-recessive haplotypes have Class II SLG and SRK genes and the others have Class I seems to be common in Brassica species.
Dominance relationships among haplotypes differ between stigma and pollen expression: for example, pollen-recessive haplotypes, S29, S40, and S44, are codominant in the stigma to many S haplotypes (Hatakeyamaet al. 1998). Pollen (but not stigma) expression correlates well with class type, with all Class II haplotypes examined showing pollen-recessivity. A current model of self-incompatibility in Brassica (Nasrallah and Nasrallah 1993) is that the SLG/SRK complex recognizes an unidentified pollen ligand, which is encoded at the S locus. A pulsed-field gel electrophoresis analysis of genomic DNA from S6 (Class I) and S2 (Class II) haplotypes of B. oleracea has revealed extensive between-class differences across the entire S locus region (Boyes and Nasrallah 1993). The complete association between Class II and pollen-recessivity suggests that some of the differences that distinguish the classes may correspond to the subregion that encodes the hypothesized pollen ligand.
The SLG29 and SRK29 genes have an in-frame stop codon, TAG, following the conserved GT motif at the 5′ end of the first intron. This in-frame stop codon was also found in the intron sequences of the SLG40 and SLG44 genes. This stop codon could be used to produce a truncated SLG-like protein from an alternative transcript that retains the first intron. The presence of transcripts of the SRK gene that retain a part or the full-length of the first intron has been reported previously (Steinet al. 1991; Girantonet al. 1995; Suzukiet al. 1996). In the case of SLG29, however, the transcript that consists of the first and the second exons was predominantly detected, whereas the alternative transcript was undetectable. Further studies are needed to examine the role of the type I transcripts, which consist of the first exon and a part of the intron.
In dendrograms reconstructed using the neighbor-joining method from the base substitutions observed in SLG, SRK, and SLG-related sequences (Hinataet al. 1995; Uyenoyama 1995; Kusabaet al. 1997), Class II SLG and SLR2 cluster together. A cDNA clone corresponding to SLR2 was isolated in addition to the PCR-amplified cDNA corresponding to SLG29 from the S29 haplotype. The deduced amino acid sequence of SLR2-S29 showed a high degree of sequence similarity (more than 85% identity) to that of three Class II SLGs, SLG29, SLG40, and SLG44 (Figure 3). Class II SLG and SLR2 likely share a common ancestor. When we aligned the deduced amino acid sequences of Class II SLGs and SLR2 that had been isolated previously (Chen and Nasrallah 1990; Boyeset al. 1991; Tantikanjanaet al. 1996; Watanabeet al. 1997) with the four clones we isolated, we found that twenty amino acids differed between the Class II SLG and SLR2 families (asterisks in Figure 3). Some of these differences, scattered throughout the amino acid sequence of Class II SLG and SLR2, may possibly reflect the different functions of the two families.
In B. rapa, the several SLG genes from Class I S haplotypes that have been isolated (Watanabeet al. 1994; Yamakawaet al. 1994; Matsushitaet al. 1996; Nishioet al. 1996; Kusabaet al. 1997) show pairwise sequence identity ranging from 78 to 98%. We observed sequence identity levels among the Class II SLGs, SLG29, SLG40, and SLG44, in excess of 95%; this apparent increase in similarity may imply that the time since divergence among haplotypes within Class II is less than that within Class I. Based on results of their simulation model, in which alleles interacted codominantly in the style and formed a dominance hierarchy in the pollen (SSIdomcod), Schierup et al. (1997) argued that loss because of drift occurs more easily for pollen-recessive than for pollen-dominant alleles, resulting in lower expected life span for recessive alleles. Our observation of lower divergence among pollen-recessive Class II SLGs may be consistent with this theoretical finding.
Acknowledgments
The authors thank Professor Teh-hui Kao, Pennsylvania State University, for his critical reading of the manuscript and help in correcting the English. We are grateful to Professor Nam-Hai Chua, The Rockefeller University, for providing us with a genomic clone of the beta subunit of the mitochondrial ATP synthase gene of Nicotiana plumbaginifolia. We thank Go Suzuki, Tohoku University, for his helpful comments and criticisms. This work was supported in part by Grants-in-Aid for Special Research on Priority Areas (nos. 07281102 and 07281103; Genetic Dissection of Sexual Differentiation and Pollination Process in Higher Plants) from the Ministry of Education, Science, Culture and Sports, Japan.
Footnotes
-
Communicating editor: M. K. Uyenoyama
- Received December 1, 1997.
- Accepted April 6, 1998.
- Copyright © 1998 by the Genetics Society of America
