Research Article

Collaborative Non-Self Recognition System in S-RNase–Based Self-Incompatibility

See allHide authors and affiliations

Science  05 Nov 2010:
Vol. 330, Issue 6005, pp. 796-799
DOI: 10.1126/science.1195243

Dissecting Self-Incompatibility

Although the pollen may be available for a flower to fertilize itself, molecular determinants on the pollen and the pistil prevent inbreeding in a process termed self-incompatibility. In the Petunia self-incompatibility, if male determinants (F-box proteins) on pollen are recognized by a female ribonuclease determinant on the pistil, the pollen tube is killed when its ribosomal RNA is digested. Outcrossed fertilizations can occur because of allelic diversity in the female that fails to recognize its male counterparts; however, the genetic diversity of the ribonuclease gene is greater than that of the known F-box gene. Kubo et al. (p. 796; see the Perspective by Indriolo and Goring) have discovered that there are several related F-box genes in Petunia, each of which brings its own allelic diversity to bear—thus, increasing the variety of potential mating partners.


Self-incompatibility in flowering plants prevents inbreeding and promotes outcrossing to generate genetic diversity. In Solanaceae, a multiallelic gene, S-locus F-box (SLF), was previously shown to encode the pollen determinant in self-incompatibility. It was postulated that an SLF allelic product specifically detoxifies its non-self S-ribonucleases (S-RNases), allelic products of the pistil determinant, inside pollen tubes via the ubiquitin–26S-proteasome system, thereby allowing compatible pollinations. However, it remained puzzling how SLF, with much lower allelic sequence diversity than S-RNase, might have the capacity to recognize a large repertoire of non-self S-RNases. We used in vivo functional assays and protein interaction assays to show that in Petunia, at least three types of divergent SLF proteins function as the pollen determinant, each recognizing a subset of non-self S-RNases. Our findings reveal a collaborative non-self recognition system in plants.

Self-incompatibility (SI) is an intraspecific reproductive barrier adopted by angiosperms that allows the pistil to distinguish between self (genetically related) and non-self (genetically unrelated) pollen. In most cases, this self/non-self discrimination is controlled by male- and female-specificity determinants (pollen-S, style-S) encoded by multiallelic genes at the S locus (1, 2). Self-incompatible species in Solanaceae, Rosaceae, and Plantaginaceae use extracellular S-RNase as style-S (3). If the S haplotype of pollen matches either S haplotype of the style, S-RNase exerts cytotoxicity inside the self-pollen tube to inhibit growth (3). Pollen-S was identified as an F-box protein, named S-locus F-box (SLF or SFB) (46), which may be a component of an SCF (Skp1–Cullin1–F-box) or SCF-like complex (7, 8).

A protein degradation model was proposed to explain S haplotype–specific rejection of pollen tubes by S-RNase. It predicts that an SLF allelic variant specifically recognizes its non-self S-RNases and mediates their degradation by the ubiquitin–26S-proteasome system (1, 8, 9). This model can explain competitive interaction, where SI breaks down in heteroallelic pollen carrying two different pollen-S alleles (10, 11). Each SLF allelic product in heteroallelic pollen mediates degradation of all S-RNases except its self S-RNase, and two different SLF allelic products together mediate the degradation of all S-RNases, rendering the pollen tube compatible with styles of any S genotype (3, 8). Experiments designed on the basis of competitive interaction showed that PiSLF2 (S2 allele of Petunia inflata SLF) functions as pollen-S (6). When PiSLF2 was introduced into S1S2 and S2S3 plants, it caused breakdown of SI in S1 and S3 pollen, but not in S2 pollen, as predicted by competitive interaction (6).

Thus far, PiSLF2 is the only SLF allele in Petunia shown to function as pollen-S (6, 9). SLF shows much lower allelic sequence diversity than S-RNase, and nonsynonymous substitution rates of SLFs from Antirrhinum, Petunia, and Prunus are 0.01 to 0.11, whereas those for S-RNase are 0.14 to 0.51 (12, 13). Given the large number of S haplotypes within each species, it is puzzling how an SLF allelic product could recognize a large repertoire of highly divergent non-self S-RNases to allow cross-compatible pollinations. Moreover, phylogenetic studies of SLF and S-RNase in Solanaceae and Plantaginaceae showed no evidence of coevolution, with SLF having a much shorter evolutionary history (12), which is unexpected for the “male” and “female” genes encoding proteins directly involved in self/non-self discrimination during sexual reproduction. Here, we address the question of whether the previously identified SLF in Petunia is the only protein that constitutes the pollen determinant in SI.

