Male gametogenesis in Arabidopsis results in the production of pollen grains containing a VN as well as two SCs (Twell, 2011). Transposons, which are methylated and therefore silent in most tissues, are demethylated and expressed in the VN of the pollen but not in SCs (Table 3; Slotkin et al., 2009). Decreased methylation in pollen is supported by the down regulation of RdDM components as well as the absence of DDM1 in the VN. Reduction of DNA methylation in the VN is likely an active process, as it is separated from SCs by only two cell divisions (Slotkin et al., 2009).

Slotkin et al. (2009) suspected the existence of a DNA demethylase active only in the VN, where it is responsible for removing DNA methylation from some transposable elements (TEs). It is crucial to protect the genome of SCs from transposons but not the genome of the VN, as it does not contribute genetic material to the progeny (Law and Jacobsen, 2010). In addition, an increase of 21 nt siRNAs that do not correspond to transposons expressed in SCs has been found in VN. It is thought that siRNAs produced in the VN may have the ability to travel to SCs, reinforcing silencing (Slotkin et al., 2009). The impact of methylation on TEs is discussed further in Section “Regulation of Transposable Elements Via Small RNAs.”

In addition, the DNA methyltransferase MET1 plays an important role during male gametogenesis in maintaining silencing. This has been demonstrated on genes such as FWA and FIS2, which are only expressed in the maternal genome of the endosperm. The use of maintenance methylation as opposed to de novo makes the establishment of imprints distinct in plants versus animals (Jullien et al., 2006).

In Arabidopsis, paternal imprinting results from demethylation in the CCN prior to fertilization, which combines with a SC to form the endosperm, resulting in maternal-only expression in the endosperm (Hsieh et al., 2009; Law and Jacobsen, 2010). Cases of paternal-only expression may also occur, although there are fewer examples of this (Hsieh et al., 2011). DME is essential in the CCN and seeds with a mutant maternal dme have CpG methylation levels similar to other tissues and do not complete development (Choi et al., 2002; Hsieh et al., 2009, 2011). However, the enzyme disappears following fertilization, resulting in an almost entirely demethylated maternal genome in the endosperm (Hsieh et al., 2009, 2011; Table 3). The end result is that the endosperm inherits two genomes with differing epigenetic states, a necessary requirement for imprinting (Bauer and Fischer, 2011).

So far only one imprinted gene has been found in the embryo, emphasizing that the majority of plant imprinting occurs in the endosperm. In addition, the majority of imprinting results in maternal-only expression, as was seen with 52 of the 65 imprinted genes identified in a recent study by Wolff et al. (2011). Another study by Gehring et al. (2011) found maternal-only expression for 165 of the 208 genes uncovered. Finally, work by Hsieh et al. (2011) found that among 43 imprinted genes, 34 of them were maternally expressed. Even among those that are paternally expressed, maternal demethylation may occur; for example demethylation at the tandem repeats downstream of PHE1 gene allows for PcG-mediated silencing of the maternal allele (Villar et al., 2009; Hsieh et al., 2011).

This maternal demethylation is likely the cause of the hypomethylation seen in endosperm tissue in comparison to the embryo. These findings correspond to previous observations that chromatin in the endosperm is less tightly bound, a characteristic often caused by hypomethylated DNA (Baroux et al., 2007; Gehring et al., 2009). In fact, studying differentially methylated regions (DMRs) in the endosperm and embryo has shown that approximately 90% are more methylated in the embryo. Many of these imprinted regions are those enriched for TEs, which are also the best candidates for imprinting as they show increased expression in the endosperm and low levels of transcription in other parts of the plants (Gehring et al., 2009; Kohler and Weinhofer-Molisch, 2010).

Recent work by Luo et al. (2011) in rice found that only one of 56 imprinted loci occurred in the embryo, and it also showed maternal-only expression in the endosperm. The locus identified was Os10g0750, which is a homolog of Ole e 1, the major allergen in olive and thought to control pollen tube emergence and production (de Dios Alche et al., 2004; Luo et al., 2011). Zemach et al. (2010) showed many genes that are hypomethylated and preferentially expressed in the endosperm are involved in rice endosperm biogenesis, such as starch synthesizing enzymes, indicating the crucial role of DNA methylation on gene expression. In addition, short TEs lost the most CHH methylation in the rice endosperm, while showing hypermethylation in the embryo, further suggesting the enhancement of transposon silencing via the endosperm, similar to what occurs in Arabidopsis (Zemach et al., 2010).

