The MBD–ACD DNA methylation reader complex recruits MICRORCHIDIA6 to regulate ribosomal RNA gene expression in Arabidopsis

Abstract

DNA methylation is an important epigenetic mark implicated in selective rRNA gene expression, but the DNA methylation readers and effectors remain largely unknown. Here, we report a protein complex that reads DNA methylation to regulate variant-specific 45S ribosomal RNA (rRNA) gene expression in Arabidopsis (Arabidopsis thaliana). The complex, consisting of METHYL-CpG-BINDING DOMAIN PROTEIN5 (MBD5), MBD6, ALPHA-CRYSTALLIN DOMAIN PROTEIN15.5 (ACD15.5), and ACD21.4, directly binds to 45S rDNA. While MBD5 and MBD6 function redundantly, ACD15.5 and ACD21.4 are indispensable for variant-specific rRNA gene expression. These 4 proteins undergo phase separation in vitro and in vivo and are interdependent for their phase separation. The α-crystallin domain of ACD15.5 and ACD21.4, which is essential for their function, enables phase separation of the complex, likely by mediating multivalent protein interactions. The effector MICRORCHIDIA6 directly interacts with ACD15.5 and ACD21.4, but not with MBD5 and MBD6, and is recruited to 45S rDNA by the MBD–ACD complex to regulate variant-specific 45S rRNA expression. Our study reveals a pathway in Arabidopsis through which certain 45S rRNA gene variants are silenced, while others are activated.

The METHYL-CpG-BINDING DOMAIN–ALPHA-CRYSTALLIN DOMAIN protein complex reads CG DNA methylation and recruits the effector protein MICRORCHIDIA6 to regulate variant-specific 45S rRNA expression.

IN A NUTSHELL.

Background: Eukaryotic genomes harbor thousands of 45S ribosomal RNA (rRNA) genes that are tandemly arrayed at nucleolus organizer regions (NORs). Some of these rRNA genes are silenced during development. For instance, NOR2 is silenced a few days after germination in Arabidopsis (Arabidopsis thaliana). Selective rRNA gene silencing is thought to be a manifestation of rRNA gene dosage control, a mechanism that controls the number of active rRNA genes to meet the physiological needs of the cell.

Question: Growing evidence has revealed that DNA methylation is involved in selective 45S rRNA gene silencing in both plants and animals. However, the molecular mechanisms that act downstream of DNA methylation to silence specific rRNA gene variants in Arabidopsis remain unclear.

Findings: In this study, we investigated the DNA methylation readers and effectors involved in selective rRNA gene silencing. We found that a protein complex, consisting of METHYL-CpG-BINDING DOMAIN PROTEIN5 (MBD5), MBD6, and two α-crystallin domain–containing proteins, ACD15.5 and ACD21.4, reads DNA methylation to silence specific 45S rRNA gene variants. The complex directly binds to 45S rDNA through phase separation. The four proteins in this complex are interdependent for their phase separation and function. We further demonstrated that MICRORCHIDIA6 (MORC6) is an effector that is recruited to 45S rDNA regions by the MBD–ACD complex for selective 45S rRNA silencing. Our results reveal a mechanism that functions downstream of DNA methylation to achieve selective 45S rRNA gene silencing in plants.

Next steps: We are working on elucidating the molecular mechanism underlying MORC6-mediated selective rRNA gene silencing. We will determine DNA methylation levels of 45S rRNA genes and detect NOR structures in morc6 mutants.

Introduction

In eukaryotic genomes, hundreds of nearly identical 45S ribosomal RNA (rRNA) genes are tandemly arrayed at nucleolus organizer regions (NORs). In the nucleolus, each 45S rRNA gene can be transcribed by RNA polymerase I (Pol I) to produce a single large RNA (45S pre-rRNA), which is processed into the 18S, 5.8S, and 25 to 28S (depending on species) catalytic RNAs of ribosomes (Moss et al. 2007; Kressler et al. 2010; Sáez-Vásquez and Delseny 2019). However, not all 45S RNA genes are transcribed, and they are selectively inactivated during development (Grummt and Pikaard 2003; Durut et al. 2014).

In Arabidopsis (Arabidopsis thaliana), ∼1,500 45S rRNA genes are tandemly arrayed at the northern tips of chromosomes 2 and 4, which are designated NOR2 and NOR4, respectively (Copenhaver et al. 1995; Copenhaver and Pikaard 1996a, 1996b). Despite high sequence similarity, 4 types of 45S rRNA gene variants can be defined (VAR1 to VAR4; Supplementary Fig. S1A) according to short insertion/deletions in their 3′ external transcribed spacers (3′ ETSs; Pontvianne et al. 2010; Micol-Ponce et al. 2018). VAR1, which accounts for ∼50% of all 45S rRNA genes, is localized to NOR2, while VAR2 and VAR4 are localized to NOR4. VAR3 is mainly localized to NOR4, but a small amount is also localized to NOR2 (Chandrasekhara et al. 2016; Fultz et al. 2023). All variants are expressed at early postembryonic development, but VAR1 is selectively silenced a few days after germination (Earley et al. 2010; Durut et al. 2014).

Epigenetic mechanisms, including histone modifications and DNA methylation, play a crucial role in rRNA gene silencing (Sáez-Vásquez and Delseny 2019). For example, HISTONE DEACETYLASE 6 (HDA6) is required for developmentally regulated silencing of VAR1 (Earley et al. 2010). H3K9 methyltransferases SU(VAR)3-9 HOMOLOG 5 (SUVH5) and SUVH6 play a role in selective silencing of VAR1 (Pontvianne et al. 2012). ARABIDOPSIS TRITHORAX-RELATED PROTEIN 5 (ATXR5) and ATXR6, the key enzymes involved in H3K27 monomethylation, play partially redundant roles in regulating selective rRNA gene silencing, and the subnuclear distribution of NORs is altered in atxr5 atxr6 mutants (Pontvianne et al. 2012). The initial evidence showing that DNA methylation is involved in rRNA gene silencing came from the study of nucleolar dominance, a phenomenon in which the rRNA genes derived from one parent are repressed in interspecific hybrids (Pikaard 1999; Preuss and Pikaard 2007; Costa-Nunes et al. 2010). Factors required for DNA methylation were shown to be necessary for nucleolar dominance (Preuss et al. 2008). Later, it was found that DNA methylation is important for selective rRNA gene silencing in A. thaliana (Pontvianne et al. 2013). However, the mechanism that acts downstream of DNA methylation to silence variant-specific rRNA genes is still unclear.

Cytosine DNA methylation (5mC) was originally found to be an epigenetic mechanism that plays key roles in transposable element (TE) silencing (Law and Jacobsen 2010; Matzke and Mosher 2014; Zhang et al. 2018). In plants, DNA methylation occurs in 3 sequence contexts: symmetrical CG, symmetrical CHG, and asymmetrical CHH (H represents A, T, or C) (Lister et al. 2008). DNA methylation in each sequence context is established by DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2) through an RNA-directed DNA methylation (RdDM) pathway (Cao and Jacobsen 2002; Zhong et al. 2014). Once established, CG methylation is maintained by METHYLTRANSFERASE 1 (MET1) during DNA replication, while CHG methylation is maintained by CHROMOMETHYLASE 3 (CMT3) or CMT2 (Law and Jacobsen 2010). CHH methylation at the edges of TEs and short TEs can be de novo established in each new cell cycle by the RdDM pathway (Law and Jacobsen 2010; Zhang et al. 2018), while CHH methylation at long TEs is maintained by CMT2 (Zemach et al. 2013; Stroud et al. 2014).

To interpret the information encoded by DNA methylation, eukaryotes have evolved a diverse set of DNA methylation readers, including methyl-CpG–binding domain (MBD) proteins and SET and RING finger–associated (SRA) domain-containing proteins, to specifically bind methylated DNA (Coelho et al. 2022). A. thaliana contains 13 MBD proteins, among which MBD5, MBD6, and MBD7 can bind CG methylated DNA (Zemach and Grafi 2003). MBD5 and MBD6 repress the transcription of a subset of target genes by recruiting the J-domain protein SILENZIO (SLN) during the seedling stage (Ichino et al. 2021). This repressor complex is particularly critical for TE silencing in the pollen vegetative cell, in which chromatin undergoes decompaction (Ichino et al. 2022). MBD7 facilitates active DNA demethylation by recruiting a histone acetyltransferase complex (Lang et al. 2015; Li et al. 2015; Wang et al. 2015a). A. thaliana contains 9 SRA domain-containing proteins belonging to the SUVH clade. SUVH2 and SUVH9 are required for the recruitment of RNA polymerase V (Pol V), a plant-specific RNA polymerase producing long scaffold RNAs, to target sites for RdDM (Johnson et al. 2014, 2008). SUVH2 and SUVH9 also act downstream of DNA methylation to silence TEs by remodeling the higher order chromatin structure (Liu et al. 2014; Jing et al. 2016).

The interpretation of DNA methylation also depends on effector proteins. The microrchidia (MORC) proteins are a conserved family of GHKL-type ATPases that are required for TE silencing and chromatin compaction in plants (Dong et al. 2018). A. thaliana contains 6 MORC genes (MORC1, 2, 4, 5, 6, and 7, with MORC3 being a pseudogene; Harris et al. 2016), which are redundantly required for the efficient establishment of RdDM and de novo gene silencing (Xue et al. 2021). MORC proteins also act downstream of DNA methylation for gene repression. MORC6, also known as DEFECTIVE IN MERISTEM SILENCING 11 (DMS11), forms stable heterodimers with either MORC1 or MORC2 to mediate heterochromatin condensation with the assistance of SUVH2 and SUVH9 and INVOLVED IN DE NOVO 2 (IDN2; Moissiard et al. 2014; Jing et al. 2016; Liu et al. 2016).

