Snapshot of epigenetic regulation in legumes

In the current context of food security and increase in plant protein demand, legumes have an important role in facing challenges of climate change. With the recent improvement of sequencing technologies and the emergence of new knowledge related to plant epigenetic regulation in response to developmental and environmental changes, legume epigenetics is an emerging field with high potential for improving legume crop productivity and adaptability. The objective of this review is to provide a snapshot of epigenetic studies in different legume species. We have summarized the state‐of‐the‐art regarding legume epigenetic regulation controlling or participating in developmental aspects such as nodule, flower, and seed development and related to biotic and abiotic stresses. This extensive view of the different studies on legume epigenetics provides a baseline for identifying common and distinct mechanisms, and key players in epigenetic regulation from those of model species, such Arabidopsis, and highlights the impact that a better understanding of these mechanisms in legumes could have in order to improve plant productivity and adaptability.

context, CHROMOMETHYLASE2, and 3 (CMT2 and CMT3) for the CHG context, DOMAIN REARRANGED METHYLASE2 (DRM2) and CMT2 for the CHH context. These enzymes have a role in DNA methylation maintenance, but some of them have specific role in de novo methylation establishment through the DRM2-and CMT2-dependent pathways (for review Law & Jacobsen, 2010). Indeed, the DRM2-dependent pathway or de novo RNA-directed DNA methylation (RdDM) pathway relies on small RNAs, especially small interfering RNAs (siRNAs) to serve as guides to methylate specific DNA sequences. The RdDM pathway involves two RNA polymerases specific to plants: POL IV, necessary for siRNA production (Herr, 2005;Onodera et al., 2005;Pontier et al., 2005) and POL V needed to guide ARGONAUTE4 (AGO4) to the chromatin (Wierzbicki, Ream, Haag, & Pikaard, 2009)  plex may stabilize the siRNA/scaffold RNA to interact with the SWI/SNF chromatin remodeling complex that will change nucleosome positioning to silence transcription (Finke, Kuhlmann, & Mette, 2012;Zhu, Rowley, Böhmdorfer, & Wierzbicki, 2013). It is still unclear to which extent the RdDM pathway is involved in direct gene methylation and regulation, but it is crucial for targeting specific repetitive sequences and transposable elements (TEs), which may indirectly control nearby gene activation or repression (for review Sigman & Slotkin, 2016). Partially redundant with the RdDM pathway, the CMT2-dependent pathway is involved in de novo CHH methylation of heterochromatin regions and more specifically TEs (Zemach et al., 2013). This pathway, less described than the RdDm pathway, occurs through a siRNA independent manner and relies on DECREASED IN DNA METHYLATION 1 (DDM1) chromatin remodeler.
To counterbalance methylation activity, DNA demethylation occurs either passively due to failure in DNA methylation maintenance following replication or actively by regulation of methylation  Zhang, Lang, & Zhu, 2018). Several studies showed a coordination between DNA methylation and active demethylation by an antagonistic effect of RdDM and ROS1 activity to prevent hypermethylation at specific loci (Tang, Lang, Zhang, & Zhu, 2016). A 39-nt specific regulatory element in the ROS1 promoter, called a DNA monitoring methylation sequence (MEMS), has been identified to serve as a putative sensor of MET1 and RdDM pathway activities. Indeed, high MET1 and RdDM activities lead to hypermethylation of this sensor, which activates ROS1 demethylase expression to regulate the genome-wide DNA methylation (Lei et al., 2015). DNA methylation is usually described as a repressive modification of heterochromatin and pericentromeric regions, associated with gene and transposon silencing. However, it seems to play different roles depending on methylation locations. Methylation of promoter (or intergenic) regions has been proposed to regulate gene expression by inhibiting the binding of transcriptional activators/repressors, therefore activating or repressing transcription. It was shown to completely repress gene expressions, for tissue-specific genes (Johnson et al., 2007;Zhang et al., 2006). Methylation in promoter regions has been shown to be involved in gene regulation of specific processes such as imprinting during seed development and regulation of some immune-responsive genes (for review Matzke & Mosher, 2014;Zhang, Su, Hu, & Li, 2018). Methylation promoters appeared to be the consequence of the spreading of methylation from closely located TEs. In Arabidopsis, only 5% of promoter regions are methylated, but it does not reflect the situation of plant species with larger genome, such as legume crops, which contain many transposons and repeat elements with possible impacts on nearby gene expression