The various classes of small non-coding RNAs are a fundamentally important component of the transcriptome. These molecules have roles in many essential processes such as regulation of gene expression at the transcriptional and post-transcriptional levels, guidance of DNA methylation and defence against selfish replicators such as transposons. Their diversity and functions in the sporophytic generation of angiosperms is well explored compared with the gametophytic generation, where little is known about them. Recent progress in understanding their abundance, diversity and function in the gametophyte is reviewed.
Over the last decade, the epigenetic mechanisms based on small non-coding RNAs have taken centre stage in biology. These small RNAs are a highly diverse component of the transcriptome, yet were long overlooked because of their size, leading to their description as the ‘dark matter’ of biology . Numerous recent discoveries have led to the realization that different classes of small RNA molecules are integral to many important cellular and molecular processes within plant cells. However, almost everything known about these molecules in plants has come from studies of somatic (sporophytic) material. Given the importance of small RNAs in the reproductive development of other organisms, from protozoa to vertebrates [2,3], this oversight is surprising. In the present paper, the different classes and functions of small RNAs are outlined before exploring recent progress in understanding these molecules in the context of the post-meiotic development of the gametophyte generation. Significant progress has been made not only through new generation sequencing technology, but also through the development of new cell-isolation methods.
The diversity of small RNAs
miRNAs (microRNAs), first discovered in nematodes, have been shown to be key regulators of post-transcriptional gene expression changes in plants [4,5]. In plants, miRNAs are typically generated by the nuclear processing of stem–loop or hairpin precursor transcripts to form a 21–22 nt mature RNA. The precursor transcripts are often discrete genes, usually with an intergenic location, expressed under their own promoter by the DNA-dependent RNA polymerase Pol II, although exceptions are well known, such as precursor transcripts contained in the introns of coding genes and transposon sequences. In the nucleus, the precursor is processed by specific protein machinery that includes DCL (Dicer-like) 1, SE (SERRATE), HYL1 (HYPONASTIC LEAVES1) and HEN1 (HUA ENHANCER1), with DCL1 performing the critical step of excising the miRNA from the precursor. The miRNA is excised by two cuts of the stem of the precursor and exported the cytoplasm where one strand is bound by an AGO (ARGONAUTE) protein, whereas the other strand (miR*) is typically dispensed with. Small RNAs, including miRNAs, have been shown to associate with AGO1 and also AGO10 [6,7]. AGO1 mediates miRNA function by binding mRNAs that are complementary to the miRNA sequence and inducing cleavage of the target mRNA at this site. The cleaved mRNA is subsequently degraded, resulting in down-regulation of transcript levels. AGO1, together with AGO10, also mediates cleavage-independent translational repression of target mRNAs [8,9]. Consequently, miRNAs contribute to post-transcriptional gene expression changes by two distinct mechanisms.
The majority of miRNAs target protein-coding transcripts. However, some target long non-coding RNAs, called TAS transcripts, in an extension of the miRNA pathway . Typically, these TAS transcripts have two independent binding sites for the same miRNA, although, in angiosperms, this leads to one cleavage event, rather than two [10,11]. The cleaved transcript is not degraded, but is instead processed further. An RNA-dependent RNA polymerase, RDR6, copies the cleaved TAS RNA in a 3′→5′ direction to form a dsRNA (double-stranded RNA) . This dsRNA is then repeatedly cut by the DCL4 enzyme to form a series of phased 21 nt small RNAs, named trans-acting short interfering RNAs (ta-siRNAs) [10,12]. These small RNAs in turn target coding transcripts for cleavage and degradation.
Other classes of siRNAs (short interfering RNAs) formed by plant cells are also triggered by properties that set them aside as ‘aberrant’, such as formation of dsRNA regions or lacking normal termini (such as transcripts derived from some transgenes and transposons) (e.g. [13,14]). Not only are 21 nt siRNAs formed, but also a separate 24 nt class is also produced simultaneously by the DCL3 enzyme . The 24 nt size class is associated with particular properties, notably its ability to promote de novo DNA methylation at homologous genomic loci . These RNAs are processed through a complex RdDM (RNA-dependent DNA methylation) pathway with exchange between subnuclear compartments and become associated with AGO4 protein [16,17]. In plants, two novel RNA polymerase complexes, Pol IV and Pol V, have specific roles in the production and processing of siRNAs from repeat-rich genomic regions . Through their combined activity, the dense DNA methylation pattern and transcriptional silencing is maintained in these genomic regions.