The α-crystallin domain (ACD) proteins, also known as small heat shock proteins, are ubiquitous proteins in all cellular organisms. In mammals, ACD proteins play important roles in multiple cellular processes and help protect against diseases such as cataracts and neurodegenerative disorders (Welsh and Gaestel 1998; Basha et al. 2012; Hilton et al. 2013). In Arabidopsis, IDM2 (ACD39.0) and IDM3 (ACD51.9) are components of the IDM complex that functions in active DNA demethylation and prevents transcriptional gene silencing (Qian et al. 2014; Lang et al. 2015; Li et al. 2015). ACD15.5/RDS2 and ACD21.4/RDS1 form a stable complex with MBD5 and MBD6 and are required for transcriptional silencing of reporter genes and endogenous genes (Feng et al. 2021; Ichino et al. 2021). Although many studies have proposed that these proteins may function as molecular chaperones, the molecular functions of ACD proteins remain unclear.

In this study, we used A. thaliana as a model species to investigate the DNA methylation readers and effectors involved in rRNA gene silencing. We identified a protein complex, consisting of MBD5, MBD6, ACD15.5, and ACD21.4, that regulates the selective expression of 45S RNA genes in A. thaliana. Interestingly, these proteins have liquid-like properties and undergo liquid–liquid phase separation (LLPS) in vitro and in vivo. ACD15.5 and ACD21.4 are essential for the proper localization of MBD5 and MBD6 at chromocenters and 45S rDNA regions and vice versa. Furthermore, we demonstrated that this complex could recruit MORC6 to 45S rDNA regions to regulate variant-specific 45S rRNA expression. Our results reveal a mechanism that functions downstream of DNA methylation to regulate selective expression of 45S rRNA genes in A. thaliana.

Results

MBD5 and MBD6 redundantly regulate variant-specific rRNA gene expression

To validate the role of DNA methylation in variant-specific 45S rRNA gene silencing, we amplified the transcripts of different 45S rRNA gene variants in wild-type Col-0 and met1-11 and drm1 drm2 cmt2 cmt3 (ddcc) mutants, 2 mutants that lack CG methylation and non-CG methylation (Stroud et al. 2014; Liang et al. 2022), respectively, using a pair of primers (p3/p4) that flank the 3′ ETS variable region (Supplementary Fig. S1A). In Col-0, VAR2 and VAR3 transcripts were readily detected. Although VAR1 is the most abundant 45S rRNA gene variant, its transcript was not detected (Supplementary Fig. S1B). No obvious changes in 45S rRNA gene expression were detected in ddcc (Supplementary Fig. S1B). However, VAR1 silencing was disrupted in met1-11 (Supplementary Fig. S1B). These results are consistent with the findings of a previous study (Pontvianne et al. 2013) and suggest that CG methylation is required for VAR1 silencing.

As CG methylation can be specifically recognized by MBD5, MBD6, and MBD7 in Arabidopsis (Zemach and Grafi 2003), we tested whether these MBD proteins mediate the silencing of VAR1. To this end, we generated mbd5, mbd6, and mbd7 single mutants in the Col-0 background using CRISPR/Cas9 technology or T-DNA insertion (Supplementary Fig. S1D). To determine whether MBD5 and MBD6 are functionally redundant, we also obtained mbd5-1 mbd6-1 (referred to as mbd56-1) and mbd5-2 mbd6-2 (mbd56-2) double mutants by crossing. We found that the selective expression of 45S rRNA variants was not affected by mbd5, mbd6, or mbd7 single mutations (Supplementary Fig. S1, B and C).

However, VAR1 silencing and VAR3 expression were disrupted in the mbd56 double mutants in comparison with their levels in wild-type Col-0 (Fig. 1A), although the genomic content of each VAR remained unaltered in the mutants (Fig. 1B). Thus, VAR1 replaced VAR3 as the dominant variant in the mbd56 double mutants (Fig. 1A). Complementation of the mbd56-1 double mutant with 35S promoter-driven MBD6-GFP almost fully restored VAR1 silencing and VAR3 expression (Supplementary Fig. S1, E and F), suggesting that defects in VAR1 silencing and VAR3 expression in the mbd56 double mutants are caused by the mbd56 double mutation.

Figure 1.

The MBD–ACD complex regulates variant-specific expression of 45S rRNA genes. A and B) MBD5 and MBD6 are redundantly involved in 45S rRNA gene regulation. PCR analysis of the relative abundance of 45S rRNA A) and rDNA B) variants (VAR1-VAR4) in Col-0, mbd5, mbd6, mbd56, and met1-11 plants as indicated. The ORNITHINE TRANSCARBAMYLASE (OTC) housekeeping gene was used as an internal control. C) Immunolabeling/FISH assay showing that MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP are colocalized with NORs. MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP were immunolocalized using an anti-GFP antibody, and NORs were detected by DNA-FISH in isolated nuclei of Arabidopsis seedlings expressing 35S:MBD5-GFP, 35S:MBD6-GFP, proACD15.5:ACD15.5-GFP, or proACD21.4:ACD21.4-GFP. DNA was counterstained with DAPI. Three biological replicates were performed, and similar results were obtained. Scale bars, 5 μm. D) Association of MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP with 45S rDNA. Upper panel: schematic diagram of the positions of the primer pairs (orange bars) used for ChIP qPCR. Lower panel: ChIP qRT-PCR results. ChIP was performed in 12-d-old Col-0 and the indicated transgenic plants with anti-GFP antibodies. ACTIN7 was chosen as a negative control. The ChIP signal was quantified relative to the input DNA. Data are presented as mean ± Sd (n = 3); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). E and F) ACD15.5 and ACD21.4 regulate variant-specific 45S rRNA gene expression. PCR analysis of the relative abundance of 45S rRNA E) and rDNA F) variants (VAR1-VAR4) in Col-0, acd15.5, acd21.4, acd15.5 acd21.4, mbd56-1, and met1-11 plants as indicated. The OTC gene was used as an internal control. See also Supplementary Figs. S1 to S3 and Table S1.

Previous studies revealed that MBD6 and MBD10 are required for nucleolar dominance in allotetraploid Arabidopsis suecica hybrids (Preuss et al. 2008). To determine whether MBD10 is also involved in variant-specific 45S rRNA gene silencing, we generated an mbd10 mutant (Supplementary Fig. S1D). However, there were no obvious changes in the expression of rRNA gene variants in mbd10 (Supplementary Fig. S1, B and C), suggesting that MBD10 is not required for VAR1 silencing, although we could not exclude the possibility that MBD10 may function redundantly with other MBD proteins in 45S rRNA gene regulation. Together, our results suggest that MBD5 and MBD6 are redundantly required for variant-specific rRNA gene silencing in nonhybrid diploid A. thaliana, and different sets of factors are employed to regulate nucleolar dominance in interspecific hybrids and variant-specific rRNA gene silencing in nonhybrids.

MBD5 and MBD6 interact with ACD15.5 and ACD21.4 and form a protein complex

To elucidate the molecular mechanism underlying MBD5-mediated and MBD6-mediated variant-specific rRNA silencing, we generated transgenic plants expressing GFP-tagged MBD5 and MBD6 under the control of the 35S promoter and identified MBD5-GFP and MBD6-GFP associated proteins by performing immunoprecipitation coupled with mass spectrometry (IP-MS). We found that MBD5 and MBD6 coprecipitated with each other (Supplementary Table S1), suggesting that these 2 MBD proteins may form heterodimers or oligomers. In addition to SLN, ACD15.5 and ACD21.4, 2 α-crystallin proteins, coprecipitated with MBD5 and MBD6 (Supplementary Table S1). To confirm these results, we performed IP-MS using ACD15.5-GFP or ACD21.4-GFP transgenic plants and found that MBD5 and MBD6 coprecipitated with ACD15.5 and ACD21.4 (Supplementary Table S1).

The interactions among these 4 proteins were further confirmed by yeast 2-hybrid (Y2H) assays (Supplementary Fig. S2A), bimolecular fluorescence complementation (BiFC) assays (Supplementary Fig. S2B), and coimmunoprecipitation (Co-IP) experiments (Supplementary Fig. S2C). These results are consistent with previously published results (Feng et al. 2021; Ichino et al. 2021) and indicate that MBD5, MBD6, ACD15.5, and ACD21.4 directly interact with each other and form a protein complex in Arabidopsis (Supplementary Fig. S2D).

MBD5, MBD6, ACD15.5, and ACD21.4 are colocalized with NORs and associated with 45S rDNA

To determine the subcellular localization of the MBD–ACD complex, we isolated nuclei from F1 heterozygous plants expressing GFP-tagged MBD5 or MBD6 under the control of the 35S promoter and mCherry-tagged ACD15.5 or ACD21.4 under the control of their native promoters. These 4 proteins were colocalized, and this protein complex localized to intensely DAPI-stained chromocenters (Supplementary Fig. S2E). Fluorescence in situ hybridization (FISH) experiments further revealed that the protein complex was colocalized with 45S rDNA (Fig. 1C). To further confirm that MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP associated with 45S rDNA in vivo, we performed chromatin immunoprecipitation (ChIP) combined with quantitative PCR (qPCR). Our ChIP-qPCR results showed that MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP were enriched at 45S rDNA regions, including the gene promoter region (+1), 5′ ETS, 25S coding sequences, 3′ ETS, and intergenic spacer (IGS; Fig. 1D).

ACD15.5 and ACD21.4 are indispensable for variant-specific rRNA gene silencing

To test whether ACD15.5 and ACD21.4 are also required for 45S rRNA gene silencing, we generated acd15.5 and acd21.4 single mutants (Supplementary Fig. S3A) and obtained the acd15.5-1 acd21.4-1 (referred to as acd15.5 acd21.4) double mutant by crossing. As in mbd56-1, VAR1 was highly expressed in the acd15.5 and acd21.4 single and double mutants, but VAR3 was expressed at a very low level (Fig. 1, E and F). Changes in the dominance of different rRNA gene variants in the acd15.5-1 and acd21.4-1 mutants were fully rescued by ProACD15.5:ACD15.5-GFP and ProACD21.4:ACD21.4-GFP transgenes (Supplementary Fig. S3, B and C). These results indicate that both ACD15.5 and ACD21.4 are essential for variant-specific rRNA gene silencing in A. thaliana.

