The MOM1 complex recruits the RdDM machinery via MORC6 to establish de novo DNA methylation

MORPHEUS’ MOLECULE1 (MOM1) is an Arabidopsis factor previously shown to mediate transcriptional silencing independent of major DNA methylation changes. Here we find that MOM1 localizes with sites of RNA-directed DNA methylation (RdDM). Tethering MOM1 with an artificial zinc finger to an unmethylated FWA promoter leads to establishment of DNA methylation and FWA silencing. This process is blocked by mutations in components of the Pol V arm of the RdDM machinery, as well as by mutation of MICRORCHIDIA 6 (MORC6). We find that at some endogenous RdDM sites, MOM1 is required to maintain DNA methylation and a closed chromatin state. In addition, efficient silencing of newly introduced FWA transgenes is impaired in the mom1 mutant. In addition to RdDM sites, we identify a group of MOM1 peaks at active chromatin near genes that colocalized with MORC6. These findings demonstrate a multifaceted role of MOM1 in genome regulation.

MORPHEUS' MOLECULE1 (MOM1) is an Arabidopsis factor previously shown to mediate transcriptional silencing independent of major DNA methylation changes. Here we find that MOM1 localizes with sites of RNA-directed DNA methylation (RdDM). Tethering MOM1 with an artificial zinc finger to an unmethylated FWA promoter leads to establishment of DNA methylation and FWA silencing. This process is blocked by mutations in components of the Pol V arm of the RdDM machinery, as well as by mutation of MICRORCHIDIA 6 (MORC6). We find that at some endogenous RdDM sites, MOM1 is required to maintain DNA methylation and a closed chromatin state. In addition, efficient silencing of newly introduced FWA transgenes is impaired in the mom1 mutant. In addition to RdDM sites, we identify a group of MOM1 peaks at active chromatin near genes that colocalized with MORC6. These findings demonstrate a multifaceted role of MOM1 in genome regulation.
Transcriptional silencing is critical to keep transposable elements (TEs) and DNA repeats under control in eukaryotic genomes. The process of transcriptional silencing involves several elaborate mechanisms involving many proteins as well as DNA methylation and histone modifications 1,2 . In Arabidopsis, the MORPHEUS' MOLECULE1 (MOM1) gene, which was originally identified with the phenotype of reactivation of a DNA-methylated and silenced hygromycin-resistance transgene in the mom1 mutant 3 , is a distinct component of the transcriptional silencing machinery. In the mom1 mutant, a set of TEs, mainly located in pericentromeric regions 4,5 , is robustly activated without major alteration in DNA methylation patterns [5][6][7] . In addition, no obvious visible decompaction of heterochromatin at chromocenters was observed in the mom1 mutant [8][9][10] . The mechanism of MOM1 mediated silencing remains elusive.
MOM1 encodes a large protein (2001 amino acids) with sequence homology to the ATPase domain of SWI2/SNF2 family proteins 3 .
However, this SNF2 homology sequence is largely dispensable for MOM1's silencing function 11 . Instead, the Conserved MOM1 Motif 2 (CMM2) domain, which is conserved among MOM1 orthologs, is required for the silencing function of MOM1 11 . The CMM2 domain of MOM1 multimerizes with itself and interacts with two PIAS (PROTEIN INHIBITOR OF ACTIVATED STAT)-type SUMO E3 ligase-like proteins, PIAL1 and PIAL2 5,12 . The pial1 pial2 double mutant phenotype highly resembles the endogenous TE de-repression phenotype of mom1 5 , suggesting that the PIAL proteins and the MOM1 protein function in the same pathway. However, evidence suggests that the SUMO ligase activity is not required for the transcriptional silencing by PIAL2, and the interaction of MOM1 and PIAL2 with SUMO is also not required for the silencing function of the MOM1 complex 5,13 .
RNA-seq analysis shows that the majority of up-regulated genes and TEs in the mom1 mutant and in the nrpe1 mutant (mutant of the largest subunit of Pol V) do not overlap 5,35 . In addition, some genes are exclusively up-regulated in the mom1 nrpe1 double mutant 35 , and a mutant allele of nrpe1 was identified in a screen for enhancers of the de-repression of a transgenic luciferase reporter in the mom1 background 35 . These studies suggest that, although MOM1 mediated transcriptional silencing and RdDM function as two different pathways, they also can act cooperatively to silence some endogenous and transgene targets.
