Charles Darwin made extensive observations of the pollination biology of a wide variety of plants. He carefully documented the consequences of self-pollination and described species that were self-sterile but that could easily be crossed with other plants of the same species. He believed that compatibility was controlled by the ‘mutual action’ of pollen and pistil contents. A genetic model for self-sterility was developed in the early 1900s based on studies of the compatibility relationships among, what are now referred to as, self-incompatible (SI) Nicotiana species. Today, it is believed that SI in these species is controlled by an interaction between S-RNases produced in the pistil and F-box proteins expressed in pollen and, moreover, that this S-RNase-based SI system is shared by a great diversity of other plant species. Current research is aimed at understanding how the mutual actions of these S-gene products function in the physiological context of pollen tube growth.
The 200th anniversary of Darwin's birth and the 150th anniversary of the publication of Origin of species is a good time to reflect on Darwin's broad contributions. Darwin's family was definitely on the winning side of the industrial revolution; he was a grandson of Josiah Wedgwood who developed the famous line of manufactured pottery. Given the importance of Darwin's ideas to modern biology, it is a little frightening to consider how contingent they were on the whims of history. Nevertheless, it is fortunate that an intellect of Darwin's quality ended up in the circumstances to travel the world observing and collecting, and then to spend a lifetime in further reflection and experimentation.
This review will recall Darwin's work on self- and cross-pollination, in particular, on what is now called self-incompatibility (SI). The focus will be on progress toward understanding S-RNase-based SI. Darwin's own words will be used to convey his thoughts and his fascination with many of the same questions that ignite students today. The insights are often uncanny, and researchers interested in population-level studies still find it productive to test his hypotheses and extend his studies. Darwin's insight notwithstanding, the modern understanding of inheritance—in terms of the unit of transmission, the gene, its chemical nature, the processes of gene expression, and the genomic and population-level contexts of genes—represents dramatic advances that would surely have thrilled him. Thus, this review will also highlight current SI research; insights that this author would share with him, if that were possible. The reader is referred to de Nettancourt's (1977, 2001) classic monographs and the recent volume edited by Franklin-Tong (2008a) for more in-depth treatments.
Darwin on heterostyly and self-sterility
Darwin had a very keen interest—expressed on several levels—in plant reproduction. He published two editions on pollination of orchids (Darwin 1862, 1877a), but the works most relevant here are The different forms of flowers on plants of the same species, published in 1877, and The effects of cross and self-fertilisation in the vegetable kingdom, first published in 1876 with a second edition in 1878. He was well aware of earlier works (especially by Kölreuter, Müller, and Scott) that described instances of what we now refer to as SI. An early paper on self-sterility by EM East includes a useful review of this early work (East and Park, 1917). Darwin's later work shows that he was definitely excited by the variety and the subtlety of plant reproduction.
There is hardly anything more wonderful in nature than the sensitiveness of the sexual elements to external influences, and the delicacy of their affinities.…We see how sensitive the sexual elements of those plants must be, which are completely sterile with their own pollen, but are fertile with that of any other individual of the same species. (Darwin, 1878; p. 467)
In The different forms of flowers, Darwin discusses many instances where a single species elaborates alternative floral morphologies, but plants that display, what is now referred to as, heteromorphic SI capture the most attention. Heterostyly in Primula and its relatives is the best known system. His passion for comprehensive, meticulous observation and experimentation is clearly expressed. Darwin describes the differences between floral morphs in great detail, including pollen size and appearance, organ positioning and size, and stigma papillae. He experimentally probes the function of organ positioning, but he finds, importantly, that it is not sufficient to ensure that crossing occurs only between morphs:
It follows from the position of the organs that if the proboscis of a dead humble-bee, or a thick bristle or rough needle, be pushed down the corolla, first of one form and then of the other, as an insect would do in visiting the two forms growing mingled together, pollen from the long-stamened form adheres round the base of the object, and is left with certainty on the stigma of the long-styled form; whilst pollen from the short stamens of the long-styled form adheres a little way above the extremity of the object, and some is generally left on the stigma of the other form. In accordance with this observation I found that the two kinds of pollen, which could easily be recognised under the microscope, adhered in this manner to the proboscides of the two species of humble-bees and of the moth, which were caught visiting the flowers; but some small grains were mingled with the larger grains round the base of the proboscis, and conversely some large grains with the small grains near the extremity of the proboscis. Thus pollen will be regularly carried from the one form to the other, and they will reciprocally fertilise one another. … The several foregoing facts led me to try the effects of the two kinds of pollen on the stigmas of the two forms. (Darwin, 1877b; p. 24)
His experiments revealed that intermorph (i.e. legitimate) pollinations were fully fertile while intramorph (i.e. illegitimate) pollinations set few seed. Figure 1, reproduced from Darwin's book, shows these relationships diagrammatically; this figure has been so frequently imitated by subsequent generations of scholars that any student of pollination will recognize it.
On a more theoretical level, Darwin recognized the importance of restricted mating and described it as a mechanism for dividing a population into distinct ‘bodies’ of compatible partners.
(W)e have here a case to which no parallel exists in the vegetable or, indeed, in the animal kingdom. The individual plants of the present species, and as we shall see of several other species of Primula, are divided into two sets or bodies, which cannot be called distinct sexes, for both are hermaphrodites; yet they are to a certain extent sexually distinct, for they require reciprocal union for perfect fertility. (Darwin, 1877b; p. 28)
He also understood that mating was subject to natural selection and that heterostyly had emerged independently in different plant lineages at different times. Furthermore, he conducted experiments showing the benefit of outcrossing and the ‘evil’ of selfing in many species. He perceived that the benefit of outcrossing within a species was sufficient to select for emergence of incompatibility systems. Barret and Shore (2008) emphasize that a more developed view takes account of different benefits of heterostyly per se versus intramorph incompatibility; the former reduces pollen waste by facilitating its placement on receptive stigmas while incompatibility prevents inbreeding and contributes more to maternal fitness.