To investigate whether ACD15.5 and ACD21.4 regulate the expression of other genes, we applied RNA-seq to wild-type Col-0 and acd15.5 acd21.4 double mutant seedlings. The results showed that only 29 genes and 8 TEs were upregulated and 72 genes and 7 TEs were downregulated with |log2Foldchange| ≥ 1 and Padj ≤ 0.01 in acd15.5 acd21.4 (Supplementary Data Set 1). The results are consistent with the previous finding that mutations in MBD5 and MBD6 only affect the expression of a small number of genes at the seedling stage (Ichino et al. 2021) and suggest that ACD15.5 and ACD21.4 do not play an important role in regulating the expression of other genes in young seedlings.

MBD5, MBD6, ACD15.5, and ACD21.4 undergo LLPS in vitro and in vivo

LLPS has emerged as an important mechanism responsible for the formation of cellular foci and granules (Strom et al. 2017; Larson and Narlikar 2018; Gibson et al. 2019; Sanulli et al. 2019; Wang et al. 2020). The 4 constituent proteins of the MBD–ACD complex may form foci at chromocenters by undergoing LLPS. Because LLPS is driven by multivalent interactions, which often occur between proteins with intrinsically disordered regions (IDRs; Hyman et al. 2014; Cuevas-Velazquez and Dinneny 2018; Rippe 2022), we first examined whether ACD15.5, ACD21.4, MBD5, and MBD6 contain IDRs. As predicted by PONDR (Xue et al. 2010), ACD15.5 contains no IDR, while ACD21.4, MBD5, and MBD6 each contain 1 IDR (Supplementary Fig. S4A).

We next tested whether MBD–ACD complex proteins underwent LLPS in vitro. Recombinant GFP-MBD5, GFP-MBD6, GFP-ACD15.5, and GFP-ACD21.4 did not undergo LLPS individually under the physiological salt condition (Supplementary Fig. S4, B to D); however, a large number of spherical droplets formed after the addition of 10% PEG8000 (Fig. 2A; Supplementary Fig. S4D). The number and size of the GFP-MBD6 and GFP-ACD15.5 droplets increased as the protein concentration increased and decreased as the NaCl concentration increased (Fig. 2B; Supplementary Fig. S4, E and F). After bleaching with an intense laser, the fluorescence intensity at the center of the droplets dropped to ∼20% to 30% and recovered to ∼50% to 70% of the original fluorescence intensity within 60 s (Fig. 2, C and D).

Figure 2.

MBD5, MBD6, ACD15.5, and ACD21.4 undergo LLPS in vitro and in vivo. A) Images showing that GFP-MBD5, GFP-MBD6, GFP-ACD15.5, and GFP-ACD21.4 form droplets at the concentration of 15 μM with 10% PEG8000. Scale bars, 10 μm. B) Phase diagram of the formation of GFP-MBD6 and GFP-ACD15.5 droplets at different concentrations of proteins and NaCl. Red circles indicate phase separation. The blue scale indicates CV calculated from representative images. The experiments were performed in 2 biological replicates (5 images in each biological replicate were used for quantification analyses), and similar results were obtained. C) FRAP of GFP-MBD5, GFP-MBD6, GFP-ACD15.5, and GFP-ACD21.4 droplets. Time 0 indicates the time of the photobleaching pulse. Data are representative of 5 independent experiments. Scale bars, 0.5 μm. D) Recovery curves of GFP-MBD5, GFP-MBD6, GFP-ACD15.5, and GFP-ACD21.4 after bleaching. Data are presented as mean ± Sd (n = 5). E) FRAP of MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP in Arabidopsis root tip cells of 35S:MBD5-GFP, 35S:MBD6-GFP, ProACD15.5:ACD15.5-GFP, and ProACD21.4:ACD21.4-GFP transgenic plants, respectively. Time 0 indicates the time of the photobleaching pulse. White arrows indicate bleached foci. Data are representative of 10 nuclei for each sample. Scale bars, 2 μm. F) Recovery curves of MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP after photobleaching. Data are presented as mean ± Sd (n = 10). See also Supplementary Figs. S4 and S5.

We next determined whether MBD–ACD complex proteins underwent LLPS in vivo. The results of fluorescence recovery after photo-bleaching (FRAP) experiments showed that the MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP signals in bleached areas gradually recovered over time (Fig. 2, E and F; Supplementary Fig. S5), which indicated that these proteins were redistributed into bleached areas from the surrounding areas. These results suggest that MBD5, MBD6, ACD15.5, and ACD21.4 have liquid-like properties and form phase-separated condensates in vivo.

To determine whether LLPS of any MBD–ACD complex component is dependent on the presence of the other components in vivo, we introduced the ACD15.5-GFP or ACD21.4-GFP transgene into the mbd56-1 mutant by crossing. While ACD15-GFP and ACD21.4-GFP formed foci at chromocenters in the complementation lines, GFP signals were diffused within the nucleus in the mbd56-1 mutant (Fig. 3, A and B), suggesting that MBD5 and MBD6 are required for LLPS of ACD15.5 and ACD21.4 in vivo. We also introduced the MBD5-GFP or MBD6-GFP transgene into acd15.5-1 and acd21.4-1 mutants by crossing. The MBD5-GFP and MBD6-GFP signals in acd15.5-1 and acd21.4-1 were diffused in comparison with the corresponding signals in wild-type controls and no foci were detected, indicating that ACD15.5 and ACD21.4 are required for LLPS of MBD5 and MBD6 (Fig. 3, C and D).

Figure 3.

ACD15.5 and ACD21.4 are essential for the proper localization of MBD5 and MBD6 and vice versa. A) Subnuclear localization of ACD15.5-GFP in acd15.5-1 and mbd56-1. B) Subnuclear localization of ACD21.4-GFP in acd21.4-1 and mbd56-1. Subnuclear localization of MBD5-GFP C) and MBD6-GFP D) in Col-0, acd15.5-1 and acd21.4-1. Association of ACD21.4-GFP E) and MBD5-GFP F) with 45S rDNA in the indicated genotypes as determined by ChIP-qPCR. ChIP was performed in 12-d-old Col-0 and the indicated transgenic plants with anti-GFP antibodies. ACTIN7 was chosen as a negative control. Data are presented as mean ± Sd (n = 3); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). For A to D), images are representative of 3 independent biological replicates, and at least 10 nuclei were monitored for each replicate. Scale bars, 5 μm. See also Supplementary Table S1, Data Set 2, and Fig. S6.

ChIP-qPCR results confirmed that the association of the MBD–ACD complex with 45S rDNA region is dependent on both MBD proteins and ACD proteins (Fig. 3, E and F). It is worth noting that the levels of GFP-tagged protein were comparable between different genetic backgrounds (Supplementary Fig. S6), indicating that mutations in MBD5 and MBD6 do not affect the stability of ACD15.5 and ACD21.4 proteins and vice versa. Interestingly, IP-MS results revealed that the protein–protein interaction between MBD5 and MBD6 was disrupted in the acd15.5-1 and acd21.4-1 backgrounds, while the protein–protein interaction between ACD15.5 and ACD21.4 was disrupted in the mbd56-1 mutant (Supplementary Table S1), suggesting that the MBD–ACD complex is completely disassembled when one component is absent.

Binding with CG-methylated DNA promotes phase separation of MBD5 and MBD6

MBD5 and MBD6 each contain a conserved MBD domain (Supplementary Fig. S4A). To test whether the MBD domain is important for LLPS of MBD5 and MBD6, we mutated 2 conserved arginine residues in the MBD domain to alanine (created MBD5^R1R2 and MBD6^R1R2; Supplementary Fig. S7A). These mutations did not obviously affect LLPS of GFP-MBD5 and GFP-MBD6 in vitro (Supplementary Fig. S7, B to D), although a previous study showed that these 2 arginine residues are required for CG-methylated DNA binding (Ichino et al. 2021).

Since MBD5 and MBD6 can directly bind CG-methylated DNA in vitro and in vivo (Ichino et al. 2021), we determined whether methylated DNA can promote LLPS of GFP-MBD5 and GFP-MBD6. We mixed MBD5 or MBD6 with DNA-containing methylated CG and found that the number and size of the GFP-MBD5 and GFP-MBD6 droplets increased as the concentration of CG-methylated DNA increased (Fig. 4, A and B; Supplementary Fig. S7, E and F). However, adding methylated DNA could not promote LLPS of purified GFP-MBD5^R1R2 and GFP-MBD6^R1R2, and adding nonmethylated DNA had no obvious effect on LLPS of GFP-MBD5 and GFP-MBD6 (Fig. 4C; Supplementary Fig. S7, G and H). These results suggest that binding with CG-methylated DNA can promote GFP-MBD5 and GFP-MBD6 to undergo LLPS in vitro.

Figure 4.

DNA methylation promotes MBD5 phase separation and is essential for the proper localization of the MBD–ACD complex in the nuclear. A) Fluorescent images of GFP-MBD5 droplets in the presence of methylated DNA. GFP-MBD5 and DNA concentrations are indicated; 10% PEG8000 was added to the solution. Representative images of the GFP channel are shown. Data are representative of 3 independent experiments. Scale bars, 10 µm. Phase diagram of the formation of GFP-MBD5 droplets at different concentrations of proteins and CG-methylated DNA (mCG-DNA) B) and phase diagram of the formation of GFP-MBD5, GFP-MBD5^R1R2, GFP-MBD6, and GFP-MBD6^R1R2 droplets at different concentrations of DNA C). Red circles indicate phase separation. The blue scale indicates CV calculated from representative images. The experiments were performed in 2 biological replicates (5 images in each biological replicate were used for quantification analyses), and similar results were obtained. D) Subnuclear localization of MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP after treatment with 10 μM 5-azacytidine. DNA was stained with DAPI (blue). Scale bars, 2 μm. E) Effect of DNA methylation on the association of the MBD–ACD complex with 45S rDNA. ChIP was performed in the indicated transgenic plants (12-d-old) under control conditions or after treatment with 10 μM 5-azacytidine. ChIP was performed using anti-GFP antibodies. The ChIP signal was quantified relative to the input DNA. ACTIN7 was chosen as a negative control. Data are presented as mean ± Sd (n = 3); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). See also Supplementary Fig. S7.