through promoter methylation. Role of DNA methylation within gene bodies is still unclear. In contrast to methylation of promoters, gene body methylation occurs in 30% of Arabidopsis genes but with relatively low methylation levels . Some correlations revealed that body-methylated genes were enriched in GC context and that these genes were often associated with high and/or constitutive expression of such housekeeping genes . A recent study revealed that gene body methylation levels were not associated with highly expressed genes but rather with long and slowly evolving genes (Kawakatsu et al., 2016). To date, two hypotheses regarding the role of methylation in gene bodies have been proposed: (a) it could mask cryptic transcription sites, and it could help splicing of isoforms (Neri et al., 2017), (b) it could reduce variation of gene expression by excluding H2A.Z from the nucleosome, whose binding to gene bodies is anticorrelated to methylation but correlates to gene responsiveness to the environment (Zilberman, Coleman-Derr, Ballinger, & Henikoff, 2008). The role of methylation in TEs is much clearer; it acts as a repressor of the transposition activity inducing TE silencing. TEs are heavily methylated in all contexts and methylation maintenance involves mainly MET1, CMT3, DRM2 and relies on the RdDM pathway . TEs and repetitive elements represent a large proportion of most plant genomes, active TEs could insert within or around protein sequences disrupting normal genome function and threatening genome stability. To prevent this phenomenon, hypermethylation of TEs will silence and immobilize transposons in order to prevent disruption of normal gene functions and enhance genome stability (Mlura et al., 2001;Suzuki & Bird, 2008;Sekhon & Chopra, 2009; for review Sigman & Slotkin, 2016).
PHM are a conserved epigenetic mechanism controlling recruitment of chromatin remodeling proteins via modification of the nucleosome structure. Indeed, the nucleosome is an important structure controlling access and binding of regulatory factors (Berger, 2007).
Eight histone proteins form the nucleosome with two copies of each of H2A, H2B, H3, and H4 proteins, around which is wrapped 147 BP of DNA (Peterson & Laniel, 2004). Amino acids of the N-terminal tails of histones H3 and H4 are easily modified by methylation, acetylation, phosphorylation, ubiquitination, ribosylation, or biotinylation. These modifications will affect inter-nucleosomal interactions and permit recruitment of chromatin remodeling enzymes, leading to chromatin structure change. Histone modifications can activate genes through acetylation, phosphorylation, and ubiquitination and mostly repress genes through methylation, with some exceptions (Table 1). Repressive marks such as H3K27me3, H3K9me3, H4K20me have also been associated with heterochromatin-associated histone modification (Zhao, Zhan, & Jiang, 2019). Acetylation of lysines on H3 and H4 histones is controlled by multiple histone acetyltransferases (HATs) and histone deacetylases (HDACs). Methylation of lysines on H3 and H4 histones is controlled by histone methyltransferases (HMTs) and histone demethylases (HDMs). Regarding methylation, lysine residues can be mono-, di-, or tri-methylated, which confer different transcriptional roles with marks such as H3K4me2 and H3K36me3 acting in gene activation, whereas others such as H3K27me3 and H3K9me2 are repressive (see Table 1).
Although histones are highly conserved proteins, plants have developed structurally and functionally distinct classes of Histone 2A (i.e., H2A.X, H2A.Z) and H3 (i.e., H3.3) variants, which play important roles in the dynamics of association with DNA (see review Deal & Henikoff, 2011). H2A.Z, for instance, is a variant mainly found in gene bodies and around transcriptional start site of genes, acting with the SWR1 remodeling complex, and highly responsive to heat stress, which induces nucleosome dissociation from DNA, activating gene expression (Kumar & Wigge, 2010;Sura et al., 2017).
Finally, several studies have provided evidence of the interplay between DNA methylation and modification of histone marks to modify chromatin structure. As an example, it was recently shown how DNA demethylase ROS1 is recruited to target specific loci via two bromodomain-containing proteins, essential for recruiting the SWR1 remodeling complex through recognition of histone acetylation, which enhances active demethylation and deposition of H2A.Z histone variants (Nie et al., 2019).
Starting from the state of the art, mainly obtained in Arabidopsis, several recent articles have deciphered and compared epigenetic mechanisms in other plant species. In this review, we intend to provide a snapshot of epigenetic studies in legumes with a specific focus on epigenetic roles in developmental processes such as nodule, flower, pod and seed development, and responses to biotic and abiotic stresses.