Previously studied SLF is not the sole element of pollen-S. We first cloned four additional alleles of Petunia SLF from pollen cDNA of S5, S7, S9, and S11 homozygotes by 3′- and 5′-RACE (rapid amplification of cDNA ends) with the use of primers (table S1) designed on the basis of PiSLF1, PiSLF2, and PiSLF3 sequences (6, 14). The deduced amino acid sequences of all nine of the identified SLF alleles exhibited higher sequence similarities (86.4 to 100% identity) than the corresponding nine S-RNase alleles (40.1 to 79.4% identity) (fig. S1). Because the taxa with S-RNase–based SI have multiple SLF-like genes (5, 9), we renamed SLF “type-1 SLF,” designating alleles as Sn-SLF1, with n denoting the S haplotype.

A surprising finding from the sequence comparison was that the deduced amino acid sequence of S7-SLF1 is identical with that of S19-SLF1 (previously named PaSLF19) (11), although the amino acid sequences of S7- and S19-RNase are 45% identical (fig. S1). Reciprocal pollinations between S7 and S19 homozygotes showed that all pollinations were compatible (Fig. 1A), confirming that S7 and S19 are distinct S haplotypes. The finding of the identical SLF1 in two different S haplotypes raised the possibility that SLF1 is not the sole element of pollen-S.

Fig. 1

Type-1 SLF is not the sole element of pollen-S. (A) Reciprocal pollinations between S7- and S19-homozygous plants of Petunia. Pollen tubes were stained with aniline blue and monitored by a fluorescence microscope. Presence of a large number of pollen tubes at the basal end of the style (arrowhead) indicates compatible pollination. (B to G) In vivo functional assays of S7-SLF1. Transgenic plants S5S7/S7-SLF1 (B), S7S11/S7-SLF1 (C), and S5S19/S7-SLF1 (D) retained self-incompatibility. Transgenic plant S7S9/S7-SLF1 (E) exhibited breakdown of SI. Pollen from S7S9/S7-SLF1 was compatible with wild-type S7S9 pistils (F), whereas pollen from S7S9 wild-type plants was incompatible with S7S9/S7-SLF1 pistils (G). (H) PCR genotyping of 10 progeny plants from crossing a wild-type (WT) S7S9 plant and transgenic (T) S7S9/S7-SLF1 plant. Primers specific for each transgene and for the S-RNase alleles indicated were used for PCR. All genotyping results are shown in table S3.

To address this possibility, we first examined whether S5-, S7- (= S19-), S9-, and S11-SLF1 function as pollen-S. We introduced each transgene construct (fig. S2) into appropriate S heterozygotes of Petunia (table S2) (14) and confirmed expression of each transgene in pollen by reverse transcription polymerase chain reaction (RT-PCR) (fig. S3). When an S7-SLF1 transgene was introduced into S5S7, S7S11, and S5S19 plants, all transgenic plants remained self-incompatible (Fig. 1, B to D, and table S2), indicating that it did not cause competitive interaction either in homoallelic S7 and S19 pollen or in heteroallelic S5 and S11 pollen.

When the S7-SLF1 transgene was introduced into S7S9 plants, however, they exhibited breakdown of SI (Fig. 1E), and reciprocal crosses with wild-type S7S9 plants showed that S7-SLF1 caused breakdown of pollen, but not style, function in SI (Fig. 1, F and G). To determine whether breakdown of SI resulted from competitive interaction, we examined the inheritance of the S7-SLF1 transgene and progeny S genotypes from pollination of wild-type S7S9 plants with pollen from an S7S9/S7-SLF1 plant. PCR analysis revealed that all 26 plants examined carried the transgene and were either S9S9 (14 plants) or S7S9 (12 plants) (Fig. 1H and table S3). The absence of an S7S7 genotype suggested that only S9 pollen, but not S7 pollen, carrying the S7-SLF1 transgene became compatible with S7S9 pistils. Competitive interaction was also observed when the S7-SLF1 transgene was expressed in S17 pollen (table S2); this confirms our previous conclusion that duplicated S7-SLF1 caused breakdown of SI in a naturally occurring self-compatible line of Petunia carrying an S17 haplotype (11). These results and those from other transgenic experiments (table S2) suggested that S7-SLF1 causes competitive interaction in only a subset of heteroallelic pollen (S9 and S17 pollen, but not S5 or S11 pollen) (Table 1).

