SHH1 cooperates with the DNA methylation reader MBD7 to suppress transcriptional silencing of promoter-methylated genes in Arabidopsis

Abstract

In Arabidopsis thaliana, METHYL-CpG-BINDING DOMAIN 7 (MBD7) and its associated α-crystallin domain (ACD) proteins form a complex that interprets DNA methylation to prevent the silencing of methylated luciferase (LUC) reporter transgenes. However, the mechanism by which the MBD7 complex effectively targets methylated transgenes remains largely unclear. Here, we identify a novel role for SAWADEE HOMEODOMAIN HOMOLOG 1 (SHH1), extending its function beyond the canonical RNA-directed DNA methylation (RdDM) pathway. We demonstrate that SHH1 prevents the transcriptional silencing of methylated LUC transgenes and a subset of endogenous genes by acting in concert with MBD7 within the same regulatory pathway. SHH1 co-localizes with MBD7 at nuclear foci and physically interacts with it to enhance its stability. Furthermore, SHH1 binds to methylated loci via its SAWADEE domain, which recognizes the H3K9me2 histone mark. This interaction promotes the reciprocal recruitment of SHH1 and MBD7 to methylated loci, revealing a cooperative mechanism that maintains transcriptional activity at promoter-methylated genes. Collectively, our findings unveil a dynamic, mutually reinforcing SHH1–MBD7 module that enhances the expression of promoter-methylated genes, likely by facilitating effective binding to chromatin marked by repressive epigenetic modifications. This work provides important insights into how DNA methylation fine-tunes gene expression in plants by balancing between transcriptional repression and activation.

This study demonstrates that SHH1 prevents the transcriptional silencing of promoter-methylated genes by physically interacting with MBD7. Through its SAWADEE domain, SHH1 recognizes the H3K9me2 histone mark and binds to methylated loci. This interaction facilitates the reciprocal recruitment of SHH1 and MBD7 to these loci, revealing a cooperative mechanism that maintains transcriptional activity at promoter-methylated genes.

Introduction

DNA methylation is a highly conserved chromatin modification that contributes to transcriptional gene silencing and genome stability in eukaryotes (Zhang et al., 2018). In plants, DNA methylation occurs predominantly at transposable elements (TEs) and repetitive sequences, causing transcriptional silencing of the TEs, repeats, or nearby genes (Cokus et al., 2008; Lister et al., 2008). Genome-wide DNA methylation patterns are dynamically regulated through de novo methylation, maintenance of methylation, and active demethylation (Zhang et al., 2018).

The de novo establishment of DNA methylation in plants is primarily mediated by RNA-directed DNA methylation (RdDM), which is guided by small interfering RNAs (siRNAs) and long non-coding RNAs (Zhang et al., 2018; Erdmann and Picard, 2020). In this process, DNA-DEPENDENT RNA POLYMERASE IV (Pol IV), RNA-DEPENDENT RNA POLYMERASE 2 (RDR2), and DICER-LIKE 3 (DCL3) act sequentially to generate siRNAs, which are mainly incorporated into ARGONAUTE 4 (AGO4) (Huang et al., 2021; Sigman et al., 2021; Wang et al., 2021, 2023; Loffer et al., 2022). At the same time, Pol V synthesizes long non-coding RNAs that serve as scaffolds to recruit AGO4–siRNA complexes to target loci through sequence complementarity (Lahmy et al., 2016; Sigman et al., 2021; Wang et al., 2023). The AGO4–siRNA complex then recruits the DNA methyltransferases DOMAINS REARRANGED METHYLTRANSFERASE 1/2 (DRM1/2), which catalyze de novo methylation of nearby sequences and establish transcriptional gene silencing (Zhong et al., 2014).

Once established, CG methylation is maintained by DNA METHYLTRANSFERASE 1 (MET1), whereas CHG methylation is maintained by CHROMOMETHYLASE 3 (CMT3) or CMT2. This process relies on a positive feedback loop between methylated CHG and H3K9me2 histone modifications (Zhang et al., 2018; Fang et al., 2022; Jiang et al., 2024). Maintenance of CHH methylation depends on two distinct DNA methyltransferases, DRM2 and CMT2. CHH methylation in short euchromatic TEs is maintained by DRM2 through the continuous action of the RdDM pathway, whereas in long heterochromatic TEs it is maintained by CMT2 (Zemach et al., 2013; Zhong et al., 2014).

In contrast to methylation, active DNA demethylation removes methyl groups to regulate cytosine methylation levels. In Arabidopsis thaliana (A. thaliana), this process is mediated by REPRESSOR OF SILENCING 1 (ROS1) and its three paralogs, DEMETER (DME), DEMETER-LIKE 2 (DML2), and DML3 (Zhang et al., 2022; Du et al., 2023). DME functions primarily during plant reproduction, whereas ROS1, DML2, and DML3 act during vegetative development (Liu and Lang, 2020).

The Arabidopsis genome encodes two SAWADEE HOMEODOMAIN HOMOLOG proteins, SHH1/DNA-BINDING TRANSCRIPTION FACTOR 1 (DTF1) and SHH2/DTF2, both containing a homeodomain at the N terminus and a plant-specific SAWADEE domain at the C terminus (Law et al., 2011; Zhang et al., 2013). The SAWADEE domain serves as a chromatin-binding module that functions as a reader of methylated H3K9 (Law et al., 2013; Zhang et al., 2013). Recent studies have identified SHH2 as a subunit of SWI/SNF complexes in Arabidopsis, although its precise role remains unclear (Guo et al., 2022). In contrast, the function of SHH1 has been extensively characterized. SHH1 acts upstream in the RdDM pathway by physically interacting with Pol IV, RNA-DEPENDENT RNA POLYMERASE 2 (RDR2), and the putative chromatin remodeling factors CLASSY1 (CLSY1) and CLSY2 (Law et al., 2013). It facilitates the recruitment of Pol IV to active RdDM loci, promoting the production of 24-nt siRNAs and thereby contributing to both de novo and maintenance methylation (Law et al., 2013). Additionally, SHH1 is essential for transcriptional silencing at a subset of RdDM loci, independent of DNA methylation (Zhang et al., 2013). While SHH1’s role in RdDM is well established, its possible involvement in suppressing transcriptional gene silencing remains unclear.

Although DNA methylation typically represses gene expression, its presence in gene promoter regions does not necessarily result in transcriptional silencing (Li et al., 2016). This observation suggests the existence of regulatory mechanisms that counteract the repressive effects of this modification. One such mechanism involves the DNA methylation–binding proteins SUVH1 and SUVH3 and the transcriptional activators DnaJ1 and DnaJ2, which together promote the expression of methylated genes (Li et al., 2017; Harris et al., 2018). Another key mechanism is mediated by the MBD7–ACD complex identified in our previous study. This complex contains METHYL-CpG-BINDING DOMAIN 7 (MBD7), which interacts with two α-crystallin domain (ACD) proteins, LOW IN LUCIFERASE EXPRESSION (LIL) and REPRESSOR OF SILENCING 5 (ROS5). Together with ROS4, these proteins form a DNA methylation reader complex that promotes the expression of methylated luciferase (LUC) reporter transgenes without significantly altering DNA methylation levels (Li et al., 2017). However, the precise mechanism by which the MBD7–ACD complex is targeted to methylated sites remains unclear. Because DNA methylation is often co-localized with the repressive histone mark H3K9me2, which limits DNA accessibility (Cedar and Bergman, 2009; Stroud et al., 2014; Du et al., 2015), it is unclear how the MBD7–ACD complex gains access to such chromatin to activate gene expression. In this study, we identify SHH1, a plant-specific protein containing a SAWADEE domain, as a new component of the MBD7–ACD complex through a forward genetic screen. Our findings demonstrate that SHH1 physically associates with MBD7 and plays a previously unrecognized role in suppressing gene silencing at methylated LUC reporter transgenes and a subset of promoter-methylated endogenous genes. Genetic analyses confirm that SHH1 acts within the same pathway as MBD7. Moreover, SHH1 interacts with and co-localizes with MBD7 in vivo. Notably, SHH1 binds to methylated loci via its SAWADEE domain, promoting MBD7 binding at the same sites, and vice versa, suggesting a cooperative interaction between the two proteins. We propose that SHH1 and MBD7 form a mutually reinforcing binding module that facilitates the expression of promoter-methylated genes, likely by enabling more efficient access to chromatin marked by repressive epigenetic modifications.