We may feel sure that plants have been rendered heterostyled to ensure cross-fertilisation, for we now know that a cross between the distinct individuals of the same species is highly important for the vigour and fertility of the offspring. (Darwin, 1877b; p. 258)
(W)e may confidently believe that this has been effected in order that cross-fertilisation should be assured. For the full and legitimate fertilisation of these plants pollen from the one form must be applied to the stigma of another.… There is reason to believe that the sterility of these unions has not been specially acquired, but follows as an incidental result from the sexual elements of the two or three forms having been adapted to act on one another in a particular manner, so that any other kind of union is inefficient, like that between distinct species. (Darwin, 1877b; p. 345)
Here, and in other places, Darwin seems to suggest that intramorph incompatibility must arise from interactions between the constituents of pollen and pistils. This observation seems especially prescient since he did not understand how traits are transmitted between generations. It implies that he understood that, however inheritance occurred, it was the basis for not only overt morphological variation, but also for compositional (or, as we would say today, molecular) variation. This insight anticipates the rich molecular-level studies possible today. It is noteworthy that he also suggests a fundamental similarity between intra- and interspecific mechanisms, at least in so far as these arise from the ‘constituents’ of pollen and pistils.
We must therefore look to the appearance of inner or hidden constitutional differences between the individuals of a varying species, of such a nature that the male element of one set is enabled to act efficiently only on the female element of another set.…But what the nature of the inner constitutional differences may be between the sets or forms of the same varying species, or between distinct species, is quite unknown. (Darwin, 1877b; p. 267)
While The different forms of flowers on plants of the same species deals extensively with heterostyly, and Darwin definitely regarded intramorph incompatibility as an essential aspect of such systems, he also investigated self-sterility among species that do not display distinct floral morphologies. In The effects of cross and self-fertilisation in the vegetable kingdom, Darwin describes exhaustive experiments with >50 species that provide overwhelming support for his conclusion ‘that generally cross-fertilisation is beneficial, and self-fertilisation is often injurious’ (Darwin, 1878). Several of the species investigated were fully or partially ‘self-sterile’ including species still studied today. Regarding Petunia violacea, Dingy purple variety, Darwin noted:
(P)rotected flowers with their own pollen placed on the stigma never yielded nearly a full complement of seed; whilst those left uncovered produced fine capsules, showing that pollen from other plants must have been brought to them, probably by moths. Plants growing vigorously and flowering in pots in the greenhouse, never yielded a single capsule; and this may be attributed, at least in chief part, to the exclusion of moths.
Six flowers on a plant covered by a net were crossed with pollen from a distinct plant and produced six capsules, containing by weight 4.44 grains of seed. Six other flowers were fertilised with their own pollen and produced only three capsules, containing only 1.49 grains weight of seed. From this it follows that an equal number of crossed and self-fertilised capsules would have contained seeds by weight as 100 to 67. I should not have thought the proportional contents of so few capsules worth giving, had not nearly the same result been confirmed by several subsequent trials. (Darwin, 1878; p. 188)
It is clear that Darwin found self-sterility noteworthy and appreciated its potential significance. However, lacking a genetic framework for understanding self-sterility, he seems, at times, to conflate self-sterility due to environmental circumstances with, what is now regarded as, genetically controlled SI.
We know that self-fertilised seedlings are inferior in many respects to those from a cross; and as with plants in a state of nature pollen from the same flower can hardly fail to be often left by insects or by the wind on the stigma, it seems at first sight highly probable that self-sterility has been gradually acquired through natural selection in order to prevent self-fertilisation….Nevertheless, the belief that self-sterility is a quality which has been gradually acquired for the special purpose of preventing self-fertilisation must, I believe, be rejected. (Darwin, 1878; p. 345)
In this observation, Darwin seems to make a distinction between intramorph incompatibility in heterostylous species and self-sterility. He clearly regards the former as a response to natural selection, but he has a less settled view about self-sterility, as exemplified by this quote. This is understandable since factors such as nutrient availability can affect fertility and probably explain lack of seed set in some of Darwin's crosses. Also, while different mating types are easily recognized in heterostylous species, this is not so for the SI species Darwin described in Cross and self-fertilisation; incompatible or partially compatible crosses probably were also observed, particularly if Darwin's plants contained a relatively small number of S-haplotypes. Nevertheless, his distinction has little appeal to modern readers trained to explain biological phenomena in terms of biochemistry and genetics; modern research into intramorph incompatibility proceeds from the same basic assumptions as research into SI (Barrett and Shore, 2008). Furthermore, modern views of SI are largely shaped by studies of model species specifically selected for their robust response. Darwin's plants, as well as many wild species, are not so well behaved (Good-Avila et al., 2008). Still, his persistent study of self-sterile species suggests that he certainly did not always attribute the phenomenon to environmental effects and that he appreciated the significance of specific mechanisms that control pollination. Darwin's study of Reseda odorata, common mignonette, provides an interesting example.
