Diverse cell signalling pathways regulate pollen-stigma interactions: the search for consensus


Author for correspondence:
Simon Hiscock
Tel: +44 117 9549835
Fax: +44 117 3317985
Email: Simon.hiscock@bristol.ac.uk



II.The pollen–stigma interaction289
III.Pollen–stigma interactions in species with wet stigmas290
IV.Pollen–stigma interactions in species with dry stigmas295
V.Is there any consensus among cell signalling pathways regulating pollen–stigma interactions?299
VI.Incompatibility and the pollen–stigma interaction300
VII.New directions in pollen–stigma interaction research302
VIII.Future prospects310


Siphonogamy, the delivery of nonmotile sperm to the egg via a pollen tube, was a key innovation that allowed flowering plants (angiosperms) to carry out sexual reproduction on land without the need for water. This process begins with a pollen grain (male gametophyte) alighting on and adhering to the stigma of a flower. If conditions are right, the pollen grain germinates to produce a pollen tube. The pollen tube invades the stigma and grows through the style towards the ovary, where it enters an ovule, penetrates the embryo sac (female gametophyte) and releases two sperm cells, one of which fertilizes the egg, while the other fuses with the two polar nuclei of the central cell to form the triploid endosperm. The events before fertilization (pollen–pistil interactions) comprise a series of complex cellular interactions involving a continuous exchange of signals between the haploid pollen and the diploid maternal tissue of the pistil (sporophyte). In recent years, significant progress has been made in elucidating the molecular identity of these signals and the cellular interactions that they regulate. Here we review our current understanding of the cellular and molecular interactions that mediate the earliest of these interactions between the pollen and the pistil that occur on or within the stigma – the ‘pollen–stigma interaction’.

I. Introduction

All life evolved in the oceans, and the life cycles and reproductive biology of land-dwelling organisms must be interpreted and understood in terms of these aquatic origins (Church, 1919). As Corner (1964) explains so eloquently in the classic Life of Plants: ‘even an orchid carries a recollection of its marine ancestry’. The motile male gametes (sperm) of animals and many plant groups are a legacy of this marine ancestry, which dictates that fertilization can take place only in a more-or-less aqueous environment. The success of land-dwelling organisms has therefore depended largely on coping with the reproductive limitations of this legacy. Angiosperms and some gymnosperms (e.g. pines and Gnetales) have evolved a system of sexual reproduction where water is not required at the point of fertilization, nonmotile sperm being delivered directly to the egg by a pollen tube. This key innovation, termed ‘siphonogamy’, coupled with the evolution of the carpel (gynoecium), freed angiosperms from a dependency on water for sexual reproduction (pines and Gnetales still require a watery ‘pollination droplet’ to capture airborne pollen grains) and allowed them to reproduce sexually in most terrestrial environments. Because of their superior reproductive abilities, flowering plants have been likened to the mammals of the animal kingdom (Heslop-Harrison, 2000), but not even mammals have managed to take sexual reproduction to the anhydrous limits permitted by siphonogamy in angiosperms.

Siphonogamy evolved as an inevitable consequence of drastic modifications to the land plant life cycle (alternating between haploid gametophyte and diploid sporophyte life stages) in gymnosperms and angiosperms. These modifications saw the once free-living female gametophyte reduced to a ‘parasitic’ egg-producing structure borne within an ovule on the sporophyte, and the male gametophyte reduced to a cluster of cells encased within a sporopollenin-encrusted pollen grain. This process was driven by selection for smaller and smaller gametophytes to reduce the need for water in the sexual stage of the life cycle (Niklas, 1997). In angiosperms, ovules are surrounded and enclosed by further layers of sporophytic tissue that form the gynoecium (carpel or pistil), a unique maternal structure considered fundamental to their evolutionary success (Whitehouse, 1950; Mulcahy, 1979; Crepet & Friis, 1987).

The origins of the carpel and the flower – Darwin's ‘abominable mystery’– have received much attention from molecular geneticists in recent years, with the identification of genes controlling flower initiation, organ identity, and development in the model plants Arabidopsis and Antirrhinum, inspiring hypothetical ‘evo-devo’ solutions to the ‘abominable mystery’ (reviewed by Thiessen & Melzer, 2007). A critical consequence of the gynoecium is that before a pollen tube (male gametophyte) can access an ovule to release its two sperms into the embryo sac (female gametophyte), it must first penetrate the gynoecium and navigate its way through this maternal sporophytic tissue to find an ovule. The events and interactions that occur during this prezygotic cellular and molecular ‘courtship’ between haploid pollen and diploid gynoecium have been termed the pollen−pistil interaction (Heslop-Harrison, 1975), and consist of six key stages:

  • 1pollen capture and adhesion
  • 2pollen hydration
  • 3germination of the pollen to produce a pollen tube
  • 4penetration of the stigma by the pollen tube
  • 5growth of the pollen tube through the stigma and style
  • 6entry of the pollen tube into the ovule and discharge of the sperm cells.

In a compatible pollen–pistil interaction, all six steps are completed, resulting in double fertilization to form zygote and endosperm (see Lord & Russell, 2002; Dresselhaus, 2006 for details of double fertilization). In an incompatible or ‘incongruous’ interaction (see below), one or more steps may break down such that fertilization is precluded. Most cross-pollinations involving pollen and pistil partners of the same species are compatible, however; because the majority of angiosperms are hermaphrodite, self-pollination is always a possibility. Many angiosperms have therefore evolved genetically determined self-incompatibility (SI) systems to avoid self-fertilization and its detrimental consequence, inbreeding depression. SI is usually controlled by a single, highly polymorphic locus, S, which when matched in pollen and pistil triggers pollen rejection by interfering with one or more steps of the pollen–pistil interaction (de Nettancourt, 1977; Gaude & Dumas, 1987; Wheeler et al., 2001; Hiscock & McInnis, 2003a; Nasrallah, 2003; Takayama & Isogai, 2005; McClure & Franklin-Tong, 2006). Recognition and rejection of pollen is orchestrated by a minimum of two different nonrecombining genes at the S locus, controlling pollen identity and pistil identity, respectively. The precise genetics of SI (see section VI) means that it not only prevents self-fertilization, but also prevents fertilization between different individuals that share the same S haplotype – a feature that has important consequences for mate compatibility within populations. Notwithstanding the importance of SI as an outbreeding system, many hermaphrodite angiosperms (approx. 20%) are habitual selfers, and even more (approx. 33%) reproduce through a mixture of selfing and outcrossing (mixed mating systems; Kalisz et al., 2004; Goodwillie et al., 2005).

Pollinations involving pollen and pistillate partners of different species are usually incompatible or incongruous (because of genetic divergence; Hogenboom, 1975), a prerequisite for the maintenance of biological species boundaries (Stebbins, 1950; Grant, 1981). Even so, many interspecific (and even intergeneric) pollinations may be compatible and lead to the formation of hybrids (Grant, 1981; Rieseberg, 1997; Arnold, 2006), indicating the extraordinary diversity and unpredictability of compatibility in the pollen–pistil interaction. As a general rule, however, closely related species tend to hybridize more easily than distantly related species, unless interspecific incompatibility systems, which actively reject heterospecific pollen (see section VI), preclude. The term incongruity, rather than incompatibility, is usually applied to the failure of matings between more distantly related species because it implies a lack of fit between the pollen−pistil recognition machinery of the two mating partners because of genetic and evolutionary divergence (Hogenboom, 1975).

The widespread occurrence of incompatibility and incongruity phenomena in the pollen–pistil interaction indicate the effectiveness of the angiosperm gynoecium as a ‘pollen sieve’, ensuring that only those pollen tubes carrying desirable male gametes reach the ovules and fertilize the egg. It has been suggested that the early acquisition of SI by the first angiosperms helped drive their explosive adaptive radiation during the late Cretaceous and Tertiary (Whitehouse, 1950). At another level, because more compatible pollen is usually deposited on a stigma than there are ovules available for fertilization, only the most vigorous or fastest-growing pollen tubes will reach the ovules and effect fertilization (Ottaviano et al., 1980; Hormazo & Herrero, 1992). This competition between pollen grains allows selection to act on the fitness of the male gametophyte as well as on the sporophyte, a further consequence of the pollen–pistil interaction that is thought to have had profound consequences for angiosperm evolution and diversification (Mulcahy, 1979).

Understanding the physiological and molecular basis of the cellular interactions that occur during the pollen–pistil interaction is a major goal for plant biologists, because ultimately we all depend on this fundamental biological process for our food, whether directly (through consumption of cereal crops such as wheat, rice and maize) or indirectly (through consumption of animals fed on fodder crops such as kale, maize and canola). In recent years, huge strides have been taken in identifying molecules that mediate specific events during the pollen–pistil interaction, in particular pollen adhesion to the stigma, pollen tube growth and guidance, and pollen recognition and rejection during SI (for reviews see Holdaway-Clarke & Hepler, 2003; Edlund et al., 2004; Sanchez et al., 2004; Swanson et al., 2004; Takayama & Isogai, 2005; Malho et al., 2006; McClure & Franklin-Tong, 2006; Wilsen & Hepler, 2007). Despite this recent exponential increase in the number of molecules implicated in compatible pollen–pistil interactions in different model plant species, no general consensus has yet emerged of a universal set of pollen−pistil-mediation molecules that regulate a common programme of cellular interactions necessary for compatibility (Lord, 2003). Instead, we have a mosaic of data from diverse model plants amenable to, or desirable for, studies of pollen–stigma interactions (Fig. 1). This lack of consensus is perhaps not surprising because genes with products that regulate sexual reproductive processes are known to evolve and diversify far more rapidly than those that regulate and maintain housekeeping processes (Swanson & Vacquier, 2002). Furthermore, given that molecules regulating pollen–pistil interactions must, by their nature, contribute to the definition of species boundaries (Swanson et al., 2004), it is expected that such key molecules will show extreme levels of diversification. Recent identification of highly polymorphic gene families expressed in stigmas and pollen support this prediction (Mayfield et al., 2001; Schein et al., 2004).

Figure 1.

Phylogenetic relationships among plants most widely used in studies of pollen–stigma interactions. The most commonly studied model plants are underlined. The tree is based on a simplified version of APG II (2003).

Here we provide an overview of what is currently known about the cellular and molecular interactions that mediate the earliest events of the pollen–pistil interaction on or within the stigma: the pollen–stigma interaction. For recent reviews of later events in the pollen–pistil interaction, fertilization and self-incompatibility, we recommend Sanchez et al., (2004); Dresselhaus (2006); Takayama & Isogai (2005), respectively.

II. The pollen–stigma interaction

Pollen is released from the anther as a dehydrated microspore containing the male gametophyte. The degree of pollen dehydration varies considerably between species, but the water content of pollen at anthesis is usually within the range of 15–35% of fresh weight (Stanley & Linskens, 1974; Dumas et al., 1984). Pollen is dispersed biotically (by pollinating animals) or abiotically (by wind or water), and makes contact with the stigma. The pollen–stigma interaction (Fig. 2) comprises those stages of the pollen–pistil interaction (see above) from pollen capture to passage of the growing pollen tube through the transmitting tissue of the stigma and entry into the style. These events have been particularly well studied at a cellular and molecular level in species with SI systems that trigger early arrest of incompatible pollen on the stigma, either before pollen germination or just after germination, where studies have often consisted of detailed comparisons of compatible versus incompatible pollinations. These models principally comprise Brassica species and other Brassicaceae (Dickinson & Lewis, 1973, 1975; Elleman & Dickinson, 1986, 1990; Dickinson et al., 1997), various grass species (Poaceae) (Heslop-Harrison, 1979a, 1979b, 1982; Heslop-Harrison & Heslop-Harrison, 1981) and poppy (Papaver rhoeas, Papaveraceae) (Elleman et al., 1992). Interestingly, despite having an SI recognition/rejection system active on the stigma, these species do not share the same genetic mechanism of SI. In the Brassicaceae, SI is of the sporophytic type (Bateman, 1955), whereas in the grasses and poppy, SI is of the gametophytic type (Lundquist, 1956; Hayman, 1956; Lawrence, 1975; see section VI). Despite these differences in SI, all three families share the same basic type of stigma, the so-called dry stigma type (Heslop-Harrison & Shivanna, 1977), endorsing the opinion that stigma structure and microecology are the main factors influencing the nature of events associated with any given pollen–stigma interaction, irrespective of the type of SI system (Heslop-Harrison, 2000; Hiscock, 2004; Edlund et al., 2004).

Figure 2.

Different stages of the pollen–stigma interaction. The diagram represents a typical stigma of the dry papillate type found in species from the Brassicaceae. Pollen is shown at various stages of development on the stigma and growing into the transmitting tissue of the style.

Angiosperm stigmas can be classified into two broad categories, wet and dry, depending on whether or not they possess a surface secretion (Heslop-Harrison & Shivanna, 1977; Heslop-Harrison, 1981). Although there are examples of intermediate, semi-dry stigmas (e.g. in the Asteraceae; Hiscock et al., 2002a), this broad classification into wet and dry types is useful in the context of this review because studies of pollen–stigma interactions in model species have revealed important general differences consistent with this fundamental difference in stigma structure and microecology (Heslop-Harrison, 2000; Johnson & Preuss, 2002; Lord, 2003; Edlund et al., 2004). For instance, in all wet-stigma species investigated to date, pollen capture by the stigmatic secretion is nonspecific, and hydration of pollen within the secretion appears to be passive and largely unregulated (Swanson et al., 2004). By contrast, studies in dry-stigma species, most notably from the Brassicaceae (e.g. Arabidopsis), indicate that pollen capture and adhesion to the stigma exhibit a degree of species specificity (Luu et al., 1997a; Zinkl et al., 1999; Heizmann et al., 2000a; Zinkl & Preuss, 2000), and that hydration of pollen on the stigma is a highly regulated process (Heslop-Harrison, 1979a; Dumas et al., 1984; Dickinson, 1995; Dickinson & Elleman, 1995). Epidermal cells of wet stigmas often lack a continuous cuticle, so penetration of the stigma by pollen tubes is relatively unimpeded compared with the same event in dry-stigma species, where pollen tubes secrete hydrolytic enzymes such as cutinase to breach a continuous cuticle during stigma penetration (Hiscock et al., 1994, 2002b).

