INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Self-incompatibility (SI) in flowering plants is a prezygotic reproductive barrier that enables the pistil of a flower to reject self-pollen, but to accept nonself pollen for fertilization (de Nettancourt, 1977). This inbreeding-prevention strategy is widespread in flowering plants, and recent molecular data suggest that it has evolved independently in different families multiple times during the course of evolution (for review, see McCubbin and Kao, 2000). In the simplest case the specificity of SI interactions is controlled by a multiallelic locus called the S locus. SI can be further classified into gametophytic and sporophytic types. For gametophytic (G)SI, the SI phenotype of the pollen is determined by its own S-genotype, whereas for sporophytic (S)SI, the SI phenotype of the pollen is determined by the S genotype of the pollen-producing plant.

GSI has been extensively studied in the Solanaceae, which includes Petunia, potato, tobacco, and tomato. The S-RNase gene, a highly polymorphic pistil-specific gene at the S locus, has been shown to encode the S haplotype specificity determinant of the pistil (Lee et al., 1994; Murfett et al., 1994). Comparison of the deduced amino acid sequences of Solanaceous S-RNases has revealed five conserved (C1-C5) and two hypervariable (HVa and HVb) regions (Ioerger et al., 1991; Tsai et el., 1992). Because the S-RNase gene does not control pollen function in SI, another polymorphic gene (yet unidentified) at the S locus, named the pollen S gene, is thought to play this role. For SSI, most research so far has focused on the family Brassicaceae. Three highly polymorphic genes have been identified at the S locus, SLG, SRK, and SCR, and the latter two determine the S haplotype specificity of the stigma (Takasaki et al., 2000) and pollen (Schopfer et al., 1999; Takayama et al., 2000), respectively.

The molecular mechanisms responsible for the generation of haplotype diversity at the S locus have been studied in Brassicaceae and Solanaceae SI systems. The central question is whether point mutation alone or point mutation and recombination/gene conversion has contributed to the generation of allelic diversity. Results from domain-swapping experiments suggest that point mutations in the S-RNase gene can generate new allelic specificities (Matton et al., 1997). However, so far no evidence for recombination/gene conversion at the S-RNase gene has been obtained.

In fact, the following findings have even been interpreted to mean that recombination does not contribute at all to the generation of S haplotype diversity in the Solanaceae SI system. First, the S locus contains the S-RNase gene and the pollen S gene, so that recombination would destroy the tight linkage between these two genes, and thus cause breakdown of SI by generating different haplotype specificities for the pistil and pollen. Second, the flanking regions of different alleles of the S-RNase gene share virtually no sequence similarity and contain abundant repetitive sequences (Kaufmann et al., 1991; Coleman and Kao, 1992; Matton et al., 1995), indicative of a lack of recombination in the vicinity of the S-RNase gene. Third, no evidence of intragenic recombination was found when sensitive statistical methods were used to examine cDNA sequences for 12 alleles of the S-RNase gene for clustering of polymorphic sites (Clark and Kao, 1991). Last, the S locus of Lycopersicon peruvianum has been shown by genetic mapping to be located close to the centromere of chromosome I (Bernatzky, 1993), and the S locus of Petunia hybrida has been shown by fluorescence in situ hybridization to be located close to the centromere of chromosome III (Entani et al., 1999), the centromere being a region where recombination is known to be suppressed. In the Brassicaceae SI system, extensive sequence divergence and rearrangements were observed when comparing the S locus of two different S haplotypes (Boyes et al., 1997), suggesting that recombination is likely to be suppressed at the Brassicaceae S locus as well. No recombination has been detected between SLG and SRK, even though the physical distance between these two genes may exceed 200 kb in some haplotypes (Boyes and Nasrallah, 1993; Casselman et al., 2000), nor has recombination between the pollen S gene and the S-RNase gene of the Solanaceae SI system ever been observed (de Nettancourt, 1977; Kheyr-Pour and Pernes, 1986).