Plants of the common mignonette were raised from purchased seed, and several of them were placed under separate nets. Of these some became loaded with spontaneously self-fertilised capsules; others produced a few, and others not a single one. It must not be supposed that these latter plants produced no seed because their stigmas did not receive any pollen, for they were repeatedly fertilised with pollen from the same plant with no effect; but they were perfectly fertile with pollen from any other plant. (Darwin, 1878; p. 119)
Darwin's R. odorata materials may have been segregating for SI, although he would not have understood it in this way. Notwithstanding, it can be inferred that Darwin understood the basic behaviour of species like this in the same way SI is understood today and also appreciated the implications of such a system, as suggested by the following observations in Cross and self-fertilisation:
It is an extraordinary fact that with many species, flowers fertilised with their own pollen are either absolutely or in some degree sterile; if fertilised with pollen from another flower on the same plant, they are sometimes, though rarely, a little more fertile; if fertilised with pollen from another individual or variety of the same species, they are fully fertile; but if with pollen from a distinct species, they are sterile in all possible degrees, until utter sterility is reached. We thus have a long series with absolute sterility at the two ends;—at one end due to the sexual elements not having been sufficiently differentiated, and at the other end to their having been differentiated in too great a degree, or in some peculiar manner.
The fertilisation of one of the higher plants depends, in the first place, on the mutual action of the pollen-grains and the stigmatic secretion or tissues, and afterwards on the mutual action of the contents of the pollen-grains and ovules. Both actions, judging from the increased fertility of the parent-plants and from the increased powers of growth in the offspring, are favoured by some degree of differentiation in the elements which interact and unite so as to form a new being. (Darwin, 1878; p. 455)
Darwin's insights are uncanny. He anticipates the modern view to the extent that we think about mechanisms controlling pollination in terms of the ‘constitution’ of the pollen and pistil. We also appreciate the overall function of such systems as avoiding deleterious effects from inbreeding and interspecific crosses. A genetic framework is clearly missing from Darwin's work, however. With an understanding of inheritance and with a greatly developed understanding of the ‘constitution’ of cells (i.e. biochemistry), mechanisms controlling pollination are now conceived of as a seamless continuum from the molecular to the population level.
Early genetic studies of SI
Botanists working at the end of the 19th century articulated the basic rules of Mendelian inheritance. In the first decades of the 20th century, inheritance was linked to chromosomes, and genetic maps were being developed. However, the nature of the gene was also under debate; many traits did not seem to be inherited simply. Studies of the presence of self-sterility revealed Mendelian inheritance. For example, Compton (1913) reinvestigated R. odorata and found that self-pollination of some self-fertile plants yields self-fertile and self-sterile plants in a 3:1 ratio. East and Park (1917) made the important observation that the term ‘self-sterility’ is not really apropos. By analysing large families of Nicotiana forgetiana×N. alata hybrids over five generations, they showed that the families consisted of classes that were intrasterile but interfertile. One of East's later studies revealed that the classes were produced according to specific rules.
When individuals of any two classes are crossed, two classes appear in approximately equal numbers, but the class to which the female parent belongs is never represented.
X×Y gives Y and Z Z×X gives X and Y
Y×X gives X and Z Y×Z gives X and Z
X×Z gives Y and Z Z×Y gives X and Y
At first sight one might suppose that these results, differing as they do from any hitherto obtained in genetic work, are due to a novel type of gamete distribution; yet such a supposition is rendered unlikely by two facts. Morphological characters in these varieties appear to be transmitted in the normal Mendelian manner; furthermore the ratios in which the classes appear here are strikingly like those expected when F1, monohybrids are backcrossed with pure recessives.
To be brief, we believe that the inheritance in these cases is of the ordinary type, but that there are physiological limitations to the opportunity for fertilization.
Assume that in these populations there are three allelomorphs, S1, S2 and S3; and that Class X = S1S3, Class Y = S1S2, Class Z = S2S3. Assume further that a plant affords stimulus only to pollen which bears sterility factors other than its own. Thus a Class Z plant (S2S3) affords proper stimulus only to gametes carrying factors other than S2 or S3. In other words, of the three factors present in these various populations, the Z plants afford stimulus only to S1 gametes. (East and Mangelsdorf, 1925)
Later, East and Yarnell (1929) reported 16 S-alleles in Nicotiana. Even today, the diversity of S-alleles is a subject of much research in theoretical and population biology, but, at the time, this must have seemed like a wild excursion into new territory. The nature of the gene was still controversial, and the existence of so many S-alleles contributed to the debate. Importantly, each S-allele behaved as a discrete entity that could be passed unchanged from one generation to the next. East argued, in a spirited and entertaining way, that this is not compatible with a quantitative gene concept.
It is natural to speculate as to the bearing of these facts, if any, on the nature of the gene, and in particular on GOLDSCHMIDT'S quantitative theory of heredity. We have isolated 16 allelomorphs of the S factor (including Sf). They are all point mutations since they all behave as allelomorphs to each other and since the crossover values of 3 of them with the corolla-color factor C are approximately the same. …. Now, it may be that each of these 16 allelomorphs is charactized by a pollen-tube growth-rate of velocity different from all others.…
On the other hand, one may assume that each S allelomorph differs from the next in order by a definite quantum expressed in reaction velocity and may speculate as to what conclusions follow, if any, in connection with our conception of the gene. To put the matter most simply: Does a postulate of the above type lead to the conclusion that the different velocities of the reactions directed by the various allelomorphs of a single gene are due to the presence of different quantities of the gene? Arguments in favor of such a conclusion have been presented with much vigor and clarity in the various writings of GOLDSCHMIDT. The writers have only admiration for the brilliant work of the Berlin geneticist and certainly do not wish to be numbered among those who say that his position is ‘neither sound genetics nor sound physiology’, (GOLDSCHMIDT 1928); yet they are of the opinion that a sympathetic and receptive agnosticism in regard to his quantitative theory of gene action is at present the wisest course to take. (East and Yarnell, 1929)
East's works from this period, especially the passage quoted above from East and Mangelsdorf (1925), can be said to mark the beginning of modern thinking about SI. Although he continued to use the term self-sterility, East definitively showed that the system is better characterized as a genetically controlled mechanism for differentiating families into intrasterile, interfertile compatibility classes that could be understood as distinct S-genotypes. In summary, the ‘Nicotiana type’ of SI was found to be controlled by a series of alleles of a single locus, designated the S-locus. Plants with two S-alleles in common are incompatible. When plants sharing one allele are crossed, only the pollen with the S-allele that is distinct from the pistil can effect fertilization. Since it is the S-genotype or, as it is currently understood, the S-haplotype, of the male gametophyte that determines compatibility, this type of SI is now called gametophytic SI.