Important broad correlations between stigma type and pollen morphology also exist, most notably the general occurrence of trinucleate pollen in species with dry stigmas and binucleate pollen in species with wet stigmas (Heslop-Harrison & Shivanna, 1977), implying a strict coevolution of pollen and stigma structures (Edlund et al., 2004). It is therefore appropriate to discuss events associated with pollen–stigma interactions in species with wet stigmas and those with dry stigmas separately, before attempting to draw together commonalities between the two. Because of the inherent diversity of molecules mediating sexual processes, drawing together a consensus of data from taxonomically very different study species, even within the same stigma category, is difficult (Lord, 2003; Edlund et al., 2004). Among wet-stigma species, studies of the pollen–stigma interaction have focused principally on species of Nicotiana and Petunia (Solanaceae) and Lilium (Liliaceae), whereas for dry-stigma species most studies have been conducted on species in the Brassicaceae, notably Brassica, Raphanus and Arabidopsis, and cereal crop species (Poaceae), most notably Secale (Fig. 1).

III. Pollen–stigma interactions in species with wet stigmas

Secretions on wet stigmas can be primarily lipidic (e.g. Solanaceae), containing complex mixtures of long-chain saturated and unsaturated triacylglycerides (Dumas, 1977; Wolters-Arts et al., 1998), or primarily aqueous and carbohydrate-rich (e.g. Liliaceae) (Dumas et al., 1984; Sanchez et al., 2004). Both types of secretion also contain a wide range of proteins. Few detailed studies of stigma development have been conducted, but available data indicate that deposition of secretions on stigmas occurs late in development and involves the release of fluids from internal reservoirs (Heslop-Harrison, 1976, 1981; Herrero & Dickinson, 1979) and intense vesicular activity at the stigma epidermis (Heslop-Harrison, 1976). The emergence of secretions on the stigma surface is usually associated with the breakdown or disruption of the cuticle, an event that can also be facilitated by insect activity on the stigma (Heslop-Harrison, 2000). When stigmas are mature and receptive to pollination, surface enzymes, notably peroxidase and esterase, show their highest activity levels (Dafni & Motte Maues, 1998; Fig. 3). The function of these ubiquitous stigma enzymes is not known, although roles have been suggested in signalling and recognition, and in defence (Knox et al., 1976; McInnis et al., 2005, 2006a).

Figure 3.

The dry stigma of Arabidopsis thaliana and the wet stigma of Petunia hybrida, unstained and stained for the presence of peroxidase. (a) A. thaliana unstained stigma (arrow); (b) stigma stained with 0.1 m guaiacol, 0.1 m H2O2 in 20 mm phosphate buffer pH 4.5 to visualize peroxidase activity; bar, 1 mm. (c,d) as (a,b), respectively, but for stigmas of P. hybrida; bar, 1 mm. (Figure adapted from McInnis et al., 2006a.)

Stigmatic secretions play a key role in pollen capture and adhesion (Dumas & Gaude, 1983; Dumas et al., 1984). On making contact with a wet stigma, pollen (from almost any species) is quickly trapped by surface tension and immersed within it. This first phase of the pollen–stigma interaction appears to be a passive and indiscriminate process (Dumas et al., 1984; Swanson et al., 2004). Interestingly, spores from fungi, and also bacteria that become trapped in stigmatic secretions, rarely if ever germinate and infect the stigma, suggesting that stigmatic secretions possess an effective pathogen defence system (Atkinson et al., 1993; Heslop-Harrison, 2000). Stigmas, both wet and dry, express a wide range of pathogenesis-related genes (Curtis et al., 1997; Kuboyama, 1998; Tung et al., 2005; Swanson et al., 2005) that presumably contribute to this stigmatic defence system. The recent discovery that high levels of reactive oxygen species (ROS), most notably H2O2, accumulate in stigmas (McInnis et al., 2006b) suggests a potential role for these molecules in defence (section VII.3). Such a role has also been proposed for the superoxides and other ROS that accumulate at high levels in nectar (Carter & Thornburg, 2000, 2004). The duration of stigma receptivity for pollination appears to be extremely variable between species, and can vary from a few hours to several days (Heslop-Harrison, 2000; Yi et al., 2006). Astonishingly, nothing is known about what controls the duration of stigma receptivity, despite its profound importance for pollination and the reproductive output of crop plants.

In species with wet stigmas, pollen hydration, like pollen adhesion, is thought to be largely passive and unregulated, because water is readily available within surface secretions (Swanson et al., 2004). However, because of the technical difficulties associated with studying pollen hydration, germination, and tube growth in vivo (on wet and dry stigmas), our knowledge of these processes has come largely from in vitro studies using assorted pollen germination/growth media (Heslop-Harrison, 1987; Heslop-Harrison & Heslop-Harrison, 1992). These studies have yielded tremendous insights into the cellular and molecular processes associated with the mechanism and regulation of pollen tube growth. As this subject has been reviewed so extensively elsewhere (Hepler et al., 2001; Holdaway-Clarke & Hepler, 2003; Feijo et al., 2004; Malho et al., 2006; Wilsen & Hepler, 2007), we present here just a broad overview of the processes of pollen germination and pollen tube growth, which are applicable equally to the pollen of species with wet stigmas and those with dry stigmas.

Pollen hydration is accompanied by an activation of metabolic processes and renewed protein synthesis utilizing largely preformed mRNAs (Mascarenhas, 1989, 1993). In turn, this is accompanied by reorganization of the cytoskeleton and cytoplasm to polarize the vegetative cell in preparation for its extension as a tip-growing pollen tube (Heslop-Harrison, 1987; Steer & Steer, 1989). During the process of polarization, actin filaments are focused towards the nascent pollen tube tip, and the vegetative cell nucleus is oriented into a position allowing it to enter the pollen tube ahead of the generative cell (Heslop-Harrison & Heslop-Harrison, 1989; Raudaskoski et al., 2001; Hepler et al., 2001; Lalanne & Twell, 2002).

Polarization and subsequent germination and tip growth of the pollen tube are largely mediated through a tip-focused gradient of calcium (Ca2+) ions (Holdaway-Clarke & Hepler, 2003). This tip-focused Ca2+ gradient is central to the signalling network that coordinates pollen tube growth, regulating both the speed and direction of tip growth (Malho & Trewavas, 1996; Camacho & Malho, 2003). Other ions, particularly protons (H+), potassium (K+) and chloride (Cl), have also been implicated in the regulation of pollen tube growth because, like Ca2+, their concentration gradients within pollen tubes undergo periodic oscillations during pollen tube extension, a pattern that is reflected in similar oscillations in the rate of pollen tube elongation (Holdaway-Clarke & Hepler, 2003; Feijo et al., 2004; Malho et al., 2006; Wilsen & Hepler, 2007).

The actin cytoskeleton is essential for the maintenance of polarized pollen tube growth, acting both to deliver the vesicles required for tip growth and to generate polymerized actin microfilaments, necessary for elongation (Staiger, 2000; Vidali & Hepler, 2001; Raudaskoski et al., 2001). The polymerization of new actin microfilaments is thought to be responsible for the characteristic polarization and zonation of cytoplasm at the apex of the pollen tube, where a ‘clear zone’ is created by the exclusion of large organelles (Vidali & Hepler, 2001; Cárdenas et al., 2008). Polymerization of new actin microfilaments is regulated by a number of actin-binding proteins (ABPs) which, in turn, respond to signal cues such as elevated pH and Ca2+ levels (Ren & Xiang, 2007). Numerous approaches to studying the structure of pollen tubes have revealed highly conserved actin organization. This consistent structural organization has been used to define the three main regions of the pollen tube: the shank, subapex and apex (Ren & Xiang, 2007). In the shank region, actin is organized longitudinally into bundles, which act as tracks for cytoplasmic streaming of vesicles towards the apex (Lancelle & Hepler, 1992; Vidali & Hepler, 2001). In the subapical region, fine actin (F-actin) is organized into a ‘collar’, while in the apex F-actin exists as a dynamic network of short filaments (Miller et al., 1996; Kost et al., 1998; Vidali & Hepler, 2001). Actin polymerization occurring in the subapex appears to be responsible for the ‘reverse-fountain’ cytoplasmic streaming characteristic of actively elongating pollen tubes (Cárdenas et al., 2005). The F-actin filaments at the extreme apex of the pollen tube are thought to play a role in vesicle docking and fusion (Fu et al., 2001). Studies of actin organization and activity in the apical region have indicated that actin may be involved in initiating pollen tube growth, and consequent actin polymerization in modulating oscillations of pollen tube growth and Ca2+ levels (Fu et al., 2001; Cárdenas et al., 2005, 2008; Lovy-Wheeler et al., 2007).

Tip growth appears to be coordinated through the activity of the small GTP-binding protein ROP1 (rho of plants1) which, when bound to RIC3 and RIC4 (ROP-interacting CRIB motif-containing proteins 3 and 4), localizes to the pollen tube tip and, through interactions with the actin cytoskeleton and the tip-focused Ca2+ gradient, acts as a target for delivery of secretory vesicles carrying wall components to the extending tip (Fu et al., 2001; Wu et al., 2001; Gu et al., 2003, 2005, 2006). Interactions between ROP1, Ca2+ and the actin cytoskeleton are potentially mediated through interactions with ABPs (Ren & Xiang, 2007).

Many recent studies have sought to characterize ABPs in growing pollen tubes to determine their role in the organization and dynamics of the actin cytoskeleton during tip growth (Raudaskoski et al., 2001; Hussey et al., 2006). ABPs respond to internal and external stimuli, serving as a link between signal-transduction pathways and dynamic actin changes, to determine cellular architecture (Staiger & Blanchoin, 2006). Over 70 classes of ABP are known to exist in eukaryotic cells, and many have now been identified in pollen tubes, including: myosin, profilin, ADF/cofilin, villin, Arp2/3 complex, gelosin/fragmin, formin and caldeson (reviewed by Hussey et al., 2006; Ren & Xiang, 2007). These ADPs are likely to be the targets for key signalling pathways that regulate tip growth. In addition to ionic signals, particularly Ca2+, and Rop GTPases, numerous other cell-signalling molecules have been identified in pollen, which may act as modulators of pollen tube tip growth through interactions with ADPs. These include cAMP, phosphoinositides, calmodulin, protein kinases, GABA and nitric oxide, all of which have either been demonstrated to, or proposed to, interact with the tip-focused Ca2+ gradient and/or the cytoskeleton in their modulation of tip growth (for reviews see Feijo et al., 2004; Malho et al., 2006).

From these myriad in vitro studies of pollen tube growth and development, a detailed picture is emerging of a complex signalling network regulating pollen tube growth, which has striking similarity to the analogous signalling network regulating tip growth in root hairs (Samaj et al., 2006). Both these networks appear fairly conserved among the different model plants investigated, indicating consensus in the basic internal cellular machinery regulating tip growth. However, unlike the in vitro environments used to assay root hair development, the assay environments used to investigate pollen tube growth in vitro are far removed from the in vivo environments they normally encounter. It is therefore essential to attempt to interpret in vitro data on pollen tube germination and growth in the context of the pollen tube's interaction with the physical and chemical environment of the stigma and style and, where possible, in the future to attempt similar pollen tube growth assays in vivo.

In Nicotiana, the importance of the stigmatic secretion for pollen development and pollen tube growth into the stigma was demonstrated by tissue-specific ablation studies that eliminated the stigma epidermis and its secretion (Goldman et al., 1994). These ‘stigmaless’ plants were female sterile, as their blunt, dry gynoecia failed to support compatible pollen development. However, application of freshly extracted stigma secretion from wild-type plants to these gynoecia restored fertility, showing that components of the Nicotiana stigma secretion, rather than the stigma cells themselves, were necessary and sufficient for a ‘normal’ pollen–stigma interaction (Wolters-Arts et al., 1998, 2002). Interestingly, extracted secretions from Petunia, but not lily stigmas, were also able to substitute for Nicotiana stigma secretions in restoring fertility to stigmaless plants (Wolters-Arts et al., 1998), indicating that the similar lipidic nature of the Nicotiana and Petunia (both Solanaceae) stigma secretions is crucial in resurrecting gynoecium fertility. Lipids, specifically cis-unsaturated triacylglycerides, were subsequently confirmed to be the essential component of the stigmatic secretion necessary for restoring stigma fertility (Wolters-Arts et al., 1998, 2002). Lipids are therefore necessary for correct pollen tube growth on the stigma and entry into the transmitting tissue, but it was further proposed that the lipids might be providing a directional cue to the developing pollen tubes by controlling the flow of water to the pollen (Wolters-Arts et al., 1998; Lush et al., 2000). This gradient of increasing water potential would act as the initial cue for polarization of the pollen cytoplasm to direct germination of a pollen tube towards the stigma. Elegant in vitro simulations of the stigmatic environment using oil-in-water emulsions confirmed that Nicotiana pollen tubes do indeed germinate and grow down gradients of water potential towards an aqueous phase (Lush et al., 1998, 2000). In these assays, both speed of hydration and speed of germination were proportional to the proximity of pollen grains to the aqueous phase and, following germination, emergent pollen tubes grew directly towards the aqueous phase (Lush et al., 1998, 2000). Thus it appears that the initial cue for pollen tubes to grow towards and into the stigma is simply a gradient of water potential, rather than any specific protein or other chemical cues generated by the pistil. How the pollen perceives this water ‘signal’ remains to be determined.