The Arabidopsis microrchidia (MORC) proteins were discovered as additional factors required for gene silencing downstream of DNA methylation 36 . In addition, MORCs associate with components of the RdDM pathway, are loaded onto sites of RdDM and are needed for the efficiency of RdDM maintenance at some sites [37][38][39][40] . The connection between the RdDM pathway and the MORC proteins has also been demonstrated through experiments targeting the FWA gene. In wild type plants, FWA expression is silenced in all tissues except the endosperm due to DNA methylation in the promoter 41 . In the fwa-4 epimutant (fwa), the FWA gene promoter is unmethylated leading to constitutive expression of the FWA gene and late flowering phenotype 42 . Tethering MORC proteins to the unmethylated promoter of the FWA gene in the fwa mutant via protein fusion to an artificial zinc finger protein 108 (ZF) led to efficient methylation of the promoter via recruitment of the RdDM machinery 40,43 . In addition, mutations of the MORC genes impair the efficient de novo methylation and silencing of FWA transgenes 40 .
Previous studies have identified functional similarities between MORC proteins and the MOM1 complex. Multiple screens using silenced transgene reporters have identified mutations in both MOM1 and MORC6 5,6 , suggesting that they are both required for maintaining the silenced state of these transgenes. Analysis of gene expression defects in mutants has shown that most of derepressed TEs in the morc6 mutant were also derepressed in mom1, while another group of TEs are uniquely derepressed only in the mom1 morc6 double mutant 6 . Thus, investigating the relationship between the RdDM machinery, MORC proteins and the MOM1 complex should help to understand the convergence and divergence in their functions.
In this study, by performing chromatin immunoprecipitation sequencing (ChIP-seq), we observe a strong colocalization of MOM1 complex components, with the MORC6 protein and RdDM sites. Tethering of MOM1 complex components to the FWA promoter in the fwa mutant by ZF fusion leads to the establishment of DNA methylation and silencing of the FWA gene. By transforming ZF fusions into mutants, we discover that the establishment of DNA methylation by ZF-MOM1 is not only blocked by the mutants of the downstream components of the RdDM pathway, but also blocked in morc6. Furthermore, an interaction between PIAL2 and MORC6 is detected by a yeast two-hybrid (Y2H) assay and co-immunoprecipitation (co-IP). In addition, efficient de novo methylation and silencing of an FWA transgene are impaired in the mom1 and the pial1/2 mutants. Consistent with the divergent function of the MOM1 complex and the RdDM pathway, the MOM1 complex is more enriched at TEs in pericentromeric regions, while Pol V is more enriched at TEs in the chromosome arms. MOM1 also binds to a group of RdDM independent sites, at active, unmethylated, and accessible chromatin. These results highlight the functions for MOM1 in genome regulation and help to clarify the relationship between MOM1, MORCs and RdDM.
To study the function of the MOM1 complex, ChIP-seq was performed in FLAG or Myc tagged MOM1, PIAL2, PHD1 and AIPP3 transgenic lines. Surprisingly, MOM1, PHD1, AIPP3, and PIAL2 were all highly colocalized with Pol V at RdDM sites ( Fig. 1 a and b). To further validate colocalization of the MOM1 complex with RdDM sites, we performed crosslinking IP-MS of FLAG tagged MOM1 and observed that in addition to the MOM1 complex components, several proteins in the RdDM machinery, including NRPD2 (subunit of Pol V and Pol IV), NRPE1 (subunit of Pol V), DMS3 and SPT5L (P = 0.01243) were also enriched ( Fig. 1c and Supplementary Data 2). Interestingly, we also observed a strong enrichment of MORC1 and MORC6 in the MOM1 crosslinking IP-MS ( Fig. 1c and Supplementary Data 2), suggesting that the RdDM machinery, the MORC proteins and the MOM1 complex are co-located at the same loci, either because they are crosslinked by co-bound stretches of chromatin, or because the crosslinking process enhanced relatively weak interactions between the proteins.
Further examination of the MOM1 ChIP-seq signal over the AIPP3 peaks suggested that a group of AIPP3 binding loci were not enriched a c d  for MOM1 (Fig. 1d). We named the group of AIPP3 peaks that have MOM1 ChIP-seq signal enriched as Group 1 peaks and those with no MOM1 enrichment as Group 2 peaks. Consistent with our IP-MS data suggesting that PHD1 is a MOM1 complex component, PHD1 ChIP-seq signal was predominantly enriched in AIPP3 Group1 peaks which also bound to MOM1 ( Fig. 1d and Supplementary Fig. 1). We also performed ChIP-seq with FLAG tagged PHD3 transgenic plants. In contrast to PHD1, PHD3 ChIP-seq signal was enriched in both groups of AIPP3 peaks, closely resembling the pattern of AIPP3 ChIP-seq signal ( Fig. 1d and Supplementary Fig. 1). These data further suggests that AIPP3 exists in multiple protein complexes including the MOM1 complex.