Clearly, with these advances, compatibility between classes was conceived as being controlled at the physiological level. Although no mechanism was postulated, the physiology was thought of as directly controlling the opportunity for fertilization (East and Mangelsdorf, 1925). This conclusion can be seen as a dramatic realization of Darwin's notion, quoted earlier, that pollination could be controlled by ‘mutual action of the pollen-grains and the stigmatic secretion or tissues’ (Darwin, 1878). Physiological control over compatibility occurs against the background of siphonogamy; that is, the stigma and style provide an environment for pollen to germinate and for pollen tubes to grow toward the ovary and carry the sperm cells that will fertilize the egg and central cells. Thus, today, it is believed that Darwin's ‘mutual action’ is best regarded as a series of interactions in the stigma, style, and ovary that must occur successfully for fertilization to occur. The challenge for understanding SI at the physiological level came to be understood in terms of identifying the products of the S-locus genes. In the Nicotiana system, it took 60 years to meet this challenge. Species with similar SI systems are now studied throughout the Solanaceae, Rosaceae, and Plantaginaceae. SI in Papaver has an entirely different mechanism, although the genetics are identical (Franklin-Tong, 2008b), and many other distinct SI systems have been described in other families (deNettancourt, 1977, 2001; Franklin-Tong, 2008a).
Identification and characterization of S-genes in S-RNase-based SI systems
A major goal of the last quarter century has been to identify the genes that control SI, characterize their products, and place them in a physiological framework that accounts for pollination specificity. Studies of SI also took on significance as an example of plant cell–cell communication. The single-locus genetics of SI, with its implied simplicity at the underlying biochemical level, seemed to offer both a good prospect for success and an entry point for molecular-level studies. The approaches used to identify S-gene products were based on the biological and genetic foundations laid previously.
Broadly, the stigma and style function to support the growth of desirable pollen and prevent the growth of undesirable pollen. Interactions between the pollen and pistil are mediated by the extracellular matrix (ECM) of both partners, and signals travel in both directions. The ECM contains lipids, carbohydrates, and glycoproteins. Thus, the problem has been framed as identifying S-specific products in the ECM.
Standing on the shoulders of a giant like East makes identifying S-genes easy; they are the genes that determine S-specific pollen rejection (Fig. 2). Identifying S-gene products is another matter. Bredemeijer and Blass (1981) first correctly identified S-glycoproteins in N. alata. They analysed pistil protein extracts and identified bands associated with specific S-alleles. Earlier, Pandy (1967) had used a similar approach and identified isozymes associated with S-alleles, but they did not prove to be products of the S-locus. The different results from these two studies foreshadowed a lingering problem in SI studies: SI favours outcrossing and, thus, maintains a high level of polymorphism (especially in genes closely linked to the S-locus) that complicates genetic analysis. Modern researchers do not generally have the wherewithal for studies such as East's that involve tens of thousands of crosses (East and Park, 1917).
Anderson et al. (1986) cloned the first S-glycoprotein; they obtained the sequence from protein isolated from the pistil ECM and then used this information to identify the N. alata S2-glycoprotein cDNA. McClure et al. (1989) showed that S-glycoproteins possess RNase activity, and, henceforth, these proteins have been known as S-RNases. Cloning the first S-RNase was a breakthrough that opened the door to characterizing the structure and expression of S-gene products and laid the groundwork for understanding the physiological basis for pollen rejection that Darwin sensed, but could not grasp, a century earlier. Early studies showed that S-RNase expression coincided with the developmental onset of SI and that expression was restricted to cells forming the path from the stigma to the ovary (Anderson et al., 1986; Cornish et al., 1987). Sequence analysis showed that S-RNases possess secretion signals, and immunolocalization studies provided direct evidence that they are deposited in the ECM (Anderson, 1989). The abundance of S-RNase in the ECM was studied by protein purification and immunological techniques, and both methods support concentration estimates near 1 mM (Jahnen et al., 1989). S-RNase transcripts are also extremely abundant and display very long poly(A) tails (McClure et al., 1993). All S-RNases also possess N-linked glycans (Woodward et al., 1989), and Oxley et al. (1996, 1998) have determined the structure of several glycans present on S-RNases from N. alata. Although it is clear that N-glycans are not required for S-specific recognition per se (Karunanandaa et al., 1994), it is possible that glycosylation facilitates secretion, solubility in the ECM, or some other function in SI. Requirement for high-level expression is clear (Clark et al., 1990; Murfett et al., 1994, 1995), but it is still not known why so much S-RNase is needed.
Many experiments characterizing S-RNases relied on the ability to manipulate cloned genes in vitro and to reintroduce them into transgenic plants. These experiments were challenging because of the extraordinary expression levels of S-RNases. However, experiments in Nicotiana, Petunia, and Solanum showed that transformation of a cloned S-RNase into a new genetic background is sufficient to cause rejection of the corresponding pollen S-haplotype (Huang et al., 1994; Lee et al., 1994; Murfett et al., 1994, 1995; Matton et al., 1997, 1999; Beecher and McClure, 1999). For example, expressing SA2-RNase from N. alata causes rejection of SA2-pollen (Murfett et al., 1994). Similarly, suppressing S-RNase expression prevents S-specific pollen rejection (Lee et al., 1994; Murfett et al., 1995). These experiments are important because they definitively show that the S-RNase protein contains all the information needed to determine S-specificity in the pistil. In a classic experiment, Huang et al. (1994) expressed catalytically inactive S3-RNase in P. inflata and showed that it does not cause S-specific pollen rejection. This finding provided direct evidence that RNase activity is required for pollen rejection.