Despite a decrease of methylation in the endosperm, findings by Hsieh et al. (2009) actually indicated increased levels of CpNpN methylation in both endosperm and embryo tissues in comparison to adult plant tissue. They suggested that this might be the result of RdDM activity. In fact, endosperm hypermethylation has been shown to be a highly specific process performed by targeted siRNA (Hsieh et al., 2009). It was suggested that the high levels of maternally derived siRNAs that accumulate in the endosperm as a result of hypomethylation may be generated in the CCN, and possibly responsible for local hypermethylation and silencing in the EC as well as the developing embryo, a process similar to the one described in pollen (Hsieh et al., 2009).

Two distinct nuclei exist in pollen, and DNA hypomethylation is seen in the pollen VN but not the SC. SCs maintain high levels of methylation, which prevent TEs from being activated, a mechanism that is reinforced by siRNAs from the VN, which is a terminal tissue (Slotkin et al., 2009). As DDM1 (Table 1) is downregulated in the VN of Arabidopsis, reactivation of TEs occurs. The VN degenerates in the pollen tube prior to the SCs entering the ovule and so the cells responsible for fertilization lack active TEs, resulting in paternal imprinting (Slotkin et al., 2009).

Small RNA pathway components that show a strong increase in SCs include AGO9, DDM1, DRB4, MET1, and SUVH5, and their presence in pollen is likely a result of enrichment in SCs (Borges et al., 2008). Many of the enriched transcripts are involved in RdDM including DDM1, as previously described, and SUVH5, a histone methyltransferase that helps maintain non-CG methylation (Ebbs and Bender, 2006; Borges et al., 2008; Table 3). Although DCL3 was not found to be expressed in SCs, DCL1 was, as well as the AGO1-homolog AGO5. This suggests that although RdDM and maintenance of DNA methylation are important in SCs, they likely occur as a result of a novel small RNA pathway due to the absence of certain important transcripts such as DCL3 (Borges et al., 2008).

Demethylation of the maternal genome in the endosperm supports transposon reactivation while TEs remain silenced in the embryo (Hsieh et al., 2009). However, studying the transcriptome of female gametophytes is much more difficult than it is for males, due to its location and the relative rarity of female gametophytic cells (Le Trionnaire et al., 2011). Still, there is early evidence that similar to males, components of small RNA pathways may be present in female gametophytes. It has been speculated that down regulation of MET1 early during female gametogenesis results in the production of Pol IV-dependent siRNAs, also known as p4 siRNAs (Bourc’his and Voinnet, 2010). However, it is also currently thought that DME-directed genome demethylation is what allows the transcription and production of p4 siRNAs (Mosher and Melnyk, 2010). A recent study by Mosher et al. (2011) indicated that locus 08002 in Arabidopsis does not require differential DNA methylation for maternal-specific expression of p4 siRNAs. More work is still needed to investigate this potential link.

Work in Arabidopsis revealed several different AGO transcripts (AGO1, AGO2, and AGO5), which bind to siRNAs and direct chromatin remodeling, as well as DCL1 transcripts indicating that small RNA pathways are definitely present in the female gametophyte (Wuest et al., 2010; Le Trionnaire et al., 2011). AGO9 plays a crucial role in silencing of transposons in the ovule; ago9 exhibits reactivation of transposons in the ovule (Feng et al., 2010; Olmedo-Monfil et al., 2010). While ago9 mutants can initiate gametophyte development in somatic tissue, their ortholog in maize, ago104 results in the megaspore mother cell producing unreduced gametophytes (Grossniklaus, 2011; Singh et al., 2011). For this reason, AGO104 is required for the production of male and female spores in maize. In fact, ago104 showed reduced methylation at non-CG sites as well as increased transcription of repeats (Singh et al., 2011). Finally, in rice, MEL1, another AGO protein, has a crucial role in sporogenesis that is likely achieved via small RNA-mediated gene silencing (Nonomura et al., 2007).