Small RNA pathways in the gametophyte generation
In order for small RNAs to be produced in the gametophyte, components of the biochemical pathways would need to be expressed in the cells. The gametophyte, especially the male gametophyte, may be sufficiently isolated to prevent effective transfer from soma of small RNAs that are known to be systemic . There is also no unequivocal evidence of inheritance of small RNAs through meiosis . Early work examined the expression levels of genes involved in small RNA pathways using data from microarrays of Arabidopsis pollen . This analysis concluded that the general trend was for down-regulation of transcript levels during development, and transcripts were absent from mature pollen. On microarray evidence alone, it was suggested that small RNA pathways became inactive in late pollen development. However, recent work has shown this is not the case. Grant-Downton et al.  showed that the dynamics of expression patterns were more complex through gametophyte development, with mature pollen showing expression of a significant number of key genes such as AGO1, AGO4, DCL1 and RDR6. Others were maintained until pollen mitosis II, making it feasible that recently translated protein products are persistent until maturity. Although this study  examined developmental changes in expression, the recent technical advance of sperm cell isolation in Arabidopsis  allowed analysis of expression in the gametes alone. Several small RNA pathway genes are expressed in sperm, including DCL1, AGO6 and RDR2. Strikingly, two AGO family members of currently undetermined function in Arabidopsis, AGO5 and AGO9, and a dsRNA-binding protein, DRB4, are significantly enriched. AGO5 is particularly interesting as a close homologue in rice, MEL1 (MEIOSIS ARRESTED AT LEPTOTENE1), plays an essential role in determining cell identity in pre-meiotic reproductive development . Collectively, these data support the notion that the small RNA pathways in the male gametophyte are maintained even after gametogenesis and show significant differences from somatic cells. Given the unique profile of small RNA pathway gene expression in sperm cells, it is possible that they possess distinct epigenetic systems based on a novel small RNA transcriptome .
In the female gametophyte, there is much less work supporting small RNA pathway expression patterns. No doubt this is due to the difficulty of isolating female gametophyte material, let alone isolated cells. Through subtractive analysis of gene expression using wild-type and mutants defective in embryo sac formation, two AGO family members, AGO4 and AGO5, were shown to be expressed in the embryo sac . The up-regulation of AGO9 in the ovule was also identified . In maize egg cells, an AGO10 homologue can be detected . Undoubtedly, the machinery for miRNA synthesis and miR-directed gene regulation is present in the developing embryo sac as artificial miRNA constructs work efficiently to down-regulate their targets . Evidently, much further work is required to characterize expression of small RNA pathway genes in the female gametophyte.
Small RNA diversity in the gametophyte
Compared with somatic cells, the focused exploration of small RNA diversity in the gametophyte was made more challenging because of its small size and difficulty of isolation, especially the female. Initial indications that endogenous small RNAs were present in the gametophyte have come from studies of in situ expression patterns of miRNAs using locked nucleic acid probes. Several known miRNAs are clearly localized in gametophytic tissues of Nicotiana, with miR164 and miR171 in pollen and miR160, miR164, miR167, miR171 and miR319 in the ovules . In Arabidopsis, further evidence of gametophytic expression of miRNAs can be found with expression of miR164 in pollen  and miR160 and miR167 in ovules . Later work confirmed that many miRNA families could be detected in mature pollen using RT (reverse transcription)–PCR, including their precursor RNAs , and this observation has been extended further by more comprehensive surveys. One study confirms, using reporter constructs, the de novo expression of miRNAs in the gametophyte, showing that the miR319 precursor transcript is indeed expressed in developing pollen . RT–PCR and miRCURY™ microarrays as well as Illumina™ sequencing data have been used to significantly increase the list of known miRNAs expressed in mature pollen . Although perfect agreement between different detection methods was not achieved, this work illustrates the potential diversity of miRNAs in just the two cell types of mature pollen. Use of 454 sequencing to survey small RNA populations from mature pollen found 33 families of known miRNAs . One important question is whether individual miRNA levels are comparable between sporophytic and gametophytic tissues. qRT-PCR (quantitative real-time PCR) has been used to compare inflorescence and mature pollen levels of miRNAs , whereas another study compared leaf and mature pollen levels . Whereas one study found that levels of almost all miRNAs tested were significantly lower , the other found that some miRNAs were enriched in mature pollen . Indeed, the 454 sequencing approach revealed several candidate novel miRNAs, including one which appeared, if not specific, at least highly enriched in the male gametophyte and targets a sperm-expressed F-box family transcript (At3g19890) for cleavage . The existence of such miRNAs was already hinted at by a study of male gametophytic mutants where some were mapped to intergenic regions containing no known genes .
In addition to miRNAs, several ta-siRNAs (and their initiating miRNAs) have been identified in mature pollen . Furthermore, recent work by Slotkin et al.  shows that the Arabidopsis male gametophyte actively produces siRNAs derived from TE (transposable element) sequences. In the vegetative cell, many TEs were found to be derepressed, and this may be due to loss of the chromatin-remodelling factor DDM1 specifically in this supportive cell. These transcripts were, unusually, processed into 21 nt siRNAs like ta-siRNAs rather than 24 nt siRNAs. These 21 nt siRNAs would appear to be exported from the vegetative cell into the sperm cells to reinforce silencing in the gametes. The exact mechanism of their biogenesis and cell–cell movement remains to be elucidated. As that study examined only the terminal developmental stage (i.e. mature pollen and isolated sperm cells), a complete understanding of siRNA dynamics and silencing will only be achieved by analysis of earlier developmental stages. Intriguingly, in maize sperm cells, the silencing of some TE sequences appears to be lifted, with TE transcripts accounting for a remarkably high proportion of the transcriptome .