To determine whether binding with methylated DNA is required for the MBD–ACD complex to form foci in vivo, we treated the MBD5-GFP, MBD6-GFP, ACD15.5-GFP, and ACD21.4-GFP transgenic plants with 5-azacytidine, an inhibitor of DNA methylation, and found that the MBD–ACD complex was not associated with chromocenters and dispersed within the nucleus in these plants (Fig. 4D). ChIP assay revealed that occupancy of the MBD–ACD complex at 45S rDNA was significantly reduced after 5-azacytidine treatment compared with the control (Fig. 4E), suggesting that binding of the MBD–ACD complex to 45S rDNA is dependent on DNA methylation.

The ACD is critical for LLPS and the proper functioning of ACD proteins

ACD15.5 and ACD21.4 each harbor a conserved ACD at the C-terminus (Supplementary Fig. S8A), which is demonstrated to be essential for the proper functioning of ACD proteins in plants (Qian et al. 2014). To determine whether the ACD is important for LLPS of ACD15.5 and ACD21.4, we replaced a conserved glycine residue in the ACD with aspartic acid (created ACD15.5^G54D and ACD21.4^G119D; Supplementary Fig. S8B). Purified GFP-ACD15.5^G54D and GFP-ACD21.4^G119D had substantially reduced ability to form droplets in vitro even after the addition of 10% PEG8000 (Fig. 5A). FRAP experiments revealed that GFP-ACD15.5^G54D and GFP-ACD21.4^G119D signals in bleached areas could not recover over time (Fig. 5, B and C). Furthermore, although it was expressed at a similar level to the wild-type protein (Supplementary Fig. S6C), ACD15.5^G54D-GFP was not associated with chromocenters and dispersed within the nucleus (Fig. 5D). The ChIP-qPCR results confirmed that ACD15.5^WT-GFP was associated with 45S rDNA, while ACD15.5^G54D-GFP had significantly reduced ability to associate with 45S rDNA (Fig. 5E). These results suggest that the conserved glycine in the ACD is essential for LLPS of ACD15.5 and ACD21.4 in vitro and in vivo.

Figure 5.

The ACD is critical for the proper functioning of ACD15.5 and ACD21.4. A) Images showing that GFP-ACD15.5, GFP-ACD15.5^G54D, GFP-ACD21.4, and GFP-ACD21.4^G119D form droplets at the concentration of 10 μM with 10% PEG8000. Scale bars, 10 μm. B) FRAP of GFP-ACD15.5^G54D and GFP-ACD21.4^G119D droplets. Time 0 indicates the time of the photobleaching pulse. Data are representative of 5 independent experiments. Scale bars, 0.5 μm. C) Recovery curves of GFP-ACD15.5^G54D and GFP-ACD21.4^G119D after bleaching. Data are presented as mean ± Sd (n = 5). D) Subnuclear localization of ACD15.5-GFP and ACD15^G54D-GFP in acd15.5-1. Scale bars, 5 μm. E) Association of ACD15.5^WT-GFP and ACD15.5^G54D-GFP with 45S rDNA as determined by ChIP-qPCR. The ChIP signal was quantified relative to the input DNA. ACTIN7 was chosen as a negative control. Data are presented as mean ± Sd (n = 3); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). F) Y2H assays. Combinations of constructs were transformed into yeast strain AH109 and assayed on medium lacking Trp and Leu (SD-TL), as well as medium lacking Trp, Leu, and His with 5 mM 3-amino-1,2,4-triazole (SD-TLH + 3AT). AD, activating domain; BD, binding domain; Data are representative of 2 independent experiments. G) The ACD is essential for the proper functioning of ACD15.5. PCR analysis of the relative abundance of 45S rDNA and rRNA variants (VAR1-VAR4) in the indicated genotypes. G54D: ACD15.5^G54D-GFP. The OTC gene was used as an internal control. See also Supplementary Figs. S8 and S9 and Supplementary Table S1.

To understand whether the ACD mediates MBD/ACD protein interaction to enable LLPS of ACD proteins, we examined whether mutations in this domain affected the interactions between MBD and ACD proteins. Our Y2H, BiFC, and Split-luciferase (Split-LUC) results revealed that ACD15.5^G54D and ACD21.4^G119D were unable to interact with MBD5 and MBD6 (Fig. 5F; Supplementary Fig. S9). The mutations also disrupted the self-interaction in ACD15.5 or ACD21.4 (Fig. 5F; Supplementary Fig. S9A). IP-MS using ACD15.5^G54D-GFP/acd15.5-1 transgenic plants further confirmed that the MBD–ACD protein complex was completely disassembled when ACD15.5 harbored the G54D mutation (Supplementary Table S1). These results suggest that the conserved glycine in the ACD is essential for the interactions between MBD and ACD proteins that lead to formation of the complex. Therefore, the ACD of ACD proteins promote LLPS of the entire MBD–ACD complex.

As the MBD–ACD protein complex was disassembled in ACD15.5^G54D-GFP acd15.5-1 transgenic plants and failed to undergo LLPS, the expression levels of 45S rRNA gene variants were similar to those in acd15.5-1, as expected (Fig. 5G). Taken together, these results suggest that the ACD is essential for LLPS and the proper functioning of ACD proteins and the entire MBD–ACD complex.

MORC6 is required for variant-specific 45S rRNA silencing

Previous studies revealed that MBD5 and MBD6 silence genes through SLN (Ichino et al. 2021). To determine whether the MBD–ACD complex also regulates variant-specific 45S rRNA gene expression through SLN, we assessed the expression of VAR1, VAR2, and VAR3 in the sln mutant. We found that the sln mutation did not change the expression patterns of 45S rRNA gene variants (Supplementary Fig. S10, A and B), suggesting that the MBD–ACD complex regulates variant-specific rRNA gene expression independently of SLN.

To identify proteins that act downstream of the MBD–ACD complex to regulate variant-specific rRNA gene expression, we obtained a panel of mutants for epigenetic regulators and detected the expression patterns of 45S rRNA gene variants in these mutants. No obvious changes in the expression of rRNA gene variants were detected in morpheus’ molecule1-2 (mom1-2), morc1-4, or suvh2 suvh9, but VAR1 silencing was disrupted in axe1-5 (hda6) and ddm1-10 (Supplementary Fig. S10, C to F). The results are consistent with the previous finding that mutations in HDA6 disrupt VAR1 silencing (Earley et al. 2010). Interestingly, like in mbd56-1, acd15.5-1, and acd21.4-1, VAR1 was highly expressed, but VAR3 was expressed at a very low level in the morc6-3 and morc6-5 mutants (Fig. 6A; Supplementary Fig. S10, C and D). Changes in the dominance of different rRNA gene variants in the morc6-3 mutant were fully rescued by the ProMORC6:MORC6-Flag transgene (Fig. 6A). These results suggest that MORC6 is required for variant-specific rRNA gene silencing in A. thaliana.

Figure 6.

The MBD–ACD complex recruits MORC6 to 45S rDNA to regulate variant-specific rRNA gene expression. A) MORC6 regulates variant-specific rRNA gene expression. PCR analysis of the relative abundance of 45S rRNA variants (VAR1-VAR4) in the indicated plants. The OTC gene was used as an internal control. B) ACD15.5 and ACD21.4 interact with MORC6 in a Y2H assay. Yeast cells expressing the indicated proteins from the pGAD-T7 (AD) and pGBK-T7 (BD) vectors were plated onto medium lacking Leu and Trp (SD-TL) or medium lacking Trp, Leu, and His with 5 mM 3-amino-1,2,4-triazole (SD-TLH + 3AT). Vec, empty vector; AD, activating domain; BD, binding domain. Data are representative of 2 independent experiments. C) Domain structure of MORC6. The boundaries of the domains are labeled on top of the diagram to match the constructs used in this study. D) Split luciferase complementation assays showing that ACD15.5 and ACD21.4 interact with MORC6 in N. benthamiana leaves. Positive interaction is highlighted with red front. Similar results were obtained from 3 biological replicates. E) Co-IP of ACD15.5 and ACD21.4 with MORC6 in transgenic plants. MORC6-Flag transgenic plants and F1 offspring seedlings coexpressing ACD15.5-GFP or ACD21.4-GFP with MORC6-Flag were used for Co-IP. Input, total protein before IP. Data are representative of 2 independent experiments. F) Effect of DNA methylation on the association of MORC6-Flag with 45S rDNA. ChIP was performed in MORC6-Flag/morc6-3 plants under control conditions or after treatment with 10 μM 5-azacytidine. G) Association of MORC6-Flag with 45S rDNA is dependent on the MBD–ACD complex. The MORC6-Flag transgene was introduced into the indicated mutant backgrounds by crossing. ChIP was performed using 12-d-old MORC6-Flag transgenic plants in the indicated mutant backgrounds. In F) and G), ChIP was performed using anti-Flag antibodies. The ChIP signal was quantified relative to the input DNA. ACTIN7 was chosen as a negative control. Data are presented as mean ± Sd (n = 3); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). See also Supplementary Figs. S10 to S13.