MOM1-ZF triggers DNA methylation at the FWA promoter
Since MOM1 localized at RdDM sites, and ZF fusions of RdDM components have been shown to silence FWA expression in the fwa mutant 43 , we investigated whether tethering the components of the MOM1 complex could also lead to the silencing of FWA expression. We created ZF fusion proteins with MOM1, MOM2, PIAL1, PIAL2, AIPP3 and PHD1 and transformed them into the fwa mutant. ZF fusion of MOM1, MOM2, PIAL1, PIAL2 and PHD1 restored the early flowering phenotype (Fig. 2a, Supplementary Fig. 2a), repressed FWA expression (Fig. 2b), and induced DNA methylation at the FWA promoter region as detected by bisulfite amplicon sequencing analysis (BS-PCR-seq) (Fig. 2c). The DNA methylation induced at the FWA promoter region was retained in the transgene-free T2 plants, showing that the newly established DNA methylation was heritable (Fig. 2c). PIAL1-ZF was somewhat less efficient at restoring the early flowering phenotype in the T1 population ( Supplementary Fig. 2a). However, reduced FWA mRNA levels and increased FWA promoter DNA methylation were detected in some PIAL1-ZF T1 plants ( Supplementary Fig. 2b), and plants with similar flowering time to the Col-0 were observed from three T2 populations of the earliest flowering T1 plants (Fig. 2a, Supplementary Fig. 2a). In addition, DNA methylation at the FWA promoter region was retained in T2 plants free of PIAL1-ZF transgenes, showing that PIAL1-ZF can also induce heritable DNA methylation (Fig. 2c). AIPP3-ZF led to a slightly early flowering time in the T1 population compared to the fwa control population, however, zero T1 transgenic plants and very few T2 plants flowered as early as the Col-0 control plants ( Supplementary Fig. 2a, c). A low level of DNA methylation in the FWA promoter region, mainly methylation in the CHH sequence context, was detected in the AIPP3-ZF T2 plants which were positive for the transgene ( Supplementary  Fig. 2d). However, no DNA methylation was detected in transgene-free  T2 plants segregating in the same T2 populations ( Supplementary  Fig. 2d). These data suggests that the establishment of DNA methylation by AIPP3-ZF is much weaker compared to other MOM1 complex components. Previous work reported that, in addition to the designed binding site in the FWA promoter, ZF also binds to many off-target sites in the genome 43 . Whole genome bisulfite sequencing (WGBS) showed that MOM1-ZF, MOM2-ZF, PIAL1-ZF, PIAL2-ZF and PHD1-ZF also enhanced DNA methylation at ZF off-target sites (Fig. 2d, Supplementary Fig. 3a, b). Overall, these results show that ZF fusions of the components of the MOM1 complex are able to trigger the establishment of DNA methylation and silence FWA expression in the fwa mutant, as well as establish methylation at other ZF off-target sites. The CMM2 domain has been shown to be essential for the transcriptional gene silencing function of the MOM1 protein 11,12 . We found that a ZF fusion with the CMM2 domain together with a nuclear localization signal (called miniMOM1) 11 was efficient at targeting heritable FWA methylation ( Supplementary Fig. 3c, d). We performed IP-MS with a miniMOM1-FLAG line and found peptides for MOM2, PIAL1, and PIAL2, but not for AIPP3 or PHD1 (Supplementary Data 1). These results suggest that AIPP3 and PHD1 may be dispensable for the targeting of methylation to FWA promoter.
To begin to genetically dissect the requirements for MOM1mediated establishment of FWA methylation and silencing, we first transformed MOM1-ZF and PHD1-ZF into mom1 fwa and phd1 fwa mutant backgrounds ( Supplementary Fig. 4a). MOM1-ZF was able to trigger early flowering in phd1 fwa, positioning MOM1 downstream of PHD1 ( Supplementary Fig. 4a). Consistent with this order of action, the mom1 mutant blocked PHD1-ZF activity ( Supplementary Fig. 4a). PHD1-ZF activity was also blocked in the aipp3 fwa mutant ( Supplementary  Fig. 4a). These results are consistent with IP-MS result showing that the MOM1-PHD1 interaction was abolished in the aipp3-1 mutant (Supplementary Data 1).
