Molecular mechanisms and biological functions of active DNA demethylation in plants

Abstract

DNA methylation is a conserved epigenetic modification that plays important roles in silencing transposable elements, regulating gene expression, and maintaining genome stability. In plants, DNA methylation is de novo established by the RNA-directed DNA methylation pathway and maintained during each cell cycle. It can be actively removed by the REPRESSOR OF SILENCING 1/DEMETER family proteins through the base excision repair pathway. Active DNA demethylation is essential for plant growth, development, reproduction and stress adaptation. During the past two decades, significant progress has been made in our understanding of active DNA demethylation. In this review, we will discuss the molecular mechanisms, regulation, and biological functions of active DNA demethylation in plants.

Introduction

DNA methylation, also known as 5-methylcytosine (5mC), is one of the most conserved and heritable epigenetic modifications in eukaryotes. It plays crucial roles in regulating gene expression, maintaining genome stability, and thereby ensuring normal growth and development in both plants and mammals. In plants, DNA methylation occurs in CG, CHG (where H represents A, C, or T), and CHH sequence contexts [1, 2]. Among these, CG methylation is the most abundant [3]. While CG and CHG methylation are primarily identified in heterochromatic transposable elements (TEs) and repetitive sequences [4, 5], CHH methylation tends to decorate euchromatic short TEs and the edges of long TEs [6, 7]. CG methylation may also occur in gene body regions, which is a conserved feature in plants. Gene body methylation (GBM) is predominantly found in moderately and constitutively expressed housekeeping genes [4, 8, 9]. Recent studies showed that GBM can reduce the probability of allele-specific expression, and thereby stabilize gene expression [10, 11]. In addition, dynamic GBM has been shown to regulate gene expression plasticity [12]. Moreover, GBM is also involved in the response to drought stress and may contribute to modulating the inducibility of immune responses in plants [13, 14].

The de novo establishment of DNA methylation primarily depends on the RNA-directed DNA methylation (RdDM) pathway. In RdDM, SAWADEE HOMEODOMAIN HOMOLOGUE 1 (SHH1) recognizes dimethylated histone H3 lysine 9 (H3K9me2) and recruits the plant-specific protein RNA polymerase IV (Pol IV). Pol IV, in collaboration with chromatin remodelers CLASSY1 (CLSY1) or CLSY2, then produces single-stranded non-coding RNAs [15, 16], which are converted into double-stranded RNAs by RNA-DEPENDENT RNA POLYMERASE 2 (RDR2). These double-stranded RNAs are processed into 24-nucleotide (nt) small interfering RNAs (siRNAs) by DICER-LIKE PROTEINS (DCLs) [17, 18]. The 24-nt siRNAs are then loaded onto ARGONAUTE 4 (AGO4) or AGO6, and guide these proteins to the scaffold RNAs generated by Pol V through complementary base pairing. Finally, DOMAINS REARRANGED METHYLASE 1/2 (DRM1/2) are recruited to catalyze DNA methylation in all sequence contexts [19–21]. Once established, DNA methylation can be maintained through three distinct pathways: CG methylation is preserved by METHYLTRANSFERASE 1 (MET1); CHG methylation is maintained by CHROMOMETHYLASE 3 (CMT3); and CHH methylation is maintained by CMT2 or RdDM, depending on the chromatin environment [22–27].

DNA methylation is reversible. In plants, DNA demethylation can be either actively removed by DNA demethylases or passively through DNA replication when the maintenance of DNA methylation is absent. DNA demethylation is essential for plant reproduction, development [28–31] and stress responses [32, 33]. During the past two decades, significant progress has been made in our understanding of active DNA demethylation. This review provides a comprehensive summary of the discovery and characterization of the DNA glycosylases, the molecular mechanisms underlying active DNA demethylation, and its biological functions in plant growth and development.

Passive DNA demethylation

Passive demethylation occurs during DNA replication when unmodified cytosines are incorporated into newly synthesized DNA strands, due to the inactivation of DNA methyltransferases or a deficiency of methyl donors. Over successive cell cycles, methylated DNA is progressively diluted, and the failure to maintain methylation during replication leads to passive demethylation [34]. DNA methyltransferase genes are highly expressed in dividing cells, but their expression is significantly reduced in quiescent cells due to the activity of the DREAM [Dimerization partner (DP), RB-like, E2F, and MuvB] complex [35–37]. The DREAM complex binds to the promoters of DNA methylation maintenance genes, including MET1, CMT3, and DECREASE IN DNA METHYLATION 1 (DDM1), repressing their transcription and thereby preventing DNA hypermethylation [38]. Moreover, the DREAM complex inhibits SU(VAR) HOMOLOGUE 4 (SUVH4), the H3K9 methyltransferase, which forms a tight feedback loop with DNA methylation. Therefore, passive demethylation is regulated by the DREAM complex. In addition to the activities of DREAM complex, passive DNA demethylation also occurs in the absence of DNA methyltransferases, for instance, MET1 and CMT2/3 [26, 39–41]. In addition, the chromatin remodeler DDM1 plays a crucial role in maintaining both DNA methylation and histone modifications, particularly within heterochromatic regions [27, 42]. Similarly, mutation of DDM1 results in passive DNA demethylation at specific TEs [27, 43]. Passive DNA demethylation can also be triggered by mutations in genes involved in methyl donor metabolism, such as HOMOLOGY-DEPENDENT GENE SILENCING 1 (HOG1), METHYLENETETRAHYDROFOLATE DEHYDROGENASE/METHENYLTETRAHYDROFOLATE CYCLOHYDROLASE 1 (MTHFD1), and FOLYLPOLYGLUTAMATE SYNTHETASE 1 (FPGS1) [44–46]. Lastly, chemical inhibitors can be used to induce passive DNA demethylation, among which 5-azacytidine (5-Az) and zebularine (ZEB) are the most widely used. They are incorporated into newly synthesized DNA strands to inhibit DNA methyltransferase activities [47].