However, these aforementioned results do not completely rule out the involvement of recombination. First, even though no recombination at the S locus has ever been observed in genetic studies, it does remain possible that recombinants are generated, but at a very low frequency, and that those that become self-compatible have been eliminated from the population. Second, the high degree of sequence diversity in the flanking regions of the S genes might prevent intergenic, but not intragenic recombination. Moreover, the S-RNases available to Clark and Kao (1991) for examining the possibility of intragenic recombination by statistical methods are too divergent in sequence to allow detection of any ancient recombination events because accumulation of neutral mutations over long evolutionary time would have obscured the exchanges.

In another more recent study of the Brassicaceae SI system, Kusaba et al. (1997) compared the sequences of six class I alleles of the SLG gene of Brassica oleracea and obtained different phylogenetic tree topologies for different regions of the gene, thus providing the first evidence for the occurrence of intragenic recombination among the alleles of an S locus gene. Awadalla and Charlesworth (1999) subsequently analyzed linkage disequilibrium within the SLG genes of B. oleracea and Brassica campestris, and found that it declines with distance within the SLG gene, providing further support for the existence of recombination in the Brassicaceae SI system.

These findings of possible intragenic recombination at the Brassicaceae S locus prompted us to reexamine this possibility for the S locus of the Solanaceae. In this study we used plants derived from seeds collected from a natural population of Petunia inflata in Argentina for the cloning of S-RNase sequences. Plants from the same population are more likely to contain alleles that are closely related and share recent progenitor alleles. We identified 19 different S haplotypes in this population and used reverse transcription (RT)-PCR to obtain cDNAs corresponding with 15 alleles of the S-RNase gene. Phylogenetic analyses using the deduced amino acid sequences of these alleles and a number of previously identified alleles showed that five of the alleles identified from this natural population and two previously identified alleles of P. inflata form a major clade. These closely related alleles allowed us to more precisely assess the occurrence of intragenic recombination between different alleles of the S-RNase gene and also to identify some other interesting features about the Solanaceae SI system.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Identification of S Haplotypes and Cloning of cDNAs for S-RNases of P. inflata from a Natural Population in Argentina

One hundred SI plants were raised from seeds randomly collected from a natural population of P. inflata found in Argentina, and 36 of them were used in this study. Nineteen different S haplotypes were identified based on the results of genetic crosses described in the "Materials and Methods." One of these 19 S haplotypes was designated S₁ because its SI behavior was identical to that of one of the three previously identified S haplotypes of P. inflata, S₁, S₂, and S₃ (Ai et al., 1990). All the other 18 S haplotypes were novel and thus designated S₄ through S₂₁. Table I lists the S genotype of each of the 36 plants and indicates the plants from which each of the 19 S homozygotes were obtained by bud-selfing.

Plants homozygous for all except S₁₄ and S₁₈ haplotypes were used for cloning cDNAs of S-RNases. We designed a degenerate primer based on the amino acid sequences of the conserved region C2 of solanaceous S-RNases (Ioerger et al., 1991; Tsai et al., 1992). Separate RT-PCR was performed using this primer and an oligo (dT)₁₇ primer on total pistil RNA isolated from plants homozygous for each of the 17 S haplotypes. A major DNA band of approximately 550 to 600 bp in size was obtained for each of the 17 S haplotypes (results not shown). This size range was what would be expected of cDNAs containing the coding sequence for S-RNases (except for the region upstream of C2) plus the 3'-non-coding sequence and the poly(A) tail. This major DNA band for each S haplotype was cloned and individual colonies from each transformation were analyzed by DNA gel blotting using a radiolabeled probe containing cDNAs for S₁-, S₂-, and S₃-RNases of P. inflata (Ai et al., 1990). Under low stringency membrane washing conditions, cDNA clones for eight of the 17 S haplotypes (S₁, S₅, S₆, S₈, S₁₂, S₁₃, S₁₅, and S₁₉) hybridized to the probe, whereas for the remaining nine S haplotypes (S₄, S₇, S₉, S₁₀, S₁₁, S₁₆, S₁₇, S₂₀, and S₂₁), none of the cDNA clones hybridized to the probe. The failure to hybridize could possibly be due to greater sequence diversity between the cDNAs for the S-RNases of these haplotypes and the cDNAs for S₁-, S₂-, and S₃-RNases.

