How far are we from unravelling self-incompatibility in grasses?

Authors


Author for correspondence:
Susanne Barth
Tel: +353 (0) 5991 70290
Fax: +353 (0) 5991 42423
Email: susanne.barth@teagasc.ie

Summary

The genetic and physiological mechanisms involved in limiting self-fertilization in angiosperms, referred to as self-incompatibility (SI), have significant effects on population structure and have potential diversification and evolutionary consequences. Up to now, details of the underlying genetic control and physiological basis of SI have been elucidated in two different gametophytic SI (GSI) systems, the S-RNase SI and the Papaver SI systems, and the sporophytic SI (SSI) system (Brassica). In the grass family (Poaceae), which contains all the cereal and major forage crops, SI has been known for half a century to be controlled gametophytically by two multiallelic and independent loci, S and Z. But still none of the gene products for S and Z is known and only limited information on related biochemical responses is available. Here we compare current knowledge of grass SI with that of other well-characterized SI systems and speculate about the relationship between SSI and grass SI. Additionally, we discuss comparative mapping as a tool for the further investigation of grass SI.

Self-incompatibility in grass species

The grass family (Poaceae) is the fourth largest family of flowering plants, comprising > 10 000 species of economic and ecological importance (Watson & Dallwitz, 1992). Grasses include all the cereal crops and 75% of the cultivated forage crops (Nelson & Moser, 1995). They are more extensively adapted to all temperature ranges and rainfall habitats than any other family of flowering plants (Watson & Dallwitz, 1992). Self-incompatibility (SI) has been known in flowering plants for over a century since Darwin's description in 1876 (Darwin, 1876). SI was defined by deNettancourt (1977) as ‘the inability of a fertile hermaphrodite seed plant to produce zygotes after self-pollination’. The pistil is able to distinguish self and same-species nonself pollen with common S alleles, and prevent fertilization, thus circumventing the tendency towards self-fertilization and consequent inbreeding. The promotion of outbreeding through SI has played a significant role in the evolution and diversification of grass species (Pandey, 1977). The effectiveness of SI in grasses limits to a certain extent the efficient production of inbred lines and hybrids in plant breeding but also ensures the maintenance of heterozygosity in wild populations and, potentially, could contribute to adaptive success. SI is widely distributed in grasses and has been reported in at least 16 genera within five of the thirteen subfamilies of the Poaceae (Connor, 1979) (Table 1).

Table 1.  Grass subfamilies and tribes where self-incompatibility (SI) has been identified and overview of some incompatible and compatible species (after Connor, 1979; Li et al., 1997; Baumann et al., 2000)
SubfamilyTribeSpeciesSelf-incompatibility
PooideaePoeaeBriza mediaYes
 Briza minorNo
 Cynosurus cristatusYes
 Dactylis aschersonianaYes
 Festuca pratensisYes
 Lolium perenneYes
 Lolium multiflorumYes
 Lolium rigidumNo
AveneaeAnthoxanthum odoratumYes
 Avena barbataNo
 Phalaris coerulescensYes
TriticeaeSecale cerealeYes
 Hordeum bulbosumYes
 Hordeum vulgareNo
 Triticum aestivumNo
PanicoideaeAndropogoneaeSorghastrum nutansYes
 Themeda australisNo
 Zea maysNo
ChloridoideaeChlorideaeChloris gayanaYes
 Chloris striateNo
ArundinoideaeDanthonieaeDanthonia linkiiNo
 Molinia caeruleaYes
BambusoideaeOryzeaeOryza barthiiYes
 Oryza sativaNo

Evolutionary properties of self-incompatibility systems

Classic genetic studies in the early 20th century revealed two major classes of SI systems, gametophytic and sporophytic (deNettancourt, 1977). In many cases, SI is controlled by a single multi-allelic locus, the S locus, but complementary multiple-locus systems also exist. In gametophytic SI (GSI), the SI phenotype of pollen is determined by its own (haploid) S genotype. The pollen is rejected when the S haplotype of the haploid pollen matches either of the two S haplotypes of the diploid pistil. In sporophytic SI (SSI), the SI phenotype of pollen is determined by the S genotype of its diploid parent, and thus dominance interactions between S alleles are possible. When both pollen and stigma S alleles are co-dominant, the pollen is recognized as self and rejected if either of the two S haplotypes of its parent matches one of the two S haplotypes of the pistil. However, interactions of S alleles can occur independently for pollen and stigma, leading to complicated compatibility/incompatibility patterns and differences in reciprocal compatibility. Recent molecular studies on dominance have grouped haplotypes of SSI pollen and stigma determinants into two classes, class I and class II (Shiba et al. 2002). On the pollen side, class I haplotypes are generally dominant or co-dominant, whereas class II haplotypes are recessive to class I (Kusaba et al., 2002; Shiba et al., 2002). In the stigma, however, dominance interactions are not as common as in pollen and might be regulated by downstream factors during the SI response (Hatakeyama et al., 2001).