Active DNA demethylation

The discovery of DNA glycosylase

To understand the maternal regulation of embryo and endosperm development, Choi et al. [28] conducted a mutant screening in Arabidopsis and identified the dme mutant. The dme mutant exhibited abnormal floral organs and seed abortion phenotypes [28]. DME contains a DNA glycosylase domain and a nuclear localization signal [28]. It functions in DNA demethylation and regulates the expression of maternally imprinted genes in the central cell [48]. Arabidopsis has three DML genes: DML1 (also known as ROS1), DML2, and DML3, all of which encode proteins containing a DNA glycosylase domain [28].

In the same year, Gong et al. [49] independently identified ROS1 through a genetic screening using the ProRD29A::Luciferase (LUC) reporter system. The RD29A promoter responds to various stresses, including drought, salt and cold [50]. By performing ethyl methanesulfonate (EMS) mutagenesis and mutant screen with ProRD29A::LUC expressing plants, they identified ros1 [49]. In the ros1 mutant, RD29A promoter became hypermethylated and remained silent under any stress treatment, indicating that ROS1 functions in DNA demethylation [49]. ROS1 was subsequently cloned through map-based cloning and was found to contain an atypical DNA glycosylase domain [49].

Subsequent studies revealed that DME and ROS1 are bifunctional DNA glycosylases capable of excising 5mC and cleaving the DNA backbone, two essential steps in active DNA demethylation [51]. The ROS1/DME family of DNA glycosylases demethylates substrates via a 5mC-specific base excision mechanism. A critical histidine residue in the base-binding pocket determines the specificity for 5mC over the structurally similar T:G mismatch substrate [52].

Base excision repair pathway

Reactive oxygen species (ROS) generated during cellular metabolism can oxidize DNA bases, representing a common form of DNA damage in organisms [53, 54]. The bifunctional formamidopyrimidine-DNA glycosylase (FPG) cleaves the oxidized DNA backbone, generating nicks with 3′-phosphate (3′-P), and mediates BER pathway for DNA damage repair [55, 56]. Active DNA demethylation is considered a specific type of DNA damage repair. In mammals, 5mC must undergo oxidation or deamination before it can be removed by the monofunctional thymine DNA glycosylases (TDG) or METHYL-CPG-BINDING DOMAIN-CONTAINING PROTEIN 4 (MBD4) [57]. In contrast, plants possess bifunctional DNA glycosylases that can directly recognize and excise 5mC. These enzymes hydrolyze the glycosidic bond between the base and the deoxyribose with the glycosylase activity and cleave the DNA backbone with their apurinic/apyrimidinic (AP) endonuclease activity [48] This cleavage generates a single-nucleotide nick with either 3′-P or 3′-phosphor-α,β-unsaturated aldehyde (3′-PUA), resulting in single-strand breaks (SSBs) with 3′ blocked ends (DNA 3′-blocks) [48]. The nick generated by the β-elimination reaction has a 3′-PUA, while the β,δ-elimination reaction produces a nick with a 3′-P [48, 51, 58]. Both 3′-PUA and 3′-P must be converted to 3′-OH before the gap can be filled by DNA polymerase and DNA ligase. This processing is primarily carried out by AP ENDONUCLEASE (APE)/ZINC FINGER DNA 3’-PHOSPHOESTERASE (ZDP) through the BER pathway [48, 51, 59, 60] (Fig. 1).

Fig. 1.

Base excision repair pathway. The bifunctional glycosylases, ROS1/DME, are capable of directly recognizing and excising 5mC. The cleaved DNA presents nicks of the 3ʹ-PUA or 3ʹ-P terminus types, which are generated via the β-elimination reaction and the β,δ-elimination reaction, respectively [48, 51, 58]. Prior to the repair of the gap by DNA polymerase (POL?) and DNA ligase LIG1, the 3ʹ-PUA and 3ʹ-P termini must be converted to 3ʹ-OH by ZDP/APE [58, 68]. In zdp/ape2 mutant, defective processing leads to the accumulation of DNA 3′-blocks, thereby triggering the DDR. The kinase ATR detects these lesions and activates the transcription factor SOG1 via phosphorylation [71], which in turn induces DNA hypermethylation via the RdDM pathway [182]