Results

SHH1 is required for suppressing transcriptional silencing of the YJ reporter

To identify novel factors involved in suppressing DNA methylation-mediated gene silencing, we conducted a forward genetic screen using an ethyl methanesulfonate-mutagenized population of the YJ reporter line. This reporter line was introduced into the rdr6-11 mutant background to bypass post-transcriptional silencing (Li et al., 2016). In this system, LUC expression is driven by a dual cauliflower mosaic virus 35S promoter (d35S), which is prone to epigenetic regulation due to the presence of both DNA methylation and siRNAs. Using this reporter, several anti-silencing factors have been identified, including SUVH1, MBD7, LIL, and SAC3B (Li et al., 2016, 2017; Yang et al., 2017). In this study, we isolated a new recessive mutant that exhibited undetectable luminescence (Figure 1A). RT-qPCR analysis confirmed a strong reduction in LUC expression in this mutant (Figure 1B). Map-based cloning and whole-genome resequencing identified a G-to-A mutation resulting in a premature stop codon at Tyr-60 in SHH1/DTF1 (At1g15215). This mutant was designated YJshh1-2 (Figure 1C). To confirm that the SHH1 mutation was responsible for the loss of luminescence, genomic SHH1 fragments driven by the native promoter and fused to eGFP, 3×FLAG, or 3×MYC were introduced into the YJshh1-2 mutant. All three constructs fully rescued the mutant phenotype (Figure 1D and 1E), demonstrating that SHH1 is essential for LUC expression in the YJ reporter system. Given that SHH2 is a homolog of SHH1 (Law et al., 2011), we next examined whether SHH2 has a similar regulatory function. To this end, we generated shh2 mutants in the YJ background using CRISPR–Cas9 and analyzed their luminescence phenotype and LUC expression. However, shh2 mutations did not reproduce the phenotype observed in YJshh1-2 (Supplemental Figure 1), indicating a functional distinction between SHH1 and SHH2.

Figure 1.

Characterization of shh1 mutants.

(A) Luciferase (LUC) luminescence in 6-day-old YJ and YJshh1-2 seedlings.

(B) RT–qPCR quantification of LUC transcript levels in YJ and YJshh1-2 seedlings.

(C) Diagram of the SHH1 gene structure showing the shh1-2 mutation. Boxes and lines denote exons and introns, respectively.

(D) LUC luminescence of YJ, YJshh1-2, and YJshh1-2 complemented with SHH1 fused to eGFP, 3×FLAG, or 3×MYC under the native promoter.

(E) RT–qPCR quantification of LUC transcript levels in YJ, YJshh1-2, and complementation lines.

(F) LUC luminescence of YJ and YJshh1-2 seedlings with mock or 5-aza-2′-deoxycytidine (5-Aza-dC) treatment.(G) RT–qPCR quantification of LUC transcript expression levels in YJ and YJshh1-2 with mock or 5-Aza-dC treatment.

Error bars indicate standard deviation (SD) from three technical replicates; statistical significance was determined by Student’s t-test. ∗p < 0.05; ns, not significant. Results were reproduced in three biological replicates.

To determine whether SHH1 regulates LUC expression via the DNA methylation pathway, both YJshh1-2 and wild-type YJ plants were treated with the cytosine methylation inhibitor 5-aza-2′-deoxycytidine (5-Aza-dC), as previously described (Li et al., 2017). Luminescence imaging and RT-qPCR analysis showed that the reduced LUC expression in YJshh1-2 was fully restored after 5-Aza-dC treatment (Figure 1F and 1G). This result indicates that the removal of DNA methylation from the YJ transgene reverses the molecular phenotype of the YJshh1-2 mutant. Therefore, SHH1 regulates LUC expression through the DNA methylation pathway, highlighting its role in transcriptional regulation.

SHH1 dysfunction leads to decreased CHH methylation and H3K9me2 levels at the YJ reporter

To determine whether the reduced LUC expression observed in the YJshh1-2 mutant is associated with changes in DNA methylation at the d35S promoter, we analyzed DNA methylation levels by whole-genome bisulfite sequencing. The results revealed a slight overall reduction in DNA methylation at the d35S promoter in YJshh1-2 compared with YJ control plants (Figure 2A). This reduction was most pronounced at asymmetric CHH sites, where methylation levels decreased by approximately 40% (Figure 2A and 2B). In contrast, CG methylation at the d35S promoter showed a slight increase, whereas CHG methylation remained largely unchanged, likely due to reduced ROS1 expression in the shh1 mutant (Zhang et al., 2013) (Figure 2A and 2B and Supplemental Figure 2). To further assess the effect of the YJshh1-2 mutation on DNA methylation, we performed a PCR-based methylation assay using McrBC, a restriction enzyme that specifically digests methylated DNA. Consistent with the bisulfite sequencing results, no significant difference in d35S promoter amplification was observed between YJshh1-2 and YJ plants (Figure 2C), indicating that overall methylation levels were only minimally affected in the mutant.

Figure 2.

Effect of the SHH1 mutation on DNA methylation and H3K9me2 levels at the d35S promoter.

(A and B) CG, CHG, and CHH methylation levels at the d35S promoter in YJ, YJshh1-2, and SHH1-3×MYC/YJshh1-2 by whole-genome bisulfite sequencing (A), with distribution patterns shown in a genome browser view (B).

(C) DNA methylation analysis by McrBC-PCR using genomic DNA with (+) or without (mock) McrBC digestion. Unmethylated Actin served as the loading control.

(D) ChIP–qPCR analysis of histone H3K9 dimethylation (H3K9me2) levels at the d35S promoter. Signals are shown relative to input DNA. The promoter of Tubulin 8 (TUB8), which has a low H3K9me2 level, was used as a non-specific control region. The RdDM target retrotransposon AtSN1 served as a positive control. Error bars, SD of three technical replicates; statistical significance was determined by Student’s t-test. ∗p < 0.05; ns, not significant. Results were reproduced in three biological replicates.

We next examined the levels of H3K9me2, a key epigenetic mark associated with transcriptional silencing and strongly correlated with DNA methylation in Arabidopsis. Chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) revealed a significant reduction in H3K9me2 at the d35S promoter in YJshh1-2 compared with YJ control plants (Figure 2D). This finding suggests that the shh1 mutation disrupts both CHH methylation and H3K9me2 deposition at the YJ reporter locus.

Together, these results support the role of SHH1 as a component of the RdDM pathway, as the shh1 mutation led to reduced CHH methylation and H3K9me2 levels at the d35S promoter. However, the decrease in these repressive marks contrasts with the reduced luminescence observed in YJshh1-2, suggesting that SHH1 has a specialized function in counteracting DNA methylation-mediated gene silencing at the YJ transgene, beyond its canonical role in RdDM.

Silencing of the YJ reporter in shh1 mutants is largely independent of ROS1 expression

Previous studies have shown that mutations in certain RdDM components, such as the Pol IV subunits NRPD1 and RDR2, lead to silencing of 35S-NPTII transgenes due to reduced ROS1 expression (Li et al., 2012). Because SHH1 was originally identified as a Pol IV-associated factor (Law et al., 2011, 2013; Zhang et al., 2013), we examined whether silencing of the d35S-LUC transgene in the shh1 mutant might also result from ROS1 downregulation. To test this, we overexpressed ROS1 in the YJshh1-2 mutant. However, ROS1 overexpression caused only a slight increase in LUC luminescence and transcript levels (Supplemental Figure 2). These results indicate that SHH1 promotes the expression of methylated transgenes largely independently of ROS1 regulation.