ACD15.5 and ACD21.4 interact with MORC6

To determine whether MORC6 functions downstream of the MBD–ACD complex, we tested whether MORC6 interacts with MBD5 and MBD6 or ACD15.5 and ACD21.4, although MORC6 peptides were not identified in the MBD–ACD protein complex IP-MS datasets. Our Y2H and Split-LUC results showed that MORC6 did not interact with MBD5 and MBD6 (Supplementary Fig. S11). However, full-length MORC6 protein interacted with ACD15.5 and ACD21.4 but not ACD15.5^G54D or ACD21.4^G119D in Y2H assays (Fig. 6B), suggesting that MORC6 interacts with ACD15.5 and ACD21.4, and the ACD is required for these interactions.

MORC6 protein contains a GHKL at the N-terminus, an S5 fold ATPase domain in the middle of the protein, and a CC domain at the C-terminus (Fig. 6C; Moissiard et al. 2012). To determine which part of MORC6 is required for its interaction with ACD15.5 and ACD21.4, we generated constructs expressing truncated MORC6 for Y2H assays (Fig. 6C). We found that all of the truncated versions of MORC6 with the GHKL domain (GHKL and GHKL-S5) interacted with ACD proteins, whereas the truncated versions of MORC6 without the GHKL domain (S5, CC, and S5-CC) failed to interact with ACD proteins (Fig. 6B), suggesting that the GHKL domain mediates the interaction between MORC6 and ACD proteins.

To confirm the protein–protein interactions between ACD proteins and MORC6, we performed Split-LUC assays, which revealed that ACD15.5 and ACD21.4 interacted with MORC6 (Fig. 6D). Moreover, we performed Co-IP experiments using transgenic plants coexpressing MORC6-Flag and ACD15.5-GFP or ACD21.4-GFP under the control of their native promoters, and the MORC6/ACD15.5 and MORC6/ACD21.4 interactions were readily detected (Fig. 6E). These results demonstrate that MORC6 directly interacts with ACD15.5 and ACD21.4.

The MBD–ACD complex recruits MORC6 to 45S rDNA

Next, we tested whether MORC6 was associated with 45S rDNA. ChIP-qPCR results showed that MORC6 was associated with 45S rDNA under normal growth conditions, but MORC6 occupancy at 45S rDNA was significantly reduced when the MORC6-Flag transgenic plants were treated with 5-azacytidine compared with the control, suggesting that MORC6 binding to 45S rDNA is dependent on DNA methylation (Fig. 6F). Given that the MBD–ACD complex recognizes DNA methylation and directly interacts with MORC6, we tested whether MORC6 is recruited to 45S rDNA by the MBD–ACD complex by introducing the MORC6-Flag transgene into mbd56-1, acd15.5-1, and acd21.4-1. ChIP-qPCR results showed that MORC6-Flag failed to accumulate at 45S rDNA in mbd56-1, acd15.5-1, and acd21.4-1 (Fig. 6G), suggesting that MORC6 is recruited to 45S rDNA by the MBD–ACD complex. Accordingly, the expression of MORC6-Flag did not restore VAR1 silencing and VAR3 expression in the mbd56-1, acd15.5-1, and acd21.4-1 mutant backgrounds (Supplementary Fig. S12).

To understand whether MORC6 regulates the activity of the MBD–ACD complex, we transiently expressed MBD5-GFP, MBD6-GFP, ACD15.5-GFP, or ACD21.4-GFP in Col-0 and morc6-3 protoplasts. The subnuclear localization of the MBD–ACD complex remained unaltered in morc6-3 protoplasts (Supplementary Fig. S13A). FRAP experiments further showed that the morc6-3 mutation does not affect LLPS of the MBD–ACD complex (Supplementary Fig. S13B). Taken together, our results suggest that MORC6 acts downstream of the MBD–ACD complex.

The MBD–ACD–MORC6 complex regulates condensation of NORs

We next explored the molecular mechanism through which the MBD–ACD complex and MORC6 regulate variant-specific 45S rRNA gene expression. DNA methylation and histone deacetylation are crucial for VAR1 silencing during plant development (Earley et al. 2010; Pontvianne et al. 2013). We thus tested whether the DNA methylation status of 45S rRNA gene promoter regions (positions −340 to +243 relative to the transcription start site, defined as +1) was affected in their mutants (Supplementary Fig. S14A; Pontvianne et al. 2013). Bisulfite sequencing results revealed that the levels of CG, CHG, and CHH methylation at 45S rRNA gene promoter regions in mbd56-1, acd15.5-1, acd21.4-1, and morc6-3 remained unaltered relative to those of Col-0 (Supplementary Fig. S14, B and C).

As the DNA methylation level we observed is the total level and VAR1 and VAR3 cannot be distinguished in this assay, we were unable to exclude that there might be opposite DNA methylation changes at VAR1 and VAR3. To further test whether VAR3 silencing in mbd56, acd15.5, acd21.4, or morc6 is dependent on DNA methylation and histone deacetylation, we treated mbd56-1, acd15.5-1, acd21.4-1, and morc6-3 mutants with 5-azacytidine or TSA, a histone deacetylation inhibitor, and found that VAR3 was derepressed (Fig. 7, A and B), suggesting that the MBD–ACD–MORC6 complex could be epistatic on DNA methylation and histone acetylation.

Figure 7.

The MBD–ACD–MORC6 complex regulates condensation of NORs. PCR analysis of the relative abundance of 45S rDNA variants (VAR1-VAR4) in reverse-transcribed RNA A) and genomic DNA B) from the indicated plants under control conditions or after treatment with 10 μM 5-azacytidine or 1 mg/L trichostatin A (TSA). The OTC gene was used as an internal control. C) FISH for 45S rDNA loci (NORs) in leaf interphase nuclei. Nuclei were counterstained with DAPI. Representative images of 1-4 NOR DNA FISH foci are shown. Scale bars, 5 μm. D) The proportion of 4 typical NOR-FISH signals observed per nucleus in Col-0 and the indicated mutants. Data are presented as mean ± Sd (n = 3 biological replicates; in each biological replicate, at least 100 nuclei per genotype were examined); different letters indicate significant differences (P < 0.01, One-way ANOVA along with LSD multiple comparison test, Supplementary Data Set 3). E) A working model for the regulation of variant-specific expression of 45S rRNA genes by the MBD–ACD–MORC6 complex. The MBD–ACD complex binds CG-methylated DNA and directly associates with 45S rDNA. MORC6 is recruited by the MBD–ACD complex to the 45S rDNA region, where it regulates variant-specific rRNA expression by chromatin remodeling. See also Supplementary Fig. S14.

MORC6 has been shown to play a crucial role in transcriptional gene silencing by enforcing heterochromatin condensation (Moissiard et al. 2012; Jing et al. 2016). To test whether the MBD–ACD–MORC6 complex regulates variant-specific 45S rRNA gene expression through modulating chromatin structures, we performed DNA-FISH to identify alterations in the chromatin structure of 45S rRNA genes in wild-type Col-0 and several mutants (Fig. 7C). Under normal conditions, inactive rRNA genes are highly condensed and sequestered in heterochromatin at the periphery of the nucleolus (Pontvianne et al. 2013). The bright FISH signals detected in the experiments correspond to the portions of the NORs consisting of these inactive rRNA genes.

Active rRNA genes are transcribed in the nucleolus, and they are visualized as black holes (low DAPI signals) in the nuclei when cells are stained with DAPI due to the paucity of their DNA (Fig. 7C). As NORs often coalesce, we frequently observed 2 or 3 NOR-FISH signals in each nucleus in Col-0 (Fig. 7D). There was no difference in the proportions of nuclei containing 2 NOR-FISH signals among Col-0 and the mbd56-1, acd15.5-1, acd21.4-1, and morc6-3 mutants. However, in mbd56-1, acd15.5-1, acd21.4-1, and morc6-3, the proportion of nuclei containing 1 NOR-FISH signal was significantly increased in comparison with Col-0, while the proportion of nuclei containing 3 NOR-FISH signals was significantly decreased (Fig. 7D). The proportion of nuclei containing 4 NOR-FISH signals was also significantly decreased in acd21.4-1 and morc6-3 compared with Col-0. Taken together, our results suggested that disruption of some NOR portions or more frequent coalescence of NORs in the groups with mutations in MORC6 and the MBD–ACD complex components.

Discussion

Variant-specific rRNA gene expression is thought to be a manifestation of rRNA gene dosage control, a mechanism that controls the number of active rRNA genes to meet the physiological needs of the cell (Lawrence and Pikaard 2004). Epigenetic mechanisms, including DNA methylation, regulate variant-specific rRNA gene expression during plant development (Pontvianne et al. 2012; Sáez-Vásquez and Delseny 2019). However, the mechanisms operating downstream of DNA methylation remain unclear.

In this study, we identified a protein complex, the MBD–ACD complex, that acts downstream of DNA methylation to regulate variant-specific 45S rRNA expression. The MBD–ACD complex can bind CG-methylated DNA and directly associates with 45S rDNA (Fig. 7E). We found that MBD5, MBD6, ACD15.5, and ACD21.4 proteins undergo LLPS, which corresponds with binding of the MBD–ACD complex to 45S rDNA. The MBD–ACD complex can recruit MORC6 to 45S rDNA. MORC6 facilitates proper condensation of NORs to ensure selective expression of certain 45S rRNA gene variants (Fig. 7E). Previous studies reported that the MBD–ACD complex is required for gene silencing due to its role in recruiting SLN (Ichino et al. 2021). However, we found that this complex regulates rRNA expression independently of SLN (Supplementary Fig. S10A). These results suggest that the MBD–ACD complex regulates the expression of protein-coding genes and 45S rRNA via different mechanisms.

VAR1 is activated in mbd56, acd15.5, and acd21.4, while VAR3 is silenced in these backgrounds (Fig. 1, A and E), suggesting that MBD–ACD inactivates VAR1 but chooses VAR3 for activation. Based on our results, MORC6-mediated chromatin remodeling should play an important role in inactivation of VAR1 and activation of VAR3 in wild-type plants (Fig. 7E). So far, it remains unknown why changes in VAR1 and VAR3 expression in the mutants show the opposite trend. We propose that because VAR1 and VAR3, which are mapped to 2 different NORs (Chandrasekhara et al. 2016; Sims et al. 2021), have different chromatin contexts, MORC6-mediated chromatin remodeling induces different effects on them.