In plants, APE1L, APE2, and APURINIC ENDONUCLEASE-REDOX PROTEIN (ARP) are homologous to AP endonucleases found in bacteria, yeast, and animals [61]. APE1L and ARP are capable of processing nicks with 3′-PUA in vitro [58]. Since APE1L exhibits higher activity than ARP, it is likely more important in converting 3′-PUA to 3′-OH. Mutations in APE1L result in large-scale alterations in DNA methylation, and disrupt the expression of maternally imprinted genes, such as FLOWERING WAGENINGEN (FWA) and MEDEA (MEA) during endosperm and seed development [7, 58]. ZDP/APE2 can process DNA with 3′-P, with ZDP playing a more dominant role [59, 60]. It should be noted that APE2 and ZDP not only participate in DNA demethylation but are also important in DNA damage repair. While mutations in zdp or ape2 alone do not result in obvious developmental defects, the zdp ape2 double mutant accumulates DNA damage, leading to sustained activation of the DNA damage response (DDR) and developmental defects [59].

The 3′-OH nicks are repaired by DNA polymerase and DNA ligase (Fig. 1). Although the specific DNA polymerase involved in this process remains unclear, DNA LIGASE 1 (LIG1) has been identified as a key player in the final ligation step. Arabidopsis contains four DNA ligases: LIG1, LIG1a, LIG4, and LIG6 [62]. LIG1 is primarily responsible for the ligation during DNA replication and BER [63]. LIG1a may not be expressed [64]. LIG4 is involved in non-homologous end joining (NHEJ) of DNA double-strand breaks (DSBs) and is crucial for the survival of Arabidopsis thaliana under genotoxic stress [65, 66]. LIG6 plays a role in DSB repair and is essential for seed germination under both normal and stress conditions [67]. Recent studies suggest that LIG1 is likely the sole ligase responsible for the final ligation step in DNA demethylation. In the lig1 mutant, DNA hypermethylation occurs at multiple loci. LIG1 co-localizes with ROS1, ZDP, and APE and is crucial for the expression of maternally imprinted genes such as FWA and MEA [68].

Typically, the BER pathway is highly efficient, ensuring prompt repair of DNA damage [69]. Inefficient repair leads to neurological diseases and cancers in animals [70]. In plants, if DNA 3′-blocks remain unrepaired, the DDR will be triggered. Two protein kinases, ATAXIA-TELANGIECTASIA MUTATED (ATM), and ATM and RAD3-related (ATR) are involved in detecting DNA damage and initiating the DDR [71]. ATM is activated by DNA DSBs, while ATR responds to replication fork stalling and DNA SSBs [72, 73]. SUPPRESSOR OF GAMMA RESPONSE 1 (SOG1) transcription factor regulates plant responses to DNA DSBs, polymerase and helicase mutations, aluminum stress, and plastid genome instability [74–78]. In the zdp ape2 mutant, the accumulation of DNA 3′-blocks is detected by ATR, which activates SOG1 through phosphorylation [71].

The regulatory and recruitment mechanisms of DNA glycosylases

The regulation of DNA glycosylases

ROS1, DML2, and DML3 are expressed in vegetative tissues, whereas DME is predominantly expressed in the central cell of the female gametophyte and the vegetative cell of the male gametophyte [79, 80]. Recent studies have demonstrated that DME is capable of performing DNA demethylation not only in gametophytes but also in somatic cells [81, 82]. Compared to the Col-0 and the ros1 dml2 dml3 (rdd) mutant, dme ros1 dml2 dml3 (drdd) mutant exhibits DNA hypermethylation in leaves, suggesting that members of the DME/ROS1 family function redundantly in somatic cells [31].

Recent studies have revealed the regulatory mechanisms of DME/ROS1 at transcription level. Specifically, the sequence extending 46 bp downstream of the transcription start site is sufficient for the proper expression of DME in both central and vegetative cells [83]. Additionally, an enhancer located within the minimal promoter has been identified [83]. While DNA methylation at promoter regions typically silences nearby genes, in the case of ROS1, DNA methylation promotes transcription. The expression of ROS1 is significantly reduced in the RdDM mutants, as well as met1, but significantly increased in ros1 [84]. The ROS1 promoter contains a 39 bp DNA methylation monitoring sequence (MEMS), which is capable of sensing DNA methylation status, and thereby modulates the expression of ROS1 to finetune genome-wide DNA methylation levels [85]. SUVH1 and SUVH3 bind to methylated MEMS and activate ROS1 expression [86]. Disrupting this DNA methylation-sensing circuit, for example, by restoring ROS1 expression to wild-type levels in the rdr2 mutant through remethylation of the endogenous ROS1 promoter, results in widespread DNA hypomethylation and developmental defects, which progressively worsen over generations [87]. Additionally, the expression of ROS1 can be inhibited by certain regulatory factors. DNA DAMAGE-BINDING PROTEIN 2 (DDB2) forms a complex with ROS1 and AGO4, which regulates the DNA methylation level at the ROS1 locus and suppresses its transcription [88].