SHH1 functions in the same genetic pathway as MBD7 to suppress transcriptional gene silencing at the YJ transgene and a subset of endogenous genes

Two DNA methylation reader complexes, SUVH1/3–DNAJ1/2 and MBD7–ACD, are known to enhance the expression of promoter-methylated genes (Li et al., 2016; Harris et al., 2018). Given SHH1’s role in promoting LUC expression, we hypothesized that SHH1 might act within the same genetic pathway as either SUVH1/3–DNAJ1/2 or MBD7–ACD. To test this possibility, we generated the double mutants YJshh1-2 suvh1-1, YJshh1-2 mbd7-4, and YJshh1-2 lil-1 through genetic crossing. Both LUC luminescence and transcript levels were reduced in YJmbd7-4, YJlil-1, and YJsuvh1-1 (Figure 3A–3D and Supplemental Figure 3), indicating that the YJ transgene is positively regulated by both the MBD7–LIL and SUVH1/3–DNAJ1/2 pathways.

Figure 3.

Genetic relationship between SHH1 and the MBD7–LIL pathway.

(A) LUC luminescence of YJ, YJshh1-2, YJmbd7-4, and their double mutant YJshh1-2 mbd7-4.

(B) RT–qPCR of LUC transcript levels in YJ, YJshh1-2, YJmbd7-4, and YJshh1-2 mbd7-4 seedlings.

(C) LUC luminescence of YJ, YJshh1-2, YJlil-1, and the double mutant YJshh1-2 lil-1.

(D) RT–qPCR of LUC transcript levels in YJ, YJshh1-2, YJlil-1, and YJshh1-2 lil-1 seedlings.

Error bars indicate SD from three technical replicates; statistical significance was determined by Student’s t-test. ns, not significant. Results were reproduced in three biological replicates.

Next, we compared LUC luminescence and transcript levels in the double mutants with their respective single mutants. The YJshh1-2 suvh1-1 double mutant showed significantly additive effects, whereas YJshh1-2 mbd7-4 and YJshh1-2 lil-1 did not display additive effects relative to the corresponding single mutants (Figure 3A–3D and Supplemental Figure 3). These results suggest that SHH1 most likely functions within the MBD7–ACD pathway to suppress transcriptional gene silencing at the YJ transgene.

Because both MBD7 and SHH1 are highly expressed in embryos (Supplemental Figure 4), we performed mRNA sequencing (mRNA-seq) of YJ, YJshh1-2, and YJmbd7-4 embryos to further explore their roles in counteracting endogenous gene silencing. Compared with YJ, 2294 and 1244 genes were significantly downregulated in YJshh1-2 and YJmbd7-4, respectively (false discovery rate [FDR] < 0.05 and |log₂FC| > 1). Of these, 927 downregulated genes were shared between the two mutants, and their reduced expression levels were validated by RT-qPCR (Figure 4A and 4B). To further assess the genetic relationship between SHH1 and MBD7, we analyzed 10 representative genes in the YJshh1-2 mbd7-4 double mutant. No additive reduction in expression was observed compared with the single mutants, supporting the idea that SHH1 and MBD7 function in the same pathway to prevent transcriptional silencing at a subset of endogenous genes (Figure 4B). To test whether increased DNA methylation contributes to the downregulation of DEGs, we compared methylation levels within 1-kb promoter regions and gene expression levels among the 927 shared downregulated DEGs. Scatterplot analysis revealed no clear anticorrelation between promoter methylation and gene expression, as indicated by the low coefficient of determination (Figure 4C). Among these, three genes (AT3G30460, AT5G05250, and AT5G33395) showed high promoter methylation and were selected for detailed presentation and subsequent experiments (Figure 4D). Consistently, ChIP-qPCR analysis showed that these promoter regions were enriched for H3K9me2 (Supplemental Figure 5).

Figure 4.

SHH1 and MBD7 function in anti-silencing of endogenous genes.

(A) Venn diagram of downregulated differentially expressed genes (DEGs; FDR < 0.05 and |log2FC| > 1) in YJshh1-2 (2294 genes) and YJmbd7-4 (1244 genes) mutants identified by mRNA-seq.

(B) RT–qPCR of 10 downregulated DEGs shared between YJshh1-2 and YJmbd7-4 mutants in YJ, YJshh1-2, YJmbd7-4, and YJshh1-2 mbd7-4 embryos. UBQ5 served as the loading control. Bars with different letters indicate significant differences at p < 0.05 (one-way ANOVA followed by Dunnett’s multiple-comparisons test).

(C) Scatterplots of 1-kb promoter methylation ratio versus expression fold change for 927 shared downregulated DEGs in YJshh1-2 and YJmbd7-4. The x-axis shows CG/CHG/CHH methylation ratios of YJmbd7-4 or YJshh1-2 relative to YJ, and the y-axis represents the corresponding fold change in expression. A linear regression line is shown, with the coefficient of determination (R²) indicating the extent to which promoter methylation explains expression variation.

(D) Integrative Genomics Viewer (IGV) tracks of mRNA-seq expression and whole-genome bisulfite methylation at three representative shared DEGs (locus IDs in red) for YJ, YJshh1-2, and YJmbd7-4.

Together, these findings indicate that SHH1 and MBD7 act within the same pathway to suppress transcriptional silencing at methylated transgenes and a subset of promoter-methylated endogenous genes.

SHH1 interacts with MBD7 in vivo

To further investigate the mechanism underlying SHH1 function, we performed pairwise yeast two-hybrid assays using full-length SHH1 as bait, which demonstrated a direct interaction with MBD7 (Figure 5A and 5B). SHH1 contains a cryptic homeodomain and a SAWADEE domain, whereas MBD7 comprises three MBD domains and a serine/threonine kinase-like (STKC) domain (Figure 5A). To map the interaction domains of both proteins, we tested various fragments of SHH1 and MBD7 in yeast two-hybrid assays. The results showed that full-length SHH1 and its C-terminal SAWADEE domain (residues 125–264) interacted with MBD7, whereas the truncated SHH1 fragment containing only the N-terminal homeodomain (residues 1–75) did not (Figure 5B). This indicates that the C-terminal SAWADEE domain is crucial for SHH1 interaction with MBD7. Similarly, the truncated MBD7 construct containing only the STKC domain failed to interact with SHH1, whereas both full-length MBD7 and the construct containing all three MBD domains successfully interacted with SHH1 (Figure 5B). These findings demonstrate that the MBD domains of MBD7 are essential for its interaction with SHH1. The interaction was further confirmed by split-LUC complementation assays in Nicotiana benthamiana leaves and by pull-down assays (Figure 5C and Supplemental Figure 6).

Figure 5.

SHH1 physically interacts with MBD7.

(A) Schematics of SHH1 and MBD7 protein domains.

(B) Yeast two-hybrid assays testing SHH1–MBD7 interaction. Full-length SHH1 and truncated forms (homeodomain, SAWADEE) were fused to the GAL4 activation domain (AD). Full-length MBD7 and truncated forms (3MBD and STKC) were fused to the GAL4 binding domain (BD).

(C) Split-LUC complementation in tobacco leaves testing SHH1–MBD7 interaction. Luciferase activity was detected at 48 h post-infiltration.

(D and E) Co-IP analysis of the interaction between SHH1 and MBD7. Co-IP was performed using anti-FLAG beads to examine the interaction between MBD7 and SHH1 (D) or SHH1 variants carrying individual point mutations in the homeodomain (SHH1^K72A) or the SAWADEE domain (SHH1^Y140A or SHH1^Y212A) (E) and MBD7. Immunoblotting was conducted using anti-MYC and anti-FLAG antibodies. “Input” indicates total protein extracts before immunoprecipitation. Ten-day-old seedlings co-expressing SHH1-3×MYC or its mutant variants with MBD7-3×FLAG were used for co-IP analysis.

(F) Split-LUC assay showing interactions between MBD7 and SHH1 variants with a point mutation (SHH1^K72A, SHH1^Y140A, and SHH1^Y212A) in Nicotiana benthamiana leaves co-expressing the indicated constructs. Luciferase activity was detected 48 h after infiltration. The interaction between SHH1 and MBD7 served as a positive control, whereas that between SHH1 and LIL served as a negative control.