ACD15.5 and ACD21.4 have been implicated in transcriptional silencing of protein-coding genes (Feng et al. 2021; Ichino et al. 2021). However, we did not identify many differentially expressed genes (DEGs) and TEs in the acd15.5 acd21.4 double mutant seedlings under normal growth conditions (Supplementary Data Set 1). More DEGs were identified when using mature pollen of the acd15.5 acd21.4 double mutant (Boone et al. 2023), suggesting that ACD15.5 and ACD21.4 regulate the expression of protein-coding genes mainly in specific tissues or organs. In whole seedlings, we found that an important role of ACD15.5 and ACD21.4 is to regulate rRNA gene expression. ACD15.5 and ACD21.4 cooperate with MBD5 and MBD6 to achieve this.

Interestingly, ACD15.5 and ACD21.4 can undergo LLPS in vitro and in vivo (Fig. 2). A point mutation in the conserved ACD of ACD proteins, which abolished protein interactions among the components of the MBD–ACD complex, can disrupt LLPS of ACD15.5 and ACD21.4 (Fig. 5). Moreover, we find that LLPS of ACD and MBD proteins is dependent on each other in vivo. These results suggest that multivalent protein interactions among the MBD–ACD complex are important for LLPS (Fig. 3). Thus, we have demonstrated that ACD proteins perform their functions through phase separation. Given the high level of conservation of the ACD, we speculated that IDM2 and IDM3 may undergo LLPS by interacting with other proteins in the IDM complex (Qian et al. 2014; Lang et al. 2015; Li et al. 2015), and their functions in active DNA demethylation may be closely linked to LLPS. Studies focused on determining whether ACD proteins in mammals also regulate diverse cellular processes via phase separation could reveal additional regulatory mechanisms.

MORC6, together with MORC1 and MORC2, mediate heterochromatin condensation and transcriptional silencing of TEs and protein-coding genes (Lorković et al. 2012; Moissiard et al. 2012; Xue et al. 2021). Our results indicate that MORC6 can be recruited by the MBD–ACD protein complex to 45S rDNA loci to regulate the expression of different 45S rRNA gene variants (Fig. 6). Defective recruitment or loss of MORC6 disrupted the formation of NORs (Fig. 7, C and D), suggesting that MORC6 facilitates proper formation of NORs to ensure selective expression of 45S rRNA gene variants. Although loss of function of MORC6 caused changes in selective rRNA gene expression, loss of function of MORC1 did not, suggesting that MORC1 alone is not required for proper 45S rRNA gene expression. However, it remains to be tested whether MORC1 and MORC2 play redundant roles in the regulation of selective rRNA gene expression.

Previous studies reported that SUVH2 and SUVH9 were able to bind methylated DNA and recruit MORC6 to target loci to silence TEs (Jing et al. 2016; Liu et al. 2016); however, we found no significant changes in the expression patterns of 45S rRNA gene variants in the suvh2 suvh9 double mutant (Supplementary Fig. S10, E and F), suggesting that SUVH2 and SUVH9 are not involved in selective 45S rRNA expression. Thus, MORC6 functions in 2 independent pathways: one acting through the SUVH2 and SUVH9 complex to mediate TE silencing and the other acting through the MBD–ACD complex to mediate chromatin remodeling and regulate selective 45S rRNA gene expression.

Taken together, our findings reveal a pathway via which 45S rRNA gene variants are selectively silenced by DNA methylation in A. thaliana. Our results demonstrate that MBD and ACD proteins form a complex via LLPS, which is dependent on multivalent protein interactions and the presence of both ACD components and their ACDs. Upon formation, the MBD–ACD complex recruits MORC6 to facilitate proper condensation of NORs, thus ensuring selective expression of rRNA gene variants. Our findings also provide useful insights on the molecular mechanism of nucleolar dominance, another form of rRNA gene dosage control, in interspecific hybrids.

Materials and methods

Plant materials and growth conditions

The A. thaliana T-DNA insertion mutant lines mbd6-1 (Salk_043927), mbd7-1 (GK_067A09), ddm1-10 (Salk_093009), morc6-3 (GABI_599B06), mom1-2 (Sail_610_G01), morc1-4 (Sail_1239_C08), and sln (Salk_090484) were obtained from the European Arabidopsis Stock Centre and confirmed by PCR and reverse transcription PCR (RT-PCR). The met1-11 (Liang et al. 2022), ddcc (Stroud et al. 2014), axe1-5 (hda6) (Murfett et al. 2001), morc6-5 (Jing et al. 2016), atxr5 atxr6 (Jacob et al. 2009), and suvh2 suvh9 (Kuhlmann and Mette 2012) mutants were described previously. All A. thaliana materials used in this study were generated in the Columbia-0 (Col-0) accession.

After being treated at 4 °C for 2 d, sterilized seeds were grown on 1/2 MS solid medium with 1% sucrose (m/V) at 22 °C in a growth chamber under long-day conditions (16 h of light [Philips; TLD, 36 W/865] and 8 h of darkness). The seedlings (12-d-old) were then harvested for further experiments or transplanted into soil and grown at 22 °C with the same photoperiod. Nicotiana benthamiana was grown in soil at 28°C with the same photoperiod.

Generation of mbd and acd mutants by CRISPR/Cas9 technology

The egg cell–targeting CRISPR/Cas9 system (Wang et al. 2015b) was used to generate the mbd5, mbd6, mbd10, acd15.5, and acd21.4 single mutants. All CRISPR/Cas9-generated mutants were genotyped by Sanger sequencing of the PCR-amplified genomic regions surrounding the guide targets. Cas9 construct-free mbd5, mbd6, mbd10, acd15.5, and acd21.4 mutants that were homozygous for the mutations were then obtained and used for further experiments. The mbd56 and acd15.5 acd21.4 double mutants were generated by crossing mbd5 with mbd6 and acd15.5 with acd21.4, respectively. All primers used in this study are listed in Supplementary Data Set 2, and the locations of sgRNAs are indicated in Supplementary Figs. S1D and S3A.

Plasmid construction and generation of transgenic plants

To generate Pro35S:MBD5-GFP and Pro35S:MBD6-GFP constructs, the genomic regions of MBD5 and MBD6 were amplified from Col-0 genomic DNA by PCR and cloned into the Pst I/Spe I site of pCAMBIA1300 vector by In-Fusion (Vazyme, C112). To generate ProMBD5:MBD5-GFP, ProMBD6:MBD6-GFP, ProACD15.5:ACD15.5-Myc/GFP/mCherry, and ProACD21.4:ACD21.4-Myc/GFP/mCherry constructs, MBD5, MBD6, ACD15.5, and ACD21.4 genomic DNA fragments with approximately 2-kb promoter regions were amplified and cloned into the KpnI I/Xba I site of pCAMBIA1300 vector by In-Fusion (Vazyme, C112). The ACD15.5^G54D mutation was introduced into the ProACD15.5:ACD15.5-GFP construct through site-directed mutagenesis with the Fast Mutagenesis System kit according to the manufacturer's instructions (TransGen, FM111). Agrobacterium tumefaciens strain GV3101 carrying various constructs was used to transform the wild-type (Col-0) or mutant plants via the standard floral dipping method (Clough and Bent 1998). Primary transformants were selected on 1/2 MS plates containing 30 mg/L hygromycin or 50 μg/mL carbenicillin. Homozygous transgenic lines were used for phenotypic analysis and other experiments. Crossing was used to generate ProACD15.5:ACD15.5-GFP and ProACD21.4:ACD21.4-GFP transgenic plants in the mbd56-1 background, as well as Pro35S:MBD5-GFP and Pro35S:MBD6-GFP transgenic plants in the acd15.5-1 or acd21.4-1 background. ProMORC6:MORC6-Flag transgenic plants in the morc6-3 background were generated as described previously (Liu et al. 2014). ProMORC6:MORC6-Flag transgenic plants in the mbd56-1, acd15.5-1, or acd21.4-1 background were generated by crossing.

Total RNA isolation and RT-PCR

Total RNA was extracted using TRIzol reagent (Thermo, 15596026) according to the manufacturer's protocol. For RT-PCR, contaminating DNA was removed by DNase I (Thermo, EN0521). Using random primers, approximately 1 μg of total RNA was reverse-transcribed into cDNA with the PrimeScript II First-Strand Synthesis kit (Takara, 6210A). The conditions for the amplification of 45S rRNA gene variants were as follows: 26 to 28 cycles of 30 s at 94 °C, 30 s at 53 °C, and 60 s at 72 °C. PCR primers are listed in Supplementary Data Set 2.

Affinity purification and mass spectrometry

Each 2-g sample from 12-d-old 35S:GFP, MBD5-GFP, MBD6-GFP, ACD15.5-GFP, or ACD21.4-GFP transgenic seedlings was flash frozen and ground in liquid N₂. The resulting fine powder was suspended in 15 mL lysis buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 5 mM MgCl₂, 10% glycerol [v/v], 0.2% NP-40 [v/v], 0.5 mM DTT, 1 mM PMSF, and protease inhibitor cocktail) and then incubated at 4 °C for 30 min with rotation. After centrifugation for 15 min at 18,500   g at 4°C, the supernatant was incubated with 30 μL GFP-Trap agarose (Chromotek, gta-20) for 3 h at 4 °C. The agarose beads were washed 4 times with 1.5 mL lysis buffer. The agarose bead-bound proteins were separated by SDS-PAGE, and the proteins were digested in a gel with trypsin (0.5 ng/μL). The extracted peptides were separated by HPLC and sprayed into an LTQ Orbitrap Elite System mass spectrometer (Thermo Scientific). Database searches were performed on the MASCOT server (Matrix Science Ltd., London, UK) against the IPI (International Protein Index) Arabidopsis protein database.