In addition to transcriptional regulation, the protein stability and enzymatic activity of the 5mC DNA glycosylases are tightly regulated in plants. The catalytic activity of DME is determined by its C terminus, while its N terminus is likely responsible for recruiting chromatin remodeling factors that facilitate DNA demethylation in heterochromatin [89]. The N-terminal region is unique to flowering plants [89]. Studies show that the C-terminal domain of ROS1 may function as a reader of the N-terminal tail of histone H3, enabling its recruitment to specific chromatin loci [90]. Furthermore, the conserved amino acid residues within the C-terminal domain are essential for ROS1’s DNA binding ability and catalytic activity [90]. Additionally, the small ubiquitin-like modifier (SUMO) E3 ligase SIZ1 interacts with ROS1 and promotes its stability through SUMOylation [91]. Lastly, ROS1 contains an iron-sulfur (Fe–S) cluster binding motif. The conserved Fe–S cluster assembly protein MET18 interacts with this motif to ensure proper ROS1 function in DNA demethylation [92].

The recruitment mechanisms of DNA glycosylase

DME preferentially binds to small, AT-rich transposons in euchromatic regions, and promotes the expression of nearby genes [93]. While some euchromatic targets are easily accessible, demethylation of the inaccessible targets requires chromatin remodeling by the FACILITATES CHROMATIN TRANSACTIONS (FACT) complex [94]. The FACT complex, composed of two subunits—STRUCTURE SPECIFIC RECOGNITION PROTEIN 1 (SSRP1) and SUPPRESSOR of TY16 (SPT16)—is highly conserved in eukaryotes [95–97]. It interacts with the nucleosome to facilitate transcription initiation and elongation and also possesses chromatin remodeling activity, enabling DME access to heterochromatin [94, 98–100].

ROS1, on the other hand, primarily targets TEs and intergenic regions. TEs targeted by ROS1 are often located near protein-coding genes, suggesting that ROS1 may help establish boundaries between TEs and genes by preventing the spread of DNA methylation from TEs to adjacent genic regions [101]. Many of the TEs targeted by ROS1 are enriched in H3K18Ac and H3K27me3, and are depleted of H3K27me1 and H3K9me2 [92]. Several studies focused on the recruitment mechanism of ROS1, leading to the identification of the INCREASED DNA METHYLATION (IDM) complex (Fig. 2). This complex consists of IDM1, a histone acetyltransferase that binds to DNA methylated regions lacking H3K4me2/me3 [102]; IDM2, a member of the highly conserved α-crystallin domain (ACD) family [103]; IDM3 and MBD7, anti-silencing factors, with MBD7 specifically recognizing CG-rich, hypermethylated chromatin [104]; HARBINGER TRANSPOSON-DERIVED PROTEINS 1 (HDP1) and HDP2, derived from the Harbinger transposon [105]. This protein complex is crucial for recognizing DNA methylated regions and marking the chromatin with histone acetylation. MBD9 and NUCLEAR PROTEIN X1 (NPX1), a protein homologous to yeast BROMODOMAIN FACTOR 1 (ScBDF1), subsequently recognize acetylated histones and recruit the SWR1 complex [106]. SWR1, in turn deposits H2A.Z at the target sites of IDM1 [106, 107] (Fig. 2). Finally, H2A.Z interacts with ROS1 to initiate the process of active DNA demethylation [106]. Recent studies have shown that H3K4me3 can also participate in DNA demethylation by recruiting the DNA demethylase ROS1 [108] (Fig. 2).

Fig. 2.

Recruitment of ROS1 by the IDM complex. IDM1, IDM2, IDM3, MBD7, and HDP1/2 form a histone acetyltransferase complex and bind to the DNA methylation region. Acting as a histone acetyltransferase, IDM1 catalyzes the acetylation of histone H3 lysine 18 (H3K18Ac) and lysine 23 (H3K23Ac) [101]. Subsequently, MBD9 and NPX1 recognize these histone acetylation marks, then recruit the SWR1 complex to deposit H2A.Z [106]. Finally, H2A.Z recruits ROS1 to the target sites through direct interaction. The deposition of H3K4me3 promotes the targeting of H2A.Z and ROS1 [108]

DNA demethylation in plant sexual reproduction

In animals, germ cells are specified early during development, whereas in plants, they differentiate later from somatic cells. In Arabidopsis, microspores and megaspores undergo multiple rounds of division to give rise to the male and female gametophytes, respectively. During double fertilization, one sperm cell fuses with the egg cell to form the zygote, while the other one fuses with the central cell to form the endosperm. Finally, the zygote develops into the embryo, while the endosperm supplies nutrients to support embryo development. Active DNA demethylation plays a crucial role in these processes.

The erasure and re-establishment of DNA methylation across generations is known as DNA methylation reprogramming [109]. In mammals, this process occurs in the germline and fertilized zygotes and is critical for the erasure of parental imprints and the restoration of pluripotency [110–113]. In plants, DNA methylation reprogramming occurs during gametogenesis. Genome-wide methylation analyses revealed dynamic changes in DNA methylation in both sperm and vegetative cells [93, 114]. While CG demethylation occurs in vegetative cells [114], CHH methylation levels decrease significantly in microspores and sperm cells, likely due to passive demethylation resulting from defects in methylation maintenance during meiosis [114].