| DEVELOPMENTAL PROCESSES
Nodule development is mainly controlled by nodule-specific genes including cysteine-rich genes (NCRs), which are specific to legumes producing indeterminate nodules, such as Medicago. In this species, nodule zones represent the temporal developmental stages and are composed of the meristem (or apical meristem, ZI), the invasion zone (ZII), and the nitrogen-fixing zone (ZIII), which display specific ploidy levels ranging from 2C/4C (ZI), 4C/8C (ZII), and up to 32C/64C (ZIII; Vinardell et al., 2003). Moreover, these ploidy levels are correlated with expression of NCR genes in different nodule zones (Nagymihály et al., 2017). The proportion of these zones will define nodule maturity from immature nodule, when ZI and ZII are predominant, to mature nodule, when ZIII is well expanded. The first correlative evidence of the importance of DNA methylation for nodule development was a differential expression of methylases and demethylases between nodule zones, with higher expression of DNA methylase genes such MET1, CMT2, and CMT3 in the nodule apex (ZI) and in contrast, demethylation genes, such as DME, which was more expressed in proximal part of invasion zone (ZIIp; Satgé et al., 2016).
To validate the role of DNA methylation in nodule development, DME was silenced by RNA interference. This led to abnormal development of the nitrogen-fixing zone, which was unable to fix nitrogen, indicating that DME control of demethylation is required for forming a functional nodule. DNA capture was performed to detect regions with high gene expression in immature and mature nodules. Four hundred seventy-four of highly expressed regions showed a correlated variation of methylation levels in CG and CHG contexts, whereas the level of methylation in CHH context was stable along nodule development T A B L E 1 Summary of some major histone modifications with their preferential binding locations and their transcriptional roles Activation -

H4K16ac
Activation - (Satgé et al., 2016). This result was confirmed by methylation level analyses between 4C and 32C cells, where 79% of DMR and 74% of DMR-associated to genes were found in CG context (Nagymihály et al., 2017). Interestingly, both studies showed that CG-DMRassociated to genes were located in NCR genes, which were more expressed in mature nodule. In parallel, Nagymihály et al. (2017) showed that 11% of coding genes were differentially methylated between 32C and 4C (6% hypermethylated and 5% hypomethylated).
These hypomethylated genes were overrepresented by NCR genes and nodule-specific genes. Methylation analyses of 375 NCR genes showed that 44% were hypomethylated and 4% hypermethylated, but unfortunately, no correlation between the level of methylation and the expression of NCR genes was observed in this study. Indeed, methylated NCR genes were expressed at the same level as hypomethylated NCR genes, indicating that methylation might be involved in activation but another mechanism such as histone modifications and/or chromatin accessibility could be implicated in gene expression.
To investigate this hypothesis, DNA accessibility during nodule development was performed by ATAC-seq analyses. It revealed that high DNA accessibility was correlated to high expression of NCR genes but only for the late stages of development and independently of methylation state, indicating that chromatin opening can occur without gene demethylation (Nagymihály et al., 2017). The same study analyzed the presence of the repressive histone mark H3K27me3 and the active mark H3K9ac on NCR genes. They showed that H3K27me3 was massively present on the promoter and gene body of NCR genes, which were not expressed and correlated with high level of chromatin compaction. Inversely, highly expressed NCR genes associated with H3K9ac on gene body and coincided with opened chromatin.
The initiation of flower and pod development is highly regulated and depends on environmental and developmental cues. In Arabidopsis, a strong epigenetic control has been highlighted for flower development (Groszmann et al., 2011;Liu & Wendel, 2003;Saze, Scheid, & Paszkowski, 2003;Simpson, 2004; for review Whittaker & Dean, 2017).
Several studies also confirmed the role of epigenetics in legume flower development, highlighting some similarities and differences with Arabidopsis, despite the fact that in soybean, flower initiation is known to be induced by short day conditions and does not need vernalization, unlike Arabidopsis. Liew, Singh, and Bhalla (2013) (Benhamed, Bertrand, Servet, & Zhou, 2006). Twenty-four histone deacetylases (HDACs) were identified, distributed in three classes (i.e., HD2, SIR2, and HDA with, respectively, 6, 4, and 14 genes).