In lily, the other model wet-stigma species, the stigma secretion is aqueous and carbohydrate-rich (Dumas et al., 1984), but there is no evidence that water gradients direct pollen tube growth on the stigma. Instead, two peptides, chemocyanin (a member of the plantacyanin protein family of unknown function) and stigma/stylar cysteine-rich adhesin (SCA) present on the stigma appear to act in concert as chemotropic attractants directing growing lily pollen tubes towards and into the hollow style (Kim et al., 2003; Park & Lord, 2003). In vitro assays of pollen tube growth showed that lily pollen tubes navigate towards lily stigma extracts. The chemotactic substance responsible for this in vitro pollen tube attraction was purified and identified as chemocyanin, a blue copper protein of the plantacyanin family (Kim et al., 2003). Interestingly, the attractive effect of chemocyanin was enhanced by SCA, even though SCA by itself was not chemotactic (Kim et al., 2003).

Wet-stigma species usually lack a continuous stigmatic cuticle, so once the pollen tube has germinated and started to grow towards the stigma, its entry and passage into the stigma is largely unimpeded, and in the case of hollow-styled species such as lily, is not impeded at all. In such hollow-styled species, stigmatic secretions are continuous with secretions lining the style, and pollen tubes grow within these secretions in close contact with stylar epidermal cells, following what has been described as a ‘facilitating mechanical pathway’ (Dumas et al., 1984; Gaude & Dumas, 1987). In species of Solanaceae, which have solid stigmas and styles, pollen tubes grow into the stigma through the intercellular spaces (extracellular matrix, ECM) between the cells of the secretory zone (de Nettancourt et al., 1974; Heslop-Harrison, 1976; Herrero & Dickinson, 1979, 1981). Once inside the stigma, pollen tubes continue to grow within the ECM of the transmitting tissue, which is continuous with the style. Enzymes probably facilitate entry of the pollen tube into the stigma by loosening the ECM and/or modifying the pollen-tube wall as it extends by tip growth. A number of pollen- and stigma-derived enzymes have been implicated in this process in both wet- and dry-stigma species, including pectin esterases (Mu et al., 1994; Jiang et al., 2005; Tian et al., 2006), polygalacturonases (Kim et al., 2006a), and glucanases (Kotake et al., 2000). Expansins have also been implicated in these early penetration events, as Nicotiana exudate was shown to have cell wall-loosening activity (Nieuwland, 2004; Nieuwland et al., 2005). However, the β-expansin-like protein pistil pollen allergen-like (PPAL), present in Nicotiana stigma secretions, did not show cell wall-loosening activity in an in vitro expansin assay (Sanchez et al., 2004). Nevertheless, using the same expansin assay system Nieuwland et al. (2005) then showed that a lipid transfer protein (LTP), the most abundant protein present in the tobacco stigma exudate, was responsible for the majority of the cell wall-loosening capacity of the exudate. While it is not clear how the Nicotiana stigma LTP functions as an expansin, the highly diverse nature of this class of protein suggests that LTPs do not necessarily function exclusively in lipid metabolism. As well as functioning as a chemotropic signal, the LTP SCA also functions as an adhesin in the lily style (see below), and some nonspecific LTPs have been implicated in defence against pathogens (Garcia-Olmedo et al., 1995), a function entirely consistent with a location in stigma secretions.

Entry into and growth within the stigma must be a tightly regulated process, and involve an almost continuous exchange of signals between the pollen tube and cells of the stigma and transmitting tissue. Nicotiana and other members of the Solanaceae possess solid stigmas and styles with a distinct tract of secretory transmitting tissue running through them, and it is through this tissue that pollen tubes grow on their journey to the ovary (Heslop-Harrison, 1976; Herrero & Dickinson, 1979, 1981; Fig. 2). In lily, however, the stigma and style are hollow, and pollen tubes grow over the surface of the epidermal cells lining this hollow passageway en route to the ovary. It is possible that this fundamental structural and microecological difference between the pistils of Solanaceae and lily translates into correspondingly different types of cellular and molecular interaction between their pollen tubes and pistil cells. Supporting this assumption, molecules identified to date that appear to regulate pollen tube growth in vivo in species of Solanaceae and lily are quite different (Lord, 2003; Sanchez et al., 2004).

Studies in tomato (Lycopersicon, Solanaceae) have highlighted receptor kinases and their ligands as potential mediators of pollen tube growth and pollen–stigma communication (Muschietti et al., 1998; Kaothien et al., 2005). LePRK1, LePRK2 and LePRK3 (Lycopersicon esculentum protein receptor kinases 1–3) are members of the leucine-rich repeat class of receptor kinase, and are localized to growing pollen tubes (Muschietti et al., 1998). Yeast two-hybrid screens for potential LePRK-interacting proteins expressed in pollen yielded numerous potential ligands, including small cysteine-rich proteins (molecular mass 6–16 kDa), one of which was the previously characterized LAT52 (LATE ANTHER TOMATO 52; Twell et al., 1989), which interacted specifically with LePRK2. In Lycopersicon, LAT52 is essential for pollen hydration and germination in vitro and for normal pollen tube growth in vitro, suggesting that its interaction with LePRK2 may be important for pollen development on the stigma in vivo. One possibility is that LePRK2 and LAT52 constitute an autocrine-signalling system regulating the initiation and maintenance of pollen tube growth (Tang et al., 2002). Subsequent studies showed that LePRK1 and LePRK2 form a complex in pollen and when expressed together in yeast, and that this complex could be disrupted in vitro by the addition of stylar extracts (Wengier et al., 2003). The active fraction of these stylar extracts (molecular weight range 3–10 kDa) also promoted dephosphorylation of LePRK2 (Wengier et al., 2003). Further analysis of the pool of LePRK interactors identified in yeast two-hybrid screens identified LeSTIG1 as a potential candidate ligand in the active stylar fraction (Tang et al., 2004). Like LAT52, LeSTIG1 is a small, cysteine-rich protein, but unlike LAT52, LeSTIG1 interacts with both LePRK1 and LePRK2 in vitro and is able to displace LAT52 from its interaction with LePRK2, suggesting that LePRK1 and LePRK2 interact with different ligands at different stages of pollen tube growth (Tang et al., 2004). Furthermore, in vitro bioassays showed that LeSTIG1 could promote tomato pollen tube growth in vitro (Tang et al., 2004), strengthening the probability that its in vitro interactions with LePRK1 and LePRK2 reflect an important signalling role for these proteins in pollen–pistil interactions. One suggestion is that LeSTIG1-induced dissociation of LePRK2 and LAT52 and the concomitant formation of the LeSTIG−LePRK2 complex acts as a checkpoint for pollen tube growth (Johnson & Preuss, 2003) during the pollen–stigma interaction, perhaps as a means of sensing a pollen tube's entry into the intercellular environment of the stigma.

In Nicotiana tabacum, there is evidence that a pistil-specific arabinogalactan protein (AGP) termed transmitting tissue-specific (TTS) AGP acts as a signal directing pollen tube growth towards the ovary (Wang et al., 1993). In vitro assays showed that TTS, which binds to the pollen cell wall and can enter growing pollen tubes, stimulates pollen tube growth and attracts growing pollen tubes (Wu et al., 1995). Meanwhile, evidence that TTS is required for pollen tube growth in vivo came from transgenic (antisense-suppression) assays in which reduced levels of TTS in the pistil were correlated with reduced rates of pollen tube growth in the style and reduced seed set (Cheung et al., 1995). Equating these observations with a role for TTS in pollen tube guidance in vivo has proved problematic, because TTS is found throughout the transmitting tissue. Nevertheless, there is a gradient in the level of TTS glycosylation down the pistil, with the highest levels of glycosylation detected at the ovarian end, reflected by a corresponding gradient of increasing acidity down the pistil (Wu et al., 1995). It is proposed that pollen tubes grow through the transmitting tissue of the pistil following this gradient of increasing TTS glycosylation/acidity and perhaps utilizing the sugar moieties of TTS as a source of nutrients on the way, because pollen tubes grown in vitro were able to deglycosylate TTS (Wu et al., 1995). In turn, if pollen tubes were able to deglycosylate TTS in vivo, this would enhance the TTS gradient differential in the vicinity of the growing pollen tube.

Nevertheless, the proposed role of TTS in pollen tube guidance has proved contentious. This is because estimates of the theoretical maximum distance over which chemical cues could guide pollen tubes (approx. 1.2–9.3 mm) does not equate to the approx. 50 mm length of the Nicotiana pistil (Lush, 1999; Lush et al., 2000). Also a homologue of TTS, galactose-rich style glycoprotein (GaRSGP) from Nicotiana alata, which is 97% identical to TTS at the amino acid level, did not attract growing pollen tubes in an in vitro assay (Sommer-Knudson et al., 1998). However, it was shown subsequently that GaRSGP is not the true orthologue of TTS when NaTTS was identified (99.9% identical to TTS) and demonstrated to attract growing pollen tubes of N. alata in in vitro bioassays (Wu et al., 2000). So far, however, TTS homologues have not been identified outside the genus Nicotiana, despite AGPs being a ubiquitous component of angiosperm pistils. This suggests that the TTS guidance system may be specific to the genus Nicotiana. Two further Nicotiana AGPs, the 120-kDa glycoprotein from N. alata (Lind et al., 1996, and pistil-specific extensin-like protein III (PELPIII) from N. tabacum (de Graaf et al., 2003), are like TTS in that they accumulate in the transmitting tissue ECM. However, neither of these AGPs establishes a glycosylation gradient in the style, is deglycosylated by pollen tubes, or attracts pollen tubes in in vitro assays (Sanchez et al., 2004). Interestingly, recent studies have suggested that TTS, the 120-kDa glycoprotein and PELPIII might be involved in the S-RNase-mediated GSI response of SI Nicotiana species (see section VI), as they associate as protein complexes with S-RNase in vitro (Cruz-Garcia et al., 2003). It is suggested that this protein complex may facilitate uptake of S-RNases by pollen tubes. Nevertheless, the role of GaRSGP, the 120-kDa glycoprotein and PELPIII in compatible pollen−pistil events in species without SI (if any) is not known. One possible role for these proteins, along with other AGPs, could be simply to act as lubricants or nutrients to create an environment favourable for pollen tube growth within the ECM (Sanchez et al., 2004).

Lily has a hollow style with an opening in the centre of the stigma, and pollen tubes grow towards this opening through the stigmatic secretion. As discussed above, SCA, a small cysteine-rich LTP, and chemocyanin have been implicated in guidance of pollen tubes towards the entrance to the stylar canal (Park et al., 2000). While chemocyanin acts as a chemotactic guidance cue, SCA functions as a haptotactic guidance cue (Park et al., 2000). SCA, which exists in three isoforms (Chae et al., 2007), is abundant in the stigma and transmitting tract, where it exists either free or bound to pectin. The free form of SCA, perhaps acting with chemocyanin, is thought to form a concentration gradient on the stigma focused on the entrance to the stylar canal, which is responsible for attracting pollen tubes (Lord, 2003). In the style, SCA is largely pectin-bound, where it has been proposed to facilitate adhesion-mediated haptotactic guidance of pollen tubes growing on the surface of the transmitting tract epidermis towards the ovary (Lord, 2003). In support of this model, Mollet et al. (2000) showed that SCA bound to a pectic polysaccharide from the lily style could adhere to pollen tubes in an in vitro assay. Recent studies suggest a more complex mechanism for SCA-mediated pollen tube adhesion because, rather than functioning as an extracellular pollen ‘glue’ in the presence of pectin, SCA is taken up by growing pollen tubes in vitro and in vivo (Kim et al., 2006b), where it is presumed to exert its effects internally within the pollen cytosol, perhaps as a signalling molecule (ligand) that activates an as yet uncharacterized adhesion pathway via its cognate receptor (Ravindran et al., 2005; Kim et al., 2006b). Interestingly, endocytosis of SCA (presumably bound to a SCA receptor) by pollen tubes is enhanced by free ubiquitin, which also enhances pollen adhesion when combined with SCA/pectin mixtures in the in vitro adhesion assay (Kim et al., 2006b). How SCA mediates pollen tube adhesion from within a pollen tube, and how extracellular ubiquitin participates in the adhesion process, remain to be determined.

This brief survey of molecules and molecular interactions implicated in pollen–stigma interactions in species with wet stigmas highlights the fact that, as yet, few commonalities have been identified between the Solanaceae and lily systems. This is probably a consequence of their evolutionary divergence and their fundamental differences in stigma morphology and associated mechanisms of pollen–stigma interactions. Alternatively, more concerted efforts may be needed to identify the specific homologues of the Solanaceae proteins in lily, and vice versa. More molecular studies are needed in these and other species with wet stigmas, bridging the monocot−eudicot divide to determine how conserved these two systems are among angiosperms.

IV. Pollen–stigma interactions in species with dry stigmas

For nearly 30 yr, Brassica has been the model for studies of pollen–stigma interactions on the dry stigma. Studies of the ‘default’ (compatible) condition of pollen–stigma interactions in Brassica have frequently been combined with studies of sporophytic self-incompatibility (SSI), because incompatible pollen is recognized and rejected at the stigma surface (Dickinson & Lewis, 1973; Heslop-Harrison, 1975; Roberts et al., 1980). More recently, studies of compatibility have shifted to the model brassica Arabidopsis thaliana and its SI relatives, A. lyrata and A. petraea, because of their more highly tractable functional genomics made possible by the availability of a fully sequenced A. thaliana genome.

