The epigenetic state of an organism is the outcome of different molecular mechanisms such as DNA methylation, histone modifications and variants, as well as non-coding RNA changes 2,14,26,29. Among these molecular mechanisms, DNA methylation at the 5’ position of cytosine has been most well studied so far especially from the perspective of inheritance mechanism and variation 14,30,31. DNA methylation is established, maintained and removed by a series of pathways 14, and most of the proteins that are involved in these pathways have already been identified for rice 5, which provides a good starting point for exploiting the breeding potential of rice’s epigenetic pathways. The recent progress about DNA methylation pathways in rice has been reviewed by Deng et al. 2016 29 and Lanciano & Mirouze 2017 5, therefore here we will only give a brief summary of it.

Epigenetic pathways are relatively conserved between different plant species 2. As in Arabidopsisthaliana 14 rice de novo establishment of DNA methylation is carried out by the RNA-directed DNA methylation (RdDM) pathway 5. There is also RdDM in animals though with a much less important role 32,33. During RdDM, small interfering RNAs (siRNA) are first produced, via several steps, from the target locus, and they then, with the help of a set of proteins, head to the target locus, where the target locus gets eventually methylated by an enzyme that is called domain rearranged methyltransferase (DRM) 5,14. There are three DRM-encoding genes that have been identified in rice so far, OsDRM2, OsDRM1a and OsDRM1b, with the last two not being expressed 34. RdDM takes on two different forms in plant: canonical and non-canonical. The main difference between these two forms lies in the production of siRNA. Canonical RdDM is a proactive de novo methylation process, during which siRNAs production starts from the transcription of target locus by RNA polymerase IV (POL IV) 5,14. In contrast, the non-canonical RdDM is a rather passive remedial strategy that takes action when regular transcription of the target locus (by POL II) has already occurred, and it is the mRNAs produced via regular transcription that will be processed into siRNA 14,35,36. Apart from OsDRM2, POL IV and POL II that have been mentioned above, there are a bunch of other proteins that have also been identified to be possibly involved within the rice RdDM pathways: OsRDR1, OsRDR2, OsRDR6, OsDCL3a, OsDCL3b, OsAGO4a, OsAGO4b, OsAGO1a, OsAGO1b, OsAGO1c, OsAGO1d, OseIF4A, POL V, WAF1 and OsDCL4 etc. 5.

Different from animals where DNA methylation mainly occurs at the CG dinucleotide sequence context, plant DNA methylation takes place at three different sequence contexts: CG, CHG and CHH (H = A, T or C) 37. Plant DNA methylation can be maintained by different mechanisms depending on the sequence context 14. In rice, CG methylation is mainly maintained by the methyltransferase OsMET1-2, that is a possible ortholog of the mammalian DNMT1 enzyme 5. DNMT1 recognizes hemi-methylated CG during DNA duplication and methylates the unmethylated cytosine in the newly-synthesized daughter DNA strand 38. Apart from OsMET1-2, another methyltransferase, OsMET1-1, has also been identified in rice, and OsMET1-1 may have a minor and/or redundant role in maintaining the CG methylation 5. The maintenance of rice CHG methylation is mostly the responsibility of the plant specific chromomethyltransferase OsCMT3a 5. Another rice chromomethyltransferase, OsCMT3b, is also involved in CHG methylation but with a much smaller role. A third rice chromomethyltransferase (OsCMT2) has also been identified, however the function of OsCMT2 is still unknown 5. Rice CHH methylation is mainly maintained by OsDRM2 39 that is also involved in the RdDM pathways 40. The A. thaliana ortholog of OsDRM2, DRM2, maintains CHH methylation through RdDM at RdDM target regions that include certain transposons and repeat sequences 14. In addition, two rice chromatic remodeling proteins OsDDM1a and OsDDM1b also play roles in maintaining DNA methylation; OsDDM1a/b act at CG and CHG methylation within both euchromatins and heterochromatins, but at CHH methylation of small TEs that are mainly located in euchromatins 39.