CG demethylation in vegetative cells is primarily mediated by DME [29]. In somatic nuclei, heterochromatin is highly condensed, and heterochromatic TEs are silent. However, in vegetative cells, heterochromatin decondenses, likely due to the absence of the chromatin remodeler DDM1 and linker histone H1 [115, 116]. This facilitates DME binding to heterochromatin regions, activating the expression of TEs and genes [40]. DME-mediated DNA demethylation is essential for male gametophyte fertility and genome stability [29, 115, 117]. In the dme mutant, pollen tube guidance is impaired, and sperm cells fail to reach the female gametophyte, reducing the male gamete transmission rate [29]. The widespread CG demethylation in vegetative cells is accompanied by the reactivation of TEs, which are processed by miRNAs and AGO proteins to generate 21–22 nt small RNAs (easiRNAs) [114, 118] (Fig. 3A). Following fertilization, the zygote undergoes DNA methylation reprogramming during early embryogenesis which is mediated by methyltransferases. The embryo exhibits significantly higher levels of DNA methylation compared to the endosperm [119]. Previous studies have suggested that, the high level of de novo methylation, at least immediately after fertilization, contributes to the maintenance of genome integrity in the embryo [109, 120, 121]. Additionally, easiRNAs can reflect chromosome number and contribute to the establishment of hybridization barriers, as demonstrated in studies on hybridization between parents with different ploidy levels [122].

Fig. 3.

Classical functions of active DNA demethylation in plants. A DNA demethylation during plant sexual reproduction. In the flower of Arabidopsis thaliana, an asymmetric division of the microspore give rise to a vegetative cell and a generative cell. The generative cell subsequently divides into two sperm cells. In the vegetative nucleus, DME activates TEs. This activation of TEs promotes the generation of siRNAs, which can move into the sperm cell, thereby safeguarding genome integrity [115, 117]. B DNA demethylation during fruit ripening. During tomato ripening, as the transcriptional level of SlDML2 increases, the DNA methylation level gradually decreases. The transcription factor RIN regulates tomato ripening, and demethylation of the promoter regions of its target genes is crucial for its binding [134]. C DNA demethylation during mycorrhizal symbiosis. Rhizobia establish a symbiotic relationship with Medicago truncatula, which leads to the development of nodules. MtDME medicated active DNA demethylation is crucial for nodule development [173]

Imprinted genes, in which the two alleles are differentially expressed in a parent-of-origin-specific manner, are also regulated by DNA demethylation. DNA demethylation is crucial in the central cells before fertilization, leading to a lower DNA methylation level compared to the sperm cell [123, 124]. This demethylation is essential for the expression of maternally expressed genes (MEGs) in the endosperm. After fertilization, the endosperm MEGs are activated through DNA demethylation, while paternal alleles are silenced through DNA methylation or H3K27me3 modifications [125, 126]. In Arabidopsis, DME regulates the expression of MEGs such as MEA and FWA. In the dme mutant, the failure to express MEGs leads to embryo abortion [126]. In maize (Zea mays), loss of ZmROS1ab results in increased DNA methylation at thousands of genomic loci, affecting proper expression of both endosperm-specific and imprinted genes [127]. In contrast, paternally expressed genes (PEGs) in maize are decorated by H3K27me3 on their maternal alleles and H3K36me3 on their paternal alleles in the endosperm [128].

DNA demethylation not only influences embryo and endosperm development through imprinting but also regulates transcription factors involved in seed development. In rice (Oryza sativa), the endosperm consists of the outer aleurone layer and the inner starchy endosperm. The aleurone layer contains proteins, vitamins, and minerals, while the starchy endosperm primarily serves as a storage site. OsROS1-mediated DNA demethylation regulates the transcription factors associated with aleurone layer differentiation and limits the number of aleurone layer cells in rice grains [129]. In the thick aleurone 2–1 (ta2-1) mutant, a point mutation in the 14th intron of OsROS1 induces alternative splicing of OsROS1. Consequently, this leads to the hypermethylation of genes associated with aleurone layer differentiation, ultimately altering the number of aleurone layer cells in rice grains [129]. In addition, downregulation of DNA demethylases, such as ROS1 and DME, has been implicated in the regulation of seed size in peanut (Arachis hypogaea L.) [130]. In soybean (Glycine max [L.] Merr.), GmDMEa has also been identified as a negative regulator of seed size. It acts through demethylation of AT-rich regions and subsequent activation of abscisic acid (ABA) signaling genes [131]. Collectively, these studies suggest that targeted editing of DNA demethylases may offer a promising strategy to improve the seed size and nutritional value in crops.