The stigma of Brassica and Arabidopsis consists of a dome of epidermal papillae covered by a continuous cuticle. This cuticle protects the stigma from pathogens and forms a major barrier to pollen tube penetration, although it appears to be freely permeable to water and larger molecules such as proteins as a consequence of its globular structure (Roberts et al., 1984a; Elleman et al., 1988). Permeability of the stigma cuticle to water makes it unlike the cuticle of other plant epidermal cells, which, by their nature, must insulate the plant against water loss. Mutations that perturb cuticle development often have a profound effect on its ability, or not, to support pollen development. This is exemplified by the fiddlehead mutant, which has unusually permeable cuticles that allow pollen grains to hydrate and germinate on nonstigmatic surfaces, including leaves (Lolle & Cheung, 1993). FIDDLEHEAD encodes a β-ketoacyl CoA synthase essential for the synthesis of long-chain lipids (Pruitt et al., 2000). Whether the permeability of the stigma cuticle to water is modulated by pollen signals, or whether it is constitutively permeable, have yet to be fully resolved, but a large body of evidence points to the former. For instance, the widespread ‘mentor’ pollen effect, whereby compatible pollen can induce hydration and germination of normally incompatible pollen in mixed pollen populations (Knox et al., 1987, Preuss et al., 1993, Hulskamp et al., 1995), is often cited as evidence for pollen's ability to affect cuticle permeability. However, this is not consistent with the results of micromanipulation experiments in which compatible and incompatible pollen grains of Brassica oleracea were placed next to each other on a single stigmatic papilla: while compatible grains hydrated, incompatible grains hydrated only partially, if at all (Dickinson, 1995; Dickinson et al., 1997). It is likely that these apparently contrasting observations reflect density dependence in the mentor effect because conspecific pollen-germination rates have been shown to increase with increasing pollen density on the stigma − the ‘pollen population effect’. This effect is associated with the pollen secreting a small (five amino acid) peptide, phytosulphokine, which in purified form can induce elevated levels of pollen germination in low-density pollen populations (Chen et al., 2000).

On the cuticle sits a thin, membrane-like layer of protein, the pellicle, which acts as the first site of molecular contact between the stigma and alighting pollen grains (Mattsson et al., 1974; Heslop-Harrison, 1978; Dumas et al., 1984; Fig. 4). An intact pellicle appears to be essential for compatible pollen–stigma interactions: disruption or removal of the pellicle with dilute detergents or proteases prevents pollen tube entry into the stigma (Heslop-Harrison, 1975, 1977; Heslop-Harrison & Heslop-Harrison, 1975; Heslop-Harrison et al., 1975; Knox et al., 1976; Shivanna et al., 1978; Stead et al., 1980; Roberts et al., 1984b; Hiscock & Dickinson, 1993). Following such treatments, however, stigmas regain their fertility within a few hours (Stead et al., 1980; Roberts et al., 1984b), presumably as a consequence of the deposition of a new pellicle. Despite its importance in the pollen–stigma interaction, the protein composition of the pellicle is poorly understood. Cytochemical studies have shown that esterases and glycoproteins are the major components of the pellicle of Brassica, Malva, Hibiscus, Silene, Gladiolus and ryegrass (Secale) (Heslop-Harrison et al., 1975; Heslop-Harrison, 1978; Knox et al., 1976; Hiscock, 1993), but the function of these proteins in pollen–stigma interactions is unknown.

Figure 4.

The dry stigma of Brassica oleracea. (a) Stigmatic papillae stained with α-naphthyl actetate and fast blue RR salt to visualize nonspecific esterase activity; bar, 10 µm. (b) TEM of oblique section through cell wall of stigmatic papilla stained with α-naphthyl actetate and hexazatized pararosanalin to visualize esterase activity of the pellicle (P); bar, 0.2 µm. (c) The pollen foot. TEM of oblique section through pollen–stigma interface 1 h after compatible pollination: Po, pollen grain; Ft, pollen foot; S, stigmatic papilla; bar, 2 µm. (d) Brassica pollen tubes unable to penetrate the stigma following treatment of stigma with dilute pronase solution. Pollen tubes were visualized with aniline blue; bar, 20 µm.

In the absence of a stigmatic secretion, initial adhesion of pollen to the dry stigma appears to be largely a function of the pollen wall. Structural studies in Brassica implicated the lipid-rich pollen coating (tryphine) in adhesion to the dry stigma. This was largely because, immediately after making contact with the stigma, pollen coating is extruded from the bacculae of the wall exine onto the surface of the stigma, where it forms an appresoria-like ‘foot’ that appears to glue the pollen to the stigma (Elleman & Dickinson, 1986; Fig. 4). Consistent with this conclusion is the observed change in the physical and chemical properties of the pollen coating (coat conversion) that forms the foot, particularly its insolubility in cyclohexane (Elleman et al., 1987, 1989).

Later studies of pollen adhesion using the Arabidopsis ECERIFERUM (CER) mutant cer6-2, which fails to produce a pollen coating, revealed that adhesion of this ‘coatless’ pollen to the stigma was not significantly affected (Zinkl et al., 1999; Zinkl & Preuss, 2000). This somewhat unexpected observation indicated that the sporopollenin exine must also have adhesive properties. Indeed Arabidopsis less adherent pollen (lap) mutants that show aberrant patterns of exine development were also compromised in pollen adhesion (Zinkl & Preuss, 2000). Furthermore, purified Arabidopsis exine fragments were able to bind to the surface of the stigma (Zinkl et al., 1999). The adhesive function of the exine could be a function of the exinic outer layer (EOL) (Gaude & Dumas, 1984), a thin surface layer surrounding the exine, probably proteinaceous in composition but which, like the stigmatic pellicle, has yet to be characterized. Transmission electron microscopy (TEM) studies in Brassica, where esterase activity was localized at points of contact between pollen and stigma, suggest that on contact, the pollen EOL fuses with the stigmatic pellicle (Hiscock, 1993). Both the pellicle and EOL stain positively for esterase activity, a property that allows these very thin layers to be visualized more precisely than with conventional TEM staining, and at the point of contact between the pollen exine and the wall of a stigmatic papilla, just a single esterase reaction layer is present (Hiscock, 1993). This raises the intriguing possibility that pollen and stigma esterases may form a ‘recognition complex’, as first suggested by Knox et al. (1976) based on studies of early pollination events in self-compatible Gladiolus. According to the model of Knox et al. (1976), interaction between surface esterases of pollen and stigma is a species-specific recognition event required for activation of a ‘cutinase complex’ (presumably made up of different pollen- and stigma-derived subunits) that digests the cuticle, allowing entry of the pollen tube into the stigma. An appealing feature of this model is that it can explain why removal of the stigmatic pellicle prevents pollen tube penetration (Fig. 4). However, in light of the studies by Zinkl et al. (1999), it is perhaps more likely that a pellicle−EOL fusion complex is involved in adhesion, and then indirectly in pollen tube penetration. It would be interesting to test this hypothesis using assays of pollen adhesion on stigmas with their pellicles removed.

Two stigma-localized members of the Brassica S-(self-incompatibility) gene family (see section VI) have been implicated in pollen adhesion: S-locus-related-1 (SLR1) and S-locus glycoprotein (SLG). SLR1 is not linked to the S-locus, but has high sequence similarity to SLG, which is located at the S-locus. Gene knockouts of SLR1 resulted in reduced adhesion of pollen on transgenic stigmas at late, but not early time points after pollination, consistent with the pollen exine (EOL) mediating the earliest events in pollen adhesion (Luu et al., 1997a, 1997b). Pretreatment of stigmas with antisera to SLR1 or SLG also resulted in reduced pollen adhesion, indicating that both SLR1 and SLG function in pollen–stigma adhesion processes (Luu et al., 1999; Heizmann et al., 2000b). Ascribing a definitive role for SLG in pollen adhesion is, however, difficult because some Brassica S-haplotypes do not possess SLG (Suzuki et al., 2000) and experience no apparent reduction in their ability to adhere pollen. SLG may therefore have multiple functions in pollen–stigma interactions, one of which is an indirect role in adhesion. Importantly, SLR1 and SLG each bind specific members of a group of pollen coat proteins (PCPs) (Doughty et al., 1993; (Hiscock et al., 1995b). SLR1 binds to SLR1-BP (SLR1-binding protein; Takayama et al., 2000) and SLG binds to PCP-A1 (Doughty et al., 1998). These interactions are suggested to be important for strengthening adhesive bonds between pollen and stigma after the initial pollen exine-mediated binding (Heizmann et al., 2000b; Swanson et al., 2004). For this reason, Swanson et al. (2004) have described adhesion involving SLR1 and SLG and their respective pollen ligands as ‘cross-linking adhesion’. This seems a logical distinction to make, because changes in the physical and chemical properties of the pollen coating appear to be associated with the formation of the pollen foot. This phenomenon, described as ‘coat conversion’ (Elleman & Dickinson, 1986), is characterized by a reorganization of components of the pollen coating, and a change in its appearance, from electron-translucent to electron-opaque. The pollen coating forming the foot then loses its solubility in cyclohexane; treating pollinated stigmas with cyclohexane removes the pollen coating from all regions of the pollen exine except the foot, which remains in its converted state, attaching the pollen grain to the stigma (Elleman et al., 1987, 1989). Although the biochemical changes associated with coat conversion and foot adhesion are unknown, it is tempting to speculate that interaction between SLR1 and SLR1-BP, perhaps in association with SLG-PCP-A1 interaction, are involved.

Notwithstanding its role in pollen adhesion, the principal role of the pollen coating in species with dry stigmas appears to be in the mediation and regulation of pollen hydration. In contrast to species with wet stigmas, water uptake by the variously dehydrated pollen is not passive; instead, pollen hydration is a highly regulated process (Dumas & Gaude, 1983; Dickinson, 1995). There is also strong evidence to suggest that incompatibility and incongruity phenomena can manifest themselves in perturbed rates, or even complete failure, of hydration (Roberts et al., 1980; Dumas & Gaude, 1983; Dickinson et al., 1985). In Arabidopsis, for instance, the stigma will permit hydration and germination of pollen from most other species of Brassicaceae (Hiscock & Dickinson, 1993; Hiscock, 1993) and even species from the sister family Capparaceae, but will not permit hydration of pollen from species of Resedaceae, Scrophulariaceae, Urticaceae or Primulaceae (Hulskamp et al., 1995).

CER male-sterile mutants of Arabidopsis revealed the importance of the pollen coating for pollen hydration. CER mutants are defective in pollen hydration and consequently do not produce a pollen tube on the stigma; for instance, cer6-2 pollen, which has virtually no pollen coating, cannot stimulate water release from the stigma, but germinates readily in liquid medium in vitro, and in vivo under high humidity (Hulskamp et al., 1995). All cer mutants (e.g. cer1, cer3, cer6) are unable to synthesize long-chain lipids. As well as causing male sterility, mutants also show alterations in stem and leaf cuticle (Hannoufa et al., 1993; Jenks et al., 1995). CER1, mutants of which lack C29 alkanes, encodes a protein with similarity to the GLOSSY1 gene of maize (Aarts et al., 1995), whereas CER3 (mutants of which are deficient in C29 alcohols and ketones in addition to C29 alkanes) encodes a nuclear E3 ubiquitin ligase (Hannoufa et al., 1996), and CER6 encodes a long-chain fatty acid-condensing enzyme (Fiebig et al., 2000). Mentor pollen assays carried out with combinations of the various cer mutant pollens did not restore fertility, but all cer mutants could be rescued by wild-type pollen (Hulskamp et al., 1995), suggesting that cer mutants lack a single class of essential long-chain lipid. This observation is somewhat at odds with the fact that all cer mutants can be rescued by exogenous application of various triacylglycerides, mineral oil and a nonspecific alkane (Wolters-Arts et al., 1998; Zinkl & Preuss, 2000), suggesting that simply the presence of any abundant lipid at the pollen–stigma interface can facilitate pollen hydration. Under such circumstances, it would be interesting to determine whether rates of pollen hydration differ significantly between rescued cer mutants and wild type, because one might expect that the normally highly regulated passage of water from stigma to pollen, characteristic of wild type, would be eliminated or perturbed in the rescued mutants because of the absence of a normal coat. Lipids in the pollen coating of species with dry stigmas may therefore play an analogous role in pollen development to the lipids in the stigmatic secretions of species of Solanaceae (Wolters-Arts et al., 1998; Sanchez et al., 2004), with lipids in the pollen foot establishing a gradient of water potential between the pollen grain and the stigma, that acts a guide to the growing pollen tube. In support of this hypothesis, newly emerged pollen tubes of Brassica always grow towards and then through the pollen foot, where they penetrate the stigma at the base of the pollen foot (Elleman et al., 1988, 1992).

Regulated hydration of pollen is clearly an essential element of normal pollen development on stigmas of Brassica and Arabidopsis, and it has been suggested that while lipids provide the channel for water uptake by pollen, proteins play the key role in regulating this process (Swanson et al., 2004). In Brassica, regulated hydration of compatible pollen grains on the stigma is dependent on continued protein synthesis, as treating stigmas with cycloheximide before and during pollination results in elevated rates of hydration (Sarker et al., 1988). Under these experimental conditions, it is likely that this highly soluble and mobile protein synthesis inhibitor is exerting its effect in the pollen as well as in the stigma.

The pollen coating of Brassica and Arabidopsis is rich in proteins. One protein in particular, GRP17, a glycine-rich oleosin-domain protein (GRP), present in the pollen coating of Arabidopsis has been shown to be necessary for normal hydration (Mayfield & Preuss, 2000). GRP17 belongs to a family of GRPs present in the pollen coat of Arabidopsis and Brassica. Oleosins form coatings around oil bodies in seeds, regulating their size and preventing aggregation (Murphy, 2006). In Arabidopsis, the GRP family is arranged in a tandem array of eight genes, which show high levels of variation between species and even ecotypes, indicating that this gene cluster evolves rapidly (Mayfield et al., 2001). It has been suggested that this variable group of GRPs, and a group of similarly variable pollen coat lipases, may form species-specific or sub-species-specific pollen recognition ‘cassettes’ (Mayfield et al., 2001; Fiebig et al., 2004; Schein et al., 2004). How these species-specific pollen tags are read by the stigma remains to be elucidated, but an interaction with proteins of the pellicle, particularly members of the abundant nonspecific esterase class (which may also be lipases), is appealing as it is consistent with the esterase−esterase interaction model developed by Knox et al. (1976).