To test whether SHH1 and MBD7 also interact in vivo, we performed co-immunoprecipitation (co-IP) assays using hybrid plants harboring functional transgenes SHH1-3×MYC/YJshh1-2 and MBD7-3×FLAG/YJmbd7-4. SHH1-3×MYC co-immunoprecipitated with MBD7-3×FLAG (Figure 5D), providing strong evidence of an in vivo interaction. Given that both mutations in the SAWADEE domain of SHH1^Y140A and SHH1^Y212A, which disrupt the H3K9me2 binding pocket, cause severe DNA methylation defects, reduced Pol IV occupancy, and decreased siRNA levels (Law et al., 2013), we examined their effect on the SHH1-MBD7 interaction. We generated SHH1^K72A, SHH1^Y140A, and SHH1^Y212A mutants and expressed them under the native promoter in YJshh1-2 at levels comparable to wild-type SHH1 (Supplemental Figure 7). Co-IP experiments using hybrid plants expressing SHH1^K72A-3×MYC/YJshh1-2, SHH1^Y140A-3×MYC/YJshh1-2, or SHH1^Y212A-3×MYC/YJshh1-2 together with MBD7-3×FLAG/YJmbd7-4 showed that mutations in the SAWADEE domain abolished the SHH1–MBD7 interaction in vivo, whereas the homeodomain mutation (K72A) did not (Figure 5E). These results were further validated by split -LUC and pull-down assays (Figure 5F and Supplemental Figure 6).

Because the SAWADEE domain forms the binding pocket for H3K9me2, we next asked whether binding to H3K9me2 is required for the SHH1–MBD7 interaction. Binding assays performed in the presence of exogenous H3K9me2 peptides (Supplemental Figure 6) showed that H3K9me2 had no detectable effect on the interaction between SHH1 and MBD7. These findings suggest that the Y140A and Y212A mutations disrupt the SHH1–MBD7 interaction not by blocking H3K9me2 recognition but likely by altering the conformation or aggregation state of the SAWADEE domain, which is essential for mediating their physical association.

Collectively, these results confirm that SHH1 and MBD7 interact in vivo and highlight the essential role of the SAWADEE domain in facilitating this interaction.

SHH1 co-localizes with MBD7 and is essential for MBD7 localization and stability

The interaction between SHH1 and MBD7 prompted us to examine whether the two proteins co-localize in the nucleus in vivo. We performed immunofluorescence analysis using F₁ hybrid plants expressing SHH1-3×MYC and MBD7-3×FLAG. Both proteins displayed similar subnuclear localization patterns, with strong round-shaped nucleolar signals and one or two distinct nucleoplasmic foci (Figure 6A). Consistent with their physical interaction, merged signals showed clear co-localization of SHH1 and MBD7 within the nucleus, particularly at the nucleolar and nucleoplasmic foci, in all analyzed cells (100%, n = 100) (Figure 6A).

Figure 6.

SHH1 co-localizes with MBD7 and influences MBD7 protein stability.

(A) SHH1 and MBD7 were visualized by immunofluorescence in F₁ hybrids expressing SHH1-3×MYC (red) and MBD7-3×FLAG (green). Merged images show bright yellow signals resulting from the overlap of the red and green channels. Red arrows indicate nucleolar foci, and yellow arrows indicate nucleoplasmic foci. Scale bars, 10 μm.

(B) Immunofluorescence detection of MBD7 (green) using an anti-FLAG antibody in YJshh1-2 and YJmbd7-4. YJ serves as the negative control. The numbers next to the images indicate the proportion of nuclei displaying the fluorescence pattern shown on the right. Scale bars, 10 μm.

(C) Semi-quantitative western blot analysis of MBD7 protein levels in YJ, YJmbd7-4, and YJshh1-2. Protein extracts were normalized (100% lanes) before loading; 50% lanes were loaded with half the total protein. Histone H3 served as the loading control.

To test whether MBD7 localization depends on SHH1, we introduced the MBD7-3×FLAG transgene into the homozygous YJshh1-2 mutant by genetic crossing. Fluorescence microscopy revealed that 85% of nuclei in the YJshh1-2 mutant lacked both nucleolar and nucleoplasmic foci, while only 15% retained the wild-type localization pattern (n = 100) (Figure 6B). This result suggests that a small subset of cell types in YJshh1-2 may maintain normal MBD7 localization. Overall, these data indicate that SHH1 is required for proper MBD7 localization in vivo. We next asked whether the loss of MBD7 localization in YJshh1-2 was associated with reduced protein stability. Western blot analysis showed a marked decrease in MBD7-3×FLAG protein levels in YJshh1-2 compared with wild-type plants (Figure 6C). This reduction was not due to transcriptional repression, as MBD7-3×FLAG transcript levels were comparable between the two genotypes (Supplemental Figure 8). These findings demonstrate that SHH1 is required not only for the correct nuclear localization of MBD7 but also for its stability in vivo.

The SAWADEE domain is critical for the function of SHH1

Because the SAWADEE domain is essential for the interaction between SHH1 and MBD7, we next investigated whether this domain is also required for SHH1 function in suppressing transcriptional gene silencing. We tested whether point mutations within the SAWADEE domain could restore the mutant phenotypes of YJshh1-2. Luminescence imaging and RT-qPCR analyses showed that both the wild-type SHH1 construct and the SHH1^K72A mutant effectively restored luminescence, as well as the expression of LUC and three representative endogenous genes (AT3G30460, AT5G05250, and AT5G33395) in the YJshh1-2 background (Figure 7A and 7B). In contrast, the SHH1^Y140A and SHH1^Y212A mutant versions failed to complement these defects (Figure 7A and 7B). These results demonstrate that the SAWADEE domain is required for SHH1 function in vivo.

Figure 7.

The SAWADEE domain of SHH1 is essential for chromatin binding.

(A) Luminescence imaging (left) and RT–qPCR analysis of LUC transcripts (right) in YJ, YJshh1-2, and YJshh1-2 complemented with SHH1 variants carrying point mutations in the SAWADEE domain (top-left diagram). LUC expression in mutants and complemented lines was statistically compared with YJ by Student’s t-test.

(B) RT–qPCR quantification of three representative endogenous genes in YJ, YJshh1-2, and YJshh1-2 embryos complemented with SHH1 variants carrying different point mutations. Relative expression levels were statistically compared with YJ using Student’s t-test.

(C) Diagram showing the positions of endogenous genes and nearby transposons in the promoter regions; horizontal black bars indicate ChIP–qPCR amplicons.

(D) ChIP–qPCR analysis showing the enrichment of SHH1 and its variants carrying point mutations at the d35S promoter that drives LUC expression, as well as at transposon regions within the promoters of three selected endogenous genes. Enrichment levels of the mutants were statistically compared with those of SHH1 by Student’s t-test.

(E) ChIP–qPCR analysis of MBD7 enrichment in YJmbd7-4, YJshh1-2, and YJshh1-2 expressing SHH1 SAWADEE-domain mutants (SHH1^Y140A and SHH1^Y212A) at the d35S promoter that drives LUC expression and at transposons within the promoters of three selected endogenous genes. MBD7 enrichments in YJshh1-2 mutants or mutant lines were statistically compared with those in YJmbd7-4 by Student’s t-test.

(F) ChIP–qPCR analysis of SHH1 enrichment at the d35S promoter that drives LUC expression and at the promoters of the three selected endogenous genes in YJmbd7-4 and YJshh1-2. SHH1 enrichments in YJmbd7-4 mutants were statistically compared with those of MBD7 in YJshh1-2 by Student’s t-test.

(G) Model illustrating that DNA methylation and H3K9me2 marks at endogenous promoters recruit SHH1, which binds to H3K9me2, and MBD7, a DNA methylation reader. SHH1 associates with MBD7 to form the SHH1–MBD7–ACD complex, which simultaneously recognizes both methylated cytosines and H3K9me2, thereby promoting gene expression at these loci.

Error bars indicate SD from three technical replicates; statistical significance was determined by t-test. ns, not significant. All results were validated in three biological replicates.

Previous work has shown that SHH1 binds to specific chromatin regions through its SAWADEE domain (Law et al., 2013). To determine whether SHH1 directly binds to the methylated promoters of the transgene and the three selected endogenous genes, we performed ChIP assays using transgenic lines expressing either wild-type or mutant SHH1. Both wild-type SHH1 and the SHH1^K72A mutant showed significant enrichment at these methylated loci compared with the no-antibody control (Figure 7C and 7D). In contrast, SHH1^Y140A and SHH1^Y212A mutants exhibited a marked loss of occupancy at these regions (Figure 7D). These findings indicate that SHH1 directly binds methylated loci through its SAWADEE domain, which is indispensable for this interaction.