Y2H assay

The coding sequences of MBD5, MBD6, ACD15.5, ACD21.4, and MORC6 were amplified by PCR and then cloned into the NdeI I/Kpn I site of pGADT7 or pGBKT7 vector (Clontech). The ACD15.5^G54D and ACD21.4^G119D mutations were introduced by site-directed mutagenesis. For protein interaction analysis, 2 combinatory constructs were cotransformed into yeast (Saccharomyces cerevisiae) strain AH109. The cotransformed yeast clones were first grown on SD medium without Trp and Leu (SD-TL), and they were subsequently plated on SD medium without Trp, Leu, or His, which was supplemented with 5 mM 3-AT (SD-TLH + 3AT).

BiFC and Split-LUC assays

For BiFC assays, the full-length coding sequences of MBD5, MBD6, ACD15.5, and ACD21.4 were fused with N-YFP or C-YFP in the Not I/Asc I site of pCAMBIA1300 vector to generate N-terminal or C-terminal YFP-fusion constructs, respectively. For the Split-LUC assay, the full-length coding sequences of MBD5, MBD6, MORC6, ACD15.5, and ACD21.4 were fused with N-LUC or C-LUC in the BamH I/Sal I site of pCAMBIA1300 vector to generate N-terminal or C-terminal LUC-fusion constructs, respectively. All mutations were introduced by site-directed mutagenesis. For protein interaction analysis, A. tumefaciens GV3101 carrying different constructs was cultured overnight. After resuspension in buffer containing 10 mM MgCl₂, 150 μM acetosyringone, and 10 mM MES (pH 5.7) at an OD₆₀₀ of 1.0 and incubation for 3 h at room temperature in the dark, equal amounts of the culture were mixed in different combinations, and the mixture was infiltrated into N. benthamiana leaves. To prevent silencing of these genes, a construct encoding the viral p19 protein (Shamloul et al. 2014) was infiltrated at the same time. Two days after infiltration, luciferase activity was detected with a luminescence imaging system (Princeton Instrument), and YFP images were captured with a Zeiss LSM710 with a 40 × /1.0 objective. The YFP was excited at 514 nm and detected at 519 to 620 nm with a laser intensity of approximately 8% and a gain value of around 745.

Coimmunoprecipitation

Each 1-g sample from 12-d-old seedlings (ACD15.5-Myc and ACD21.4-Myc transgenic plants, F1 hybrids of ACD15.5-Myc and MBD5-GFP or MBD6-GFP, F1 hybrids of ACD21.4-Myc and MBD5-GFP or MBD6-GFP, and F1 hybrids of MORC6-Flag and ACD15.5-GFP or ACD21.4-GFP) was ground into a fine powder in liquid N₂. The fine powder was suspended in 15 mL lysis buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 5 mM MgCl₂, 10% glycerol [v/v], 0.2% NP-40 [v/v], 0.5 mM DTT, 1 mM PMSF, and protease inhibitor cocktail) and then incubated at 4 °C for 30 min. After centrifugation for 15 min at 18,500   g at 4°C, the supernatant was filtered through 4 layers of Miracloth. GFP-Trap agarose (Chromotek, gta-20) was prewashed twice with lysis buffer and added to the sample lysate, and the resulting mixture was incubated for 3 h at 4 °C. The agarose beads were washed 4 times with 1.5 mL lysis buffer and then boiled in 50 μL SDS-sample buffer. The immunoprecipitates were subjected to immunoblot analyses using anti-GFP (EASYBIO, BE-2002, 1:5000), anti-Myc (CWBIO, CW0299, 1:5000), and anti-Flag (Thermo, F1804, 1:5000) antibodies as primary antibodies.

RNA-seq and data analysis

For RNA-seq, frozen samples of wild-type Col-0 and acd15.5 acd21.4 were used for total RNA extraction with RNeasy Plus Mini kit (Qiagen, 74034). RNA-seq libraries were constructed using a Truseq PE Cluster v3-cBot-HS kit (Illumina) according to the manufacturer's instructions (Jiang et al. 2021). The libraries were sequenced on an Illumina Novaseq platform, and 150-bp paired-end reads were generated. RNA-seq was performed in 2 biological replicates for each genotype.

For RNA-seq data analysis, after removing adapter sequences and poor-quality reads using fastp (version 0.12.4), the clean reads were mapped to the A. thaliana reference genome (TAIR10, https://www.arabidopsis.org) using Hisat2 (version 2.0.5). The read counts of genes and TEs were calculated by featureCounts. DEG analysis between 2 groups was performed using the DESeq2 (v1.20.0). TEs and genes with an adjusted P-value ≤ 0.01 and |log₂Foldchange| ≥ 1 identified by DESeq2 were assigned as DEGs (Liao et al. 2014).

In vitro phase separation assay

The constructs for in vitro protein expression were generated by inserting the coding sequences of MBD5, MBD6, ACD15.5, and ACD21.4 into the EcoR I/Sal I site of a modified pET11 expression vector (Novagen) (Fang et al. 2019). MBD5^R1R2, MBD6^R1R2, ACD15.5^G54D, and ACD21.4^G119D mutations were introduced by site-directed mutagenesis. All proteins were expressed in the Escherichia coli BL21 (DE3) strain and purified with a Ni-NTA column (QIAGEN, 1018244). The N-terminal MBP tag was cleaved with TEV protease overnight, and GFP-tagged proteins were repurified with a Ni-NTA column (QIAGEN, 1018244). In vitro phase separation assays were performed in a buffer containing 20 mM Tris-HCl (pH 7.5) and 50 mM NaCl. PEG8000 was added to a final concentration of 10% (w/v) to induce GFP-MBD5, GFP-MBD6, GFP-ACD15.5, and GFP-ACD21.4 droplet formation. To test the phase separation capabilities of GFP-MBD5, GFP-MBD6, GFP-MBD5^R1R2, and GFP-MBD6^R1R2 in the presence of CG-methylated DNA or nonmethylated DNA, DNA fragments were added to a final concentration of 5 to 20 μM. To test the phase separation capabilities of GFP-MBD6 and GFP-ACD21.4 in the presence of ions, NaCl was added to a final concentration of 50 to 250 mM. For phase diagrams, images (.tif) were baseline corrected, and the mean pixel intensity and standard deviation measurements were calculated using ImageJ Software and used to determine the coefficient of variation (CV; McGregor et al. 2023).

Fluorescence recovery after photo-bleaching

FRAP experiments were performed in vivo and in vitro using a Zeiss LSM800 confocal laser-scanning microscope. The FRAP mode of the Zeiss 800 ZEN software was used to perform 2 prebleach scans, one bleach scan, and time-lapse postbleach scans for the indicated time periods. An Airyscan detector was used for detection, and images were processed using the ZEN Airyscan processing method. The bleaching laser intensity was set to 100% at 488 nm. Zeiss ZEN image analysis software and ImageJ processing software were used to acquire images and perform fluorescence intensity analysis.

Protoplast isolation and plasmid PEG transfection

To isolate protoplasts from Arabidopsis, fresh leaves from 4-wk-old plants (∼0.5 g) were cut into 0.5 to 1.0 mm strips or slices using fresh surgical blades on sterile filter paper. The samples were then transferred to 10 mL of 0.22 μm filter-sterilized enzyme solution (20 mM KCl, 20 mM MES, 0.4 M D-mannitol, 1% cellulase [m/V, Onozuka R-10], and 0.25% macerozyme [m/V, R-10]) immediately and incubated at 28 °C in the dark for 1 h with rotation. The released protoplasts were then harvested. The enzyme mixture containing protoplasts was diluted with 5 mL of W5 wash solution (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, and 2 mM MES). The protoplast-containing solution was filtered through a 100 μm nylon mesh into a 50 mL round-bottomed centrifuge tube and centrifuged at 100   g for 3 min to pellet the protoplasts. The protoplasts were then resuspended in W5 solution and incubated on ice for 30 min. The supernatant was carefully removed, and the protoplast pellet was resuspended in MMG solution (0.4 M mannitol, 15 mM MgCl₂, and 4 mM MES), and the yield and viability of protoplasts were estimated.

For transfection, plasmids were introduced into protoplasts using the plasmid-PEG-calcium transfection method as previously described (Yoo et al. 2007). Specifically, 10 μg of each construct was introduced into 200 μL protoplasts solution with 210 μL of PEG solution in a centrifuge tube and then incubated at room temperature for 15 min. The transfection mixture was then diluted with 800 μL of W5 solution and centrifuged at 100   g for 2 min at room temperature, and the supernatant was removed. The protoplasts were resuspended gently with 1 mL W5 solution and incubated for 16 h in dark conditions. The transformed protoplasts were observed with a confocal microscope (LSM800, Zeiss).

Extraction of nuclei

For extraction of nuclei, 0.5-g samples from 12-d-old seedlings were fixed in 30 mL 4% paraformaldehyde in 1 × PBS buffer and washed twice with ice-cold 1 × PBS buffer. The seedlings were chopped in 2 mL NEB1 (10 mM Tris-HCl [pH 7.5], 10 mM KCl, 2 mM MgCl₂, 10 mM spermine, 500 mM sucrose, and 0.1% Triton X-100 [v/v]) on a petri dish to obtain a fine homogenate on ice. The homogenate was filtered through 2 layers of Miracloth and centrifuged at 600   g at 4 °C for 3 min. The pellet was resuspended in 300 μL NEB2 (10 mM Tris-HCl [pH 7.5], 10 mM KCl, 2 mM MgCl₂, 10 mM spermine, 125 mM sucrose, and 0.1% Triton X-100 [v/v]), which was carefully added to the top of 300 μL NEB3 (10 mM Tris-HCl [pH 7.5], 10 mM KCl, 2 mM MgCl₂, 10 mM spermine, 850 mM sucrose, and 0.1% Triton X-100 [v/v]). The sample was centrifuged at 1,600   g for 30 min to precipitate the nuclei. The 2 layers of supernatant were carefully removed, and the final pellet was gently resuspended in 30 μL NEB1.