Two functionally uncharacterized paralogs of PAZ/Piwi-domain encoding genes, AT5G21150 and AT5G21030 have also been uncovered, providing further evidence that small RNAs play an important role in female gametogenesis, potentially protecting against TEs in a manner similar to that which occurs in males (Wuest et al., 2010; Le Trionnaire et al., 2011).

Histones are an extremely important form of epigenetic control, and specific histone H3 variants characterize the angiosperm male germline. Despite a single cell division separating the VN from the GN, they show significant differences in chromatin structure, a differentiation that depends on histones (Ueda and Tanaka, 1995).

For example, Lilium longiflorum has been shown to possess three novel histone genes; gH2A, gH2B, and gH3; that only recognize the nuclei of the male gametic (both generative and sperm) cells and appear to be correlated to chromatin structure due to their association with the nucleosome (Ueda and Tanaka, 1995; Ueda et al., 2000). In addition, a gradual decrease in histone H1 levels has been shown to exist in only the VN of Lilium resulting in mature pollen where the VN contains very little histone H1. In comparison, histone H2B is maintained at a similar level in both the VN and GN (Tanaka et al., 1998). It is thought that histone H1 is involved not only in chromatin structure, but also the condensation of chromosomes during the mitotic phase (Woodcock et al., 2006). Therefore it is possible that the decrease of H1 in the VN helps to specify the developmental fate of a cell that, unlike the GN, does not divide again (Tanaka, 1997; Tanaka et al., 1998).

In Arabidopsis, the H3 variant HTR10 is expressed in the germline and mature sperm, but is completely replaced by other H3.3 variants in the zygote upon fertilization (Ingouff et al., 2007; Table 3). The impact of this replacement is not yet fully understood, however it is hypothesized that it is a part of global reprogramming events, similar to the H3.3 replacement that occurs in the mammalian germline. It is unique from DNA methylation reprogramming because it occurs in the zygote, as opposed to being limited to accessory cells in plants (Hajkova et al., 2008; Feng et al., 2010; Twell, 2011). This is likely an active process as paternal HTR10 signals are completely removed from the SC that fertilizes the egg only hours before S phase of the first zygote division, suggesting it occurs in a replication-independent manner (Ingouff et al., 2007; Feng et al., 2010).

Histone modifications in female gametes remain significantly less well understood than in males. However, an epigenetic dimorphism between the EC and central cell is known to exist, including the level of specific H3.3 isoforms found in each cell.

Only one H3.3 isoform (HTR5) present in SCs has also been found in ECs, while two – HTR8 and HTR14 – have been found in central cells. This dimorphism appears to be eliminated following fertilization, potentially indicating that it is established in order to distinguish the EC from the CCN during gametogenesis. The resetting of H3 variants in zygote chromatin suggests that these epigenetic marks are not transmitted to progeny. However, it is still possible that some loci may escape remodeling following fertilization, resulting in a maternal imprint, and further work is needed to determine whether or not this occurs (Ingouff et al., 2010; Baroux et al., 2011).

Recent work has provided further evidence of histone modifications resulting in epigenetic dimorphism between EC and CCN late during female gametogenesis in Arabidopsis. EC accumulates high levels of histone H3 lysine 9 dimethylation (H3K9me2) and LIKE HETEROCHROMATIN PROTEIN1 (LHP1) localization in comparison to the CCN. In addition, there is selective depletion of histone H2B in the EC and synergids of mature gametophytes (Pillot et al., 2010). These changes result in differences in transcriptional activity, due to the association of LHP1 with loci enriched in H3K27me3 and of a transcriptionally repressed state (Exner et al., 2009; Pillot et al., 2010).

In maize, variation in histone marks including H3K9ac and H3K4me3, both of which are associated with active transcription states, showed a much higher level of repressed chromatin in the EC versus the CCN, consistent with the pattern of the repressive H3K9me (Garcia-Aguilar et al., 2010). In fact, the ability of the RdDM pathway to silence gene expression is reinforced by methylation of H3K9me in Arabidopsis, indicating a link between changes in histones and the DNA methylation pathways previously discussed (Jackson et al., 2002).