DNA demethylation during fruit ripening and senescence

Fruit ripening is a complex process in which DNA demethylation plays a critical role. Tomato (Solanum lycopersicum) is used as a model for studying fruit ripening due to the ease of genetic transformation and the availability of high-quality reference genomes. Studies showed that the COLORLESS NON-RIPENING (CNR) triggers DNA demethylation in the promoter regions of the transcription factor SQUAMOSA PROMOTER BINDING PROTEIN (SBP). Mutation in CNR leads to DNA hypermethylation and inhibition of SBP expression [132]. RIPENING-INHIBITOR (RIN), another transcription factor, also plays an important role in tomato fruit ripening [133, 134]. Reduced DNA methylation facilitates the binding of RIN to its target gene promoters and subsequently enhances gene expression [134] (Fig. 3B). In the sldml2 mutant, the tomato homologue of Arabidopsis ROS1, 29,764 hypermethylated regions (hyper-DMRs) were identified, and 605 of these hyper-DMRs were associated with fruit ripening genes, including transcription factors [135]. Consistently, expression of these genes is downregulated in the sldml2 mutant [135]. Mechanisms regulating the expression of SlDML2 have been identified in tomatoes. For instance, HIGH MOBILITY GROUP A 3 (SlHMGA3) binds to the promoter and activates the expression of SlDML2 [136]. Additionally, the m⁶A demethylase ALKB HOMOLOG 2 (SlALKBH2) mediates m⁶A demethylation of SlDML2 mRNA, enhancing its stability and thus promotes fruit ripening [136, 137]. Recent studies show that SlALKBH2 is sensitive to oxidative modification by H₂O₂, with Cys39 identified as the key residue in the process [138]. Oxidation of SlALKBH2 enhances its protein stability and increases its activity on target transcripts, such as SlDML2 [138]. Consistently, application of H₂O₂ has been shown to significantly accelerate tomato fruit ripening [138]. JUMONJI C DOMAIN-CONTAINING PROTEIN 7 (SlJMJ7), a histone H3K4 demethylase, functions as a negative regulator in fruit ripening. In the sljmj7 mutant, expression of SlDML2 is upregulated, leading to DNA hypomethylation and accelerated fruit ripening [139]. Moreover, DNA demethylation plays a crucial role in the removal of toxic and bitter steroidal glycoalkaloids (SGAs) during tomato fruit ripening [140]. In particular, DML2 regulates the expression of GLYCOALKALOID METABOLISM (GAME) genes, including GAME31, GAME40, and GAME5, which are involved in the metabolic conversion of α-tomatine into the nontoxic esculeoside A [140]. Similarly, DNA demethylation during fruit ripening has been observed in peaches (Prunus persica) and apples (Malus domestica). In peaches, the transcription factor PpNAC1 activates PpDML1, facilitating DNA demethylation and the expression of ripening-associated genes [141, 142]. In apples, reduced methylation in the promoter region of AMINOCYCLOPROPANE-1-CARBOXYLATE SYNTHASE 3A (MdACS3a) has been reported during ripening [143].

DNA demethylation also plays a critical role in plant senescence. In Arabidopsis shoot tissue, a global decrease in DNA methylation is observed during the aging process. This is accompanied by the upregulation of key DNA demethylases, including ROS1, DME, and DML2/3 [144]. Several studies have also investigated the role of DNA demethylation in leaf senescence. For example, DML3 has been shown to promote leaf senescence by demethylating the promoters of senescence-associated genes [145]. Vatov et al. [146] proposed that the downregulation of DNA methylation levels during the early stages of leaf senescence may be caused by the inhibition of cytosine methylation maintenance.

In summary, these studies underscore the conserved role of DNA demethylation in the regulation of fruit ripening and senescence in plants.

DNA demethylation during vegetative growth of plants

Stomata on the plant epidermis function in gas exchange between plants and the external environment. Both the distribution and density of stomata are tightly regulated [147]. The small peptides EPIDERMAL PATTERNING FACTOR 1/2 (EPF1/2) and STOMAGEN are key regulators of stomatal density. While EPF1/2 inhibit stomatal development, STOMAGEN promotes it [148–152]. Recent studies indicate that the expression of EPF2 is regulated by DNA demethylation. In both ros1 and dme mutants, decreased expression of EPF2 results in an overabundance of stomatal lineage cells [81, 153]. In addition to stomatal development, DNA demethylation is important in xylem development. During the differentiation of xylem tracheary elements, DNA demethylation occurs predominantly in the CHH context near centromeres, alongside CG and CHG demethylation in other genomic regions [154]. In ros1 and rdd mutants, differentiation of tracheary elements is disrupted, leading to an increased frequency of protoxylem discontinuities in roots [154]. Due to the lethality of most homozygous dme mutations, study on DME in somatic cells remains limited. However, recent studies have identified a non-lethal homozygous dme mutant allele, revealing that DME regulates key biological processes such as seed germination, root hair growth, and cellular proliferation and differentiation.

Importantly, DNA demethylation also influences flowering time. In the drdd mutant, hypermethylation at the FLOWERING LOCUS C (FLC) locus leads to transcriptional repression and early flowering [31]. In Chrysanthemum lavandulifolium, CIROS1 is highly expressed in leaves and floral tissues. Overexpression of CIROS1 in Arabidopsis thaliana leads to decreased methylation at the CONSTANS (AtCO) promoter and early flowering [155].

Altogether, these findings demonstrate that DME is essential for proper sporophyte development in Arabidopsis thaliana [81, 82].