Self-incompatible pollen is rejected at some point in the pollination process, depending on the species. This may happen at pollen hydration or germination, during growth through the style, in the ovule, or even post-fertilization. The general relationships among pollen type, stigma type and pollen inhibition site in SI species have been reviewed by Heslop-Harrison & Shivanna (1977), as ‘stigmatic inhibition–dry stigma–tri-nucleate pollen’ and ‘stylar inhibition–wet stigma–bi-nucleate pollen’. The variety of mechanisms involved in different SI systems has led to the suggestion that SI systems may have evolved independently many times during the diversification of flowering plants (Steinbachs & Holsinger, 2002). The evolutionary properties of SI systems have been revealed using both empirical advances and theoretical developments following Wright's theory of evolutionary dynamics of SI alleles (Wright, 1939). Several evolutionary features are shared by different SI systems, including: high allelic richness at the S locus within a species; high levels of nucleotide diversity among S alleles within a species; and low nucleotide diversity among different gene copies of the same allelic specificity. With a recent accumulation of biochemical and molecular studies in different SI mechanisms, effects concerning the evolutionary process of SI systems, such as frequency-dependent selection and the evolution of S-locus linked genes, have also been investigated and comprehensively reviewed by Castric & Vekemans (2004).

How much is known of the underlying genetic control?

Studies in Secale cereale by Lundqvist (1954) and Phalaris coerulescens by Hayman (1956) showed that SI in grasses is controlled gametophytically by at least two multiallelic and independent loci, S and Z. The incompatibility phenotype of the pollen grain is determined by its haploid genome and depends upon the combination of S and Z alleles in the pollen grain. A pollen grain is incompatible when both its S and Z alleles are matched in the pistil (Fig. 1). In addition to S. cereale and P. coerulescens, the S-Z system has been characterized in many other grass species (reviewed in Li et al., 1997). Studies on the breakdown of SI in grasses have revealed the existence of self-compatible pollen mutants and the involvement of an additional locus in at least three species, which comparative mapping suggests is probably the same locus: a self-compatibility (SC) locus independent of the S and Z loci has been reported in Lolium perenne (Thorogood & Hayward, 1991; Thorogood et al., 2005); at least three self-compatible mutants at the S and Z loci and at least a third locus (T) have also been reported in P. coerulescens (Hayman & Richter, 1992); in S. cereale, three incompatibility loci, S, Z and S5 (analogous to T in Phalaris) have been identified (Voylokov et al., 1998).

Figure 1.

Genetic control of gametophytic self-incompatibility (GSI) by two multiple-allelic loci S and Z. When both pollen S and Z alleles are matched in the pistil, incompatibility occurs, and pollen growth is inhibited. Otherwise pollen is compatible. The degree of compatibility can be 0, 50, 75 and 100% compatible, depending on the genotypes of pollen and stigma. Reciprocal crosses (marked with ‘*’) between plants of the genotypes S1S2Z1Z2 and S1S1Z1Z2 produce pollen of different proportions of compatible pollen. If a S1S1Z1Z2 genotype is the pollen donor and crossed with S1S2Z1Z2 genotype pistils, the pollen is incompatible, while, if a S1S2Z1Z2 genotype is the pollen donor and pollinated on S1S1Z1Z2 pistils, 50% of the pollen is compatible.

Compared to the Poaceae, single-locus (S) controlled gametophytic SI (GSI) is more widespread, present in many families, including the Solanaceae, the Liliaceae, the Rosaceae, and the Papaveraceae. In basal eudicot and monocot species, complex four-locus gametophytic control has been reported, for example in sugarbeet (Beta vulgaris) (Larsen, 1977), Ranunculus acris (Østerbye, 1975), Ranunculus repens (Lundqvist, 1990) and Lilium martagon (Lundqvist 1991), although more persuasive molecular and genetic evidence than the existing diallel cross-based breeding results are required for the confirmation of the four-locus SI system (Mayo & Leach, 1993). A joint origin of the incompatibility systems in monocots and basal eudicots has been proposed (Østerbye, 1975) and it is thought that the one-locus SI systems may have been derived from the complex multi-locus systems (Østerbye, 1975; Heslop-Harrison, 1978). However, the identification of the self-compatibility locus in the grass species L. perenne (Thorogood & Hayward, 1991; Thorogood et al., 2005), P. coerulescens (Hayman & Richter, 1992) and S. cereale (Voylokov et al., 1998) and the existence of a fourth locus in L. perenne that interacts with the S locus, preventing the transmission of specific S-locus–‘fourth-locus’ allele combinations (Thorogood et al., 2002), is consistent with the hypothetical close evolutionary relationship of grass SI systems with the multi-locus system of basal eudicots.

As a result of the extra locus, the S-Z two-locus SI system in grasses has features distinct from those of single-locus GSI systems. These include: the maintenance of SI at the polyploid level (Lundqvist et al., 1973), although it has been demonstrated that the breakdown of SI in tetraploid Prunus cerasus (Rosaceae) is not caused by polyploidy but by the presence of two or more nonfunctional S haplotypes within an individual (Hauck et al., 2006); the differences between two individuals in the degree of compatibility: the percentage of compatible pollen can be 0, 50, 75 or 100%, depending on the genotypes (Fig. 1); the existence of reciprocal differences in crosses between two plants. For example, if a plant with the genotype S1S2Z1Z2 crosses with S1S1Z1Z2 as the pollen donor, the pollen grains will be incompatible, while, in the reciprocal cross, 50% of the pollen will be compatible (Fig. 1); and with the two-locus system, homozygosity is possible at one of the loci, in contrast to the single-locus system, where homozygosity is theoretically impossible. S-allele homozygosity is, however, possible in SSI systems when a pollen recessive allele matches a stigma recessive allele.