Genomic DNA-blot analysis was carried out to further ascertain whether the cDNA obtained by RT-PCR for each S haplotype encoded an S-RNase. If a cDNA encodes an S-RNase, it should reveal an S haplotype specific restriction fragment-length polymorphism (RFLP) and hybridize most strongly to the genomic fragment(s) from the same S haplotype. This is because different alleles of the S-RNase gene are divergent in sequence not only in the coding region, but also in the flanking regions. Blots containing EcoRI digests of genomic DNA isolated from plants homozygous for different S haplotypes were hybridized separately with the cDNAs isolated from the 17 S haplotypes. All except the cDNAs for S₄ and S₅ haplotypes revealed a hybridization pattern expected of cDNAs encoding S-RNases. The results of genomic DNA blotting using the cDNA for S₈ haplotype as a probe are shown in Figure 1. The cDNA hybridized to an EcoRI fragment of different size in each of the S₈S₈, S₁₂S₁₂, S₁₃S₁₃, and S₂₀S₂₀ homozygotes, and did not hybridize to the genomic DNA digests of the other eight S homozygotes on the blot. Moreover, the intensity of the hybridization to the DNA fragment of S₈S₈ homozygote was the strongest.

View larger version (71K): [in this window] [in a new window]   Figure 1.   DNA-blot analysis of genomic DNA digests of 12 S homozygotes of P. inflata. EcoRI digests of genomic DNA (5 µg) from each of the S homozygotes indicated were hybridized with radiolabeled cDNA that had been obtained by RT-PCR from S8 haplotype and cloned into pGEM-T.

The cDNAs for all the 17 S haplotypes were completely sequenced, and all except the cDNAs for S₄ and S₅ haplotypes showed sequence similarity with the sequences of known solanaceous S-RNases. Thus, cDNAs for a total of 15 alleles of the S-RNase gene identified from a natural population of P. inflata were cloned. Because the upstream primer used in RT-PCR corresponded to the C2 region of S-RNases, the cDNAs obtained encoded approximately 167 amino acids of S-RNases, with approximately 35 amino acids at the N-terminal end not included. It is interesting that the nucleotide sequence (both coding and the 3'-non-coding) of the S₁-RNase gene identified in this work is completely identical to that of the previously identified S₁-RNase gene. Moreover, when the genomic DNA of these two independent sources of S₁ haplotype was digested with several restriction enzymes and hybridized with the full-length cDNA for S₁-RNase, no RFLP was observed (results not shown), indicating that there is little, if any, sequence difference in the immediate flanking regions of the S-RNase gene. However, when the same genomic DNA blot was hybridized with cDNAs for two pollen-expressed genes that are linked to, but located at as yet undetermined distances from the S-RNase gene (McCubbin et al., 2000), polymorphism was detected (results not shown).

Sequence Comparison among S-RNases

Deduced amino acid sequences of the 15 S-RNases of P. inflata identified from this natural population, as well as those of S₂- and S₃-RNases of P. inflata were aligned by CLUSTAL X (Thompson et al., 1997), and the alignment is shown in Figure 2A. Three of the five conserved regions and both hypervariable regions are contained in the aligned sequences. The pairwise amino acid sequence identity ranges from 38% to 92%; this wide range of sequence diversity is similar to that found from all previous comparisons of solanaceous S-RNases. Two pairs, S₆-RNase and S₉-RNase, and S₈-RNase and S₁₅-RNase, show the highest degree of sequence identity. The nucleotide sequences of S₆-RNase and S₉-RNase are completely identical in the HVa region, and only differ by four nucleotides in the HVb region, resulting in two different amino acids (Fig. 2B).

View larger version (49K): [in this window] [in a new window]   Figure 2.   Alignment of deduced amino acid sequences of 17 S-RNases of P. inflata (A) and of S6- and S9-RNases (B). At each aligned position, amino acids that are identical to the amino acid of the reference sequence, S1-RNase in A and S6-RNase in B are indicated with dots, and only those amino acids that are different are labeled. Gaps, indicated with dashes, are introduced to maximize similarity. The two hypervariable regions (HVa and HVb) and three conserved regions (C3, C4, and C5) are shaded. The GenBank accession numbers for these sequences are as follows: S1, M67990; S2, AF301533; S3, M67991; S6, AF301167; S7, AF301168; S8, AF301169; S9, AF301170; S10, AF301171; S11, AF301172; S12, AF301173; S13, AF301174; S15, AF301175; S16, AF301176; S17, AF301177; S19, AF301178; S20, AF301179; and S21, AF301180.