Interdependent roles of SHH1 and MBD7 in targeting methylated loci

Our previous work demonstrated that MBD7 is a key component of the MBD7 complex, responsible for recruiting this complex to the methylated d35S promoter. The findings presented here show that SHH1 interacts with MBD7 and binds to methylated loci via its SAWADEE domain, which recognizes H3K9me2. This suggests a cooperative mechanism in which SHH1 and MBD7 jointly mediate the efficient recruitment of the MBD7 complex to methylated regions. To test this hypothesis, we performed ChIP assays. ChIP of FLAG-tagged MBD7 revealed a marked decrease in MBD7 occupancy at the methylated promoters of the transgene and three selected endogenous genes in the YJshh1-2 mutant, as well as in plants expressing the SHH1^Y140A and SHH1^Y212A mutant versions of SHH1, compared with at least a two-fold enrichment in wild-type plants (Figure 7E). These results indicate that SHH1 is essential for the efficient recruitment of MBD7 to the methylated loci. Similarly, ChIP analysis of MYC-tagged SHH1 showed a comparable decrease in SHH1 binding at methylated loci in the YJmbd7-4 mutant (Figure 7F), demonstrating that MBD7 is also required for efficient SHH1 recruitment.

Together, these findings indicate that SHH1 and MBD7, by jointly recognizing the dual marks of DNA methylation and H3K9me2, act cooperatively to promote the binding of the MBD7 complex to methylated chromatin.

Discussion

DNA methylation is widely associated with transcriptional gene silencing, yet promoter methylation does not always lead to gene repression (Rishi et al., 2010; Li et al., 2016, 2017; Wan et al., 2017; Harris et al., 2018). This observation suggests that DNA methylation can also contribute to gene activation under specific conditions. In previous work, we identified the MBD7–ACD complex as a DNA methylation reader that promotes the expression of promoter-methylated transgenes without significantly altering their methylation levels (Li et al., 2017). However, the mechanism by which this complex is recruited to methylated promoters remained unclear. The present study provides new insight into this process, showing that SHH1 interacts with MBD7 and cooperates to recruit the MBD7–ACD complex to methylated promoters of transgenes and a subset of endogenous genes. This interaction counteracts DNA methylation-mediated silencing and promotes transcription. These findings reveal a previously unrecognized role of SHH1 and expand our understanding of how DNA methylation fine-tunes gene expression, emphasizing the key function of SHH1 in modulating the activity of methylation readers such as MBD7.

SHH1 was previously identified as an essential component of the canonical RdDM pathway (Law et al., 2011, 2013; Zhang et al., 2013). Consistent with these earlier studies, we observed that both CHH methylation and H3K9me2 levels at the d35S promoter were significantly reduced in the YJshh1-2 mutant. Despite this reduction, SHH1 dysfunction resulted in undetectable luminescence and a sharp decrease in LUC expression. Moreover, 5-Aza-dC treatment restored LUC activity only when the promoter was methylated, suggesting that SHH1 acts as an anti-silencing factor downstream of DNA methylation. Previous studies have shown that mutations in some RdDM components, including the Pol IV subunits RDR2 and NRPD1, cause transgene silencing of 35S-NPTII due to the loss of ROS1 expression (Li et al., 2012). In contrast, overexpression of ROS1 in the YJshh1-2 mutant only slightly enhanced luminescence and LUC expression, indicating that SHH1 suppresses methylated transgene silencing through a mechanism distinct from ROS1 regulation.

Our conclusion that SHH1 and MBD7 function within the same genetic pathway to suppress DNA methylation-mediated transcriptional gene silencing is supported by several lines of evidence. First, genetic analyses revealed no additive effects on YJ transgene silencing in the YJshh1-2 mbd7-4 and YJshh1-2 lil-1 double mutants, suggesting that SHH1 mediates its anti-silencing function through the MBD7 pathway. Second, comparison of RNA-seq and whole-genome bisulfite sequencing data identified 927 commonly downregulated genes in both mutants, none of which showed increased methylation. Moreover, the expression levels of these endogenous genes in the YJshh1-2 mbd7-4 double mutant were comparable to those in the single mutants, reinforcing the idea that SHH1 and MBD7 act within the same regulatory pathway. Third, yeast two-hybrid, split-LUC, pull-down, and co-IP assays confirmed a physical interaction between SHH1 and MBD7, with the SAWADEE domain playing a central role in this interaction. Finally, immunofluorescence analysis showed that SHH1 and MBD7 co-localize in the nucleus, particularly in nucleolar and nucleoplasmic foci. Together, these results demonstrate that SHH1 and MBD7 form a stable complex and function cooperatively to suppress DNA methylation-mediated transcriptional silencing. Interestingly, in YJshh1-2, MBD7 lost its nuclear localization and exhibited markedly reduced protein levels, suggesting that SHH1 contributes to MBD7 stability, likely by promoting subnuclear targeting or anchoring to H3K9me2-marked chromatin. Although this could imply that SHH1 acts indirectly through MBD7 stabilization, several observations indicate that this is not the main mechanism. Transcriptome analysis revealed limited overlap in DEGs between YJshh1-2 and YJmbd7-4 in seedlings (Supplemental Figure 9), indicating that the loss of SHH1 is not equivalent to the loss of MBD7. Furthermore, ChIP analysis showed reciprocal reductions in SHH1 and MBD7 occupancy in each other’s mutants. Collectively, these findings support a model in which SHH1 and MBD7 reinforce each other’s chromatin association, functioning primarily in the recruitment of the complex to methylated loci rather than solely through protein stabilization.

DNA methylation and H3K9me2 are generally closely associated, jointly restricting DNA accessibility and maintaining genome stability (Chan et al., 2005; Cedar and Bergman, 2009; Bell et al., 2023). When present near genes, these repressive modifications can lead to transcriptional inhibition, as seen when TE insertions into promoters alter the expression of adjacent genes (Zhang et al., 2018). Thus, anti-silencing mechanisms involving reader proteins that recognize epigenetic marks are essential for overcoming repression and ensuring precise gene regulation. While MBD7 binds methylated DNA via its MBD motifs, the mechanism by which H3K9me2 is recognized has remained unclear. Given that the SAWADEE domain of SHH1 specifically binds H3K9me2, we propose a two-step model explaining how SHH1 collaborates with MBD7 to suppress promoter-methylated gene silencing (Figure 7G). In the first step, SHH1 associates with MBD7 to simultaneously recognize both methylated cytosines and H3K9me2, facilitating the recruitment of the MBD7–ACD complex to target loci. This association forms a mutually reinforcing binding module: the absence of either SHH1 or MBD7 disrupts efficient recruitment of the MBD7–ACD complex to methylated chromatin. This indicates that SHH1 or MBD7 alone is insufficient to overcome restricted DNA accessibility imposed by repressive epigenetic marks and that dual recognition of DNA methylation and histone modification is required for stable chromatin association and transcriptional activation. Consistent with this model, point mutations within the SAWADEE domain (e.g., SHH1^Y140A and SHH1^Y212A) failed to rescue the YJshh1-2 phenotype and caused a pronounced loss of SHH1 occupancy at target regions. This underscores the critical role of the SAWADEE domain in both SHH1 binding and its function in suppressing promoter-methylated gene silencing. Moreover, MBD7 binding at methylated loci was markedly reduced in YJshh1-2 and in plants expressing SHH1^Y140A or SHH1 ^Y212A, whereas SHH1 binding was likewise impaired in the absence of functional MBD7. These observations indicate that SHH1 and MBD7 binding to methylated loci is interdependent. However, MBD7 binding in YJshh1-2 was not completely abolished, suggesting that additional factors may assist in SHH1 or MBD7 recruitment to methylated sites. In the second step, MBD7 interacts with the ACD proteins LIL and ROS5. Because ACD proteins form multimeric complexes (Sun and MacRae, 2005), we propose that the SHH1–MBD7–ACD platform recruits multiple ACD proteins, thereby amplifying anti-silencing activity and promoting the expression of promoter-methylated genes.