Among the four SIR2 members, two showed homologies with AtSRT2 and two others with AtSRT1. Interestingly, SRT2 genes in soybean were more expressed in leaf whereas SRT1 genes were more expressed in SAM, suggesting different and specific regulations.
Regarding the HD2 class of histone deacetylases, in Arabidopsis, high levels of ABA repress HD2 expression (Luo et al., 2012). In soybean, the two HD2 orthologs of AtHD2 showed the same behavior as in Arabidopsis being highly expressed in the SAM before short-day condition, followed by decreased expression at the onset of short-day treatment, which induces ABA production (Wong, Singh, & Bhalla, 2013). Finally, among the HDA class of histone deacetylases, 14 genes were identified in soybean. The most represented, with four members, was the HDA6 family, which is known to deacetylate at various lysine residues (Chen & Wu, 2010;Krogan, Hogan, & Long, 2012;Zhou, Zhang, Duan, Miki, & Wu, 2005). Regarding seed development, An et al. (2017) analyzed the DNA methylation pattern in cotyledon during three stages of seed maturation in soybean, from early (S2) to middle (S6) and late seed maturation (S8). Global DNA methylation was mainly identified in CG (66%), then CHG (45%) and CHH (9%) contexts. However, the global CG and CHG levels were unchanged during seed maturation, whereas CHH level increased during seed development from 6% in S2 to 9% in S8. Lin et al. (2017) confirmed these previous observations by a more comprehensive analysis of DNA methylation during soybean seed development and within dissected seed tissues from postfertilization to germination. Indeed, global methylation levels in CG and CHG contexts were slightly decreased but changed little during the studied stages, whereas CHH methylation level greatly increased during seed maturation (between early-maturation stage and late-maturation), then dropped drastically at germination. Although the average global CHH methylation level across all samples was globally low (2%) compared to CG (57%) and CHG (37%), it increased more than three-fold during seed maturation then dropped by almost two-fold during germination. The mechanism involved in the variation of CHH methylation during seed development is still unclear because the authors did not observe any changes in MET1, CMTs, or DRMs expression. To have a better understanding of the role of CHH methylation during seed development, they analyzed the Arabidopsis ddcc quadruple mutant (i.e., drm1drm2cmt2cmt3), which is deficient in all methyltransferases involved in CHH and CHG methylation. Interestingly, the ddcc mutant did not show any major seed developmental defect or major changes in gene expression, suggesting that CHH does not play a fundamental role in proper embryo or seed development. This hypothesis was confirmed by the analysis of methylation levels within or closely related to genes essential in seed development and seed germination such as storage proteins, oil biosynthesis, master regulators of seed development, and germination-enhanced proteins. The authors revealed that almost 50% of these genes were localized in regions poor in methylation, called demethylated valleys (DMVs). The seed DMVs represented 21% of the genome and appeared to be consistent in all tested seed tissues. These DMVs also appeared to be enriched in transcription factors (TFs, with 46% of them), and in genes involved in embryo formation and seed development (e.g., WOX, CUC, CLAVATA, PIN1 genes). These hypomethylated regions did not show any variation in methylated state during seed development, whereas the genes contained in these regions were highly transcriptionally active and tightly regulated. To explain this gene regulation, they observed that the repressive histone mark H3K27m3 and the bivalent marks H3K27me3/H3K4me3 showed some modulation in these regions during seed development that appeared to be correlated with TF gene expression, suggesting a regulation of these TF expressions via histone mark modifications rather than DNA methylation. Finally, in the same study, they revealed that CHH methylation and also the CHG methylation were concentrated upstream and downstream of the coding sequence and within TEs, but very low in gene bodies; in contrast to CG methylation, which was mainly located in gene bodies. Changes in differentially methylated regions between the three developmental stages appeared to be enriched in CHH methylation sites. Indeed, 97% of DMRs were linked to CHH context, and 65% of these CHH-DMRs were found to be differentially methylated between the three stages and located close to transcribed genes. In the ddcc mutant, transposases of 106 TEs were upregulated and showed a high density of CHH methylation sites, suggesting that CHH methylation could play a role in repression of TEs during seed development. Although most of these results were obtained in Arabidopsis, the authors mentioned that the overall regulation of methylation during seed development seems highly similar in Arabidopsis and soybean. Therefore, results obtained from the Arabidopsis ddcc mutant could be extrapolated to soybean seed development.