Molecular Phylogeny of S-RNases

A phylogenetic tree of the 15 S-RNases of P. inflata whose sequences were determined in this work and 19 previously reported solanaceous S-RNases was constructed by the neighbor-joining method based on the proportion of amino acid differences (Fig. 3). The tree was rooted with a fungal RNase (RNase T2) and two S-like RNases (le and lx) from tomato. The tree shows that five of the 15 P. inflata S-RNases from this natural population (S₈, S₁₂, S₁₃, S₁₅, and S₁₉) and the previously identified S₂- and S₃-RNases form a major clade, and the other 10 are grouped with one another in the same cluster with S-RNase(s) from other species of the Petunia genus, or with S-RNase(s) from another genus of the Solanaceae. This pattern of trans-species polymorphism, first observed by Ioerger et al. (1990), supports the hypothesis that S haplotype polymorphism has an ancient origin and has been maintained by balancing selection.

View larger version (24K): [in this window] [in a new window]   Figure 3.   A phylogenetic tree of solanaceous S-RNases. The GenBank accession numbers for 17 S-RNases of P. inflata (shown in bold) are listed in Figure 2. The accession numbers for the other S-RNases are as follows: P. hybrida, S1, U07362; Sx, M81685. Petunia axillaris, S1, AF239908; S13, AF239909; S15, AF239910. L. peruvianum, S3, X76065; S5, S61768; S6, Z26583; S7, Z26582; S11a, S65047; S12, D17324; S13, D17325. Nicotiana alata, S2, U08860; S3, U66427; S6, U08861; and S7, U13255; Sa2, U45957. Solanum chacoense, S11, L36464. Solanum tuberosum, S2, X62727. The tree was constructed by the neighbor-joining method based on the proportion of amino acid differences. It is rooted by RNase T2 from Aspergillus oryzae (accession no. X61086) and two S-like RNases of tomato, le (accession no. X79337), and lx (accession no. X79338). Numbers on the internal branches indicate bootstrap percentages that are higher than 50, and the scale below the tree is the measure of branch length.

Different Evolutionary Histories of the 5' and 3' Regions of the S-RNase Gene

To ascertain the possibility that intragenic recombination could have generated closely related alleles of the S-RNase gene we compared the seven closely related alleles mentioned above to see whether different regions of the S-RNase gene have taken the same evolutionary pathway. If allelic diversity of the S-RNase gene is entirely due to the accumulation of point mutations over evolutionary time, different regions of the gene would be expected to have consistent genealogies because the entire gene would have evolved as a unit. On the other hand, if recombination among different alleles of the S-RNase gene has occurred, it would shuffle polymorphic sequences among alleles, causing different regions of the S-RNase gene to have different evolutionary histories.

As an exploratory tool we plotted p_s (the proportion of synonymous differences per synonymous site) and p_n (the proportion of non-synonymous differences per non-synonymous site) in a sliding window of 30 codons for all possible pairwise comparisons among the aforementioned seven alleles of the S-RNase gene. The plots for the pair of the S₂-RNase gene and the S₁₉-RNase gene (Fig. 4A) show that they are more similar to each other at synonymous and non-synonymous sites in the 5' region than in the 3' region of the gene. The plots for the pair of the S₈-RNase gene and the S₁₉-RNase gene (Fig. 4B) show that they are more similar to each other at synonymous and non-synonymous sites in the 3' region than in the 5' region of the gene. The maximum chi-square method gave statistically significant evidence (P = 0.007) of a recombination event giving rise to different evolutionary histories in the 5' region (codons nos. 1-68; refer to Fig. 2A for the numbering of amino acids) and the 3' region (codons nos. 69-168) of the S-RNase gene.