In the female gametophyte, similar analyses of small RNA diversity are lagging behind, although future improvements in isolation techniques and sequencing technology may remedy this situation.
Small RNA function in the gametophyte
Understanding the function of small RNA systems in the gametophyte is rather challenging, as plants homozygous for strong alleles of genes such as AGO1 and DCL1 have such major effects on all aspects of sporophytic development that normal reproductive structures cannot form. Use of heterozygous plants carrying a strong ago1 allele showed that the transmission of this allele through the pollen was severely reduced compared with the wild-type allele . This demonstrates that AGO1 has a major role in male gametophyte development, although the precise nature and timing remains unknown. As yet, there are no confirmed reports for similar segregation distortion effects at other loci. Evidence for AGO1 function in pollen is reinforced by the demonstration that miRNAs in pollen are capable of cleaving their target mRNAs . miR168 cleaves AGO1 transcripts in the auto-regulatory loop that maintains steady-state AGO1 levels [37,38], suggesting that tight regulation of AGO1 remains important in the gametophyte. Cleavage of ARF (auxin-response factor) transcripts ARF16 and ARF17 by miR160 also indicated that miRNAs in pollen may promote rapid clearance of selected transcripts . Levels of both transcripts are very high before pollen mitosis II, but then suddenly fall below detectable levels. Whether AGO1 and other family members (e.g. AGO10) that are expressed in the male gametophyte contribute to transcript storage and translational repression would appear very likely, but this has yet to be determined experimentally.
Regulation of ARF transcripts by miRNAs may also be linked to patterning of the female gametophyte. Patterning of the syncytial female gametophyte has its basis in asymmetric auxin distribution that is generated by local differences in auxin biosynthesis . Uniform expression throughout the embryo sac of an artificial miRNA that targets ARFs generally results in defective embryo sac development . However, endogenous miRNAs that target ARFs (miR160 and miR167) appear to have significant expression levels in the ovule, with miR160 itself shown to have a strongly asymmetric distribution in the ovule, with the highest expression in the basal part of the ovule , indicating another level of complexity with interactions between auxin-based patterning and gradients of miRNAs.
The siRNA-based silencing of TEs in Arabidopsis pollen appears to have evolved to utilize the down-regulation of epigenetic control in the terminal vegetative cell to produce siRNAs that are subsequently used to reinforce silencing in the gametes. Indeed, altered epigenetic regulation in the gametophyte to allow ‘resetting’ of silencing in the gametes had already been predicted . An intriguing corollary is the potential for natural variation mediated by relaxation of TE regulation in the gametophyte. In maize, it would appear that sperm cells are associated with relaxation of TE regulation , and this could lead to new genomic insertions and higher rates of mutation. By restricting this event to the male gametes, novel variation could be generated, yet lethal or deleterious genotypes could be screened out by ‘soft’ selection in the competitive haploid phase of pollen tube development before fertilization  or during early post-fertilization development.
In the female gametophyte, small RNA systems are likely to have essential functional roles as two proteins with RRM (RNA-recognition motif) RNA-binding domains, FCA and FPA, are necessary for the production of functional gametophytes that develop normally upon fertilization . FCA and FPA may recognize aberrantly processed transcripts and siRNA-associated transcripts in their nascent form, and consequently target the corresponding locus for DNA methylation and chromatin silencing.
Until recently, study of small RNAs in the highly reduced angiosperm gametophyte remained intractable. Now, a wealth of new studies employing cutting edge genomics, transgenic and cell-isolation techniques are beginning to reveal small RNAs as integral components of the gametophyte transcriptome. Already it has been demonstrated that small RNAs from various classes are surprisingly diverse in the gametophyte transcriptome and that some small RNAs are enriched in the gametophyte. Many future challenges exist in this field, from describing small RNAs at different developmental stages and in the female gametophyte to understanding the functions of these molecules in the gametophyte. A critical issue is whether small RNAs regulate or contribute to the significant changes to chromatin, DNA methylation and, ultimately, cell fate that manifest during gametophytic cell divisions (e.g. [42–44]). Whether, as in animals, small RNAs are transgenerationally heritable (e.g. [45,46]) with effects in the two products of double fertilization, such as regulating parental conflict and post-fertilization development, remains another question open for future investigation.
Cell–Cell Communication in Plant Reproduction: A Biochemical Society Focused Meeting held at University of Bath, Bath, U.K., 14–16 September 2009. Organized and Edited by James Doughty (Bath, U.K.) and Rod Scott (Bath, U.K.).
I thank Professor Hugh Dickinson and Professor David Twell for comments on the paper, and the organizers of the Cell–Cell Communication in Plant Reproduction Meeting.
This work was funded by the Biotechnology and Biological Sciences Research Council [grant numbers BB/F008694/1 and BB/F007558/1].