Further experiments addressed how S-specificity is encoded in the S-RNase sequence. Sequence comparisons showed that S-RNases contain five conserved regions, C1–C5, that altogether comprise ∼40 residues (Ioerger et al., 1991; Ishimizu et al., 1998). Histidine residues implicated in catalysis are located in C2 and C3 (Ishimizu et al., 1995). Structural studies show that C1, C2, and C5 form three strands of β-sheet at the protein core; one of the catalytic histidines projects from C3 toward one face of the sheet, and C4 contributes hydrophobic residues that interact with the other side of the sheet (Ida et al., 2001). However, sequence variation is the most striking feature of S-RNases. Even in the conserved regions, there are only a small number of highly conserved residues, and the rest of the protein (i.e. ∼160 residues) is subject to variation (Ishimizu et al., 1995). Domain swap experiments tested whether a specific S-RNase region determines S-specificity. Zurek et al. (1997) made a series of nine chimeric constructs exchanging sequences between N. alata SA2- and SC10-RNases. Although all these chimeric proteins were active RNases, none was able to cause rejection of either SA2- or SC10-pollen. Similar results were obtained with two constructs in Petunia (Kao and McCubbin, 1996). These results suggest that residues that determine S-specificity are not restricted to a particular region of S-RNase. Nevertheless, sequence analyses suggest that two regions, often referred to as HVa and HVb, are especially variable (Ioerger et al., 1991). Experiments in Solanum showed that exchanging just four of these residues between two highly similar proteins is sufficient to change S-RNase specificity (Matton et al., 1997). Thus, these regions are important for pollen recognition but are not sufficient in all cases.
Non-S-specific factors also contribute to SI on the pistil side. Unlinked modifiers of S-RNase-based SI have been noted in numerous studies (Anderson and de Winton, 1931; East, 1932; Martin, 1961, 1968; Ai et al., 1991; Bernatzky et al., 1995), but only two have been cloned. The HT-B gene from N. alata encodes a 101 residue asparagine-rich protein expressed in the pistil (McClure et al., 1999). Suppression of HT-B in Nicotiana and Solanum interferes with S-specific pollen rejection (McClure et al., 1999; O'Brien et al., 2002). Although the exact function of HT-B is not known, immunolocalization experiments show that pollen tube uptake of S-RNase occurs normally in plants with undetectable HT-B protein (Goldraij et al., 2006). Thus, HT-B is required only after S-RNase has entered the pollen tube. In Solanum, two very similar genes, HT-A and HT-B, have been identified, but RNA interference (RNAi) results only implicate HT-B in SI (O'Brien et al., 2002). Further studies in the tomato clade of Solanum suggest that mutations in HT-B genes might have been important in the transition from SI to self-compatability (SC) (Kondo et al., 2002a, b). Other HT-family genes have been identified in Petunia and Nicotiana, and sequence analysis of the larger family suggests similarity to a class of membrane-associated glycine-rich proteins (Sassa and Hirano, 2006; Kondo and McClure, 2008). Interestingly, HT-B is selectively degraded after compatible pollinations in N. alata (Goldraij et al., 2006).
The 120 kDa glycoprotein (120K), an abundant arabinogalactan protein present in the N. alata pistil ECM (Lind et al., 1994), has also been implicated in SI (Hancock et al., 2005). 120K is an S-RNase-binding protein (Cruz-Garcia et al., 2005) that is taken up by pollen tubes growing in the transmitting tract's ECM (Lind et al., 1996; Goldraij et al., 2006). S-RNase is taken up normally in 120K-suppressed plants, suggesting that, as with HT-B, 120K plays a role in a later step in the SI response. Yeast two-hybrid (Y2H) experiments identified pollen proteins that interact with 120K, one of which, interestingly, is NaSBP1 (N. alataS-RNase-binding protein 1), a protein first identified in P. hybrida that interacts with S-RNase (Sims and Ordanic, 2001) and the S-locus F-box protein (Hua and Kao, 2006). 120K also interacts with a pollen-specific C2-domain-containing protein, NaPCCP (Lee et al., 2008), which may have a role in transporting proteins from the pistil's ECM to the pollen tube's endomembrane compartments. NaPCCP's C-terminal domain binds to 120K, while the C2 domain specifically binds to phosphatidylinositol-3-phosphate (PIP3; CB Lee et al., 2009).
The determinant of S-specificity on the pollen side is referred to as pollen S. Unlike S-RNase, which was identified using protein-based approaches, pollen S was identified by searching for pollen-expressed genes linked to the S-RNase gene. Lai et al. (2002) were the first to identify an F-box protein gene linked to S-RNase. F-box proteins are best known for their functions in the ubiquitin–proteasome system. In this context, they usually function to bind a client protein to an SCF (Skp1–Cullin–F-box) E3 ubiquitin ligase complex that transfers ubiquitin from an E2 carrier to the client (Moon et al., 2004; Smalle and Vierstra, 2004). The polyubiquitylated client is subject to degradation by the 26S proteasome. Although the Antirrhinum S-linked F-box protein gene did not, at first, appear to be sufficiently polymorphic to determine S-specificity, later studies strongly support that F-box proteins do function as pollen S genes in S-RNase-based SI. These genes are referred to as S-locus F-box (SLF) or S-locus F-box (SFB) in different species. Strong evidence implicating F-box proteins in SI was obtained in Prunus species with relatively compact genomes (reviewed in McClure, 2004). Sijacic et al. (2004) definitively showed that the SLF2 gene from P. inflata has an S-specific effect on pollination, providing direct evidence that this gene determines S-specificity. An Antirrhinum SLF gene has also been implicated in SI, but S-specificity could not be directly tested (Qiao et al., 2004a).