View larger version (22K): [in this window] [in a new window]   Figure 4.   Plots of ps and pn for pairwise comparison between S2 and S19 alleles of the S-RNase gene (A) and between S8 and S19 alleles of the S-RNase gene (B). Solid lines are plots of ps and dotted lines are plots of pn. Both plots were obtained using a sliding window of 30 codons.

However, a significant result of the maximum chi-square method alone is not definitive evidence of recombination because several alternative hypotheses might explain the observation of greater similarity between alleles in one region of a gene than in another region. To decide between the hypothesis of recombination and alternative explanations we constructed separate phylogenetic trees for the 5' (codons nos. 1-68) and 3' (codons nos. 69-168) regions of the S₂-, S₈-, S₁₃-, and S₁₉-RNase genes. For each region we constructed two separate trees on the basis of p_s and p_n. Recombination is expected to produce similarity at synonymous and non-synonymous sites, whereas certain other factors can yield similarity at synonymous sites alone or at non-synonymous sites alone. For example, a shared G + C content bias can produce an unexpectedly high similarity at synonymous sites, whereas a shared functional constraint can produce an unexpectedly high similarity at non-synonymous sites (Hughes, 1991). The phylogenetic trees show identical topologies with high bootstrap values (Fig. 5, A and B), strongly supporting the conclusion drawn from the results shown in Figure 4, A and B that the 5' region of the S₁₉-RNase gene is closely related to the S₂-RNase gene, but the 3' region is closely related to the S₈-RNase gene.

View larger version (21K): [in this window] [in a new window]   Figure 5.   Neighbor-joining trees for the 5' and 3' regions of two sets of alleles of the S-RNase gene. For each set of alleles, separate trees were constructed based on the ps and pn. Numbers on the branches are bootstrap percentages, and the scale below each tree is the measure of branch length for that tree. For S2, S8, S13, and S19 alleles, the 5' region (A) spans amino acid numbers 1 to 68, and the 3' region (B) spans the rest of the protein (amino acid nos. 69-168). For S1, S3, S12, and S15 alleles, the 5' region (C) spans amino acid numbers 1 to 62, and the 3' region (D) spans the rest of the protein (amino acid nos. 63-168). Refer to Figure 2 for the numbering of amino acids.

Another group of alleles in which we detected the existence of recombination are S₃, S₁₂, and S_15. The maximum chi-square test supported the hypothesis of recombination (P = 0.048) between the 5' region (codon nos. 1-62) and the 3' region (codons nos. 63-168) of these alleles. Phylogenetic analysis based on p_s and p_n showed that the 5' region of the S₃-RNase gene is closely related to that of the S₁₂-RNase gene, whereas the 3' region is closely related to that of the S₁₅-RNase gene (Fig. 5, C and D).

For both sets of alleles, because the regions examined are large (encoding at least 62 amino acids) and the bootstrap values are high, the observed topological differences are unlikely to be due to stochastic error. Thus, our results suggest that intragenic recombination has played a role in the generation of new allelic specificities of the S-RNase gene.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Use of RT-PCR in Cloning cDNAs for S-RNases from Natural Populations

Although cDNAs for a large number of S-RNases have been cloned and sequenced from a number of solanaceous species, only a few studies so far have used plants derived from natural populations as the source of S haplotypes. Richman et al. (1995, 1996) were the first to use RT-PCR to clone cDNAs for S-RNases from natural populations of solanaceous species; they determined the number of S haplotypes and the frequency of each S haplotype in individual populations. In these studies S genotypes of plants used were not determined genetically, and because the PCR primers were designed based on the C2 and C3 (or C4) regions of S-RNases, the cDNAs obtained encode approximately 60 or 120 amino acids out of a total of approximately 200 amino acids for full-size S-RNases.

In our study of a natural population of P. inflata identified in Argentina we first carried out genetic crosses to identify the S genotype of each of the plants used and then generated progeny plants homozygous for each of the S haplotypes identified. Using S homozygotes for RT-PCR has allowed unambiguous assignment of the resulting cDNAs to a particular S allele of the S-RNase gene. To obtain longer cDNAs for sequence comparison we chose a pair of PCR primers based on the sequence of the C2 region and the poly(A)⁺ tail; in this way, the cDNAs obtained lack only approximately 100 bp of the sequence upstream from the C2 region, and the deduced amino acid sequences downstream from the C2 region account for approximately 80% of the full-size S-RNases. We also used genomic-blotting analysis to examine the cDNAs obtained by RT-PCR to see whether each revealed an S haplotype specific RFLP. Subsequent sequencing of all the cDNAs showed that those cDNAs that revealed this pattern encode S-RNases and those that did not encode proteins unrelated to S-RNases.