Initial genetic mapping studies showed that the S locus was located on chromosome (C) 1R in Secale, by linkage to an isozyme phosphoglycoisomerase (PGI-2) and the leaf peroxidase Prx-7 genes (Gertz & Wricke, 1989). PGI-2 was also found to be linked to the S locus in L. perenne (Cornish et al., 1980). The Z locus was located on C2R in Secale by linkage with the beta-glucosidase and esterase 4/11 genes (Fuong et al., 1993). The T locus was located on C5R in rye (Secale cereale) and linked to the esterase 5–7 complex (Fuong et al., 1993). Recent mapping analysis in Secale (Voylokov et al., 1998; Hackauf & Wehling, 2005), Phalaris (Bian et al., 2004) and Lolium (Thorogood et al., 2002) has confirmed syntenic chromosomal locations of S and Z on C1 and C2, respectively, in accordance with the Triticeae consensus map (Armstead et al., 2002; Jones et al., 2002; Sim et al., 2005). There was also evidence of linkage between either the S or Z locus and the isozyme glutamate oxalo-acetate transaminase GOT/3 (Thorogood & Hayward, 1991). Distortion was observed on C3, where GOT/3 is located (Jones et al., 2002), and it was determined by Thorogood et al. (2002) that this was caused by a pleiotropic effect of S or a locus closely linked to S. Additionally, a self-fertility locus has been mapped on C5 of L. perenne, in a position that is likely to be orthologous to the Secale S5 and Phalaris T loci (Thorogood et al., 2005).

Efforts have also been made to identify the gene products of S and Z. Li et al. (1994) reported a putative pollen S-gene clone, named Bm2, identified from P. coerulescens. This gene co-segregates with the S genotype and is expressed only in mature pollen (Li et al., 1994). Further evidence to support Bm2 as the pollen S gene was the structure of the gene. It was identified as having a variable N terminus with potential allelic specificity and a conserved catalytic domain in the C terminus with significant homology to the thioredoxin H protein. Thioredoxin activity of Bm2 was functionally demonstrated by Li et al. (1995) through expression in Escherichia coli of the C-terminal protein, which acted as a substrate for E. coli thioredoxin reductase. It was speculated that Bm2 was involved in the SI response through the post-translational modification of other proteins by the thioredoxin proteins (Li et al., 1995). However, the expression of the Bm2 gene was barely detectable in other SI grass species such as S. cereale, Hordeum bulbosum and L. perenne (Li et al., 1997) and later studies, using large populations segregating for S and Z, revealed that Bm2, while closely linked to the S locus, was at a distance of approx. 1 cM (Baumann et al., 2000). Comparison of the structure of Phalaris Bm2 with that of its homologues from S. cereale, H. bulbosum and L. perenne suggested that the proposed N-terminal allelic domain of Bm2 was actually the untranslated region of the gene (Baumann et al., 2000). It is now clear that Bm2 represents a thioredoxin-like gene, but not the pollen S gene. The role of Bm2 in the SI response is still under investigation. Increased Bm2 expression was detected during in situ hybridization in an incompatible pollination, which, however, could also be induced by damage to the stigma by pollen rejection (Baumann et al., 2000).

Mapping the Z locus in rye (S. cereale), Hackauf & Wehling (2005) identified a putative ubiquitin-specific protease (UBP) gene showing a pistil-specific expression pattern and co-segregation with the Z locus, implying the possible involvement of protein ubiquitination in the grass SI system. The ubiquitination pathway has been shown to be involved in pollen rejection in the Brassica SI response through an E3 ubiquitin ligase, an Armadillo repeat motif-containing protein (ARC1) (reviewed in Hiscock & McInnis, 2003). ARC1 interacts with the kinase domain of the Brassica SI female determinant in a phosphorylation-dependent manner. The substrates of ARC1, potential pollen germination-related proteins, then would be ubiquitinated and undergo proteasomal degradation by the COP9 signalosome (CSN). Ubiquitination is also involved in the S-RNase mediated GSI response, but in compatible pollinations. The male determinant in the S-RNase system is an F-box gene SLF (S-locus F-box) (Sijacic et al., 2004) belonging to the F-box family proteins which can bind the S-phase kinase-associated protein 1 (Skp1) and cullin-like proteins to form an E3 ubiquitin ligase complex. It has been proposed that, through interaction of SLF and S-RNase (the female determinant) in a compatible pollination, S-RNase is ubiquitinated and then degraded by the 26S proteasome (Entani et al., 2003). It has still not been determined whether the S. cereale UBP gene is a component of the Z locus or a linked gene with suppressed recombination around the Z locus. However, the known role of ubiquitination in other SI systems, and the fact that no specificity domain is associated with the UBP gene, suggest that it is more likely to be involved in the downstream reactions of the grass SI response. To date, the gene products of S and Z, on both the pollen and the stigma sides, remain uncharacterized in any species in the grass family.

Comparison of the grass self-incompatibility system with other gametophytic self-incompatibility systems

Two other well-characterized GSI systems are known in plants. They are mechanistically different and have been extensively investigated at the molecular level. One is the S-RNase system, originally found in members of the Solanaceae and later reported in the Rosaceae and Scrophulariaceae (Lee et al., 1994; Murfett et al., 1994; Cheng et al., 2006). The second well-characterized system is found in Papaver rhoeas (Franklin-Tong & Franklin, 1992).