DNA demethylation in stress response

Recent studies suggest that epigenetic modifications, including DNA demethylation, play a significant role in plant stress adaptation. Under heat stress, DNA methylation levels in Arabidopsis thaliana decrease. The majority of the demethylated sites correspond to heat responsive genes including heat shock proteins and translation regulation factors [156]. During vegetative growth, the expression of certain imprinted genes, such as SUPPRESSOR OF drm1 drm2 cmt3 (SDC), is repressed by DNA methylation [157]. However, under heat stress, DNA demethylation activates the silenced SDC gene, facilitating stress recovery [157]. Similarly, under low-temperature conditions, ROS1 promotes the expression of ACCELERATED CELL DEATH 6 (ACD6) and ACONITATE HYDRATASE 3 (ACO3), key regulators of salicylic acid (SA) and ABA pathways, respectively, through promoter demethylation [158]. Furthermore, NICOTINAMIDASE 3 (NIC3), which is involved in ABA responses by converting nicotinamide to nicotinic acid in the NAD⁺ salvage pathway, requires ROS1 dependent promoter demethylation to be activated [159]. Consistently, ros1 mutants exhibit ABA-sensitive phenotypes [159]. Besides, DNA methylation regulates cold stress responses. For instance, increase in DNA methylation enhances the stability of chromatin structure, thereby facilitating the recruitment of C-REPEAT BINDING FACTOR 1 (CBF1) [160]. Similarly, DNA methylation levels also increase in response to salt and drought stresses [161–164].

Under biotic stresses, such as bacterial, fungal, or viral attacks, DNA methylation is remodeled to enhance defense responses. Recent studies showed that ROS1 expression is upregulated in Arabidopsis upon infection with beet severe curly top virus (BSCTV) [165]. The ROS1-mediated DNA demethylation activates stress-responsive genes, thereby enhancing adaptability to biotic stress [165]. Consistently, defects in DNA demethylases can render plants more susceptible to pathogens. For instance, loss of the demethylase activity of ROS1 renders plants more vulnerable to Pseudomonas syringae DC3000 (Pst DC3000) and the biotrophic pathogen Hyaloperonospora arabidopsidis [166, 167]. Knocking out of DME renders the rdd mutant more susceptible to Fusarium oxysporum [168]. Flg22 treatment also alters DNA methylation patterns, partly through ROS1 and DML2/3, to enhance disease resistance. Mutants of these DNA demethylases failed to initiate resistance in response to flg22 [169]. Furthermore, expression of key immune receptors, including RESISTANCE METHYLATED GENE 1 (RMG1) and RECEPTOR-LIKE PROTEIN 43 (RLP43) can be regulated at DNA methylation levels [166, 170]. In the rdd mutant, more than 200 genes are down-regulated, including biotic stress responsive genes, rendering plants more susceptible to the fungal pathogen Hyaloperonospora arabidopsidis [171]. Interestingly, some viruses exploit host plant DNA demethylases to subvert the defense system. For example, the βC1 protein encoded by the betasatellite of tomato yellow leaf curl China virus (TYLCCNV) interacts with DME in Arabidopsis thaliana and NbROS1L in Nicotiana benthamiana [172]. By hijacking the host demethylation machinery, βC1 facilitates the demethylation of the viral DNA, thereby promoting the virulence [172].

DNA demethylation in symbiosis between plants and microorganisms

In Medicago truncatula, symbiosis with rhizobia induces root nodule formation (Fig. 3C), a process involving proliferation and differentiation of both plant and bacterial cells. DNA methylation patterns undergo significant changes during root nodule development, with MtDME being highly expressed in the late differentiation zone of nodules [173] (Fig. 3C). Among the 1,425 genes regulated by MtDME, many are involved in cell differentiation [173]. Reduced MtDME expression leads to the downregulation of 400 genes, many of which are involved in nodule differentiation. This results in smaller nodules and decreased nitrogen fixation efficiency [173]. Recent studies showed that the expression of DME is also increased during the early stages of nodule formation, particularly when the nodule begins to emerge from the root [174]. These findings suggest that DNA demethylation plays a crucial role in the early stages of nodule development [174, 175].

Similarly, DNA demethylation is essential for the colonization of Bacillus megaterium strain YC4 in Arabidopsis and tomato roots. Studies showed that YC4 triggers myo-inositol secretion in Arabidopsis roots, which facilitates bacterial colonization [176]. ROS1 mediates the production of myo-inositol by regulating the expression of genes involved in its biosynthesis, such as myo-inositol-1-phosphate synthase (MIPS) genes, as well as the transcription factors FAR-RED IMPAIRED RESPONSE1 (FAR1) and FAR-RED ELONGATED HYPOCOTYL3 (FHY3). This regulation promotes plant growth and YC4 colonization. Additionally, DNA demethylation regulates the inositol-mediated interaction between YC4 and tomato [176]. The endophytic fungus Epichloë sp. LpTG-3 strain AR37 interacts with Lolium perenne, triggering DNA demethylation in the host plant [177]. Notably, the duration of the endophyte-plant association is also influenced by the global DNA hypomethylation [177]. Other examples include the plant growth-promoting bacteria (PGPB) Bacillus sp. (PGP5) and Arthrobacter sp. (PGP41) [178]. Additionally, the absence of DNA methylation can affect the symbiotic relationship between trees and mutualistic ectomycorrhizal fungi, further underscoring the importance of epigenetic regulation in plant–microbe interactions [179].