On the dry stigma, as on the wet stigma (see section III), water uptake by pollen is accompanied by reactivation of metabolism and polarization of the cytoplasm of the vegetative cell in preparation for germination (Heslop-Harrison, 1987). In Brassica, activation of protein synthesis during hydration is also accompanied by protein phosphorylation (Hiscock et al., 1995a), possibly reflecting the activation of small GTP-binding proteins such as Rops and Rabs, which mediate tip growth of pollen tubes (Wu et al., 2001; Cheung et al., 2002). As with wet-stigma species, pollen germination and pollen tube growth in species with dry stigmas have been studied largely in vitro, where these processes appear fundamentally the same (see reviews by Hepler et al., 2001; Feijo et al., 2004; Malho et al., 2006; Wilsen & Hepler, 2007; section III).

Unlike species with wet stigmas, pollen tubes of Brassica and Arabidopsis must breach a cuticle during penetration of the stigma (Dickinson & Lewis, 1973, 1975). In Brassica, an intine-held cutinase appears to be largely responsible for enzymic degradation of the cuticle in the region of the pollen tube tip (Hiscock et al., 1994). This pollen cutinase has a molecular weight similar to that of fungal cutinases (approx. 20 kDa) and, like most fungal cutinases, is inhibited by serine esterase inhibitors, such as DIPF and ebelactone B. Treatment of stigmas with these inhibitors before pollination reduced pollen tube penetration by up to 70% (Hiscock et al., 2002b). However, as both pollen and stigma contain a range of serine esterases, it is not possible to attribute this reduction in pollen tube penetration to a single enzyme (Hiscock et al., 2002b). Interestingly, these data are consistent with the Knox et al., (1976) model of a pollen/stigma ‘cutinase complex’ mediating pollen tube penetration. Cutinases have been implicated in the process of pollen tube penetration in many other species with dry stigmas, including Crocus, Gladiolus, Silene, Tropaeolum and species of Poaceae (discussed by Hiscock et al., 1994), but none has so far been fully characterized at a molecular level.

Once the cuticle has been breached, the pollen tube grows within the cell wall of a papilla cell towards the base, where it emerges and continues to grow intercellularly towards the ovary (Elleman et al., 1992). Again, pollen-held enzymes appear to be involved in these processes, most notably pectin-degrading enzymes such as pectin esterase, pectate lyase and polygalacturonase (Kim et al., 1996; Wu et al., 1996; Bosch et al., 2005) and also cellulose- and hemicellulose-degrading enzymes such as xylanases (Suen et al., 2003; Suen & Huang, 2007). One pollen-specific polygalacturonase in Brassica has been localized around pollen tube tips growing in the papilla cell wall (Dearnaley & Daggard, 2001). More recently, pectin methylesterases were shown to be essential for normal pollen tube growth in Arabidopsis in vitro and in vivo (Jiang et al., 2005; Tian et al., 2006), where they are probably involved in processes distal to the growing pollen tube tip, particularly pectin deposition associated with wall thickening (Bosch et al., 2005). Unlike Brassica, Arabidopsis pollen tubes usually penetrate papilla cells by growing directly through the cell wall and into the cytosol before forcing their way through the basal wall region and into the transmitting tissues, growth behaviour resembling that of pollen tubes penetrating immature stigmas of bud-pollinated Brassica sppp. (Elleman et al., 1992).

Pollen tube guidance in the stigma and style of species with dry stigmas is less well studied than in species with wet stigmas, and to date no functional orthologues of TTS or SCA have been identified in Brassica or Arabidopsis (Sanchez et al., 2004; Swanson et al., 2005). Recently however, a plantacyanin with similarity to chemocyanin from lily was identified in Arabidopsis and shown to be involved in reproduction, possibly by acting (like chemocyanin) as a guidance cue for growing pollen tubes (Dong et al., 2005).

Molecular characterization of the Arabidopsis male-sterile mutant pollen–pistil interaction2 (pop2) (Palanivelu et al., 2003) revealed an unexpected candidate molecule for a pollen tube guidance cue. Characterization of POP2 as a class III transaminase that converts γ-amino butyric acid (GABA) into succinic semialdehyde (Palanivelu et al., 2003) implicated GABA, a signalling molecule well known in animals as a neurotransmitter, as a potential signalling molecule in plant pollen–pistil interactions. Measurements of GABA levels in the Arabidopsis pistil identified a gradient of increasing GABA concentration through the pistil that reached its maximum in the ovary at the entrance to the ovule. Further, in vitro assays revealed that at physiological concentrations, GABA could stimulate pollen tube growth, while at higher levels it was inhibitory (Palanivelu et al., 2003). This finding correlated with observations of GABA levels in pop2 mutants, which were over 100 times higher than wild type, and resulted in disruption of the ovule-focused GABA gradient leading to aberrant pollen tube growth, loss of directionality and fertilization failure (Palanivelu et al., 2003). The next important steps will be (1) to identify the GABA receptor in Arabidopsis pollen, and (2) to determine how widespread the phenomenon of GABA-mediated pollen tube guidance is in angiosperms. Reverse-genetic approaches will certainly identify the GABA receptor. Such an approach is already helping to unravel the signalling system that operates between the synergids of the embryo sac (female gametophyte) and pollen tubes in the ovary (for review see Dresselhaus, 2006). Unlike GABA, however, guidance signals emanating from the synergids (Higashiyama et al., 2001) are unlikely to have any direct influence on pollen tube guidance in the stigma or style. Interestingly, the reverse scenario, of a stigmatic influence on the ability of pollen tubes to perceive signals released by the ovule, appears to be the case. Initial evidence for such a phenomenon came from in vitro fertilization assays in Tourenia, where pollen tubes growing in growth medium were unable to locate the micropyle of excised ovules unless they had grown through an excised upper portion of the pistil (Higashiyama et al., 1998, 2001). Using a similar in vitro assay system with Arabidopsis, Palanivelo & Preuss (2006) then demonstrated quantitatively that pollen tubes acquire the ability to perceive ovule guidance signals only after they have first grown through an excised pistil. Interestingly, in these assays some pollen grains germinated on the stigmas of excised pistils and then, rather than growing into and out of the pistil, grew onto the agarose medium, where they still managed to target excised ovules, indicating that interaction with the stigma alone is sufficient to confer the ability to perceive ovule guidance signals (Palanivelo & Preuss, 2006).

V. Is there any consensus among cell signalling pathways regulating pollen–stigma interactions?

Morphological and molecular diversity appears to be the common feature of pollen–stigma interaction ‘programs’ in angiosperms, and this diversity is likely to be important in maintaining species barriers (Edlund et al., 2004). Despite considerable conservation in the internal signalling networks regulating pollen tube growth and development in vitro in model angiosperms (see section III), no pollen–stigma signalling pathway (regulating pollen development in vivo) has yet been identified that is shared between species with wet and dry stigmas. Even within these two very broad divisions, there still appears to be much variation in the ways that pollen and stigma communicate to establish compatibility. Nevertheless, three broad areas of consensus among pollen–stigma interactions in most model systems can be identified: (1) a requirement for lipids at the pollen–stigma interface; (2) water acting as an initial directional cue for pollen tube growth; and (3) a general involvement of small cysteine-rich proteins, particularly LTPs.

The presence of lipids at the pollen–stigma interface has been demonstrated to be critical for normal development of compatible pollen on wet stigmas with a lipid-rich exudate (e.g. Solanaceae) and also on the dry stigmas of Arabidopsis and Brassica (Wolters-Arts et al., 1998, 2002; Sanchez et al., 2004). Whether lipids are essential for normal pollen–stigma interactions in species such as lily, where the stigma exudate is primarily aqueous, remains to be determined, but lipids are present in the pollenkit of lily pollen exine, so a role for lipids cannot be ruled out. The role of lipids is to establish a gradient of water potential between the pollen grain and the turgid cells of the stigma, which germinating pollen tubes sense and grow towards (Lush et al., 2000). That the lipids themselves are not involved in signalling or attracting pollen tubes is shown by the fact that various types of lipid can substitute for the stigmatic lipids of stigmaless Solanaceae female-sterile mutants (Wolters-Arts et al., 1998, 2002), and for the pollen coat lipids of coatless male-sterile cer mutants of Arabidopsis (Wolters-Arts et al., 1998, Zinkl & Preuss, 2000). Swanson et al. (2004) have questioned whether these exogenously applied lipids truly recapitulate the natural conditions on the stigma, as pollen hydration rates are often far higher than during wild-type pollinations, indicating that key regulatory processes (probably involving proteins such as GRP17 in Arabidopsis pollen coat) have been bypassed by these treatments. Irrespective of the fine-tuning, in the Solanaceae and Brassicaceae, lipids are essential to establish the conditions necessary for water to act as a guidance cue for pollen tube navigation to the stigma. It must also be the case that the gradient of water potential on the stigma governs establishment of cytoplasmic polarity in the vegetative cell of a hydrating pollen grain ahead of germination. Observations of Arabidopsis cer mutant pollen germinating on stigmas in a humid environment support this conclusion, because under these conditions (where no water potential gradient can be established because of the absence of pollen coat), pollen tubes emerge at random, indicating that they are unable to perceive the presence of the stigma (Dickinson, 1995).

As described in sections III and IV, a range of small cysteine-rich proteins from pollen and stigma have been shown, or inferred, to be involved in pollen–stigma interactions in both wet- and dry-stigma species. This class of proteins is extremely diverse, and members are involved in many different cellular processes in eukaryotes, including signalling (Nieuwland et al., 2005). The small cysteine-rich proteins LeSTIG1 and LAT52 (Kaothien et al., 2005) and their receptor kinase partner LePRK2 appear fundamentally important for pollen tube growth in Lycopersicon, where their interactions may define a checkpoint in pollen development on/in the stigma/pistil. It is therefore important to determine if these proteins have functional orthologues in other species. LeSTIG1 shows 72% amino acid identity with the stigma-specific protein STIG1 of tobacco and Petunia (Goldman et al., 1994), which was recently shown to control the regulated secretion of exudate during stigma development in these species (Verhoeven et al., 2005). So far, however, there is no evidence that STIG1 acts as a ligand for a receptor kinase(s) equivalent to a LePRK. Transcriptomic studies of Arabidopsis stigmas also failed to detect receptor kinases with significant sequence similarity to LePRKs (Swanson et al., 2005; Tung et al., 2005).

In stigma exudates of Nicotiana and Petunia, the most abundant protein is a small cysteine-rich protein, LTP (Nieuwland et al., 2005). This protein, which is unrelated to LeSTIG1 and LAT52, has cell wall-loosening activity similar to expansins, but its predicted structure places it conclusively with lipid transfer proteins (LTPs) putatively involved in acyl lipid metabolism (Nieuwland et al., 2005). Despite their assumed role in lipid metabolism, LTPs are a large and highly divergent class of proteins with potentially highly divergent functions (Kader, 1997; Nieuwland et al., 2005; Chae et al., 2007). SCA, another cysteine-rich protein involved in pollen tube adhesion and guidance in lily, also belongs to the LTP family (Lord, 2003; Chae et al., 2007). This suggests that functionally divergent LTPs may have been recruited to different roles in pollen–stigma interactions. Consistent with this prediction, LTPs and LTP-like cDNAs have been identified in the pollen coat and stigma transcriptome of Arabidopsis, respectively (Mayfield et al., 2001; Swanson et al., 2005; Tung et al., 2005), including potential SCA orthologues (Tung et al., 2005; section VII 1.). Interestingly, a plantacyanin similar to chemocyanin, which cooperates with SCA in pollen tube guidance in lily (Kim et al., 2003), was also identified in the Arabidopsis stigma transcriptome (Tung et al., 2005) and was subsequently shown to be involved in pollen tube growth, potentially as a guidance cue (Dong et al., 2005). Overexpression of the plantacyanin in transgenic Arabidopsis plants resulted in reduced seed set and aberrant pollen tube growth, leading the authors to suggest that in these plants a gradient of increasing plantacyanin concentration from stigma to ovary (as revealed by protein expression levels) was disrupted (Dong et al., 2005). Further functional analyses are now needed to determine whether this Arabidopsis plantacyanin can attract pollen tubes in in vitro bioassays, and whether this effect is enhanced by putative orthologues of SCA, as it is in lily. The discovery of an Arabidopsis plantacyanin potentially involved in pollen tube guidance raises the possibility of a consensus-signalling pathway for pollen tube guidance in a wet-stigma species (lily) and a dry-stigma species (Arabidopsis). It will also be interesting to explore the relationship between the pistil gradients of plantacyanin and GABA in pollen tube growth and reproductive success in Arabidopsis.

The pollen coating of Brassica and Arabidopsis contains a range of small cysteine-rich proteins (Doughty et al., 1998). In SI species, the S-cysteine-rich protein (SCR) is the male (pollen) determinant of SI, and interacts with its cognate stigma receptor SRK (S-receptor kinase), the female SI determinant, to trigger a signalling cascade in the stigma that results in pollen tube inhibition (Hiscock & McInnis, 2003b; Takayama & Isogai, 2005; section VI). In Brassica, another cysteine-rich protein, PCP-A1, which interacts with SLG (Doughty et al., 1998), has already been discussed in the context of a predicted role in cross-linking adhesion between pollen and stigma (section IV). Similarly, interaction between another PCP family member (SLR1-BP) and SLR1 (which is not linked to the S-locus) is also predicted to be involved in cross-linking adhesion (Heizmann et al., 2000b; Swanson et al., 2004). No orthologues of SCR or PCPs have been identified in wet-stigma species, so their function in SI and adhesion may well be restricted to species with dry stigmas, or even to just the Brassicaceae. However, the recent identification of an SCR-like protein as a potential candidate for the male determinant of SI in Ipomoea trifida (Convolvulaceae), another dry-stigma species (Rahmann et al., 2007), indicates that this class of cysteine-rich protein is not restricted to the Brassicaceae, and may have a widespread role in pollen–stigma interactions in species with dry stigmas.