Table 1

Effects of SLF transgenes on SI behavior of transgenic pollen. Abbreviations: +, breakdown of SI caused by competitive interaction; –, no breakdown of SI; (homo), homoallelic relationship between the S haplotype of the transgene and that of pollen.

View this table:

Plants carrying S5-, S9- or S11-SLF1 transgenes were similarly analyzed (tables S2 and S3); their effects on SI behavior are summarized in Table 1. As with S7-SLF1, each caused competitive interaction only in pollen of a subset of non-self S haplotypes. We further examined the function of previously identified S2-SLF1 and S3-SLF1 of P. inflata (14). Consistent with our previous finding (6, 9), S2-SLF1 caused competitive interaction in S3 pollen. However, neither S3-SLF1 nor S3-SLF1:GFP caused competitive interaction in S2 pollen, even though RT-PCR showed that both were expressed in pollen; the level of the green fluorescent protein (GFP) fusion protein S3-SLF1:GFP was comparable to that of S2-SLF1:GFP produced in a previously generated S2S3 transgenic plant, where it caused breakdown of SI (9) (fig. S4, and tables S2 and S3).

All in vivo functional assays suggested that each SLF1 allele elicits competitive interaction in pollen of a subset of its non-self S haplotypes. On the basis of the protein degradation model, this would suggest that for each S haplotype, an SLF1 allelic variant only recognizes and interacts with a subset of its non-self S-RNases and mediates their degradation. This raised the question of what other protein(s) encoded at the S locus might be required.

Pollen-S comprises multiple types of SLF genes. In P. inflata, several SLF-like genes were found tightly linked to the S locus (9). PiSLFLb-S2, PiSLFLc-S1, and PiSLFLd-S2 were studied, but these did not cause competitive interaction in their respective heteroallelic pollen (9). We hypothesized that these and additional SLF-like genes might be elements of pollen-S, but like SLF1, their function would only be revealed in pollen of appropriate S haplotypes. To test this hypothesis, we designed PCR primers (table S1) based on previously characterized SLF-like genes, and extensively amplified SLF-like cDNA fragments by 3′- and 5′-RACE from pollen of S5, S7, S9, S11, S17, and S19 haplotypes (14).

We identified 30 SLF-like sequences and classified all except S5-FBX into five subgroups (Fig. 2), named type-2 to type-6 SLF. Using the same nomenclature as for type-1 SLF, we designated alleles Sn-SLFx. The previously identified PiSLFLc-S1 and PiSLFLd-S2 (9) were renamed S1-SLF2 and S2-SLF3. We identified the alleles of all SLF types from these six S haplotypes, except that of type-2 SLF from the S9 haplotype (fig. S5). Sequence identities between allelic variants of each type were 72.2 to 87.5% (type 2), 70.3 to 99.0% (type 3), 89.6 to 96.5% (type 4), 96.8 to 98.2% (type 5), and 93.4 to 95.9% (type 6). In contrast, identities between different types of SLF were only ~50% (fig. S6).

Fig. 2

Genealogies of SLF and S-RNase from Petunia. (A and B) Phylogenetic trees of deduced amino acid sequences of SLF genes (A) and S-RNase genes (B) were created with the bootstrap neighbor-joining algorithm (14). Both trees are shown to the same scale; the bar for each tree indicates the number of amino acid substitutions per site. Numbers on the branches indicate bootstrap values when the number of bootstrap trials is 1000. S1, S2, and S3 alleles of these genes are from P. inflata (6), S17 and S19 alleles are from P. axillaris (11); all others are from P. hybrida.

All six types of SLF genes showed similar developmentally regulated, male reproductive organ–specific expression profiles: Transcripts increased during anther development, peaked before anthesis, and declined in mature pollen and pollen tubes, although the timing of increase and rate of decline differed slightly among members (fig. S7). Examination of 43 plants segregating for S5, S7, S9, and S11 haplotypes and 33 plants segregating for S17 and S19 haplotypes showed no recombination between any allele of each type of SLF and S-RNase (fig. S8), which suggests that all six types of SLF genes are linked to the S locus (within 1 cM). These properties implied that the additional five types of SLF genes could potentially be elements of pollen-S.