| STRESS AND ADAPTABILITY
3.1 | Biotic stress DNA (de)methylation was found to play critical roles in defense responses against a wide variety of pathogens (for review Deleris, Halter, & Navarro, 2016). In plant defense, resistance (R) genes encode Nucleotide binding Leucine rich Repeat (NLR) proteins, which play critical roles in effector perception for triggering immunity. Under normal growth conditions, R proteins are maintained at low steady state levels and require a high degree of control to prevent fitness costs (Shirasu, 2009). In common bean, out of 197 CG-methylated NLR genes, 172 (87.3%) showed low to undetectable expression levels, suggesting that DNA methylation could be an alternative way of transcriptionally silencing R proteins under normal growth conditions to avoid fitness cost due to their unnecessary accumulation.
NLR proteins are organized in clusters that are often located close to the terminal knobs containing the satellite DNA khipu. Following this observation, Richard et al. (2018) suggested that methylation of NLR genes could result from the spreading of DNA methylation from khipu in common bean. In addition, it was shown that 24 nt siRNAs targeted 24% of NLR genes, which were identified as methylated, validating a potential regulation of NLR expression through RNA-directed DNA methylation (RdDm) pathway (Richard et al., 2018). including RDR2, AGO7 in chickpea, DCL2, DCL4, RDRs genes in pigeon pea, and DCL2 in groundnut, (Garg et al., 2017). A specific focus has been done on these genes as Arabidopsis ago7 and rdr and Tomato dclb2 mutants were found to be more susceptible to fungal and viral pathogens (Ellendorff, Fradin, de Jonge, & Thomma, 2009;. Several studies also demonstrated that histone methylation and histone acetylation play a role in plant immunity. ChIP-seq experiments of the repressive mark H3K9me2 and active mark H4K12Ac combined with RNA sequencing in common bean at different stages of Uromyces appendiculatus infection revealed key genes related to the bean-rust interaction. Expression profiles of genes such as defense response genes (e.g., low molecular weight cysteine 68, GIGANTEA protein and DnaJ-domain chaperone superfamily), R proteins (e.g., Pleiotropic drug resistance protein 12, MATE efflux family and NB-ARC domain-containing) were correlated with changes of histone methylation and histone acetylation modification (Ayyappan et al., 2015).
Defense priming is an intrinsic protective mechanism, in which plants prime their defense mechanisms after a first attack/infection in order to defend themselves more rapidly in subsequent interactions with pathogens (Mauch-Mani, Baccelli, Luna, & Flors, 2017). It has been shown that this phenomenon is related to the dynamics of chromatin structure. In Common bean, (pre)treatment with salicylic acid analogs such as BABA or INA enhanced resistance against P. syringae pv. Phaseolicola, with a protective effect transmitted to the next generation. It has been shown that this effect was due to the induction of pathogen-associated genes such as PR1, PR4, NPR1, and WRKY29, WRKY53, WRKY6, correlated with enrichment of the active histone mark H3K4me3 at the junction between promoter and coding regions in these genes (Martínez-Aguilar, Ramírez-Carrasco, Hernández-Chávez, Barraza, & Alvarez-Venegas, 2016).
Effectors secreted by pathogens are also known to target the components of HAT or HDAC complexes, thereby manipulating plant immunity. The ADH2 and GCN5 subunits of the SAGA complex (i.e., multi-protein chromatin modifying complex) are essential for HAT activity, which activates gene expression via acetylation of H3K9. Two robust studies in soybean highlighted the action of pathogen effectors in modulating plant immunity. The Phytophthora sojae effector PsAvh23 has been shown to bind to GmGNC5, which disrupts its assembly with ADH2, thereby decreasing H3K9 acetylation and resulting in transcriptional repression of soybean defense genes (Kong et al., 2017). Similarly, PsAvh52, an effector at the early stage of infection, interacts with GmTAP1, an acetyltransferase, regulating histone acetylation and promoting expression of susceptibility genes .

| Abiotic stress
Several studies have shown correlations between changes in methylation levels and environmental stresses, suggesting potential involvement of epigenetic mechanisms in plant adaptability. For instance, drought stress in faba bean and water deficit in pea were associated with an overall increase of DNA methylation in both tolerant and sensitive genotypes (Abid et al., 2017;Labra et al., 2002). In contrast, salt stress in pigeon pea induced a global decrease of DNA methylation in shoot (Awana et al., 2019). On a longer term, continuous stress increased global DNA demethylation mainly in a tolerant soybean genotype. This increased demethylation was consistent with increased expression of DNA demethylases such as DML and ROS1.