Conclusion and perspective

Significant progress has been made in understanding the molecular mechanisms of active DNA demethylation in plants [180]. Except for DNA polymerase, components of the BER pathway have been identified and their biological functions have been elucidated. These findings suggest that active DNA demethylation resembles a specific form of DNA damage repair. Critically, growing evidence highlights a close and bidirectional relationship between DNA methylation and DNA damage. In Arabidopsis thaliana, loss of DNA methylation results in the accumulation of DNA damage, ultimately impairing plant growth and development [181]. Conversely, DNA damage can trigger the establishment of DNA methylation through the DNA damage response, with a subset of these damage-induced methylation marks being heritable [182]. These findings underscore the intimate relationship between epigenetic regulation and genome integrity. The mechanisms underlying the site-specific recruitment of DNA demethylases have remained a central question in the field of epigenetics. The IDM complex is capable of recruiting ROS1 to target sites via histone acetylation [102, 104]. However, this complex recruits ROS1 to only a subset of target sites, indicating the existence of additional mechanisms. Moreover, the recruitment mechanisms of DME remain poorly characterized. A promising strategy to elucidate these processes involves the systematic identification of ROS1- and DME-interacting proteins through integrated genetic and biochemical approaches.

The rapid development of high-throughput sequencing technologies and precise genome-editing tools has led to the comprehensive identification of genomic loci regulated by active DNA demethylation pathways [12, 29, 31, 154]. These studies have expanded our understanding of the biological functions of active DNA demethylation in plants. Notably, DNA demethylase mutants have contributed substantially to our functional understanding, as the resulting hypermethylation at target genomic loci frequently leads to transcriptional silencing, ultimately giving rise to developmental defects and compromised stress adaptation [153, 158, 159, 168–170, 183].

While the majority of research on active DNA demethylation has been conducted in Arabidopsis thaliana, the genomic architecture of this model organism presents certain limitations for comprehensive epigenetic studies. The Arabidopsis genome is characterized by a relatively simple organization, with TEs predominantly localized in centromeric regions and protein-coding genes largely concentrated on chromosome arms [184]. Therefore, Arabidopsis may not represent an ideal model for studying the biological functions of DNA methylation and demethylation. In light of these limitations, and with the rapid advancement of genome-editing and sequencing technologies, there is increasing need to understand active DNA demethylation in plants with complex genomes and traits. For instance, crops such as tomato provides an ideal system for exploring the dynamics of DNA demethylation during fruit ripening, while rice and maize represent more appropriate models for seed development. Comparative studies across diversified plants are also required to uncover both conserved and species-specific regulatory paradigms.

In the past decade, we have witnessed remarkable progress in the development of precise epigenetic editing technologies. These cutting-edge tools, including CRISPR-based targeted modulation of DNA methylation, have transformed the field by enabling locus-specific epigenetic manipulation with precision. These technology revolutions facilitated the transition from correlative descriptions between a given epigenetic feature and phenotypic consequences, to causative and mechanistic understandings. Specifically, through fusion of the catalytic domains from DNA (de)methylation enzymes with ZF proteins, researchers have achieved manipulation of DNA methylation at a subset of genomic loci. Alternatively, the SunTag system, an innovative CRISPR/dCas9-based platform, enables simultaneous recruitment of multiple epigenetic modifiers to specific target sites. The SunTag architecture, characterized by its array of peptide epitopes, facilitates the assembly of multifunctional effector complexes, thereby allowing for more robust and coordinated epigenetic remodeling at designated genomic regions. [185]. The CRISPR/Cas9 system has been adopted to insert repetitive sequences near target sites, potentially spreading the silencing effect of DNA methylation to adjacent regions [186, 187]. These epigenetic editing tools have been applied in crop improvement. For example, the Xanthomonas phaseoli pv. manihotis (Xam)-secreted transcription activator-like protein TAL20 binds to the Effector Binding Element (EBE) of the susceptible gene Sugars Will Eventually be Exported Transporters10a (MeSWEET10a) in cassava (Manihot esculenta) and activates its expression [188]. ZF-DMS3-targeted methylation of the EBE site prevents TAL20 from binding [188]. Additionally, ZF108 fused with the catalytic domain of human ten-eleven translocation (TET1cd) has been shown to target DNA demethylation with high specificity and efficiency [34, 189]. Moving forward, the development and application of epigenetic editing tools with higher specificity and efficiency will facilitate a deeper understanding of the mechanisms underlying DNA demethylation. Furthermore, tools such as ZF and CRISPR/dCas9 hold the potential to improve crop quality and enhance resistance to biotic and abiotic stresses through epigenetic editing. A major constraint in the application of genome editing technologies in crops lies in their low transformation efficiency. Interestingly, during shoot regeneration, mutation of the DNA demethylase led to the redistribution of genic CG methylation, widespread non-CG hypomethylation in pericentromeric regions and enhanced regenerative potential [190].

Therefore, understanding of the DNA demethylation regulatory and functional mechanisms holds promise for the future targeted modulation of DNA methylation for crop species.

Author contributions

R.Z., Y.X., and W.Q. designed and wrote the manuscript. R.Z. prepared the figures.

Funding

This working was supported by the National Natural Science Foundation of China (No. 32270288 to W.Q.).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.