We examined the in vivo function of the S5, S7, and S11 alleles of type-2 SLF and the S7 and S11 alleles of type-3 SLF (fig. S2) (14); the effects on SI behavior are summarized in Table 1 (see also tables S2 and S3). All except S7-SLF3 caused competitive interaction; however, they affected different subsets of non-self S haplotypes of pollen. For the six S haplotypes of pollen examined, competitive interaction was observed in all but S5 pollen. The findings that none of the three types of SLF genes of either S7 or S11 haplotype caused competitive interaction in S5 pollen suggest that some other type(s) of SLF in S7 and S11 haplotypes mediate detoxification of S5-RNase, allowing them to be compatible with S5 styles. Thus, our transformation experiments suggested that at least three types of SLF proteins function as elements of pollen-S.

Each SLF type interacts with only a subset of S-RNase. We have modified the protein degradation model to consider all types of SLFs. For example, S9 pollen is incompatible with S9 styles because none of the multiple types of SLFs produced in S9 pollen recognizes S9-RNase, whereas S9 pollen expressing S7-SLF2 is compatible with S9 styles, because S7-SLF2 recognizes S9-RNase and mediates its degradation. To test the validity of this model, we performed coimmunoprecipitation experiments. We expressed FLAG-tagged S7-SLF2 (fig. S2) in pollen of an S5S11 heterozygote, and found that it, like S7-SLF2, elicited competitive interaction in S11 pollen (tables S2 and S3), suggesting that fusion of the FLAG tag did not affect SI function of S7-SLF2.

Pollen extracts from S5S11/FLAG:S7-SLF2 and S5S11 wild-type plants were mixed with style extracts from various S homozygotes and immunoprecipitated with immobilized antibody to FLAG. When style extracts of S9 and S11 homozygotes were separately mixed with pollen extracts, strong bands of S9-RNase and S11-RNase were detected in the immunoprecipitates from transgenic pollen extracts, but only a faint S11-RNase band of background level was detected in wild-type pollen extracts. This finding suggested that S7-SLF2 interacts strongly with S9-RNase and S11-RNase (Fig. 3, A and B). When style extracts of S7 and S5 homozygotes were mixed with pollen extracts of the transgenic plants, a faint S7-RNase band of background level and no S5-RNase band were detected (Fig. 3, C and D). These results suggested that the ability of the S7-SLF2 transgene to cause competitive interaction in S9 and S11 pollen, but not in S5 and S7 pollen, is due to specific interactions between S7-SLF2 and S9- and S11-RNases.

Fig. 3

Analysis of interactions between S7-SLF2 and four allelic variants of S-RNase [(A) S9-RNase, (B) S11-RNase, (C) S7-RNase, (D) S5-RNase] by coimmunoprecipitation. Pollen extracts from a transgenic plant (TG) expressing FLAG:S7-SLF2 or from a wild-type plant (WT) were mixed with style extracts from (A) S9, (B) S11, (C) S7, and (D) S5 homozygotes and then immunoprecipitated using an antibody to FLAG. Immunoprecipitates (IP) were analyzed by immunoblotting with antibodies specific for each S-RNase; 0.1% and 1% of the style extracts (input) used for the immunoprecipitation were loaded as controls. Immunoprecipitated FLAG:S7-SLF2 was also detected by an antibody specific for S7-SLF2 using the same membranes after detection of S-RNases (lower panels). In each upper panel, arrowheads indicate S-RNase bands. The double bands in (A) and (B) represent differentially glycosylated forms of S-RNase, as a band with smaller molecular mass was observed when each S-RNase was subjected to glycosidase treatment. The asterisk in (B) indicates an unidentified S11-RNase derivative. The double asterisk in (D) indicates a dimeric form of S5-RNase, as it was not observed in higher concentration (e.g., 50 mM) of dithiothreitol than was used in the SDS–polyacrylamide gel electrophoresis shown.

A collaborative recognition model. On the basis of this study, we propose a “collaborative non-self recognition” model for S-RNase–based SI (Fig. 4). In Petunia, pollen-S comprises multiple types of SLF genes. Within an S haplotype, the product of each type of SLF interacts with a subset of non-self S-RNases, and the products of multiple types, including yet uncharacterized ones, are required for the entire suite of non-self S-RNases to be collectively recognized and detoxified. This relationship between style-S and pollen-S is radically different from that of the SI systems possessed by the Brassicaceae and Papaveraceae, where both style-S and pollen-S are single genes, and self-interaction between style-S and pollen-S triggers SI responses (1, 15, 16).