In the S-RNase system, the cytotoxic RNase function of the stigma S protein S-RNase is crucial for the rejection of incompatible pollen (Lee et al., 1994) and the pollen S protein, an F-box gene SLF, has been identified based on its S-locus location, S-haplotype-specific polymorphism and pollen-specific expression (Entani et al., 2003; Ushijima et al., 2003). These findings were recently confirmed by a transformation experiment in Petunia inflata (Sijacic et al., 2004). There is evidence that S-RNase alone is not sufficient and other stigmatic factors not linked to the S locus are required for the induction of an SI response. Group 1 factors are required for S-RNase expression (McClure et al., 2000) and group 2 factors such as a small asparagine-rich protein (HT-B), 120K and 4936-factor (McClure et al., 2000; O’Brien et al., 2002; Hancock et al., 2005) are involved in the cytotoxic activity of S-RNase. Two models have been so far proposed for the S-RNase-based inhibition of pollen in an S-haplotype-specific manner (reviewed in McClure & Franklin-Tong, 2006). The first model involves S-RNase degradation. Through interaction of S-RNase and SLF, in a compatible pollination, S-RNase is ubiquitinated and then degraded, and in an incompatible pollination, S-RNase in certain ways evades degradation and the cytotoxic activity of S-RNase then degrades the pollen rRNA, leading to the arrest of pollen tube growth. However, this model does not propose functions for other known factors (e.g. HT-B and 120K) and the evidence that no large-scale S-RNase degradation occurs in incompatible pollen tubes leads to the development of an alternative model. In this model, S-RNase, HT-B and 120K are transported into a pollen vacuole. In a compatible pollination, nonself S-RNase and SLF interaction facilitates HT-B degradation and S-RNase remains compartmentalized, resulting in compatibility. Through the interaction of S-RNase and SLF in self-pollination, the release of S-RNase from the vacuole reinforces cytotoxic activity of S-RNase, leading to pollen rejection. Differences between pollen S in Prunus (SFB) and Solanaceae species (SLF) have recently been reported (Hauck et al., 2006; Sassa et al., 2007) and several similar F-box genes, named SFBB (S-locus F-box brothers), have been identified as pollen S genes of apple (Malus domestica) and Japanese pear (Pyrus pyrifolia) (Sassa et al., 2007). These findings imply a mechanistic diversity of S-RNase-based GSI.

In Papaver, the stigmatic S proteins are small extracellular signalling molecules, interacting with the pollen component which is believed to be a plasma membrane receptor. One putative pollen receptor is an S-protein binding protein (SBP) (Hearn et al., 1996). It binds specifically to stigmatic S proteins but in a haplotype-indifferent manner, suggesting an accessory receptor role rather than it being the pollen S receptor itself. To date, the Papaver pollen S determinant has not been identified. Based on knowledge of the components involved in the SI reaction, a model for pollen tube inhibition in Papaver has been proposed (reviewed in McClure & Franklin-Tong, 2006). Inhibition of the incompatible pollen is mediated by the activation of a Ca2+-dependent signalling cascade. A rapid increase of cytosolic free Ca2+([Ca2+]i) is induced by the SI response in incompatible pollen, leading to the loss of high apical [Ca2+]i which is a key characteristic of growing pollen tubes. The increasing [Ca2+]i initiates the intracellular signalling network, inducing the following downstream reactions: phosphorylation of soluble inorganic pyrophosphatases (sPPases), which is required for pollen tube tip extension; activation of a mitogen-activated protein kinase (MAPK), p56, which might induce programmed cell death (PCD), rapid depolymerization of the pollen actin cytoskeleton through the severing activity of actin-binding proteins (ABPs), and PCD triggered by cytochrome c (cyt c) leakage-induced caspase-like activity. The incompatible pollen is in this way inhibited and does not resurrect.

Diversity of the biochemical bases of GSI responses is also reflected in the diverse physiological responses that can be seen in GSI from a number of plant families (Fig. 2). In the Solanaceae, both the compatible pollen and incompatible pollen germinate and a pollen tube grows through the transmitting tract of the style. The continued slow growth of the pollen tube in incompatible styles has been reported in Petunia hybrida (Herrero & Dickinson, 1980) and Nicotiana alata (Lush & Clarke, 1997). A polysaccharide callose deposit was observed in the walls of compatible pollen tubes in N. alata forming plugs at regular intervals, while in incompatible tubes the callose deposits become irregular with unevenly spaced plugs, the tube wall thickens and the tube tip becomes abnormal (Lush & Clarke, 1997). The growth of the incompatible pollen tube can be arrested in either the upper or lower parts of the style, ranging from 2 to 50 mm (McClure et al., 1990; Lush & Clarke, 1997). The callose deposits were reported to swell and, in some cases, burst within the style (deNettancourt, 1977; Lush & Clarke, 1997). In P. rhoeas, by comparison, rejection of pollen takes place on the stigmatic surface, before or immediately after germination (Franklin-Tong & Franklin, 1992). P. rhoeas also has a dry stigmatic surface compared to the wet, lipid-rich exudate found on the surface of the Solanaceae (Elleman et al., 1992). Inhibition of incompatible pollen in P. rhoeas is rapid, taking place on a time scale of minutes, compared with the relatively slow inhibition in the Solanaceae. Three possible phases have been proposed for the Papaver SI response. First, a very rapid but reversible inhibition of tip growth takes place, and this is followed by a ‘commitment’ phase, during which processes are triggered that lead to the irreversible degradative processes detected in the ‘late’ phase (Wheeler et al., 2001).