SLF proteins usually show less sequence polymorphism than S-RNases. A recent summary showed non-synonomous substitution rates of 0.14–0.51 for S-RNases from Antirrhinum, Petunia, and Prunus compared with 0.01–0.11 for SLF or SFB (Newbigin et al., 2008). SLF sequences from Antirrhinum are especially noteworthy for their lack of sequence polymorphism. This is surprising not only because a certain level of polymorphism is expected of recognition proteins, but also because of the contrast with S-RNase sequences. The results are difficult to reconcile with the strong expectation for a shared evolutionary history for pairs of pollen and pistil specificity genes in a given S-haplotype. Petunia SLF genes show more polymorphism than Antirrhinum sequences, and, as noted previously, these genes have been directly implicated in SI. Prunus SFB genes display a degree of polymorphism that is closer to S-RNase. The presence of multiple SLF- or SFB-like genes linked to S-RNase is a serious complication. Wheeler and Newbigin (2007) found 10 F-box protein genes linked to S-RNase in N. alata, seven of which were expressed in pollen. It is not clear which, if any, of these genes is pollen S. Further, phylogenetic analyses of Antirrhinum and Petunia sequences show that SLF genes do not display all the features expected of pollen S genes (Wheeler and Newbigin, 2007; Newbigin et al., 2008), and yet the evidence at the molecular level seems incontrovertible (Sijacic et al., 2004). In contrast, phylogenetic analyses of Prunus SFB genes do conform to expectations (Newbigin et al., 2008). Together, the results suggest that there is still much to learn about the role of SLF and SFB genes in SI.
Y2H and in vitro studies show that SLF proteins bind to S-RNase as well as to a variety of pollen proteins. Qiao et al. (2004b) reported Y2H and pull-down experiments showing that the C-terminal portion of AhSLF-S2 binds to S-RNase in a non-S-specific fashion. Subsequent studies identified a pollen-specific Skp1-like protein, AhSSK1, that binds to AhSLF proteins and also interacts with Cullin-1 (Huang et al., 2006). The authors suggest that AhSSK1, Cullin-1, and AhSLF proteins function together in an SCF-like complex. Somewhat different results have been obtained with Petunia SLF proteins. Hua and Kao (2006) specifically tested three P. inflata Skp1 homologues for interaction with PiSLF2, but reported no interaction in Y2H. This is an extremely interesting result given that the F-box specifically functions to bind an F-box protein to a Skp1-like protein (Bai et al., 1996). Nevertheless, Hua and Kao (2006) reported that, instead, PiSLF proteins bind to PhSBP1, a protein also reported to interact with S-RNase, 120K, and certain transcription factors (Sims and Ordanic, 2001; O'Brien, 2004; Ben-Naim et al., 2007; Lee et al., 2008). SBP1 proteins contain RING domains and are thought to act as E3 ubiquitin ligases (Sims and Ordanic, 2001; Hua and Kao, 2006; Hua et al., 2008). Hua and Kao (2006) also reported that PiSBP1 binds to a cullin protein and proposed that it forms an SLF–SBP1–Cullin complex that functions in SI. Other studies suggest that non-self S-RNases bind to SLF more productively than self S-RNases (Hua et al., 2007). SLF proteins are divided into three domains, FD1, FD2, and FD3. FD2 is a putative non-S-specific S-RNase-binding domain (SBD), and putative recognition regions are present in the flanking FD1 and FD3 domains. The authors proposed that, in incompatible pollinations, FD2 binds all S-RNases but that self-interactions (e.g. S1-RNase interacting with PiSLF1) prevent a productive complex from forming. A non-self interaction (e.g. S1-RNase with PiSLF2) is proposed to be stable, allowing ubiquitylation and subsequent degradation of non-self S-RNase in compatible pollinations (Hua et al., 2008). Thus, there is evidence that SLF proteins interact with other pollen proteins and with S-RNase. However, there is conflicting evidence about whether these pollen proteins form an SCF-like complex or, a novel SLF–SBP1–Cullin complex.
Darwin recognized that the constitution of the pollen and pistil control pollination. Many of the constituents required for S-RNase-based SI have been identified: S-RNase, SLF/SFB, 120K, HT-B, and perhaps SSK1, Cullin, and SBP-1. S-RNase and SLF determine S-specificity, so their interaction ultimately controls compatibility. The complexes formed with pollen proteins and their potential interactions with pistil factors such as HT-B and 120K are central to models of the physiology of SI.
Physiological models for S-RNase-based SI
The discovery that the pistil-side S-specificity determinants are S-RNases was a turning point in the field, as it suggested a physiological basis for pollen rejection. Enzymes that attack RNA function as cytotoxins in organisms as diverse as bacteria and plants. Therefore, it is reasonable to hypothesize that S-RNases have a cytotoxic effect on incompatible pollen tubes (McClure, 1989, 1990; Gray, 1991). Experimental evidence for cytotoxicity was provided from tracer experiments showing that pollen rRNA is degraded after incompatible, but not compatible, pollinations (McClure et al., 1990). Further experiments showed that S-RNases can enter pollen tubes where they can act as effective translational inhibitors (Gray et al., 1991). Thus, a physiological mechanism for pollen tube inhibition based on S-RNase-based cytotoxicity emerged: S-RNases are secreted into the ECM by pistil cells and taken up by growing pollen tubes, and incompatibility is explained by a cytotoxic effect of RNA degradation (Gray et al., 1991).