It is well established that the SHH1–Pol IV complex facilitates the recruitment of Pol IV to active RdDM loci, initiating the production of 24-nt siRNAs that promote DNA methylation and reinforce transcriptional gene silencing (Law et al., 2013). Intriguingly, our study shows that the SHH1–MBD7 complex acts in the opposite direction, counteracting DNA methylation-mediated silencing at methylated transgenes and a subset of endogenous genes, thereby enhancing their expression without altering promoter methylation levels. An important question is how SHH1 coordinates these seemingly opposing activities at the same loci. One possible explanation is that SHH1–Pol IV and SHH1–MBD7 form distinct protein complexes. After SHH1–Pol IV establishes DNA methylation, the methylated DNA may subsequently be recognized by the SHH1–MBD7 complex, which promotes the expression of nearby genes while maintaining repression of TEs. Several lines of evidence support this model. First, MBD7 overexpression does not affect Pol IV recruitment, indicating that MBD7 and Pol IV do not compete for SHH1 binding (Supplemental Figure 10). Second, CHH hypomethylated differentially methylated regions (DMRs) in YJmbd7-4 do not overlap with those of YJshh1-2 or YJnrpe1 mutants, which are hallmark features of the RdDM pathway, suggesting that MBD7 does not participate in RdDM (Supplemental Figure 11). We speculate that other components may influence whether SHH1 associates with Pol IV or MBD7. For example, RDR2 and/or CLSY1/2 might promote the formation of the SHH1–Pol IV complex, whereas ACD-containing proteins could favor the assembly of the SHH1–MBD7 complex. Alternatively, SHH1–Pol IV and SHH1–MBD7 might act at different regions within the same gene, a possibility that warrants further investigation. Future studies will be important to identify additional components of the MBD7 complex, clarify its assembly mechanism, and determine how these two SHH1-containing complexes coordinately regulate transcription at promoter-methylated genes.

Our phylogenetic analysis of SHH genes across 17 terrestrial plants reveals that SHH1 and SHH2 homologs diverged before the most recent common ancestor of seed plants, suggesting early functional differentiation (Supplemental Figure 12). The absence of phenocopy in shh2 mutants compared with YJshh1-2 supports this conclusion. Notably, SHH1 underwent duplication in eudicots, except in Arabidopsis, which retains a single copy, while monocots have retained only one SHH1 homolog (Supplemental Figure 12). This duplication event in eudicots may be associated with higher DNA methylation levels, reflecting an evolutionary adaptation for more efficient gene regulation. Our findings also suggest a new role for SHH1 in suppressing promoter-methylated gene silencing through a mutually reinforcing SHH1–MBD7 binding module. This mechanism, likely co-evolving with increased DNA methylation, enables more refined epigenetic control and enhances the regulation of promoter-methylated genes and foreign DNA in plants.

Several methyl-DNA binding protein complexes have been identified as key mediators linking DNA methylation to gene expression regulation in Arabidopsis. Complexes such as MBD5/6–SILENZIO, SUVH2/9–MORC1/2/6, SUVR2–CHR19/27/28, and the recently identified MBD2 complex have been shown to couple DNA methylation with gene silencing, thereby reinforcing transcriptional repression (Moissiard et al., 2012; Han et al., 2014; Jing et al., 2016; Ichino et al., 2021; Boone et al., 2023; Wang et al., 2024). In contrast, complexes such as SUVH1/3–DnaJ1/2 act to counter repression and promote the expression of methylated genes (Li et al., 2017; Harris et al., 2018). However, the precise mechanisms by which these complexes recognize methylated chromatin remain unclear. Our findings suggest a potential general mechanism in which SHH1–MBD7–ACD complexes recognize both methylated cytosines and H3K9me2, enabling effective binding to chromatin marked by repressive epigenetic modifications and thereby facilitating transcriptional activation. This model highlights a coordinated interaction between DNA methylation and histone modifications and provides a conceptual framework for understanding how such complexes regulate promoter-methylated gene expression in Arabidopsis.

Recent studies in mammals have identified 731 transcription factors capable of directly binding methylated DNA (Zhu et al., 2016; Wang et al., 2018). Some of these transcription factors have been shown to activate downstream gene expression, adding complexity to the regulatory landscape of methylated genes (Rishi et al., 2010; Quenneville et al., 2011; Wan et al., 2017). Future research aimed at identifying new regulatory factors and exploring their genetic interactions with known complexes will be critical for dissecting the molecular mechanisms that control methylated gene expression. Because TEs are frequently located near protein-coding genes in crops such as rice and soybean (International Rice Genome Sequencing, 2005; Schmutz et al., 2010; Zemach et al., 2010; Shen et al., 2018), understanding these regulatory mechanisms is of particular importance. Such insights will not only contribute to stabilizing transgenic traits but also deepen our understanding of how gene expression is fine-tuned to support developmental transitions and adaptation to environmental changes.

Methods

Plant materials, map-based cloning, and complementation assay

A. thaliana Columbia-0 (Col-0) was used as the wild type in this study. The LUC-based reporter line YJ; mutants YJsuvh1-1, YJmbd7-4, and YJlil-1; and MBD7-3×FLAG transgenic plants were described previously (Li et al., 2016, 2017). The shh1-2 allele was isolated from the YJ background by genetic mutagenesis screening. The YJshh2 mutant was generated via CRISPR-Cas9-mediated gene editing in the YJ background. Homozygous, Cas9-free mutants were selected for further analysis. Single-guide RNA sequences are listed in Supplemental Table 1.

For map-based cloning, the YJshh1-2 mutant (Col-0 background) was crossed with YJ in the Landsberg erecta (Ler) background (YJ-Ler). The F₂ population was used for mapping. Preliminary mapping was performed on 24 F₂ plants showing reduced LUC activity, localizing the mutation to the upper arm of chromosome 1. Whole-genome resequencing or candidate gene sequencing identified a G-to-A substitution that introduced a premature stop codon in the third exon of At1g15215, designated YJshh1-2. SSLP and dCAPS markers were designed based on polymorphisms between Col-0 and Ler (http://arabidopsis.org/browse/Cereon/index.jsp).

For complementation assays, the SHH1 genomic region, including its native promoter and coding sequence (excluding the stop codon), was cloned in-frame with eGFP, 3×MYC, or 3×FLAG tags into vectors derived from pCAMBIA2300 and pCAMBIA1300. To generate point mutants, the Y140A, Y212A, and K72A substitutions were introduced into the SHH1-3×MYC construct using the Fast Mutagenesis System kit (TransGen, FM111) according to the manufacturer’s instructions. All the constructs were transformed into YJshh1-2 for complementation testing. The 35S::ROS1 construct containing the terminator sequence was cloned into pCAMBIA1300 and transformed into YJshh1-2. The 35S::MBD7-3×FLAG constructs was also cloned into pCAMBIA1300 and introduced into YJ, generating the transgenic line named 35S::MBD7/YJ. Additionally, the pNRPD1::NRPD1-3×MYC constructs was cloned into pCAMBIA1300 and introduced into both YJ and 35S::MBD7/YJ. Primers used for plasmid construction and site-directed mutagenesis are listed in Supplemental Table 1.

Luciferase imaging and 5-Aza-dC treatment

For LUC imaging, 10-day-old seedlings were sprayed with 1 mM luciferin (Promega) containing 0.01% Triton X-100. After 5 min of incubation in the dark, luminescence was captured using a CLINX Science IVScope7500 Luminescence Imaging System with a 1-min exposure time. For 5-Aza-dC (Sigma, A3656) treatment, plants were grown on half-strength Murashige and Skoog (MS) medium supplemented with 0.8% agar, 1% sucrose, and 7 mg/mL 5-Aza-dC for 10 days.

RNA extraction and RT-qPCR

Embryos were isolated from mature seeds imbibed in water for 24 h in the dark as described by Marchetti et al. (2025). Briefly, seeds were immersed in diethyl pyrocarbonate（DEPC）-treated water and gently pressed with a plastic stick to release embryos. The mixture was centrifuged at 72 g for 10 min to separate embryos from seed coats. RNA was extracted using the NucleoSpin RNA extraction kit (cat. no. NP08011). For 10-day-old seedlings, total RNA was extracted using TRI reagent (Molecular Research Center, TR118). RNA samples were treated with DNase I (Roche, 04716728001) to remove genomic DNA contamination. First-strand cDNA synthesis was performed using the Applied Biosystems High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368814). RT-qPCR was conducted on a LightCycler Real-Time PCR System (Roche) using Universal SYBR Green Supermix (Bio-Rad, 172-5124). Quantification of gene expression was performed from three biological replicates as previously described (Liu et al., 2014).