Fig. 4

Collaborative non-self recognition model for SI in Petunia. The single S-RNase gene constituting style-S and the multiple SLFs constituting pollen-S are depicted as boxes and ovals, respectively. The locations of these SLFs relative to S-RNase are as yet undetermined, but for convenience they are placed in a cluster. For each S haplotype, the SLFs whose products are responsible for detoxifying one or more of the five allelic variants of S-RNase tested are connected by solid arrows with their target alleles. For each S haplotype, products of one or more Other SLFs (including SLF3s of S5 and S9 haplotypes) are predicted to target products of the remaining alleles of S-RNase connected by broken arrows.

In plants possessing S-RNase–based SI, increasing the repertoire of SLF genes that constitute pollen-S would be advantageous, as this would increase the number of potential mating partners by allowing pollen to recognize and inactivate more non-self S-RNases, whereas an increase in diversity of the S-RNase gene would have the opposite effect by allowing new S-RNases to escape detoxification by the existing repertoire of SLF proteins. Other self-incompatible species in Solanaceae, Plantaginaceae, and Rosaceae also have a single S-RNase gene and multiple SLF/SFB genes located at their S loci (5, 1720). For example, 10 and 12 SLF/SFB-like genes are linked to the S loci of Nicotiana alata (Solanaceae) and apple (Malus × domestica, Rosaceae), respectively (18, 20). It would be interesting to determine whether any of these S-linked F-box genes function as elements of pollen-S.

The collaborative non-self recognition model can be further tested through analysis of loss-of-function mutants of one or more types of SLF genes. So far, one such naturally occurring self-compatible mutant of Petunia has been identified, and the breakdown in SI is attributed to competitive interaction caused by duplication of an SLF1 (11). Our previous model invoking a single SLF gene for pollen-S predicted that loss of function of pollen-S would be lethal, because such mutant pollen would be incapable of detoxifying any non-self S-RNases and would thus be incompatible with styles of any S haplotype (3). However, according to our new model, loss of function in a single type of SLF would be lethal only for the haplotypes whose S-RNases are recognized by the type of SLF affected by mutation. Note that the SI behavior of a self-compatible cultivar, Osa-Nijisseiki, of Japanese pear (Pyrus pyrifolia, Rosaceae) is consistent with this prediction. This mutant has a 236-kb deletion in the S locus of the S4 haplotype, including an SLF-like gene, S4F-box0 (21), and its S4 pollen is incompatible with S1 styles, although it is compatible with the styles of other non-self S haplotypes tested (22). Our model would predict that S4F-box0 is an element of pollen-S in Pyrus that specifically recognizes S1-RNase.

SI has been compared with adaptive immunity in vertebrates (23, 24). The use of products of multiple polymorphic SLF genes to collectively recognize a suite of non-self S-RNases in order to mediate their degradation is conceptually similar to specific recognition of foreign antigens by a large repertoire of T cell receptors. Our findings provide fertile ground for studying the emergence of multiple types of SLF genes, their coevolution with a single S-RNase gene, and the biochemical basis that allows a particular type of SLF to recognize certain non-self S-RNases but not others or self S-RNase.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 to S3


References and Notes

  1. See supporting material on Science Online.
  2. We thank A. Miyawaki for Venus vectors; M. Iwano, H. Shiba, Y. Wada, P. Nakayama, M. Kakita, X. Meng, T. Tsuchimatsu, and K. K. Shimizu for discussion; and N. Katsui, M. Matsumura-Kawashima, M. Okamura, Y. Egusa, K. Kajihara, M. Bothe, and M. Goralczyk for technical assistance. Supported by Grant-in-Aid for Creative Scientific Research (A.I.) and Grant-in-Aid for Scientific Research (A) from the Japan Society for the Promotion of Science (S.T.), Grant-in-Aid for the Global Center of Excellence Program from the Ministry of Education, Culture, Sports, Science and Technology of Japan to the Nara Institute of Science and Technology, and NSF grants IOB-0543201 and IOS-0843195 (T.-h.K.). Sequence data have been deposited in GenBank under accession numbers AB568388 to AB568423.
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article