The demethylation analysis revealed that CG and CHG contexts within gene regulatory regions were more critical than CHH in soybean adaptation to stress (Liang et al., 2019). In contrast, salinity stress in Medicago truncatula, induced up to 77% of changes in CHH context, with only 9.1% and 13.9% in CHG and CG, respectively.
However, no correlation between transcript level and DNA methylation pattern of some key genes known to be involved in salinity stress was reported, implying that these genes might be regulated by other epigenetic mechanisms (Yaish, Al-Lawati, Al-Harrasi, & Patankar, 2018). In contrast, Song et al. (2012) showed that, in soybean, among four TFs induced under salt stress, three were demethylated in CG and non-CG contexts, preceding enrichments of active histone marks (H3K4me3 and H3K9ac) and decrease of the repressive mark H3K9me2, leading to gene upregulation, suggesting the possible interplay between DNA methylation and histone modification in stress response.
A growing body of evidence assigns crucial roles of histone acetylation and histone methylation in plant responses to external stress.
Changes in DNA methylation, histone methylation, and histone acetylation were observed in soybean root meristems growing at different temperatures. Immunostaining patterns indicated that 5-methylcytidine (i.e., a marker of methylated DNA) and H3K9me2, mainly located in the heterochromatin, were more abundant in soybean during chilling stress than during recovery. In contrast, H3K9ac, H4K12ac, and H3K4me, indicators of permissive chromatin, were weakly labeled in the euchromatin of stressed plants, but stronger during the recovery process (Stępi nski, 2012). Interestingly, crosstalk between histone methylation and histone acetylation was also reported in soybean subjected to salinity stress. Wu et al. (2011) F I G U R E 1 Summary of (a) developmental processes, (b) biotic, and (c) abiotic stress responses to different legumes species and their corresponding epigenetic mechanisms proposed that the salinity stress-inducible plant homeodomain TF, GmPHD5, could bind salt-induced H3K4me2 marks. This binding allowed recruitment of a complex involved in gene activation with non-histone proteins such as GmISWI, a chromatin remodeling factor and GmGNAT1, an acetyl transferase, which can preferentially acetylate H3K14 to further activate expression of salinity-induced genes.
In peanut, another study showed a regulation of gene expression of the AhDREB1 gene. The regulation, by acetylation of H3 enabled this member of the AP2/ERF TF family to positively regulate drought stress related genes, under PEG osmotic stress. Indeed, expression of AhDREB1 was showed to be higher using trichostatin (TSA), an inhibitor of HDAC (histone deacetylase), eventually inducing drought resistance (Zhang, Su, et al., 2018). Salt and drought stresses have also been showed to induce the activation of CaHDZ12, a HD-Zip TF, in chickpea, whose expression was correlated with acetylation of H3K9ac in the promotor region (Sen, Chakraborty, Ghosh, Basu, & Das, 2017).

| PERSPECTIVES
From this extensive review of legume epigenetic studies, we can clearly appreciate the importance of epigenetic regulations in developmental and stress-related processes (summarized in Figure 1). Considering legume crops not only with their large genomes containing many TEs and repeat regions but also their genes with high copy number (e.g., as described in this review with histone modifiers), their numerous small RNAs, and their specific legume processes (e.g., nodulation), there is no doubt that we will observed a growing interest in legume epigenetic studies in order to understand specific developmental processes and adaptative responses to environmental constraints in legumes. To conclude, epigenetic studies in legumes are still at an early developing stage and have been predominantly focusing on identification of key epigenetic players in different plant developmental or stress-related processes. This initial descriptive step is essential as most legume genomes are still poorly annotated and contains many genes with high copy number that could have overlapping or distinct functions. Perspectives will be an increase of functional studies of these key epigenetic players that could be enhanced by the rapid development of CRISPR technologies to generate collection of epigenetic mutants in major legume crops. From this perspective, a better understanding of epigenetic mechanisms and the identification of epialleles in legumes will potentially boost plant crop improvement and stress adaptation.

ACKNOWLEDGMENTS
This work was conducted in the framework of the regional programme "Objectif Végétal, Research, Education and Innovation in Pays de la Loire," supported by the French Region Pays de la Loire, Angers Loire Metropole, and the European Regional Development Fund.