Epigenetic inheritance refers to the transmission of modified genetic material from one generation to the next. These “epialleles” are not caused by mutations in the DNA sequence, but instead by covalent modification of chromatin and DNA, guided by developmental and environmental cues. In general, epigenetic modifications that are programmed during development must be reset in the germline, so that the zygote is restored to pluripotency and can once again initiate embryonic development. For example, imprinted genes in the mouse are expressed predominantly from either the paternal allele or from the maternal allele in the diploid embryo, and so must be reprogrammed in the germline depending on its sex (Bartolomei and Ferguson-Smith, 2011). Indeed, the mouse genome undergoes several rounds of DNA methylation, demethylation and repair as germ cells differentiate, as well as in the embryo after fertilization when imprinted genes are largely immune (Bartolomei and Ferguson-Smith, 2011; Feng et al., 2010; Popp et al., 2010). For this reason, epigenetic inheritance is thought to be rare in mammals, and is generally restricted to non-essential genes.
Flowering plants are an important exception to this rule, as epigenetic modification during development can be inherited for hundreds of generations with dramatic developmental consequences (Cubas et al., 1999). The first (and most common) examples of epigenetic inheritance in plants involved transposable elements (TE), which can regulate nearby genes, and undergo epigenetic switches during development, resulting in the inheritance of epialleles (Martienssen et al., 1990; McClintock, 1965). As in mammals, epigenetic inheritance of transposon activity in plants involves DNA methylation (Becker et al., 2011; Cubas et al., 1999; Martienssen and Baron, 1994; Schmitz et al., 2011). Imprinted genes tend to be flanked by transposable elements, whose methylation can influence their expression (Radford et al., 2011). However, imprinting in plants is largely restricted to the extra-embryonic endosperm, a terminally differentiated tissue within the seed, so that imprinted chromatin and DNA modifications need not be removed once they are established (Feng et al., 2010; Jullien and Berger, 2009; Raissig et al., 2011). The extent of reprogramming in the plant germline thus remains an important question.
Unlike mammals, which set aside their germline in early development, flowering plants give rise to germ cells during post-embryonic growth and development, in some cases many years after embryogenesis is complete. The pollen mother cell (PMC) on the paternal side and the megaspore mother cell (MMC) on the maternal side are specified from somatic cells in developing flowers (Boavida et al., 2005). In the anthers, the PMC undergoes meiosis resulting in four haploid microspores. Each microspore subsequently undergoes an asymmetric division to differentiate a larger vegetative cell and a smaller generative cell, which represents the male germline (Figure 1A). The vegetative cell exits the cell cycle into G0, while the generative cell undergoes a further symmetric division to produce two identical sperm cells that are surrounded by the vegetative cell (Berger and Twell, 2011).
The most conspicuous evidence of reprogramming in the plant germline is that the vegetative nucleus (VN) of the pollen grain has completely decondensed heterochromatin, in contrast to the tightly condensed chromatin found in sperm cell (SC) nuclei (Figure 1A). Heterochromatin in plants is mostly occupied by TEs and repeats (Lippman et al., 2004). TE repression is important for genome integrity and mutants in DDM1 (DECREASE in DNA METHYLATION 1) and MET1 (DNA METHYLTRANSFERASE 1) have reduced DNA methylation levels resulting in up regulation of TEs (Lippman et al., 2004). MET1 maintains CG methylation, and its activity in the germline impacts epigenetic inheritance (Jullien et al., 2006; Saze et al., 2003). In plants, CHROMOMETHYLASE3 (CMT3) maintains CHG methylation, guided by histone modification, and cytosines can also be methylated in an asymmetric CHH context guided by RNA interference (RNAi) (Law and Jacobsen, 2010). RNA-directed DNA methylation (RdDM) requires the DNA methyltransferase DOMAINS REARRANGED METHYLASE 2 (DRM2), and the RNA polymerase IV and V subunits NRPD1a, and NRPE1a, which are involved in production and utilization of 24nt siRNA (Haag and Pikaard, 2011). These mechanisms interact, so that RdDM is required to remethylate TEs in ddm1 mutants. TEs without matching siRNA cannot be remethylated even when DDM1 function is restored through crosses to wild-type plants (Teixeira et al., 2009).
Loss of heterochromatin in the vegetative nucleus of the pollen grain is accompanied by the loss of DDM1, the activation of TEs, and the production of a novel class of 21nt siRNAs which accumulate in sperm cells (Slotkin et al., 2009). However, while some TEs and repeats were found to be demethylated in the VN, others were hypermethylated so that the role of DNA methylation in pollen reprogramming was unclear (Schoft et al., 2011; Schoft et al., 2009; Slotkin et al., 2009). We set out to determine the dynamics of DNA methylation during pollen development, via bisulfite sequencing of genomic DNA from Arabidopsis microspores, and from their derivative sperm and vegetative cells (Figure 1A). We found that symmetric CG and CHG methylation were largely retained in Arabidopsis pollen. However, CHH methylation was lost from at least 1500 TEs, mostly long terminal repeat (LTR) retrotransposons, in microspores and sperm cells. In the VN, more than 100 DNA transposons and non-LTR retrotransposons were targeted for CG demethylation by DNA glycosylases. Many of these transposons, including those that flank imprinted genes, gave rise to 24nt siRNA in sperm cells where DNA glycosylases are not expressed. Recently discovered recurrent epialleles were pre-methylated in sperm cells guided by a similar mechanism. Thus reprogramming of DNA methylation in pollen contributes to transposon silencing, the transgenerational recurrence of epialleles, and imprinting of maternally expressed genes.