Genetically Identical S Haplotypes Identified from Different Natural Populations

When studying natural populations of SI species, questions can be asked as to whether genetically identical S haplotypes are contained in different populations and if so, whether the deduced amino acid sequences of their corresponding S locus genes are identical. If differences are found, those amino acids that are different can be eliminated from consideration as playing a role in the determinant of S allelic specificity.

Prior to this study only three S haplotypes, S₁, S₂, and S₃, had been identified from P. inflata. Although the geographical origin of the plants from which these haplotypes were identified is not known, it is unlikely to be the same as the natural population studied here. It is thus interesting to find that one of the haplotypes, S₁, was also present in this natural population. In fact, no sequence differences were found between these two S₁ alleles of the S-RNase gene in the coding region or the 3'-non-coding region. Preliminary results suggest that the sequence conservation in the S locus of these two S₁ haplotypes likely extends beyond the S-RNase gene, because no RFLP was observed when the genomic DNA digests from these two S₁ haplotypes were hybridized with the cDNA for S₁-RNase. However, the conservation appears to break down in other parts of the S locus because RFLP was revealed when cDNAs for two S-linked pollen-expressed genes were used as hybridization probes. These findings suggest that these two S₁ haplotypes may have evolved independently for some time and that the S-RNase gene may be under strict functional constraint to maintain allelic identity. Perfect amino acid sequence conservation has also been found between pistil S proteins of two genetically identical S haplotypes identified in a British and a Spanish population of Papaver rhoeas, a species that possesses a different type of GSI from solanaceous species (Walker et al., 1996).

Role of Two Hypervariable Regions in S Specificity

Among all the solanaceous S-RNases that have been reported so far, the pair of S₆- and S₉-RNases identified in this work are most similar to each other in the two hypervariable regions; they have an identical sequence in HVa and only differ by two amino acids in HVb. These two regions have recently been suggested to be the sole determinant of the S allele specificity of S-RNases (Matton et al., 1997) based on the study of hybrid S-RNases between S₁₁- and S₁₃-RNases of S. chacoense. S₁₁- and S₁₃-RNases differ by 10 amino acids, three in HVa, and one in HVb, and the S allele specificity of the S₁₁-RNase was found to be converted to that of the S₁₃-RNase simply by swapping the HVa and HVb regions. However, as pointed out by Verica et al. (1998), these results do not necessarily mean that amino acids outside the two hypervariable regions are not involved in S allele specificity, because a domain-swapping experiment can only address the role of the regions swapped, but not those that are not swapped. It remains to be determined whether the two amino acids that are different between S₆- and S₉-RNases of P. inflata are sufficient for distinguishing the S allele specificity of one S-RNase from the other; however, considering that both differences are conservative replacements, this does not seem likely.

Intragenic Recombination As a Mechanism of Generation of Allelic Diversity

The closely related alleles of the S-RNase gene that we identified in a natural population of P. inflata have allowed us to obtain the first evidence that apart from accumulation of point mutations, allelic diversity of the S-RNase gene has also been generated by intragenic recombination. This finding is quite interesting, considering that recombination has previously been thought to be absent at the S locus to maintain tight linkage between the S-RNase gene and the pollen S gene.