The fact that, with the exception of plantacyanin (and possibly SCA), relatively few specific molecular commonalities have been identified between the pathways of pollen–stigma interactions in species with wet and dry stigmas is probably a reflection of the divergent model species, but it may also, at least in part, reflect the very different experimental approaches (in vitro assays vs genetic and reverse-genetic approaches) that have been employed by different groups to investigate their preferred model species. Powerful reverse-genetic and transcriptomic approaches are now being employed to study pollen–stigma interactions in the highly tractable model Arabidopsis, and will identify new functional and candidate genes, respectively. Once a framework model has been established for the pollen–stigma signalling network in Arabidopsis, it will be possible to explore the function of these proteins in other species with dry or wet stigmas to identify consensus signalling pathways between Arabidopsis and other species with wet and dry stigmas.

VI. Incompatibility and the pollen–stigma interaction

Much of the early work on pollen–stigma interactions and pollen–pistil interactions stemmed from work aimed at characterizing the various mechanisms of self-incompatibility (SI) found throughout the angiosperms (Heslop-Harrison, 1975, 1976, 1978). No review on pollen–stigma interactions would be complete without at least a brief discussion of SI and its relationship with the events that define a compatible pollen–stigma interaction. Here, discussion of SI is brief because this vast subject area has been reviewed extensively elsewhere (de Nettancourt, 1977; Hiscock & McInnis, 2003a; Nasrallah, 2003; Takayama & Isogai, 2005; McClure & Franklin-Tong, 2006).

Self-incompatibility is a genetically determined prezygotic barrier to self-fertilization and fertilizations between individuals carrying identical haplotypes of the polymorphic S locus (de Nettancourt, 1977). The two principal genetic forms of SI, gametophytic (GSI) and sporophytic (SSI), are distinguished by the way in which the incompatibility (S) phenotype of their respective pollens is determined. In GSI the S phenotype of the pollen is determined by its own haploid genome, whereas in SSI the S phenotype of the pollen is determined by the diploid phenotype of the parent plant. This means that dominance interactions between S haplotypes are possible in SSI but precluded in GSI (de Nettancourt, 1977). Correlations have been made between the type of SI system, GSI or SSI, and the timing of the incompatibility reaction during the pollen–pistil interaction (Heslop-Harrison, 1975). GSI is associated mainly with inhibition of incompatible pollen tubes within the style, whereas in SSI incompatible pollen is usually inhibited at some point during the pollen–stigma interaction, usually before a pollen tube can penetrate the stigma. Notable exceptions to this ‘rule’ are seen in poppy (Papaveraceae) and the grasses (Poaceae), both of which have GSI and stigmatic inhibition of incompatible pollen.

In species with style-based GSI, incompatible pollen completes a ‘normal’ program of pollen–stigma interactions before it is recognized and rejected. In GSI species of the Solanaceae, Rosaceae and Plantaginaceae, stylar inhibition of incompatible pollen is mediated through an interaction between a stylar S-RNase (female determinant) and a pollen tube-borne F-box protein, SFB (male determinant) (for review see McClure & Franklin-Tong, 2006). S-RNases enter both compatible and incompatible pollen tubes growing through the style (Luu et al., 2000), but only incompatible pollen tubes are rejected, following interaction between the S-RNase and SFB (Sijacic et al., 2004). The mechanism by which this interaction determines incompatibility still remains controversial (Qiao et al., 2004; McClure & Franklin-Tong, 2006), but the ubiquitin pathway appears to be involved, possibly through the targeting of nonself but not self S-RNases for degradation (Qiao et al., 2004). Interestingly, the ubiquitin pathway has also been implicated in the mechanism of pollen tube arrest following self-recognition in Brassica and Arabidopsis, which possess sporophytic SI (Stone et al., 2003; see below).

The molecular mechanism underlying stigmatic pollen rejection in GSI grasses has yet to be elucidated, but appears not to involve RNases (Baumann et al., 2000). By contrast, molecular events associated with stigmatic pollen inhibition in GSI poppy are well characterized. An incompatible reaction is triggered when a stigma S-protein ligand encounters its cognate S-receptor in pollen of the same S-haplotype. This triggers a Ca2+-based signal transduction cascade within the pollen involving protein phosphorylation, MAP kinase activation, pyrophosphatase activity and actin alteration, which collectively lead to pollen tube arrest and eventual death of the pollen tube by programmed cell death (PCD) (for review see Bosch & Franklin-Tong, 2007). In poppy, as in the grasses, the GSI reaction is rapid; incompatible pollen rarely survives beyond germination and a short period of tube growth on the stigma surface. The poppy stigmatic S-protein is presumed to be localized to the cell wall of the papillae, where it is well placed to encounter incompatible pollen grains at the earliest stages of the pollen–stigma interaction.

SSI has been studied most extensively in species from the Brassicaceae. In Brassica and Raphanus, incompatible pollen can be arrested at any stage during the pollen–stigma interaction, from prehydration to postpenetration, although incompatible pollen tubes rarely penetrate far into the stigma, and the majority of incompatible grains cease development before or just after germination (Dickinson & Lewis, 1973). Incompatible pollen is recognized by an S-haplotype-specific protein−protein interaction between a stigma-specific S-receptor kinase, SRK (female determinant), and its cognate ligand S cysteine-rich (SCR) (male determinant). (For recent reviews of SSI see Hiscock & McInnis, 2003b; Hiscock & Tabah, 2003; Nasrallah, 2003; Takayama & Isogai, 2005). This receptor−ligand interaction bears a number of similarities to the LeRLK2−LeSTIG1–LAT52 interactions presumed to be important for compatible pollen–pistil interactions in Lycopersicon (Kaothien et al., 2005). Both SRK and LeRLK2 are receptor serine−threonine kinases (albeit from different general receptor kinase families) and both have ligands that are small cysteine-rich proteins. Far more is known about the downstream signalling events that follow the receptor−ligand interaction between SRK and SCR. Here, S-haplotype-specific protein−protein recognition involves the formation of a receptor complex involving SRK, SCR and a cytoplasmic kinase, MLPK (M-locus protein kinase), and autophosphorylation of SRK on serine and threonine residues in the kinase domain (Takayama & Isogai, 2005). This leads to phosphorylation-dependent binding of an Armadillo repeat-containing protein, ARC1, to the kinase domain of SRK; ARC1 functions as an E3 ubiquitin ligase and appears to activate a ubiquitin-based protein-degradation pathway that leads to pollen inhibition (Stone et al., 2003; Takayama & Isogai, 2005). Whether activation of the ubiquitin pathway is directly responsible for inhibition of incompatible pollen remains to be determined, because the protein targets for ubiquitin-mediated degradation have yet to be identified (Stone et al., 2003). Identification of these target proteins has been facilitated by the finding that the ‘inhibitory machinery’ of SSI is preserved in self-compatible Arabidopsis, even though it lacks an S locus because Arabidopsis plants transformed with orthologues of SRK and SCR from a close SI relative, A. lyrata, became SI (Nasrallah et al., 2002).

In contrast to SI, far less is known about mechanisms of prezygotic incompatibility between species. Many well studied systems of interspecific incompatibility appear to utilize elements of SI systems in forming prezygotic barriers to hybridization (Hancock et al., 2003). Many of these interspecific incompatibility systems exhibit a phenomenon known as unilateral incompatibility (UI), whereby SI species reject pollen from related self-compatible (SC) species, but reciprocal SC × SI interspecific pollinations are compatible (Lewis & Crowe, 1958; Hiscock & Dickinson, 1993; Hiscock et al., 1998). This so-called UI rule holds for many plant genera with both GSI and SSI, and even for genera with hetermorphic SI (distyly) (Lewis & Crowe, 1958; de Nettancourt, 1977).

Conclusive proof that UI can be attributable to an involvement of the S-locus came from a series of elegant transgenic experiments in species of the genus Nicotiana, where Murfett et al. (1996) showed that the S-RNase was necessary for pollen rejection in some, but not all, cases of UI studied. From these studies, Murfett et al. (1996) concluded that the molecular mechanisms underlying UI are complex, some being dependent on and some independent of the SI system. This conclusion is endorsed by other studies (Liedl et al., 1996), which view UI and other interspecific incompatibility phenomena as largely a consequence of incongruity between precisely coevolved species-specific compatibility signals that underlie all successful pollen–pistil interactions.

In a recent review on pollen–stigma interactions, Swanson et al., (2004) highlighted the importance of molecular congruity for compatibility in pollen–stigma interactions, and identified particular stages of the pollen–stigma interaction where there is evidence for interspecific incongruity. Citing the study of Zinkl et al. (1999), showing differential adhesion of pollens from a variety of species to the Arabidopsis stigma, Swanson et al. (2004) suggested that differences in the adhesive ability of pollen and stigmas could be an important factor in preventing certain types of interspecific cross. It was also suggested that another pollen–stigma adhesion event involving a cross-linking reaction between the pollen coating and the stigma might also be involved in interspecific congruity/incongruity. In particular, diverse and rapidly evolving groups of LTPs and lipases (Mayfield et al., 2001; Fiebig et al., 2004; Schein et al., 2004) were identified as being potentially involved in these interactions. Differential pollen–stigma adhesion may be important in determining whether certain interspecies crosses are compatible or incompatible, particularly those involving distantly related species (Zinkl et al., 1999; Heizmann et al., 2000a), but they cannot easily explain UI where stigmas of SC species tend to accept most pollens from closely related SC and SI species, but stigmas of SI species rarely accept pollen from related SC species, while their behaviour with pollens from other SI species is unpredictable (Hiscock & Dickinson, 1993). For instance, SC A. thaliana stigmas accepted pollen (pollen tubes penetrated the stigma and grew into the style) of 20 different species of Brassicaceae (10 SC species and 10 SI species), while in the reciprocal pollinations, pollen from A. thaliana was accepted by the SC species and rejected by all but one (Capsella grandiflora) of the SI species (Hiscock & Dickinson, 1993). Studies such as this indicate that mechanisms of prezygotic interspecific incompatibility/incongruity are complex and involve multilevel failures of the pollen–stigma interaction (Hiscock et al., 1998; Hancock et al., 2003).

VII. New directions in pollen–stigma interaction research

1. Transcriptomic analysis of pollen and stigma

It is probably fair to assume that pollen and stigma possess suites of genes, the protein products of which are employed principally or exclusively in events associated with the pollen–stigma interaction. In the case of genes involved in SI, there is clear evidence that the corresponding pollen- and stigma-interacting proteins have coevolved before and subsequent to speciation (Hiscock & McInnis, 2003b). Similar coevolutionary events involving other pollen–stigma signalling molecules are expected to have occurred at different rates during angiosperm evolution (Fiebig et al., 2004; Schein et al., 2004). Transcriptomic profiling of genes with expression specific to, or upregulated in, pollen or stigmas is therefore a logical approach to identifying candidate genes involved in pollen–stigma interactions. This has been facilitated by advances in microarray technology, most notably the availability of genome-wide Affymetrix arrays for Arabidopsis and rice, and the development of reliable cDNA subtraction techniques such as suppression-subtractive hybridization (SSH). Recent studies using these approaches have opened up new avenues for pollen–pistil interaction research, particularly the opportunity to identify potential shared components of pollen–stigma signalling pathways among different species.

The pollen transcriptome  At the time of dispersal, pollen grains contain most of the mRNA transcripts required for germination and pollen tube growth (Mascarenhas, 1989, 1993). Based on this assumption, large-scale analyses of the Arabidopsis pollen transcriptome have been carried out using Affymetrix 8K Arabidopsis GeneChips that sample c. 8000 genes (Honys & Twell, 2003; Becker et al., 2003; Lee & Lee, 2003) and Affymetrix Arabidopsis ATH1 arrays that sample 22 750 genes (Honys & Twell, 2004; Pina et al., 2005). Becker et al. (2003) and Pina et al. (2005) profiled the transcriptome of mature hydrated pollen following precise fluorescence-activated cell sorting, whereas Honys & Twell (2003, 2004) profiled the transcriptome of microspores and pollen at different developmental stages, and Lee & Lee (2003) profiled the pollen transcriptome under normal and chilling conditions. As the transcriptomic profiles obtained by Becker et al. (2003) and Pina et al. (2005) give the clearest picture of the pollen transcriptome as it might be assumed to be when the pollen is in contact with the stigma (at point of hydration), we focus on these findings in the context of the pollen–stigma interaction.

The study of Becker et al. (2003) identified 1584 genes expressed in pollen (threefold fewer than four different vegetative tissues sampled), 10% of which were specific to pollen or selectively expressed. Of these genes, the main functional groupings were predicted to be in signal transduction, cell wall biosynthesis, stress response, metabolism, ion transport and cytoskeletal dynamics. The largest number of genes fell into the ‘unknown’ category. Similar findings were found for mature pollen by Honys & Twell (2003); Lee & Lee (2003).

In their more extensive survey of gene expression in mature pollen using the ATH1 arrays Pina et al. (2005) identified 6587 genes expressed in pollen, a greater than fourfold increase on the study by Becker et al. (2003). Again, pollen expressed a reduced subset of genes compared with three vegetative tissues sampled, but the pollen transcriptome contained a higher percentage of enriched (26%) and selectively expressed (11%) genes. As in the study of Becker et al. (2003), the largest group of genes identified was of unknown function, but of the functional classes identified, signalling, vesicle trafficking, membrane transport, cell wall biosynthesis and the cytoskeleton were the principal classes, indicating that the pollen transcriptome is functionally skewed towards the processes of germination and tube growth (Pina et al., 2005).