Figure 2.

Comparisons between features of the grass self-incompatibility (SI) system and those of three other well-characterized SI systems: S-RNase SI, Papaver SI and sporophytic SI (SSI). Physiological differences and similarities are shown in (a) and additional molecular genetics features are summarized in (b).

In the grass family, the incompatibility reactions are generally rapid. Knox & Heslop-Harrison (1971) reported the release of intine-held antigens by the pollen grains of Phalaris tuberosa on the stigma surface within 5–10 min after pollen–stigma contact, during which time pollen tube growth was controlled in incompatible pollination. The fact that the antigens remained spread on the stigma surface in compatible pollination implied a possible role of the antigens as compatible recognition materials (Knox & Heslop-Harrison, 1971). The grass family has an unusually fast pollen germination speed and tube growth rate. Pollen rehydration by uptake of water from the stigma is essential to germination and the feathery form of the stigma reduces the capacity of the pollen-tube transmitting tracts. It was then hypothesized that incompatibility occurs rapidly at the stigmatic surface before effective penetration. Otherwise, the transmitting tracts would be saturated and the compatible pollen tube growth would then be blocked (Heslop-Harrison, 1979b). Evidence has shown that in incompatibility reactions the water flow to pollen grains is restricted at the receptive part of the stigma papilla, leading to the inhibition of pollen germination (Heslop-Harrison, 1979a). Through observation of stained pollen tubes, incompatible responses were observed near the stigma surface in Gaudinia fragilis and S. cereale within 2 min after the recognition event of pollen and stigma (Shivanna et al., 1982), and in rye within 90 s after the contact of incompatible pollen (Heslop-Harrison, 1982). The inhibited pollen tubes are short, distorted and occluded with callose deposits, although the length of inhibited tubes is variable. In H. bulbosum, however, the rejection is much slower, after the tube tip penetrates the stigma cuticle (Heslop-Harrison, 1982), and in some individuals of Alopecurus partensis inhibited pollen tubes even reached the transmitting tracts (Shivanna et al., 1982). After penetration, the water required for the pollen tube growth was transferred from the hydrated grain but not the stigma tissues, and the enzymes, such as pectinase and cutinase, released by the emerging tube tip were responsible for the dissolving of the cuticle of stigma papillae (Heslop-Harrison & Heslop-Harrison, 1981). Heslop-Harrison (1982) has proposed that the grass incompatibility response may contain at least three elements: the self-recognition step governed by the S and Z loci, the rejection response, and an additional control that determines the rate at which the growth of the tube is arrested. But, so far, none of the genetic components of these elements has been identified. Later, Wehling et al. (1994a) reported, in rye pollen, that both protein phosphorylation and Ca2+-induced signal transduction were involved. Significant phosphorylation activity was reported in self-pollinated pollen grains and decreased activity was found associated with the loss of SI. Disruption of the SI response was observed after treating the incompatible stigmas with different protein kinase inhibitors and Ca2+ antagonists (Wehling et al., 1994a). Wehling et al. (1994b) have proposed a model for the SI reaction in rye, suggesting that the pollen S and Z determinants are possibly pollen grain plasma membrane-located protein kinases with extracellular receptor domains. The stigma S and Z determinants act as signal molecules, by interacting with ‘self’ pollen partners, inducing a Ca2+-mediated signal cascade which finally leads to pollen tube inhibition. However, this hypothesis still needs to be tested in more detail.

The grass physiological incompatibility responses appear to be more similar to the Papaver SI system, with the dry stigma type, a rapid response upon stigma–pollen interaction and possible involvement of Ca2+ signal transduction in incompatible pollen, implying mechanistic links between the grass GSI and the Papaver SI systems. However, in contrast to the depolymerization of the F-actin cytoskeleton in incompatible Papaver pollen tubes, the primary target of the incompatible response in grass pollen is the abnormal formation of the wall at the tube tip through interference of the orientation and distribution of the wall microfibrils (Shivanna et al., 1982). In an incompatible grass pollen tube, the microfibril content of the precursor particles does not flow into the extending wall to form the thin apical sheath characteristic of the normal tube tip, but aggregates to form banks or nodules. There is also evidence of a link between the grass GSI and S-RNase-based SI systems. The weak self-incompatibility response observed in H. bulbosum (Heslop-Harrison, 1982) and Alopecurus pratensis (Shivanna et al., 1982), where incompatible pollen tubes are inhibited in the transmitting tracts of the style, is similar to that of the Solanaceae. It was then suggested that similar stigmatic mechanisms exist in gametophytic SI systems and the differences in the sites of action reflect the different distributions or rates of secretion of incompatibility factors (Heslop-Harrison & Heslop-Harrison, 1982).