One can envision that point mutations would first produce highly divergent polymorphic sequence motifs, and then intragenic recombination would shuffle these polymorphic sequence motifs into different allelic combinations. If accumulation of point mutations were solely responsible for allelic diversity of the S-RNase gene, favorable mutations would have to occur consecutively in the same allelic lineage. However, the additional involvement of intragenic recombination could bring together mutations that have occurred in different allelic lineages into new allelic forms and thereby speed up the process of allelic diversity generation. This evolutionary process is very similar to the diversification of alleles of the major histocompatibility (MHC) loci, which are also involved in self/nonself recognition (Hughes et al., 1993; Hughes and Yeager, 1998). Because balancing selection has maintained polymorphism for a long time at the MHC loci and the S locus, substantial allelic diversity is available for the process of recombination. Note that although there is evidence of recombination among alleles at MHC loci of mammals, there are certain MHC loci at which recombinants are rare (Hughes and Yeager, 1998). Thus, the extent to which recombination has contributed to the polymorphism of the S-RNase gene needs further investigation. Considering the potential deleterious effect of recombination on the SI system, we think recombination does occur, but at a low rate. Deleterious recombinants most likely have been eliminated from the population by natural selection. In the case of our observed recombination events that led to the generation of the S₃ and S₁₉ alleles of the S-RNase gene, presumably they had selectively neutral or even advantageous effects, thus allowing plants harboring the recombinant alleles to remain in the population.

If recombination and/or point mutations result in a new S allele specificity of the S-RNase gene, there must be complementary changes in the pollen S gene for an SI plant to remain SI while acquiring a new haplotype specificity. How the pistil and pollen genes "communicate" with each other during the mutational process is one of the most perplexing questions in the study of SI. A solution has recently been proposed based on the finding of a hybrid S-RNase that possesses dual S allele specificity (Matton et al., 1999; see discussion in McCubbin and Kao, 2000). According to this model, mutations in the S-RNase gene could result in a protein that retains the original specificity, as well as acquires a new specificity. Subsequent mutations in the pollen S gene could render its protein product able to recognize the amino acids of the mutated S-RNase that specify the new specificity, but not those that specify the original specificity. Further mutation in the S-RNase gene with dual specificity would result in the loss of the original specificity. This process would preserve SI during the entire process of the generation of a new S haplotype specificity. However, since there have never been any reports of plants in natural populations whose pistils or pollen possess dual specificity, the validity of this model remains to be determined.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Plant Material

The natural population of Petunia inflata R. E. Fr. (Petunia integrifolia [Hook.] Schinz & Thell. subsp. inflata [R. E. Fr.] Wijsman) selected for this study was designated A9. It was located 1.6 km east of the junction of Ruta Nacional 4 and Ruta Provincial 103, Municipal Candelaria, Provincial Misiones, Argentina (27 23'13"S, 55 34'00"W). The population was comprised of approximately 100 individuals and was established at a gentle load-side slope poorly covered with vegetation. Seed capsules were randomly collected on November 28, 1990, and the seeds were mixed together and stored in a refrigerator. Thirty-six plants were raised from the seeds, and they were coded from A9-1 to A9-36. All of them were SI as judged by their failure to set seed upon selfing and by the arrest of pollen tube growth in the upper one-third segment of the style.

Determination of S Genotypes

Pollen from plants homozygous for the previously identified S₁, S₂, and S₃ haplotypes of P. inflata (Ai et al., 1990) was used to pollinate plants A9-1 through A9-36. The only pollination that did not set any capsules was between A9-4 and the S₁S₁ homozygote, indicating that one of the S haplotypes carried by A9-4 was S₁. The S genotype of A9-1 was arbitrarily assigned S₄S₅. The scheme of genetic crossing used to determine the S genotypes of plants A9-2 through A9-36 was similar to that described by Tsukamoto et al. (1999).

Isolation of Total RNA from Pistils

Pistils (without ovaries) were collected 1 or 2 d before anthesis from plants homozygous for a particular S haplotype, frozen in liquid nitrogen, and were ground to fine powder. Total RNA was extracted from the powder using TRIzol reagent (Life Technologies, Rockville, MD) according to the protocol recommended by the manufacturer. RNA was recovered by precipitation with isopropyl alcohol. After washing with 75% (v/v) ethanol, the RNA pellet was dissolved in water and stored at 80°C.

Isolation of Genomic DNA

For each S genotype, 0.2 to 0.3 g of freshly collected young leaves were pulverized in liquid nitrogen, and genomic DNA was extracted from the homogenate with Plant DNAzol Reagent (Life Technologies). Chloroform was then used to remove polysaccharides. Following chloroform extraction, plant debris was removed by centrifugation and DNA was precipitated from the supernatant with ethanol. The resulting DNA pellet was washed with ethanol and dissolved in water.