The cytotoxic model provided a neat explanation for incompatibility and, consequently, focus shifted to understanding how pollen tubes evade cytotoxic S-RNase in compatible pollinations. Two alternatives were suggested: pollen S might function as an S-specific S-RNase receptor that provides for uptake of only self S-RNase; or S-RNase uptake could be non-specific, and pollen S could function to inhibit non-self S-RNase (Thompson and Kirch, 1992). The former hypothesis received little support, but the hypothesis that pollen S functions to inhibit the action of S-RNase stimulated much research. For example, mutational studies in Nicotiana showed that all pollen-part mutants could be explained as duplications of all or parts of the S-locus (Golz et al., 2001). The absence of deletion mutants was interpreted as evidence that pollen S is an essential gene and, in turn, that it functions to inhibit S-RNase. Studies of breakdown of SI in tetraploids of solanaceous species were also consistent with pollen S acting as an inhibitor. A tetraploid such as S1S1S2S2 is typically SC. The defect is on the pollen side since it rejects pollen from SI S1S2 diploids, but the reciprocal pollination is compatible. Significantly, only heteroallelic pollen (i.e. diploid S1S2 pollen) shows SI breakdown; S1S1 and S2S2 diploid pollen are rejected normally. This is sometimes referred to as the heteroallelic pollen (HAP) effect; it is thought to reflect an inhibitory action of pollen S where the pollen S alleles provide cross-protection against S-RNases. For example, pollen S1 inhibits the action of all S-RNases except S1-RNase, and pollen S2 inhibits all S-RNases except S2-RNase. However, different results have been obtained in Prunus species that also possess S-RNase-based SI. For example, an SC mutant in Prunus avium possesses a deletion of the pollen S gene, known as SFB in this species (Sonneveld et al., 2005). Furthermore, the HAP effect does not occur in tetraploid cherry, Prunus cerasus (Hauck et al., 2006). Some authors suggest that these conflicting results reflect different underlying SI mechanisms between Prunus and solanaceous species (Hauck et al., 2006). However, neither system is sufficiently understood at the biochemical level to test this hypothesis properly.
Preferential degradation of non-self S-RNase in compatible pollen tubes could explain resistance to its cytotoxic effects (Fig. 3). This is a logical hypothesis since, as described in the previous section, there is evidence that SLF proteins participate in ubiquitin ligase complexes. In the S-RNase degradation model for S-RNase-based SI, either an SCFSLF (i.e. in Antirrhinum, Huang et al., 2006) or a PiSLF–PiSBP1–PiCUL1-G (i.e. in Petunia, Hua et al., 2008) complex provide for ubiquitylation and subsequent proteasomal degradation of non-self S-RNases in compatible pollen tubes. Hua and Kao (2007) provided support for this proposal by showing that non-self S-RNases might indeed bind to SLF proteins more tightly than self S-RNases. Finally, in vitro experiments with extracts show that S-RNase can be degraded in pollen tube extracts, although degradation is not S specific (Hua et al., 2006). Thus, many results are consistent with a non-self S-RNase degradation model to explain how compatible pollen tubes evade S-RNase cytotoxicity.
Studies of the uptake and fate of S-RNase in pollen tubes led to formulation of an alternative model, one that explains compatibility as arising from compartmentalization of S-RNase. The compartmentalization model consists of three main steps: non-S-specific uptake of S-RNase into the endomembrane system, an S-specific step that affects HT-B stability, and release of S-RNase in incompatible pollinations. In this model, compatibility is regarded as a default condition arising from sequestration of S-RNase into a vacuole where it cannot attack RNA in the cytoplasm. Incompatibility actively interferes with the default mechanisms pollen tubes use to escape S-RNase cytotoxicity; ultimately, incompatibility leads to release of S-RNase from the endomembrane system. Luu et al. (2000) presented immunolocalization results showing non-S-specific uptake of S-RNase in Solanum chacoense. They reported S-RNase in the pollen tube cytoplasm, but their fixation methods were not designed for optimum membrane preservation, and specific compartment markers were not used. Later, Goldraij et al. (2006) investigated S-RNase and 120K uptake in N. alata pollen tubes using immunolocalization and confocal microscopy. In compatible pollen tubes, the results showed S-RNase present in the lumen of pollen vacuoles and 120K present in the lumen and the surrounding membrane. Two additional markers, vacuolar pyrophosphatase and aleurain, confirmed the identity of the S-RNase-containing vacuolar compartment. Late-stage incompatible pollen tubes, in contrast, did not show S-RNase compartmentalization, but rather a signal distributed throughout the pollen tube. Clearly, compartmentalization of S-RNase away from pollen tube cytoplasm would be an effective mechanism to prevent its cytotoxicity, and release of S-RNase from the endomembrane system could account for pollen tube rejection. It has already been mentioned that HT-B, a pistil protein required for pollen rejection, is selectively degraded in compatible pollen tubes. Goldraij et al. (2006) speculated that HT-B might have a role in S-RNase release, although there are many other possibilities. The S-RNase–SLF interaction may somehow stabilize HT-B in incompatible pollinations indirectly. For example, in compatible pollinations, HT-B could be degraded by a pollen protein that, in an incompatible pollination, is destabilized by a self S-RNase–SLF interaction (Goldraij et al., 2006). The compartmentalization model is compatible with breakdown of SI in tetraploids if a biochemical interaction between allelic SLF proteins interferes with self-recognition. One possibility is that SLF proteins form inactive multimers, as has been previously proposed (Luu et al., 2001). For example, in an S1S2 diploid pollen tube, an inactive SLF1–SLF2 dimer might be incapable of inactivating the pollen protein that degrades HT-B, causing the pollen tube to revert to the default condition of compatibility.