DNA methylation could be passively lost due to failure in maintaining it, it could also be actively removed enzymatically 14. Different from DNA methylation reaction that is catalyzed by one single enzyme, active DNA demethylation reaction needs a group of enzymes to catalyze several steps of the reaction: DNA glycosylase, apurinic/apyrimidinic endonuclease, DNA phosphatase and DNA ligase etc., with the first enzyme, glycosylase, being conventionally referred to as DNA demethylase 14. So far, there has been six DNA glycosylases that have been identified in rice: ROS1a, ROS1b, ROS1c, ROS1d, DML3a, and DML3b, with ROS1a being most expressed. In addition, the loss-of-function mutation of ROS1a leads to serious developmental problems 5,29.

Comparing to A. thaliana, the consequences of DNA methylation pathway disruptions in rice tend to be more serious 5,27,29. For example, functional disruption of OsDRM2 leads to failure of RdDM, resulting in various developmental problems such as complete sterility 34,39, while the loss-of-function mutation of the orthologous gene DRM2 in A. thaliana showed almost no morphological differences from the wild type 41. This different “behavior” of rice from A. thaliana may be due to the much higher levels of DNA methylation and the greater proportion of transposable elements (TEs; generally epigenetically silenced 27,42) in the rice genome 2,27,29,43, as well as the fact that most rice genes have associated TEs 27. For example more than 85% of the African cultivated rice genes have been found to be associated with TEs 44.

Epigenetic modification can regulate gene expression and is therefore able to influence phenotypes 2,14, while the substantial TEs in rice genome act like a “loudspeaker” to amplify the phenotypic impact of epigenetic changes that could be a result of epigenetic pathway disruptions. This is, firstly, because TEs are the major targets of de novo RdDM pathways 33, which may process beyond the targeted TEs and methylate the flanking sequences 42,45,46. The spreading of DNA methylation from gene-associated TEs to gene body and/or regulatory sequences may help regulate the expression of the involved genes 42,47. For example, it has been shown that reducing the expression of the OsDCL3a gene that is involved in RdDM can cause ectopic expression of hundreds of MITE-related genes (MITE is one kind of TEs) in rice 27,48. Therefore, it has been proposed that epigenetic regulation of genes may be originally evolved for silencing invading TEs and virus 42.

Secondly, in rice, under stressful condition, TEs may be released from their epigenetic controls as a result of stress-induced epigenetic changes and become activated 27,49. This process may be facilitated by certain mutations within, or other disruptions of the epigenetic pathways 14,50. For example, an A. thaliana mutant impairing the POL IV enzyme that is important for RdDM showed boosted transposition of one specific TE (ONSEN) in response to heat stress, comparing to the wild-type plant 50. In rice, overexpression of the ROS1c/DNG701 gene that is involved in DNA demethylation caused more frequent transposition of one specific TE (Tos17) in calli (a stress-like condition) 51. The activated TEs transpose from one genomic region to another and this will not only modify genomic sequences and generate genetic diversity 27,49, but also likely bring epigenetic modification to the recipient genomic regions that may establish new TE-associated genes 50.

These TE-induced epigenetic and genetic variation may hold great potentials for creating phenotypic diversity that would contribute to rice environmental adaptability 27. For example, an artificially induced insertion of one specific TE (mping) into the intron of the rice Hd1 gene, which is important for photoperiod sensitivity, has been shown to cause flowering time change, and this may help Nipponbare adapt to different photoperiods; Nipponbare is the temperate japonica rice cultivar that flowers under longer day length than the tropical japonica rice cultivars 27,52.

From above, we know that disturbance of epigenetic pathways may create both epigenetic and genetic variation, which could modify phenotypes and might therefore help breed for climate changes. Then, where are we going to find genetic variation that can cause such epigenetic pathway disturbance? For the economically very important crop, rice, there is, luckily, a large amount of publically available genome data that we can resort to.

Rice has the smallest genome (ca. 375 mega base pairs 1) among the major cereal crops 52, therefore its genome has been extensively sequenced 44,53,54,55. Especially, the 3000 Rice Genomes Project has acquired the sequences of 3010 Asian cultivated rice (Oryzasativa L) genomes with a global representation of the genetic diversity within this rice species 1,55. Rice is well known for its substantial genetic diversity 56,57,58,59,60,61, for example, a recent analysis of the 3010 rice genomes has revealed 29 million single nucleotide polymorphisms and over 2.4 million indels 1. This enormous amount of genetic variation within the over 3000 rice genomes represents a great starting point for exploring the genes encoding proteins that are involved within the rice DNA methylation pathways in their potentials to influence epigenetic profiles.