The relation of the grass GSI system with the sporophytic self-incompatibility system

The grass SI system and sporophytic SI (SSI) system show some similarities: they are both able to maintain SI at the polyploid level; homozygosity is possible at one of the loci in the two-locus grass SI system while recessive alleles in the SSI system can also give rise to homozygotes; and dominance interactions of S alleles in the SSI system can lead to differences in reciprocal pollination, which is a characteristic of the grass SI system caused by different S- and Z-allele combinations. SSI has been identified in phylogenetically divergent families (Brassicaceae, Asteraceae, Convolvulaceae, Polemoniaceae and Malvaceae), which suggests multiple origins of SSI. There is also evidence that orthologues of Brassica stigmatic SI determinants in Ipomoea trifida (Convolvulaceae) and Senecio squalidus (Asteraceae) are not involved in the SSI responses, suggesting novel SSI mechanisms different from the SSI system of the Brassicaceae (reviewed in Hiscock & Tabah, 2003). Most studies have concentrated so far on Brassicaceae species. The Brassica SI response is focused at the pollen–stigma interface as self grains are not inhibited metabolically, but are physiologically isolated from the subjacent stigmatic papilla. Recognition and the interactions at the pollen–stigma interface lead then to the rapid inhibition of pollen hydration, pollen germination or stigma penetration within 10–20 min, during which time the inhibition response is reversible if incompatible pollen grains are transferred to a compatible stigma (reviewed in Hiscock & McInnis, 2003).

Apart from the fast pollen-tube inhibition near the stigma surface (Fig. 2), there are other physiological characteristics of grass SI that resemble features of the SSI system, including the trinucleate pollen at the time of dispersal, the short-lived pollen with a high respiratory rate and the dry stigma. Phosphorylation and a signal cascade have also been reported in the Brassica SSI system from molecular genetic studies, which are reviewed below.

In Brassica, both the female and the male determinants of SI have been identified (Nasrallah, 2002). The female determinant is the S-locus receptor protein kinase (SRK) (a single-pass transmembrane serine/threonine kinase) which is expressed in the stigma epidermis (Takasaki et al., 2000). The extracellular receptor domain of SRK is considered to be responsible for the S specificity (Hiscock & Tabah, 2003). On the basis of sequence similarity, the receptor domain can be divided into three subdomains: a mannose-binding lectin-like domain at the N-terminal, a hypervariability-containing domain in the middle region and a PAN_APPLE-like domain at the C-terminal (Naithani et al., 2007). It has been shown that a distinct region of the hypervariable subdomain of the receptor domain is responsible for SRK binding (Kemp & Doughty, 2007) and the PAN_APPLE domain plays a major role in the mediation of the ligand-independent SRK dimerization (Naithani et al., 2007). Another gene product located at the stigma S locus, the S-locus glycoprotein (SLG), has a similar extracellular domain to that of the SRK with the equivalent allele and has been reported to enhance the ability of the stigma to reject incompatible pollen (Takasaki et al., 2000). However, in Brassica oleracea, the SLG genes do not encode functional SLG proteins (Suzuki et al., 2000), and in Arabidopsis lyrata the S locus appears to lack SLG (Schierup et al., 2001), implying that SLG is not essential for SI and SRK alone determines S specificity in the stigma. The physical distance between the SRK and SLG loci was found to vary greatly in different Brassica species, spanning a length from a few to as many as several hundred kilobases (reviewed in Casselman et al., 2000). Recombination between these two loci has been reported recently and it has been suggested that the nucleotide diversity of SRK in Brassica facilitates meiotic pairing, thus enhancing recombination (Takuno et al., 2007). The male determinant, the S-locus cysteine-rich protein gene (SCR) (Schopfer et al., 1999), also designated as S-locus pollen protein 11 (SP11) (Takayama et al., 2000), is expressed mainly in the anther tapetum which breaks down to form the pollen coating, where the mature SCR/SP11 protein accumulates. SCR/SP11 and SRK are tightly linked and recombination between them is suppressed, which would otherwise lead to the breakdown of SI function (Casselman et al., 2000). No recombination has been detected between the S domain of SRK and SCR/SP11 (Takuno et al., 2007). However, the order and orientations of SRK, SCR/SP11 and other S-locus genes and the distances between them are different in different species (Fukai et al., 2003). Insertion of retrotransposons has recently been reported as the reason for the differences in S-locus lengths between Brassica oleracea and Brassica rapa (Fujimoto et al., 2006). SRK and SCR/SP11 interact with each other in a haplotype-specific manner, when the S-receptor domain of SRK and the SCR/SP11 pollen ligands are encoded by the same S-haplotype, leading to the activation of the SI response (Kachroo et al., 2001; Takayama et al., 2001). A model for the mechanism of SSI has been proposed. After pollination, the SCR/SP11 protein is delivered to the surface of a stigma epidermal cell and then transferred to the plasma membrane within the region of pollen contact. The interaction of SRK and SCR/SP11 induces transphosphorylation on serine and threonine residues in the kinase domains of presumably dimerized SRKs (Takayama et al., 2001). Two thioredoxin-H-like proteins (THL1 and THL2) (Bower et al., 1996) associate with SRKs and act as inhibitors in the absence of SCR/SP11 to prevent autophosphorylation and maintain the inactivated state of SRKs (Cabrillac et al., 2001). Activation of SRK, accompanied by SLG, triggers a signal cascade that results in the arrest of pollen tube development. The downstream signalling reactions are under further investigation and a stigma-specific Armadillo repeat motif-containing protein (ARC1) has been identified which interacts with the kinase domain of SRK in a phosphorylation-dependent manner (Gu et al., 1998). A recently identified U-box motif in ARC1 and the increased levels of ubiquitinated proteins within incompatible pollinations suggested that protein degradation by ubiquitination is involved in the Brassica SI response (Stone et al., 2003). Characterization of the SCR/SP11 high-affinity binding site revealed that the integral and membrane-anchored forms of SRK exhibited high-affinity binding to SCR/SP11, but the soluble form of SRK (eSRK) exhibited no high-affinity binding (Shimosato et al., 2007). Furthermore, the artificially dimerized form of eSRK exhibited high-affinity binding, suggesting that the membrane anchorage is necessary for SRK to obtain the high-affinity dimeric form (Shimosato et al., 2007). The mechanism of pollen inhibition remains to be characterized. Recently, actin reorganization and likely depolymerization have been found in papilla cells of Brassica rapa after self-pollination (Iwano et al., 2007). The same research determined that the dynamics of the actin cytoskeleton lead to structural changes in the vacuoles of the papilla cell, which in turn regulate hydration and germination of pollen.