There are unresolved issues in both the S-RNase degradation and compartmentalization models. The S-RNase degradation model does not easily accommodate a role for the non-S-specific factors required for SI in Nicotiana, Petunia, and Solanum (Anderson and de Winton, 1931; East, 1932; Martin, 1961, 1968; Ai et al., 1991; Bernatzky et al., 1995). Some of the molecular-level details of this model are awkward. For example, the ‘common domains’ in S-RNases, usually referred to as conserved regions C1–C5 (Ioerger et al., 1991), are located in the protein interior (Ida et al., 2001) where it is unlikely they could interact with an SBD on SLF. All other S-RNase regions are subject to sequence variation, making it difficult to envisage how numerous non-self S-RNases would always be positioned to interact preferentially with a given SLF protein. In vitro studies point to direct binding between S-RNase, SLF, and auxiliary pollen proteins, and also provide evidence for ubiquitylation (Qiao et al., 2004b; Hua and Kao, 2006, 2008; Huang et al., 2006; Hua et al., 2007). Further studies show that S-RNase is subject to degradation by the 26S proteasome in pollen extracts. However, further research is needed to provide a firm connection between these results. S-RNase degradation in vitro is not S-specific, which could be interpreted as evidence that SLF is not involved (Hua and Kao, 2006). It might also be significant that glycosylated S-RNase isolated from pistils is not degraded by pollen extracts (Hua and Kao, 2006). In yeast and mammals, PNG (peptide:N-glycanase) associates with the 26S proteasome via HR23B, where it assists with glycoprotein degradation (Katiyar et al., 2004; Li et al., 2005). Thus, the possibility that the non-S-specific degradation of S-RNase in pollen tube extracts is due to a process entirely separate from SI has not been eliminated. Similarly, SBP1–SLF binding should be interpreted cautiously. SBP1 is expressed in almost all tissues (Sims and Ordanic, 2001; O'Brien, 2004; Lee, 2008), and it appears to bind to a variety of client proteins. Several of its clients are implicated in SI [i.e. S-RNase, SLF, and 120K (Sims and Ordanic, 2001; Hua and Kao, 2006; Lee et al., 2008)], but some are not (Ben-Naim et al., 2007; Lee et al., 2008). This observation in no way excludes a role for SBP1–SLF–S-RNase interactions in SI, but it highlights the need for experiments that distinguish between S-specific processes and other cellular mechanisms.
The compartmentalization model also has deficiencies. Localization studies in Antirrhinum show that SLF is localized in the cytoplasm, with some concentrations observed near the endoplasmic reticulum (ER) (Wang and Xue, 2005). Thus, S-RNase must move to the cytoplasm to interact with SLF. To date, no data address how this occurs. One speculation is that transport of S-RNase may be similar to that of the cytotoxin ricin (McClure, 2006). Ricin is a binary ribosome-inactivating cytotoxin synthesized in cotyledons of Ricinus communis and sequestered in protein storage vacuoles. It enters mammalian cells by endocytosis, and, like S-RNase, most ricin is sequestered in the vacuole. However, a small portion of ricin moves through the Golgi and then to the ER by retrograde transport (van Deurs et al., 1988). The cytotoxic ricin A chain is thought to exit the ER using the ER-associated degradation (ERAD) system (Wesche et al., 1999). During exit, it is acted upon by PNG, and only a small portion of the ricin A chain taken up by the cell evades degradation to exert its cytotoxic effect. Ricin provides a model for movement of a cytotoxin from luminal compartments in the endomembrane system to the cytoplasm, and it is possible that S-RNase follows a similar pathway (Fig. 3). The precise role of factors such as HT-B and, in particular, the mechanism of HT-B degradation in compatible pollen tubes are not yet explained. Results clearly show that HT-B stability is decreased in compatible pollinations compared with incompatible pollinations (Goldraij et al., 2006). Thus, its stability is regulated by the S-RNase–SLF interaction. As discussed above, it is possible that this occurs by action of a pollen protein whose stability is regulated in SI. However, no data confirm this hypothesis. It is also possible that regulation of HT-B stability could be less direct.
The S-RNase degradation and the compartmentalization models make different inferences about the relationship between the S-specificity-determining S-RNase–SLF interaction and the ability of pollen to prevent the cytotoxic action of S-RNase. In the S-RNase degradation model, non-self S-RNase is degraded as a direct result of this interaction. In the compartmentalization model, inhibition arises by a non-S-specific mechanism that sequesters S-RNase into a vacuole where it cannot interact with its substrate, and the S-RNase–SLF interaction interferes with this process. Thus, the degradation model predicts that SLF is an essential gene, and this is consistent with interpretations of mutagenesis experiments in solanaceous species (Golz et al., 2001). In the compartmentalization model, SLF causes pollen rejection so is not essential, and this is consistent with identification of pollen S deletion mutants in Prunus (Sonneveld et al., 2005). Nevertheless, the models are not irreconcilable. Pollen tubes might possess a multilayered resistance to S-RNase where most non-self S-RNase is rendered harmless by compartmentalization but a small amount is degraded.
Darwin clearly showed keen insight; he inferred that pollination is controlled by mutual action between constituents of the pollen and pistil. Yet, he did not understand genetics as we do, and East's genetic model for SI might be said to have laid the groundwork for all subsequent molecular-level studies. Using increasingly subtle tools of modern biochemistry, molecular biology, and cellular imaging, modern researchers have identified several key pollen and pistil constituents and proposed ‘mutual actions’ that could control pollination. For better or worse, this is what we have to show for a century and a half. We can only hope our progress will be judged as useful in another century and a half.
I thank Melody Kroll for editorial assistance and help with the figures. I also thank Professor John van Wyhe for permission to use Fig. 1 and his work on the Darwin Online project (http://darwin-online.org.uk/). Work in the author's laboratory is supported by US National Science Foundation grants IOB 0614962 and DBI 0605200.