McrBC-PCR

Genomic DNA was extracted from 10-day-old seedlings using the DNeasy Plant Kit (Qiagen, 69104). A total of 500 ng of genomic DNA was digested with 30 U of McrBC (New England Biolabs, M0272) at 37°C for 45 min. The reaction was terminated by heating at 65°C for 20 min. A parallel mock control (no enzyme) was processed in the same manner. Quantitative PCR was then performed using Universal SYBR Green Supermix (Bio-Rad) and gene-specific primers to determine the ratio between McrBC-treated and untreated DNA, thereby assessing relative methylation levels (Li et al., 2017).

Whole-genome bisulfite sequencing and data processing

Genomic DNA was isolated from 2-week-old YJ, YJshh1-2, and YJmbd7-4 seedlings using the TIANGEN Plant Genomic DNA Kit (DP305) and from embryos using the CWBIO Plant Genomic DNA Kit (CW0571T). Samples from YJ seedlings and embryos had two biological replicates, and those from YJshh1-2 and YJmbd7-4 embryos had three replicates. Approximately 500 ng of DNA was fragmented to 200–500 bp using a Bioruptor Pico ultrasonicator (Diagenode, B01060001). Fragment size distribution was verified by 2% agarose gel electrophoresis, and DNA concentrations were measured with a Qubit 4.0 Fluorometer (Thermo Fisher Scientific, Q33238). End repair and adapter ligation (Illumina TruSeq Methylated Adapters, IDT, 15028212) were carried out using the NEBNext Ultra II DNA Library Prep Kit (NEB, E7645), followed by purification with VAHTS DNA Clean Beads (Vazyme, N411-02). Bisulfite conversion was performed using the MethylCode Bisulfite Conversion Kit (Invitrogen, MECOV-50) under the following thermal cycling conditions: 98°C for 5 min, 60°C for 90 min, 98°C for 5 min, and 60°C for 60 min. Converted DNA was amplified by PCR (10 cycles) with NEBNext 2× KAPA HiFi Uracil+ ReadyMix (NEB) and purified using VAHTS DNA Clean Beads.

Sequencing reads were aligned to the Arabidopsis TAIR10 reference genome using Bismark with default parameters allowing ≤ 2 mismatches. PCR duplicates were removed, and cytosine methylation levels were extracted using the bismark_methylation_extractor. Cytosines with < 4× coverage were excluded from downstream analysis. DMRs were identified with DMRcaller using 100-bp bins, requiring ≥ 4 cytosines per context (CG, CHG, or CHH). DMRs were called based on context-specific methylation difference thresholds (≥ 0.4 for CG, ≥ 0.2 for CHG, and ≥ 0.1 for CHH) with adjusted p ≤ 0.01 determined by Fisher’s exact test. Quality metrics, including total read numbers, mapping efficiency, and bisulfite conversion rates, are summarized in Supplemental Table 2.

mRNA-seq library construction and data processing

Total RNA from 10-day-old seedlings and embryos of YJ, YJshh1-2, and YJmbd7-4 was extracted using Trizol (Molecular Research Center, TR118), and treated with DNase I (Roche, 04716728001). Embryo samples (YJ, YJshh1-2, and YJmbd7-4) had three biological replicates, and seedling samples had two. mRNA-seq libraries were prepared using the VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, cat. no. NR604) following the manufacturer’s protocol. RNA integrity was confirmed by electrophoresis, and 1 μg of total RNA was used for library construction. mRNA enrichment was performed using mRNA Capture Beads, followed by RNA fragmentation and synthesis of the first and second cDNA strands. After adapter ligation, cDNA fragments (150–200 bp) were size-selected, amplified by PCR, and purified.

Libraries were sequenced on an Illumina HiSeq 2500 platform to generate 150-bp paired-end reads. Raw reads were quality checked using FastQC, and adapters were removed using Cutadapt. The cleaned reads were then mapped to the Arabidopsis reference genome (TAIR10) using Hisat2 (http://ccb.jhu.edu/software/tophat/index.shtml). Transcript abundance was normalized as fragments per kilobase of transcript per million mapped reads using the StringTie program (https://ccb.jhu.edu/software/stringtie/). Read counts were computed with FeatureCount, and DEGs were identified using the EdgeR package (https://bioconductor.org/packages/release/bioc/html/edgeR.html), applying a Benjamini–Hochberg adjusted p < 0.05 and |log2(fold change)| > 1 as the criteria for significance. The log2FC, FDR, and gene descriptions of DEGs in embryos and seedlings are listed in Supplemental Tables 3 and 4, respectively. The quality metrics for all sequencing libraries, including the total reads and alignment rates, are summarized in Supplemental Table 2.

Yeast two-hybrid assay

Coding sequences of SHH1 and truncated versions containing either the homeodomain or the SAWADEE domain were cloned into the pGBKT7 vector to fuse with the GAL4 DNA-binding domain (BD). Full-length MBD7 and truncated constructs containing either the three MBD domains or the STKC domain were cloned into the pGADT7 vector to fuse with the GAL4 activation domain (AD). Cloning was performed using the One Step Cloning Kit (NucleoTech, NMO0002-10). Each BD and AD plasmid pair was co-transformed into yeast strain AH109, and positive strains were used for the spot assay. As a control, each AD construct was co-transformed with the empty pGBKT7 vector, while BD constructs were co-transformed with empty pGADT7 vectors. The transformants were selected on SD/−Leu/−Trp medium for screening positive yeasts. Positive colonies were grown in liquid SD/−Leu/−Trp medium at 30°C for 1 day to an OD₆₀₀ of 1.0, and then 10 μl of yeast culture and its 10⁻¹, 10⁻², and 10⁻³ dilutions were spotted onto SD/−Leu/−Trp and SD/−Leu/−Trp/−His/−Ade media and incubated at 30°C for 2 days. Primers used for plasmid construction are listed in Supplemental Table 1.

Co-IP and western blotting

For co-IP assays, 1 g of 10-day-old seedlings from F₁ hybrids of SHH1-3×MYC YJshh1-2 and MBD7-3×FLAG YJmbd7-4, SHH1^K72A-3×MYC YJshh1-2 and MBD7-3×FLAG YJmbd7-4, SHH1^Y140A-3×MYC YJshh1-2 and MBD7-3×FLAG YJmbd7-4, or SHH1^Y212A-3×MYC YJshh1-2 and MBD7-3×FLAG YJmbd7-4 and their parental plants was ground into a fine powder in liquid N₂. The fine powder was suspended in 15 ml ice-cold lysis buffer (50 mM Na₃PO₄, 150 mM NaCl, 1% NP-40, 10% glycerol, 10 mM NaF, 2.5 mM disodium glycerol phosphate, 1 mM Na₃VO₄, 0.5 mM dithiothreitol [DTT], 0.1 mM PMSF, and protease inhibitor cocktail). Lysates were centrifuged at 16 000 g for 10 min at 4°C. After two rounds of preclearing, the supernatants were incubated with 5 μl of anti-FLAG-M2 agarose (Sigma, A2220) for 3 h at 4°C. Beads were washed six times with ice-cold lysis buffer and eluted in 50 μl SDS sample buffer. The immunoprecipitates were then detected by western blotting using either the anti-FLAG (Sigma, F3165) or the anti-MYC (Abcam, 32072) antibody as the primary antibody. Goat anti-mouse IgG (H + L)-HRP (Thermo Fisher, 31430) was used as the secondary antibody. All western blots were visualized using enhanced chemiluminescence detection.

For the MBD7-3×FLAG and MBD7-3×FLAG/shh1-2 immunoblotting experiment, nuclei were first prepared from plant tissue and then immunoblotting was performed using the primary and/or secondary antibodies.