Combining the analyses of Becker et al. (2003) and Pina et al. (2005), and re-sorting the sequences according to putative functional class using the Functional Classification SuperViewer (Provart & Zhu, 2003), gave greater resolution to the functional classes of transcript expressed in Arabidopsis pollen (Fig. 5). These data may be viewed at different levels of pollen specificity. Within the pollen-enriched data set (Fig. 5a), kinase activity represents the largest functional subclass in the broad class of signalling; among the kinases are many receptor kinase-like transcripts, including leucine-rich repeat receptor kinases, but none with significant similarity to the LePRKs of tomato. Not surprisingly, many transcripts were identified that are putatively involved in signalling events known to be associated with pollen tube tip growth, such as calcium and phosphoinositide (IP3) signalling, small GTP-binding proteins, and cytoskeleton-associated proteins. In the hydrolase category were enzymes already implicated in pollen tube penetration and pollen tube growth in the stigma, such as pectinesterase, pectate lyase and polygalacturonase.

Figure 5.

Functional classification of pollen expressed genes from the combined data sets of Becker et al. (2003) and Pina et al. (2005). (a) Classification of genes enriched in hydrated pollen grains when compared with expression in vegetative tissues. After replicates between the two data sets were removed, 1999 genes were entered into the Functional Classification SuperViewer (Provart & Zhu, 2003). Of these, 1922 were assigned a function according to the Munich Information Centre for Protein Sequences (MIPS) database. (b) Classification of genes selectively expressed in hydrated pollen grains. After replicates between the two data sets were removed, 757 genes were entered into the Functional Classification SuperViewer (Provart & Zhu, 2003). Of these, 714 were assigned a function according to the MIPS database (Mewes et al., 1999). Data were normalized to the frequency of class over all ID numbers to highlight functional classes that were overrepresented in this tissue.

As the resolution of specificity of expression was increased, the ranking of the functional classes altered (Fig. 5a). Notably, the cell wall, extracellular and transporter activity classes became more significant, reflecting the importance of proteins from these functional classes in pollen germination and tip-focused pollen tube extension. Meanwhile, the significance of other functional classes decreased, indicating that these proteins may be involved in functions not exclusive to pollen. The classes of kinase activity, cell organization and biogenesis, and transporter activity were consistently high in the rankings, highlighting their importance in cellular function in pollen, even when combined with the large number of genes involved in housekeeping activities. Similar comparisons with the full data sets (all genes present in pollen) from Becker et al. (2003) and Pina et al. (2005) confirmed these trends, indicating the importance of comparing transcriptomes at different resolutions of specificity, as well as between different tissues, to identify key functional groups of genes/proteins within a specific tissue.

Now that a clear picture is emerging of the gene expression profile of mature Arabidopsis pollen, the question of how this relates to the pollen proteome becomes important. Given that some transcripts may never become translated, and others may give rise to multiple protein isoforms, perhaps with different functions, as a result of alternative splicing and/or post-translational modification, transcriptome and proteome profiles of eukaryotic cells do not always correlate simply (Lockhart & Winzeler, 2000). Apart from a proteomic analysis of the pollen coating of Arabidopsis (Mayfield et al., 2001), no proteomic analysis has yet been carried out on Arabidopsis pollen. Nevertheless, an analysis of the rice pollen proteome (Dai et al., 2006) identified some important correlations with the pollen transcriptome of Arabidopsis. Using state-of-the-art matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) and electrospray ionization quadrupole-time of flight mass spectrometry (ESI Q-TOF MS/MS) following 2-DE, Dai et al. (2006) were able to identify 507 protein spots, 307 of which were unique. Importantly, the main overrepresented functional classes of proteins were signal transduction, cell wall remodelling/metabolism, and protein synthesis, assembly and degradation, which together accounted for 35% of all proteins identified. These findings correspond broadly with those from the transcriptomic analysis of Arabidopsis pollen (Becker et al., 2003; Pina et al., 2005) confirming that, like the transcriptome of Arabidopsis, the pollen proteome of rice is geared towards germination and tube growth. Importantly, the proteomic study showed for the first time that many of the proteins probably required for germination and tube growth are preformed in mature pollen, and are not simply the de novo translation products of dormant transcripts. The preformation of many proteins involved in protein synthesis and protein assembly is nevertheless testimony to the fact that the mature pollen is also ‘primed’ for a concerted burst of translation from pre-existing mRNA transcripts at the point of pollen hydration and germination on the stigma. Unravelling the function and cellular interactions of these proteins will be the next crucial step.

The stigma transcriptome  Two recent studies used the Affymetrix ATH1 Arabidopsis GeneChips to profile the stigma transcriptome of Arabidopsis. Tung et al. (2005) compared the transcriptional profiles of stigmas and ovaries from wild-type Arabidopsis and from transgenic plants in which the cells of the stigma and transmitting tract were ablated by cell-specific expression of the diphtheria toxin subunit A (DT-A) fused to the ψSRK promoter. ψSRK is the A. thaliana orthologue of the Brassica and A. lyrata SRK that is expressed in stigmas and cells of the pistil-transmitting tissue (that is, all cells lining the pathway taken by pollen tubes on their way to the ovary). Swanson et al. (2005) also compared the transcriptomes of Arabidopsis stigmas and ovaries, but combined this analysis with a cDNA subtraction (SSH) to generate a cDNA library enriched for stigma transcripts.

Tung et al. (2005) identified a total of 14 000 genes expressed in wild-type stigmas, and a similar figure for ovaries. Comparisons between wild-type and transgenic plants identified 115 putative stigma-specific transcripts and 35 putative transmitting tract-specific transcripts. Because this screening approach was focused on identifying genes expressed specifically in the stigma and transmitting tissue, all stigma-expressed and transmitting tissue-expressed genes that are expressed in other parts of the pistil were excluded from the final data set. These included POP2, the transaminase implicated in GABA-mediated pollen tube guidance (Palanivelu et al., 2003), and a group of putative Arabidopsis orthologues of Brassica genes involved in SI signalling. Among the stigma-specific transcripts the largest group was of unknown function (unclassified), with the remaining transcripts falling into the functional classes of metabolism, signalling, transport, protein fate, stress and defence, and cell wall biogenesis. These included genes previously implicated in pollen–stigma interactions such as the orthologue of SLR1, and putative orthologues of chemocyanin and SCA from lily (Tung et al., 2005). Interestingly, no potential orthologue of LeSTIG1 was identified. The finding of three LTPs potentially orthologous to SCA and a plantacyanin potentially orthologous to chemocyanin is important because it suggests a possible mechanistic similarity between pollen adhesion and guidance processes in lily (wet stigma) and Arabidopsis (dry stigma).

Swanson et al. (2005) identified 11 404 genes expressed in Arabidopsis stigmas, and comparisons with ovaries (12 670 genes expressed) and seedlings (13 122 genes expressed) revealed 317 of these transcripts to be specifically elevated in the stigma transcriptome. Of the 30 stigma transcripts most enriched in stigmas relative to ovary, 22 were recovered in the subtracted stigma cDNA library, providing strong independent support for the array data set. Stigma-enriched transcripts fell into putative functional classes similar to those identified by Tung et al. (2005), with the overrepresented classes of cellular communication and signal transduction, cell wall metabolism and miscellaneous enzymes being the largest classes (Swanson et al., 2005).

The combined pistil-enriched data sets of Tung et al. (2005) and Swanson et al. (2005) are shown in Fig. 6. Many of the transcripts identified by Swanson et al. (2005) were identical or similar to stigma-specific transcripts identified by Tung et al. (2005), most notably peroxidases, kinases, LTPs and other putative lipid metabolism proteins, and enzymes involved in pectin metabolism. The Swanson et al. (2005) and Tung et al. (2005) pistil-enriched data sets were combined and checked for replicates (leaving 801 genes). After normalization, the largest five classes were cell wall, endoplasmic reticulum (ER), response to stress, plastid, and response to abiotic or biotic stimulus (Fig. 6a). Classification of genes selectively expressed in the pistil gave a similar result, the top five classes being cell wall, ER, response to stress, plasma membrane and response to abiotic or biotic stimulus (Fig. 6b). Notably, ER and cell wall were consistently the largest classes, even when dramatically different numbers of genes at different resolutions of pistil specificity were used. This signifies the importance of these classes of genes in pistil-specific activities, such as secretion and pollen reception. One important difference between the two data sets was the placement of the plastid-related genes in the ranking. In the pistil-enriched data set, this class appeared to be overrepresented, and was the fourth largest class (Fig. 6a). In the pistil-selective data set, this class appears to be underrepresented, suggesting that although important in cellular function, this class of genes is not required for pistil-specific processes (Fig. 6b).

Figure 6.

Functional classification of pistil-expressed genes from the combined data sets of Tung et al. (2005) and Swanson et al. (2005). (a) Classification of genes enriched in pistil tissue. After replicates between the two data sets were removed, 801 genes were entered into the Functional Classification SuperViewer (Provart & Zhu, 2003). Of these, 783 were assigned a function according to the Munich Information Centre for Protein Sequences (MIPS) database. (b) Classification of genes selectively expressed in pistil tissue. The 149 pistil-specific gene list from Tung et al. (2005) was added to the top 52 genes most enriched in pistil tissue compared with ovary, as identified by Swanson et al. (2005). After replicates between the two data sets were removed, 185 genes were entered into the Functional Classification SuperViewer (Provart & Zhu, 2003). Of these, 184 were assigned a function according to the MIPS database (Mewes et al., 1999). Data were normalized to the frequency of class over all ID numbers to highlight functional classes that were overrepresented in this tissue.

Recently, the first microarray analysis of the rice stigma transcriptome was reported (Li et al., 2007). As with the proteomic analysis of rice pollen, the findings of this analysis were remarkably similar to those from analyses in Arabidopsis, even though rice, which like Arabidopsis has a dry stigma, is a monocot. Li et al. (2007) identified 453 transcripts upregulated in rice stigmas relative to ovary, which fell into the same functional classes as those identified in Arabidopsis stigmas. Cell wall-related, signal transduction, lipid metabolism, transport, and unclassified were once again the major functional classes of transcript, with kinases, lipid metabolism proteins, pectin metabolism enzymes, and peroxidases again well represented in their appropriate classes (Li et al., 2007). With both the stigma and pollen transcriptomes of Arabidopsis well characterized, experiments can now be designed to test specific hypotheses about molecular interactions and signalling within and between these two tissue types during pollen–stigma interactions.

2. Signalling network analyses

Post-genomic technology and advances in analytical bioinformatics are beginning to allow gene networks to be predicted, based on known protein−protein interactions in model organisms (animals, fungi and plants). These predictions, combined with supporting coexpression and colocalization data, provide strong evidence for the identification of novel gene networks in Arabidopsis (Geisler-Lee et al., 2007). This technique has been applied successfully to identify protein−protein interactions in humans, allowing predictions of connected disease-associated genes that form subnetworks. This approach may be utilized when performing a genome-wide search to identify novel candidate disease genes (Gandhi et al., 2006). Various methods of gene network modelling have been used in Arabidopsis to interlink known interactions and create new hypotheses (Espinosa-Soto et al., 2004). To investigate the potential for such technology to predict gene networks active during the pollen–stigma interaction, we used the arabidopsis interaction viewer (Geisler-Lee et al., 2007) to analyse the combined pollen transcriptome (Becker et al., 2003; Pina et al., 2005; Fig. 5) and combined stigma transcriptome (Swanson et al., 2004; Tung et al., 2005; Fig. 6) to identify potential gene networks in pollen and pistil, respectively. This program generates both predicted and confirmed interactions based on data from the Biomolecular Interaction Network Database (BIND), high-density Arabidopsis protein microarrays (Popescu et al., 2007) and other literature sources. Subcellular localization was determined by information from the Arabidopsis Subcellular Database (SUBA; Heazlewood et al., 2007). This provides a potentially powerful tool for studying protein interactions both within and between tissues.

The combined gene lists of pollen-enriched genes from Becker et al. (2003) and Pina et al. (2005) generated several linked protein networks (Fig. 7a). The largest of these was centred on a single gene, At2g22290 (Fig. 7b), which encodes an Arabidopsis Rab GTPase (AtRABH1d). The Rab GTPases are a family of small GTP-binding proteins that regulate a range of cellular processes, including intracellular membrane trafficking, signal transduction and cytoskeletal organization (Vernoud et al., 2003). The high number of predicted interactions between this protein and others demonstrates its potential involvement in many cellular activities in pollen. One protein in this family, a Rab2 GTPase (AtRAB2), has been shown to be important in pollen tube growth, via the regulation of vesicle trafficking between the endoplasmic reticulum and the Golgi bodies in the pollen tube (Cheung et al., 2002). At2g22290 does not correspond to AtRAB2, so is likely to fulfil a different cellular role important for pollen function. It will be interesting to determine the exact role of At2g22290 in pollen.

Figure 7.

(a) Predicted interactions between protein products of genes enriched in hydrated pollen grains. Data were obtained from the gene lists of Becker et al. (2003) and Pina et al. (2005). The interaction network was generated by the Arabidopsis Interaction Viewer (http://www.bar.utoronto.ca). Details of protein localization and coexpression are shown in the key. Each interaction has been assigned a confidence value (CV) based on the number of organisms and supporting experiments (Geisler-Lee et al., 2007). (b) Enlargement of the cluster of interacting genes seen in (a).

We used the same approach to search for potential gene networks among the sequences in the combined pistil transcriptome data set. However, because of the extremely high number of potential interactions generated, we filtered the data to identify only those interactions predicted to occur among the target (pistil) genes entered. This proved a more useful way to view the data, and the combined pistil-enriched gene list (801 genes) from Swanson et al. (2005); Tung et al. (2005) produced 90 different potential interactions (Fig. 8a). A number of these potential interactors clustered together to form a discrete group of interacting proteins (Fig. 8b). One protein, encoded by At5g09230, appeared to be particularly important in this network, potentially interacting with four other pistil-enriched proteins. This protein, also called SRT2, is hypothesized to function in gene silencing (Brachmann et al., 1995), so could be involved in regulation of the four pistil genes with which it is predicted to interact. Interestingly, another gene identified in this pistil cluster is AtRABH1e, a homologue of the pollen AtRABH1d (Vernoud et al., 2003; see above). The presence of closely related copies of the same gene in the enriched fractions from pollen and pistil may indicate a conserved function for the homologues AtRABH1d and AtRABH1e in secretory processes in reproductive tissues.