RT-PCR, Cloning, and Sequencing of cDNAs Encoding S-RNases

To amplify cDNA for each S-RNase, first-strand cDNA was synthesized at 42°C for 60 min in a 20-µL reaction buffer (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl₂, and 10 mM dithiothreitol) that contained 1 to 5 µg of total RNA, 500 ng of oligo (dT)₁₇, 0.5 mM dNTPs, and 200 units of SuperScript II reverse transcriptase (Life Technologies). After the reaction, a 0.5-µL aliquot of the single-stranded cDNA was used as a template for PCR amplification in a 20-µL reaction buffer (20 mM Tris-HCl, pH 8.4, 2 mM MgCl₂, and 50 mM KCl) containing 2.5 µM upstream primer, 2.5 µM oligo (dT)₁₇, 0.1 mM dNTPs, and 2 units of Display Taq DNA polymerase (PGC Scientific, Gaithersburg, MD). The upstream primer was a degenerate primer, 5'-ATTCAYGGGCTTTGGCCC-3' (Y indicating C or T), which was designed based on the sequence of the conserved region C2 of S-RNases. The reaction mixture was denatured at 94°C for 2 min and then cycled as follows: 94°C for 30 s, 40°C for 2 min, and extension at 72°C for 30 s. After the first five cycles the annealing condition was changed to 45°C for 90 s. The reaction was continued for 25 more cycles with an additional 5-min extension after the final cycle. Aliquots of all PCR reaction mixtures were analyzed on 1% (w/v) agarose gels. The remainder of each PCR reaction mixture was purified, ligated into pGEM-T vector (Promega, Madison, WI), and transformed into Escherichia coli. Plasmid DNA was prepared from multiple colonies of each transformation using a NucleoSpin Plasmid Miniprep Kit (CLONTECH, Palo Alto, CA). Sequencing was carried out on an ABI 60 sequencer at the Pennsylvania State University Life Sciences Consortium's Nucleic Acid facility. To eliminate potential errors introduced by PCR, two or more clones obtained from independent runs of RT-PCR were sequenced for each S haplotype.

DNA-Blot Analysis

Plasmid DNA (100 ng) was digested with EcoRI (1 unit) and genomic DNA (5 µg) was digested with EcoRI (20 units); the digests were separately electrophoresed on 0.7% (w/v) agarose gels at 25 V overnight. The gels were blotted to Biodyne B nylon membranes (Life Technologies). The DNA fragments were fixed on the membranes by baking at 80°C for 2 h. The membranes were pre-hybridized at 65°C for 1 h in a solution containing 10% (w/v) Dextran sulfate, 1 M NaCl, and 1% (w/v) SDS. Radiolabeled DNA probes were prepared by using an RTS RadPrime DNA Labeling kit (Life Technologies). Hybridization was performed at 65°C overnight in the pre-hybridization solution plus radiolabeled probe. The membranes were washed twice in 0.1× SSC, 0.1% (w/v) SDS for 30 min each (high stringency) or in 2× SSC, 0.1% (w/v) SDS for 30 min each (low stringency), and exposed to phosphoscreen for 8 h. The screen was scanned by a phosphorimager (Molecular Dynamics, Sunnyvale, CA).

Statistical Methods

Nucleotide and deduced amino acid sequences of cDNAs for S-RNases were aligned using CLUSTAL X (Thompson et al., 1997). The alignment of nucleotide sequences was further modified according to the corresponding amino acid sequence alignment. Phylogenetic trees were constructed by the neighbor-joining method (Saitou and Nei, 1987) using the Molecular Evolution Genetic Analysis package (Kumar et al., 1993). The method of Nei and Gojobori (1986) was used to estimate the p_s and the p_n. The reliability of clustering patterns of the phylogenetic trees was tested by bootstrapping (Felsenstein, 1985). A test for the occurrence of patterns of polymorphic sites suggestive of recombination was conducted using the MAXCHI program (McGraw et al., 1999), which implements a version of Smith's (1992) maximum chi-square method.