Split-luciferase assay

For split-LUC assays, the full-length coding sequences of SHH1 and its mutant versions (K72A, Y140A, and Y212A) were fused with N-LUC in the JW771 vector. Full-length coding sequences of MBD7 and LIL were fused with C-LUC in the JW772 vector. Agrobacterium tumefaciens strain GV3101 carrying different construct pairs was infiltrated into 4-week-old Nicotiana benthamiana leaves. As a negative control, SHH1-nLUC was co-transformed with LIL-cLUC or cLUC empty vectors, while MBD7-cLUC was co-transformed with nLUC empty vector. Two days after infiltration, leaves were sprayed with 2 mM luciferin (Promega) in 0.01% Triton X-100 solution, and LUC activity was detected using a luminescence imaging system (NEWTON 7.0). Primers used for plasmid construction are listed in Supplemental Table 1.

Immunofluorescence assay

Isolation of nuclei from seedlings of F₁ hybrids between SHH1-3×MYC/YJshh1-2 and MBD7-3×FLAG/YJmbd7-4 or MBD7-3×FLAG/YJshh1-2 and immunofluorescence analyses were performed as described (Li et al., 2008). Briefly, F₁ hybrid seedlings’ nuclei were fixed in 4% formaldehyde and incubated overnight at 4°C with primary antibody anti-FLAG (Sigma, F3165) and anti-MYC (Sigma, 3460468). Secondary antibodies anti-rabbit IgG H&L (Alexa Fluor 594) (Abcam, 150080) and anti-mouse IgG H&L (Alexa Fluor 488) (Abcam, 150113) were used at a dilution of 1:200. For MBD7-3×FLAG/YJshh1-2 immunofluorescence assay, nuclei were fixed in 4% formaldehyde and incubated overnight at 4°C with primary antibodies for anti-FLAG (Sigma, F3165). Secondary anti-mouse IgG H&L (Alexa Fluor 488) (Abcam, 150113) was used at a dilution of 1:200. Chromatin was counterstained with 4′,6-diamidino-2-phenylindole. Images were captured with a Nikon A1R confocal microscope.

ChIP-qPCR assay

ChIP was performed as previously described (Liu et al., 2014). SHH1-3×MYC/YJshh1-2, SHH1^Y140A-3×MYC/YJshh1-2, SHH1^K72A-3×MYC/YJshh1-2, SHH1^Y212A-3×MYC/YJshh1-2, MBD7-3×FLAG/YJmbd7-4, MBD7-3×FLAG/YJshh1-2, MBD7-3×FLAG/YJSHH1^Y140A, MBD7-3×FLAG/YJSHH1^Y212A, SHH1-3×MYC/YJmbd7-4, YJshh1-2, and YJ lines were ground in liquid nitrogen and cross-linked with 1% formaldehyde (Sigma-Aldrich) for 10 min on ice. Chromatin complexes were isolated and sonicated into DNA fragments of ∼500 bp. Lysates were precleared and then incubated with anti-MYC (Sigma, 3460468), anti-FLAG (Sigma, F3165), and anti-H3K9me2 (Abcam, 1220) overnight at 4°C with gentle rotation. Bound chromatin was purified using columns from the Qiagen plasmid extraction kit. Real-time PCR was conducted on the input, a no-antibody control, and antibody-bound DNA in triplicate to quantify the levels of immunoprecipitated genomic fragments of interest. Primers are listed in Supplemental Table 1.

Protein purification and GST pull-down assay

Coding sequences of SHH1 and its point mutants (SHH1^Y140A, SHH1^Y212A) were cloned into pET-21a for His tagging, and that of MBD7 was cloned into pGEX for GST tagging. All constructs were expressed in Escherichia coli BL21(DE3). Cells were grown in LB medium at 37°C to an OD₆₀₀ of 0.6. Protein expression was induced with 0.2 mM isopropyl β-D-1-thiogalactopyranoside（IPTG） at 16°C for 20 h. Harvested cell pellets were lysed by sonication in appropriate buffers: lysis buffer for His-tagged proteins (SHH1-His, SHH1^Y140A-His, and SHH1^Y212A-His) contained 20 mM Tris–HCl (pH 8.0), 500 mM NaCl, and 5 mM imidazole; lysis buffer for GST-MBD7 contained PBS (pH 7.4), 10 mM DTT, and 1× protease inhibitor. Cleared lysates were obtained by centrifugation. His-tagged proteins were subjected to affinity chromatography using Ni-NTA agarose (Qiagen). Bound proteins were eluted with buffer containing 500 mM imidazole. GST-MBD7 was purified using Glutathione Sepharose 4B beads (Cytiva) and eluted with elution buffer (100 mM GSH, 50 mM Tris–HCl [pH 8.0], 10 mM DTT). Purified proteins were analyzed by SDS–PAGE and quantified for downstream assays.

For GST pull-down assays, purified GST-MBD7 (1 μg) or GST alone (1 μg, negative control) was immobilized on a 20-μl slurry of Glutathione Sepharose 4B beads (Cytiva) in 500 μl binding buffer (20 mM Tris–HCl [pH 7.5], 150 mM NaCl, 0.1% NP-40, 1 mM DTT, 1× protease inhibitor) for 1 h at 4°C with rotation. Beads were washed three times with binding buffer to remove unbound protein. Purified prey proteins (SHH1-His, SHH1^Y140A-His, or SHH1^Y212A-His; 1 μg each) were added to the immobilized bait (GST-MBD7 or GST) in fresh binding buffer and incubated for 30 min at room temperature with rotation. To assess the effect of H3K9me2 on the MBD7–SHH1 interaction, purified SHH1-His (1 μg) was first preincubated with different concentrations of modified H3 peptide H3(1–15)K9me2 (0 μM, 5 μM, 10 μM, 20 μM, and 100 μM) in binding buffer for 15 min at room temperature. This mixture was then added to the immobilized GST-MBD7 beads and incubated for 30 min at room temperature with rotation. Following all incubations, the beads were washed five times with binding buffer. Bound proteins were eluted by boiling the beads in 1× SDS–PAGE loading buffer at 100°C for 5 min. Input samples (representing starting material) and eluted samples were resolved by SDS–PAGE. Protein interactions were detected by western blotting using anti-His antibody (Huaxingbio, cat. no. HX1804; 1:2000 dilution) to detect SHH1 proteins and anti-GST antibody (Huaxingbio, cat. no. HX1807; 1:2000 dilution) to detect GST-MBD7 or GST, followed by incubation with goat anti-rabbit IgG (H + L)-HRP (Huaxingbio, cat. no. HX2031) as secondary antibody for chemiluminescence detection.

Phylogenetic analysis

Seventeen species with publicly available genomes, including A. thaliana and Physcomitrella patens, spanning the major terrestrial plant lineages, were used for phylogenetic analysis (Supplemental Table 5). Protein sequences homologous to A. thaliana SHH1 and SHH2 were identified using Diamond with an E-value threshold of 1e−5. The identified sequences were aligned using MAFFT, and poorly aligned regions were trimmed with trimAl v1.4 (-gt 0.9 -cons 50). Phylogenetic inference was conducted using IQ-TREE2 under the optimal model, with bootstrap analysis set to 1000 replicates.

Funding

This work was supported by the National Natural Science Foundation of China (32160143, 31960135, and 31801064), the Inner Mongolia Autonomous Region Natural Science Fund (2021ZD04), the Central Government Guiding Special Funds for the Development of Local Science and Technology (2020ZY0005), and the Young Science and Technology Talents Cultivation Project of Inner Mongolia University (21400-5223711) awarded to D.L.; by the Inner Mongolia Autonomous Region Natural Science Fund (2023QN03024) and the National Natural Science Foundation of China (32560913) awarded to X.Y.; by the National Natural Science Foundation of China (32470303) awarded to X.L.; and by the Inner Mongolia Autonomous Region Natural Science Fund (2025QN03037) awarded to S.J.

Acknowledgments

No conflict of interest is declared.

Author contributions

D.L., X.Y., and J.L. conceived the study. Y.L., S.J., R.G., M.W., H.W., S.G., Z.L., Q.M., C.L., H.L., J.G., X.L., and T.W. performed the experiments. Y.L., S.J., R.G., D.L., X.Y., H.L., J.G., and J.L. analyzed the data. Y.L., S.J., R.G., J.L., X.Y., and D.L. wrote the manuscript.

Published: October 8, 2025