Figure 8.

(a) Predicted interactions between protein products of genes enriched in pistil tissue. Data were obtained from the gene lists of Swanson et al. (2005) and Tung et al. (2005). The interaction network was generated by the Arabidopsis Interaction Viewer (http://www.bar.utoronto.ca). Details of protein localization and coexpression are shown in the key. Each interaction has been assigned a confidence value (CV) based on the number of organisms and supporting experiments (Geisler-Lee et al., 2007). (b) Enlargement of the cluster of interacting genes seen in (a).

Our simple network analyses of the pollen and pistil trancriptome data sets highlight the power of new interaction-modelling software for identifying gene/protein networks from raw transcriptome data. An interesting and logical extension to this work would be to examine potential interactions between different tissues. An initial screen of potential interactions between combined pollen and pistil transcriptome data sets identified 15 possible interactions (unpublished data). These included a ubiquitin ligase complex and an interaction between a protease in the pistil with its potential inhibitor in pollen. The ubiquitin pathway interaction is intriguing, because there is growing evidence that the ubiquitin/proteasome pathway is involved in pollen tube growth in both angiosperms (Actinidia deliciosa, kiwi fruit; Speranza et al., 2001; Scoccianti et al., 2003) and gymnosperms (Picea wilsonii; Sheng et al., 2006), and in lily, free ubiquitin has been shown to enhance pollen tube adhesion (Kim et al., 2006b). Although these interaction data must be treated with great caution (because they are simply predictions), they can form a basis for generating novel and testable hypotheses for protein−protein interactions during pollen–stigma interactions.

3. ROS and NO: new signalling molecules in pollen–stigma interactions

Reactive oxygen species and reactive nitrogen species, particularly NO, are key components of diverse signalling networks in animals and plants. In plants, ROS signalling is involved in many cellular processes, including stomatal closure, adventitious root development, root hair growth and gravitropism, as well as in responses to pathogen attack (Laloi et al., 2004). NO is involved in many of the same plant signalling networks as ROS, notably stomatal closure, root growth and programmed cell death (Neill et al., 2002). A number of recent studies have implicated ROS and NO as signalling molecules involved in plant reproductive processes such as pollen tube growth (Prado et al., 2004; Cárdenas et al., 2006; Potockýet al., 2007) and pollen–stigma interactions (McInnis et al., 2006a; 2006b; Hiscock et al., 2007).

In growing pollen tubes of tobacco, ROS produced by a pollen NADPH oxidase is localized to the tip of the pollen tube, where it is proposed to be important for the maintenance of polarized tip growth (Potockýet al., 2007) in a similar way to that proposed for tip-focused ROS in root hair growth (Foreman et al., 2003). Studies in lily showed that NO is excluded from the growing pollen tube tip and localizes in subapical regions of the pollen tube, where it is proposed to play a role in directional growth of the tube, because external sources of NO repel growing tube tips and cause a reorientation of tip growth in vitro (Prado et al., 2004). The mutually exclusive location of ROS and NO in growing pollen tubes suggests the possibility of a coordinated role for these signalling molecules in pollen tube growth (Potockýet al., 2007), although the exact placement of ROS and NO in the complex regulatory signalling network that controls pollen tube growth remains to be established.

That ROS and NO may be involved in pollen–stigma interactions was a chance discovery arising from an investigation into the possible function of a stigma-specific peroxidase (SSP) identified in Senecio squalidus (Oxford ragwort; McInnis et al., 2006a, 2006b). Peroxidases, like nonspecific esterases, are ubiquitous enzymes of the angiosperm stigma (Dafni & Motte Maues, 1998; McInnis et al., 2006a; Fig. 3), but their function, like that of stigmatic esterases, is unknown. SSP is the first, and so far the only, stigmatic peroxidase to be characterized (McInnis et al., 2005). SSP is expressed exclusively in the papillae of the stigma, where the SSP protein localizes to discrete regions of the cytosol and the surface pellicle (McInnis et al., 2006a). Transformation of Senecio is not yet possible, so because no obvious potential orthologues of SSP could be identified among the 73 annotated peroxidase genes in the genome of Arabidopsis (Valério et al., 2004), a functional analysis of SSP was not possible. McInnis et al. (2006b) therefore sought substrates for SSP, ROS and H2O2 in stigmas of S. squalidus. Confocal microscopy with the ROS-sensitive dye DCFH2-DA revealed that stigmas of S. squalidus accumulate ROS to relatively high levels compared with base levels of ROS in other plant tissues, and that ROS accumulation was confined primarily to the stigmatic papillae (McInnis et al., 2006b; Fig. 9). The hydrogen peroxide scavenger sodium pyruvate reduced ROS levels in papillae, indicating that H2O2 is the principal ROS of the stigma. The presence of high levels of ROS/H2O2 in Senecio stigmas was unexpected, and mirrored the cellular location of SSP. Further investigations identified similarly high levels of ROS/H2O2 in stigmas of A. thaliana, which prompted a more extensive survey of angiosperm stigmas with the H2O2-sensitive dye TMB to determine how widespread ROS accumulation in stigmas is among angiosperms. A screen of 20 species from a range of angiosperm families, from basal monocolpates to monocots and eudicots, that included species with wet and dry stigmas, revealed accumulation of ROS in stigmas of all the species tested (McInnis et al., 2006b; Fig. 10), suggesting that ROS accumulation in stigmas may be a general feature of angiosperms.

Figure 9.

Detection of ROS/H2O2 in stigmas of Senecio squalidus. (a) Stigma stained with guaiacol to visualize peroxidase activity. The distribution of peroxidase activity mirrors that of ROS/H2O2 in (b); bar, 500 µm. (b). Stigma stained with tetramethylbenzidine (TMB) to visualize ROS/H2O2; bar, 500 µm. Arrows in (a,b) indicate nonreceptive terminal pseudopapillae that stain less intensely than receptive columnar papillae. (c) Confocal image of stigma treated with 50 µm DCFH2-DA to visualize accumulation of ROS; arrows indicate intensely fluorescing columnar papillae; bar, 60 µm. (d) Untreated (control) stigma showing background fluorescence; bar, 60 µm. (Figure adapted from McInnis et al., 2006a.)

Figure 10.

Stigmas from a range of angiosperms stained with TMB to visualize ROS/H2O2 (blue staining). Basal monocolpates: (a) Magnolia sp.; (b) Drimys winteri. Monocots: (c) Lilium longiflorum; (d) Juncus sp.; (e) Carex sp. Basal eudicot: (f) Ranunculus sp.; Eudicots: (g) Ilex aquifolium; (h) Pittosporum crassifolium; (i) Apium sp.; (j) Menyanthes trifoliata; (k) Ipomoea sp.; (l) Echium canariensis. Bars: (a,b,j) 1 mm; (c,e,f,h–l) 2 mm; (d,g) 500 µm. (Figure from McInnis et al., 2006b.)

McInnis et al., (2006b) also screened stigmas for NO accumulation using the dye DAF2-DA. Low levels of fluorescence were observed in stigmas but far higher levels of fluorescence were observed in pollen (McInnis et al., 2006b), which prompted a more extensive survey of pollen from different species to determine whether pollen produces NO. A survey of 10 different species from monocot and eudicot families confirmed that pollen does produce NO (Hiscock et al., 2007; J. Bright et al., unpublished), which may be quickly converted to nitrite (J. Bright et al., unpublished).

What, then, could be the biological function of the ROS/H2O2 in stigmas and NO in pollen? One intriguing possibility is that they may be involved in some form of signalling cross-talk during the pollen–stigma interaction (McInnis et al., 2006b; Hiscock et al., 2007). Three pieces of evidence support this suggestion: levels of DCFH2-DA fluorescence in stigmas are reduced significantly by NO; levels of DCFH2-DA fluorescence in stigmas are reduced where pollen grains adhere and develop; and NO is released by pollen. These data suggest that pollen NO may cause a localized reduction in ROS/H2O2 levels in the stigma. If the main function of stigmatic ROS/H2O2 is in defence against pathogens (toxic levels of ROS have been proposed to protect nectar from bacterial and fungal infection (Carter & Thornburg, 2000, 2004), then compatible pollen may breach this defence system through a localized signal, perhaps in the form of NO or nitrite. Clearly this is a highly speculative proposition, but the possibility of a signalling role for ROS and NO in pollen–stigma interactions offers exciting possibilities for new avenues of research in this field of plant reproductive biology.

VIII. Future prospects

As more is becoming known about the molecular mechanisms of SI, greater interest is being focused on molecules that mediate the ‘default’ state of the pollen–stigma interaction: compatibility. Most of the data discussed in this review are from studies of just a few model angiosperms from the two great divisions of stigma morphology, wet (species of Solanaceae and lily) and dry (Arabidopsis and Brassica). These studies have revealed the importance of lipids at the pollen–stigma interface of species with wet and dry stigmas in setting up a gradient of water potential that orients the pollen and directs pollen tube growth towards the stigma. Critical insights into pollen adhesion and hydration on dry stigmas have been obtained from studies of Arabidopsis male-sterile cer mutants. The importance of small, cysteine-rich proteins in diverse aspects of pollen–stigma interactions, from SI recognition in the Brassicaceae, pollen adhesion and pollen tube guidance in lily, to cell wall loosening in Petunia, is becoming apparent. The list of signalling molecules involved in pollen–stigma interactions is ever increasing, and the identification of plantacyanins as pollen guidance cues in lily and Arabidopsis suggests a common pathway of pollen−pistil communication in species with wet and dry stigmas and offers great scope for future research into this group of plant cell wall molecules. The function of which has proved elusive for so long. The identification of GABA as a potential pollen tube guidance cue in the pistil of Arabidopsis is also exciting, as a model proposed for its action is similar to that proposed for plantacyanin. The recent findings that ROS and NO may also be involved in cell signalling on the stigma offers further possibilities for new research in pollen–stigma interactions. Future studies must now confirm and clarify the signalling roles of these molecules: plantacyanins, SCA, GABA, ROS and NO, in pollen–stigma interactions. Even though in vitro studies of pollen tube growth have revealed astonishing insights into the complexity of the internal signalling network that regulates tip growth, we know very little about the signals exchanged by pollen and stigma in vivo. It is therefore vital that future studies of pollen tube germination and development attempt to explore these processes in vivo, for instance using the ever-improving imaging technology afforded by confocal microscopy.

Despite the major recent progress in pollen–stigma interaction research outlined above, we still know very little about some of the most fundamental aspects of pollen–stigma interactions. For instance, we still know little more than we did 10–20 yr ago about the function and protein composition of the pellicle of dry stigmas. What are the functions of the esterases and peroxidases that define both dry and wet stigmas at their peak of receptivity to pollen, and what is the molecular identity of the cutinase or cutinases needed for pollen tube penetration of the dry stigma? These are key questions and research areas that need concerted reinvestigation. We have known for over 30 yr about the pellicle, and that it contains nonspecific esterases, but no pellicle esterases have yet been characterized. Perhaps they hold the key to the ‘cutinase conundrum’ of why removal of the stigmatic pellicle prevents pollen tube penetration. Knox et al. (1976) attempted to explain this by hypothesizing that pellicle esterases combine with pollen esterases (cutinases) to form a species-specific cutinase complex needed for pollen tube penetration. Only when the esterases present in the pellicle and in pollen diffusates have been fully characterized will we know the answer. Screens of sterile or partially sterile Arabidopsis mutants, looking in particular for aberrant pollen tube penetration phenotypes, would be one way towards identifying esterases/cutinases involved in pollen tube penetration.

Genomic technologies offer great hope for unraveling cell signalling networks involved in pollen–stigma interactions. Recent microarray analyses in Arabidopsis have provided long lists of genes expressed in pollen and stigma, many of which have been shown to be specific to either of those tissues. It is now essential to identify which genes from these lists are good candidates for full characterization and functional analysis to investigate their potential roles in pollen–stigma interactions. One way of doing this is to analyse transcriptome data sets using new predictive protein−protein interaction software to formulate hypotheses about potential protein-interaction networks. Here we explored the power of using this approach to identify pollen and stigma proteins predicted to form interacting networks in pollen and stigma only, and also in pollen and stigma combined – a ‘virtual’Arabidopsis pollen–stigma interaction. Our simple ‘test-drive’ of this new technology highlighted its utility for such an analysis, and identified several novel candidate proteins worth exploring further. As pollen and stigma transcriptomic data become available for more model plants, particularly with wet stigmas, this approach could certainly be a quick route for identifying pools of suitable candidate genes for more detailed functional analyses.

Finally, it should be remembered that approx. 95% of the material covered in this review has come from studies of just six species from only three plant families: Solanaceae, Brassicaceae and Liliaceae. Once pollen–stigma signalling networks become better elucidated in these model systems, and more consensus pathways appear (or not), it will be important to expand future studies to investigate pollen–stigma interactions in more diverse groups of flowering plants. In particular, more studies should be carried out on basal angiosperms to determine the extent of variation or conservation in basic aspects of the pollen–stigma interaction among the earliest diverging lineages of angiosperms, and how these pollen–stigma interactions compare with those of their model relatives, Nicotiana, Lilium and Arabidopsis.


We thank Gordon Breen for valuable help with the signalling network analyses, and Matthew Hegarty, Adrian Brennan, Christopher Thorogood and Stephanie McInnis for helpful discussions. We also thank Alistair Hetherington and three anonymous referees for constructive advice on improving an earlier version of the manuscript. Work in S.J.H.'s laboratory is supported by the Natural Environment Research Council (NERC).