DNA fish

To produce the probe, 1 μg 45S rDNA plasmid was digested by DIG-Nick translation mix (Roche, 11745816910) and incubated at 15 °C for 1 h. One microliter of 0.5 M EDTA was added to the mixture, and the reaction was stopped by incubation at 65 °C for 10 min. Three-microliter probe was mixed with 14 μL buffer (50% deionized formamide [v/v], 2 μg/μL salmon sperm DNA, 10% dextran sulfate [m/V], and 2 × SSC buffer) and subjected to 2 cycles of thermal treatment (95 °C for 5 min and 4 °C for 5 min). The seedlings (12-d-old) were fixed in 70% Carnoy buffer (ethyl alcohol:acetic acid = 3:1) and used to make microscope slides. Seventy percent deionized formamide was added to the samples, followed by cover slipping and incubation at 85 °C for 3 min. The samples were dehydrated in precooled 70%, 90%, and 100% ethyl alcohol for 3 min, respectively, and incubated with 20 μL 45S rDNA probes at 37 °C overnight. Next, the slides were washed with 2 × SSC buffer (0.3 M NaCl and 0.03 M sodium citrate) 3 times for 8 min each time. After blocking with 100 μL 5% BSA blocking buffer at room temperature for 10 min, the samples were incubated with 20 μL anti-digoxigenin-rhodamine (Roche, 11207750910) at 37 °C for 1 h, after which the slides were washed with 2 × SSC buffer 3 times for 8 min each time. The slides were filled with DAPI and observed with a Zeiss LSM800.

Microscopy analysis

Microscopy images were collected with a Zeiss LSM800 confocal microscope with a 63 × /1.4 oil immersion objective. For the in vivo experiments, Arabidopsis seedlings were grown vertically on 1/2 MS plates with 1% (w/v) sucrose and 1.2% plant agar. Root meristem cells or cell nuclei from 5-d-old transgenic seedlings were used to examine subcellular protein localization. At least 20 cells from different roots were analyzed for each sample. For the in vitro experiments, at least 6 droplets were analyzed for each sample. The DAPI was excited at 405 nm and detected at 400 to 495 nm with a laser intensity of approximately 8% and a gain value of around 850. The GFP was excited at 488 nm and detected at 490 to 575 nm with a laser intensity of approximately 1% and a gain value of around 746. The mCherry was excited at 561 nm and detected at 570 to 700 nm with a laser intensity of approximately 5% and a gain value of around 850.

Bisulfite sequencing

Genomic DNA was extracted from 12-d-old seedlings using the DNase Plant Kit (Tiangen, DP320), and approximately 200 ng DNA was treated with the Bisulflash DNA Modification Kit (Epigetek, P-10026) following the protocol supplied by the manufacturer. Pure DNA was PCR-amplified using Epi Taq HS (Takara, R110), and the amplification products were cloned into the T-easy vector (CT111, TransGen) for sequencing. At least 20 independent clones of each sample were sequenced to calculate the relative DNA methylation level.

ChIP assay

For ChIP assays, each 2-g sample from 12-d-old seedlings was ground into a powder in liquid N₂ and cross-linked in cold ChIP extraction buffer I (10 mM Tris-HCl [pH 7.5], 10 mM MgCl₂, and 400 mM sucrose containing 1% formaldehyde [v/v]) at 4 °C for 10 min. The cross-linking reaction was quenched by the addition of glycine to a final concentration of 0.125 M. The homogenates were filtered through a cell strainer (Falcon, 352340) and pelleted by centrifugation at 2,000   g for 20 min at 4 °C. The precipitates were washed several times with ChIP extraction buffer II (10 mM Tris-HCl [pH 7.5], 10 mM MgCl₂, 250 mM sucrose, and 1% Triton X-100 [v/v]) until they became white. The nuclei were suspended in 100 μL nuclear lysis buffer (50 mM Tris-HCl [pH 8.0], 10 mM EDTA, and 1% SDS [m/V]) and incubated for 30 min at 4 °C. After the addition of 200 μL ChIP dilution buffer (16.7 mM Tris-HCl [pH 8.0], 1.2 mM EDTA, 1.1% Triton X-100 [v/v], and 167 mM NaCl), the nuclei were sonicated for 24 cycles (Diagenode, UCD-200) to yield 200 to 500 bp DNA fragments. After centrifugation, the supernatant was diluted with 700 μL ChIP dilution buffer. For ChIP-qPCR, the samples were incubated with anti-GFP (Engibody, AT0044) or anti-Flag (Thermo, F1804) overnight at 4 °C. After washing, elution, and cross-linking reversal, DNA was recovered by phenol/chloroform extraction and ethanol precipitation. Pure DNA was suspended in 50 μL ddH₂O and diluted 5 times, and a 1 μL aliquot was used for qPCR reactions. The primers used for qPCR are listed in Supplementary Data Set 2.

Statistical analysis

For all multiple comparisons, the significance of the difference between different groups was analyzed by 1-way ANOVA along with Lsd multiple comparison test at a significance level of 0.01 using SPSS 24.0. Different lowercase letters above the bars indicate significantly different groups (P < 0.01; Supplementary Data Set 3).

Accession numbers

Sequence data from this article can be found in the TAIR libraries under the following accession numbers: MBD5 (AT3G46580), MBD6 (AT5G59380), MBD7 (AT5G59800), ACD15.5 (AT1G76440), ACD21.4 (AT1G54850), MORC6 (AT1G19100), HDA6 (AT5G63110), MET1 (AT5G49160), DRM1 (AT5G15380), DRM2 (AT5G14620), CMT2 (AT4G19020), CMT3 (AT1G67990), MOM1 (AT1G08060), DDM1 (AT5G66750), MORC1 (AT4G36290), OTC (AT1G75330), and SLN (AT5G37380). The RNA-seq data of Col-0 and acd15.5 acd21.4 were deposited at National Genomics Data Center under PRJCA020893. Statistical data are shown in Supplementary Data Set 3.

Author contributions

Z.R. and W.Q. designed the research; Z.R., R.G., W.Z., Z.C., Y.C., X.Y., Y.M., Y.W., Y.L., and W.Q. performed the experiments and analyzed the data; Z.R. drafted the manuscript; Yi.W., L.-M.F., X.W.D., and W.Q. revised the manuscript.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. DNA methylation regulates the expression of rRNA gene variants.

Supplementary Figure S2. MBD5 and MBD6 interact with ACD15.5 and ACD21.4.

Supplementary Figure S3. ACD15.5 and ACD21.4 are involved in rRNA gene regulation.

Supplementary Figure S4. Purification of MBD5, MBD6, ACD15.5, and ACD21.4 and phase separation assays.

Supplementary Figure S5. Subnuclear localization and FRAP of ProMBD5:MBD5-GFP and ProMBD6:MBD6-GFP in the indicated transgenic plants.

Supplementary Figure S6. Detection of protein abundance in the transgenic plants by immunoblot.

Supplementary Figure S7. DNA methylation promotes phase separation of MBD5 and MBD6 in vitro.

Supplementary Figure S8. Domain structure of ACD15.5 and ACD21.4 and amino acid sequence alignment of the ACD.

Supplementary Figure S9. The ACD is critical for protein–protein interaction.

Supplementary Figure S10. 45S rRNA variant expression patterns in the indicated mutants.

Supplementary Figure S11. MORC6 does not interact with MBD5 and MBD6.

Supplementary Figure S12. Expression of MORC6 cannot restore the expression patterns of 45S rRNA gene variants in the indicated mutants.

Supplementary Figure S13. Mutation of MORC6 does not affect the subnuclear localization and phase separation of the MBD–ACD complex.

Supplementary Figure S14. The MBD–ACD–MORC6 complex regulates the expression of 45S rRNA variants independently of DNA methylation changes.

Supplementary Table S1. Identification of Arabidopsis MBD–ACD complex components by IP-MS.

Supplementary Data Set S1. List of DEGs and TEs in acd15.5 acd21.4.

Supplementary Data Set S2. Primers and sgRNAs used in this study.

Supplementary Data Set S3. Tables for statistical analysis.

Funding

This study was supported by the National Natural Science Foundation of China (32270288 to W.Q., 32170285 to L.-M.F., and 31925005 to Y.W.).

Data availability

All data are available in the main text or Supplemental materials. The genetic materials supporting the findings of this study are available from the corresponding author upon reasonable request.

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• SUVH2 Gramene: AT2G33290

• SUVH2 Araport: AT2G33290

• SUVH9 Gramene: AT4G13460

• SUVH9 Araport: AT4G13460

• MBD10 Gramene: AT1G15340

• MBD10 Araport: AT1G15340

• SUVH5 Gramene: AT2G35160

• SUVH5 Araport: AT2G35160

• SUVH6 Gramene: AT2G22740

• SUVH6 Araport: AT2G22740

• ATXR5 Gramene: AT5G09790

• ATXR5 Araport: AT5G09790

• ATXR6 Gramene: AT5G24330

• ATXR6 Araport: AT5G24330

• EDTA CHEBI: CHEBI:4735

• MBD5 Gramene: AT3G46580

• MBD5 Araport: AT3G46580

• MBD6 Gramene: AT5G59380

• MBD6 Araport: AT5G59380

• MBD7 Gramene: AT5G59800

• MBD7 Araport: AT5G59800

• ACD15.5 Gramene: AT1G76440

• ACD15.5 Araport: AT1G76440

• ACD21.4 Gramene: AT1G54850

• ACD21.4 Araport: AT1G54850

• MORC6 Gramene: AT1G19100

• MORC6 Araport: AT1G19100