The involvement of phosphorylation activity and actin reorganization in both Papaver SI and Brassica SSI responses implies a possible mechanistic relationship between these two SI systems, especially as they are both characterized by dry stigmas and a rapid incompatibility response at the stigma surface (Fig. 2). Grass SI and Brassica SSI share an extra common feature, the tri-nucleate pollen. These similarities will be explored further in the following section.

Similarities between grass self-incompatibility and sporophytic self-incompatibility; do their genetic control mechanisms show common features?

Although gametophytic and sporophytic systems are clearly different, the SSI and the grass SI system have common features suggesting the possibility that the fundamental molecular mechanisms are similar or related. The basic difference between the sporophytic and gametophytic systems lies in the specific pollen regions where incompatibility is expressed in the pollen. If incompatibility is expressed in the outer pollen wall, the pollen exine, SI would act sporophytically as a result of the diploid tapetum origin of pollen exine (Bedinger, 1992). If incompatibility was expressed in the inner pollen wall, the intine formed after meiosis and thus SI was phenotypically determined by the genotype of the pollen grain (Bedinger, 1992), SI would act gametophytically. The Brassica pollen SI protein SCR/SP11 emerges in the tapetum and accumulates in the pollen coat (Takayama et al., 2000). It is possible that a SCR/SP11 or a protein with similar function is located only in the coat of grass pollen, thus leading to a ‘SSI-like’ GSI reaction. The fact that a relative of SCR/SP11, PCP-A1 (pollen coat protein, class A, 1) is expressed gametophytically in Brassica, accumulating in pollen grains at the late binucleate/trinucleate stage of development (Doughty et al., 1998), adds weight to the hypothesis of a pollen coat location of the grass male SI determinant. On the stigmatic side, similar to the Brassica stigmatic SRK protein, a membrane-bound protein kinase with receptor domains has been speculated to be involved in the grass SI responses (Wehling et al., 1994b; Baumann et al., 2000). These speculations were supported by the suppression effects of protein kinase inhibitors on the SI response in rye (Wehling et al., 1994a). Our preliminary results from investigation of differentially expressed genes during the SI response in L. perenne also identified three SI-associated genes with homology to protein kinases. Further investigations on the functions and structures of these genes are in progress.

Approaches to studying the self-incompatibility reaction in grasses

Currently, the downstream reactions of grass SI remain unknown. Biochemical studies such as in Papaver (Franklin-Tong & Franklin, 2003; de Graaf et al., 2006) via in vitro germination of pollen offer an alternative approach to identify candidate and hypothetical genes through protein functions; for example, the identified protein kinase genes in L. perenne. If the grass SI system and SSI system are highly similar and if either the pollen or the stigma SI genes in grasses can be identified, it would be possible to find the partner determinant genes through sequencing of neighbouring genomic sequences, as previously demonstrated in Brassica (Schopfer et al., 1999; Takayama et al., 2000). However, caution is required here, for the reason that the order and orientations of SRK, SCR/SP11 and other S-locus genes and the distances between them have been shown to vary considerably among species (Fukai et al., 2003). Furthermore, after identification of the grass SI gene products, in vivo approaches such as gain- and loss-of-function are necessary to ascertain the true identity of pollen or pistil S and Z genes with an efficient transformation and regeneration system.

A comparative mapping approach to identifying linked markers and candidate genes

Comparisons of the genetic maps of grasses that exhibit SI with the genetic maps of other Poaceae species, such as wheat (Triticum aestivum), oat (Avena sativa) and rice (Oryza sativa), have revealed significant regions of conserved genetic synteny and gene order (colinearity) (Jones et al., 2002; Alm et al., 2003; Sim et al., 2005). Studies exploiting these conserved syntenic relationships have also suggested that the grasses might, ancestrally, share a common incompatibility system (see Thorogood et al., 2002). Therefore, one avenue through which to clarify the genetics and genomics of incompatibility is by cross-species comparisons between SI grasses and model species such as rice and maize (Zea mays), exploiting the considerable genomic resources that are available for the model plant species. Although these are self-fertile species, there is no reason to assume that the SI loci are no longer present. It is more likely that the SI genes are represented by self-compatible variants. By applying comparative genomics, we have made progress which is described below in the identification of molecular markers and potential candidate regions for genes associated with the SI response in L. perenne.

The S and Z loci in L. perenne map to C1 and C2, respectively. Previous comparative mapping work has shown that these regions of L. perenne C1 and C2 have a degree of conserved synteny with rice chromosomes 5 (R5) and R4, respectively (Jones et al. 2002) (Fig. 3). Following the reasoning of Lundqvist (1954) that the two-locus SI system in grasses evolved by gene duplication with similar gene structure and function, it was considered possible that the presence of similar genes (i.e. potentially duplicated genes) in the regions of R4 and R5 (syntenous to the S and Z regions on C1 and C2 of L. perenne) may represent candidate genes for incompatibility systems. BLAST searches of coding sequences from R4 (syntenic with L. perenne C2) against R5 (syntenic with L. perenne C1) identified two highly conserved genes (G10-proteins; LOC_Os05g37390 and LOC_Os04g55290 according to The Institute for Genomic Research (TIGR) rice gene models) physically located in regions that would be expected to correspond to the positions of the S and Z loci on L. perenne C1 and C2 (Fig. 3). In addition, there were two further highly conserved G10-type proteins on R1 and R12, which corresponded to syntenic regions of L. perenne on C3 and C5, respectively, which also have associations with SI (Fig. 4a). As mentioned above, Thorogood et al. (2002) found distorted segregation of loci on C3, caused by association with specific S alleles on C1, and Thorogood et al. (2005) reported that C5 contained a major quantitative trait locus (QTL) for self-fertility which, in terms of comparative genetics, could be associated with the positions of the T locus in rye. The finding of G10 proteins reinforces the possibility that the S-Z system evolved from a multi-locus system. To date, only one of the putative G10 loci has been directly mapped in L. perenne (C3), but the mapping of flanking markers based on R5 and R4 sequences around S and Z on L. perenne C1 and C2 suggests that the G10 loci are at least likely to be linked to S and Z.

Figure 3.

Comparative maps of Lolium perenne linkage groups (C) 1 and 2 with rice C5 and 4. Markers (bold type) on L. perenne C1 and C2 were linked to their physical position on the rice (Oryza sativa) R5 and R4 pseudomolecules. Two G10 proteins, LOC_Os05g37390 and LOC_Os04g55290 (on the right of the rice chromosomes), are physically located in regions that would be expected to correspond to the positions of the S and Z loci (bold italics and underlined) on L. perenne C1 and C2.

Figure 4.

Sequence alignments of The Institute of Genomic Research (TIGR) rice (Oryza sativa) gene models identified as potential candidates for involvement in the self-incompatibility response in Lolium perenne through BLAST searches and genetic synteny. (a) G10 proteins: LOC_Os04g55290 corresponds to the region of the S locus on L. perenne C1; LOC_Os05g37390 corresponds to the region of the Z locus on L. perenne C2; LOC_Os01g63890 is possibly in the region of the S distortion locus on L. perenne C3 and LOC_Os12g05410 is possibly in the region of the self-fertility locus on L. perenne C5. These four sequences show high similarity and have possible association with self-incompatibility loci in L. perenne. (b) ‘Companion’ proteins: BLAST result from TIGR showing similar ‘next door’ gene proteins on rice chromosomes 1, 4 and 5. No equivalent was identified on chromosome 12 or other locations.

It has been suggested and shown that genes encoding the pollen and pistil specificity determinants are tightly linked at the S locus to maintain SI (Sijacic et al., 2004, Takuno et al., 2007). Therefore, additional examination of the rice genomic sequence around the G10-protein gene models was carried out and conserved ‘expressed’ proteins were identified on R1, R4 and R5 (Fig. 4b), although no equivalent conserved sequence was found associated with LOC_Os12g05410 on R12. While the degree of conservation is far less for these ‘companions’ than it is for the G10 proteins themselves, in a theoretical GSI haplotype, these could represent the pollen/pistil-specific counterparts to the G10 proteins. G10 proteins were first identified as being recruited for translation during Xenopus oocyte maturation (McGrew et al., 1989). They are highly conserved in a wide range of eukaryotic species. As yet, neither the G10 proteins nor the associated similar ‘expressed’ proteins have defined functions in animal or plant systems. However, the C terminal of G10 is rich in cysteine, similar to the Brassica male determinant SCR (S-locus cysteine-rich protein), which makes the G10 gene a promising grass SI candidate for further investigation.

Conclusions

Stebbins (1974) has proposed that the transition from SI to predominant self-fertilization was one of the most common transitions in plant reproductive systems, and studies in the Solanaceae have also demonstrated the essentially irreversible nature of the transition from SI to self-compatibility (Igic et al., 2004, 2006). Different results were obtained in the Asteraceae, but this may be because of the different genetic control between the GSI in Solanaceae and SSI in Asteraceae (Ferrer & Good-Avila, 2006). Subsequently, the exploitation of self-compatibility has become fundamental to modern-day plant breeding practices in the form of hybrid production, heterosis and the fixation of desirable gene combinations. Among the economically important grasses, L. perenne is unusual in still retaining effective SI. As such, L. perenne is a valuable resource for developing experimental models that will clarify both the underlying genetic control and the associated physiological and biochemical responses involved in determining SI. To date, in addition to the well-studied physiological relationships of different incompatibility responses, some mechanistic links have also been established, including the ubiquitin-mediated protein degradation between the S-RNase SI and Brassica SI systems, the actin reorganization feature between the Papaver SI and Brassica SI systems, and the phosphorylation activity in the Papaver, Brassica and grass SI systems. Through unravelling SI in grasses, a better understanding of the diversity and evolution of different SI systems can be attained.

Acknowledgements

BY is supported by a Teagasc Walsh Fellowship. We are grateful to two anonymous referees for important comments and insightful suggestions on the manuscript.

Ancillary