To further dissect the hierarchy of action of MOM1 components, we transformed PIAL2-ZF into aipp3 fwa, phd1 fwa, mom2 fwa and mom1 fwa mutant backgrounds and found that PIAL2-ZF triggered an early flowering phenotype in all mutant backgrounds ( Supplementary  Fig. 4b), suggesting that PIAL2 might act at the most downstream position within the MOM1 complex. However, we also transformed MOM1-ZF into aipp3 fwa, mom2 fwa and pial1/2 fwa, and found that MOM1-ZF was also able to trigger early flowering in all these mutant backgrounds ( Supplementary Fig. 4a), suggesting that MOM1 acts at a step parallel with PIAL1/2 in targeting DNA methylation. We did however observe that MOM1-ZF showed a lower efficiency of triggering early flowering in the pial1/2 fwa mutant compared to wild type or the other mutants (Supplementary Fig. 2a and 4a), suggesting that PIAL1/2 is required for the full functionality of MOM1-ZF. We also transformed MOM2-ZF into aipp3 fwa, phd1 fwa, mom1 fwa, and pial1/2 fwa, and MOM2-ZF was able to trigger early flowering in all the mutants except in the pial1/2 fwa background ( Supplementary Fig. 4b). As a control, we compared the flowering time in the mutant backgrounds without transgenes. mom1 fwa flowered at a similar time as compared to fwa, while pial1/2 fwa and aipp3 fwa flowered slightly earlier (Supplementary Fig. 4c), suggesting that the deficiency in triggering early flowering by ZF fusion proteins in these backgrounds is not due to differences in flowering time of mutant backgrounds. In summary, these results suggest that MOM1, and especially PIAL1/PIAL2 are acting as the most downstream factors in the MOM1 complex for establishing DNA methylation at the FWA promoter.

The MOM1 complex recruits the Pol V arm of the RdDM machinery via MORC6
Because the tethering of RdDM components to FWA has been previously shown to efficiently establish methylation of FWA 27,43 , we hypothesized that MOM1-ZF established FWA DNA methylation by recruiting the RdDM machinery. To test this hypothesis, we transformed PIAL2-ZF and MOM1-ZF into fwa backgrounds in which RdDM mutations had been introgressed, including nrpd1, suvh2/9, dms3, drd1, rdm1, nrpe1, and drm1/2 43 . PIAL2-ZF and MOM1-ZF were still capable of triggering an early flowering phenotype in nrpd1 fwa (the largest subunit of Pol IV), suggesting that Pol IV mediated siRNA biogenesis was not needed for methylation targeting by the MOM1 complex (Fig. 3a). These fusions were also capable of triggering silencing in the suvh2/9 fwa mutant background (Fig. 3a), showing that the SUVH2 and SUVH9 factors that normally recruit the DDR complex and Pol V to chromatin were not needed for silencing. However, silencing activity of PIAL2-ZF and MOM1-ZF was blocked by DDR component mutations (dms3, drd1, and rdm1) as well as by mutations in the largest subunit of Pol V (nrpe1) and the DRM de novo methyltransferases (drm1/2), as no PIAL2-ZF or MOM1-ZF T1 plants showed early flowering in these backgrounds (Fig. 3a). Some of MOM1-ZF T1 plants displayed intermediate flowering time (20-30 true leaves) in drm1/2 fwa, dms3 fwa, drd1 fwa and rdm1 fwa backgrounds (Fig. 3a). However, FWA gene expression was not decreased in the six MOM1-ZF T1 plants in the drm1/2 fwa background which had the earliest flowering time (23-25 true leaves) from this population ( Supplementary Fig. 5), suggesting that the intermediate flowering phenotype is likely due to other factors such as plant stress rather than FWA silencing. Overall, these results place the action of PIAL2-ZF and MOM1-ZF upstream of the DDR complex.
Interestingly, it was previously shown that MORC6-ZF showed an identical pattern of triggering FWA methylation in wild type, nrpd1, and suvh2/9, but not in dms3, drd1, rdm1, nrpe1, or drm1/2 43 . This similarity prompted us to test the targeting of PIAL2-ZF, MOM1-ZF, MOM2-ZF, PIAL1-ZF and PHD1-ZF in the morc6 fwa genetic background. Interestingly, we found that all these ZF fusions failed to trigger FWA silencing in morc6 fwa ( Fig. 3a and Supplementary Fig. 4d), suggesting that the MOM1 complex acts upstream of MORC6. To further confirm this order of action we transformed MORC6-ZF into fwa backgrounds in which the mom1-3, mom2-1, pial1/2, phd1-2 and aipp3-1 mutants had been introgressed. We found that MORC6-ZF could successfully target silencing of FWA in all these backgrounds ( Supplementary Fig. 4d), confirming that MORC6 acts downstream of the MOM1 complex in the targeting of FWA silencing. We also performed ChIP-seq of Myc-tagged MORC6 in the morc6-3 mutant background. Similar to the MOM1 complex reported here, and MORC4 and MORC7 proteins reported previously 40 , we observed that MORC6 was highly colocalized with Pol V at RdDM sites (Fig. 3b, c).
Given that PIAL1/PIAL2 and MOM1 appeared to be the most downstream critical components of the MOM1 complex required for triggering FWA methylation, and that ZF fusions of these proteins failed to trigger methylation in a morc6 mutant, we reasoned at least one of these components might physically interact with MORC6. Indeed, we found that PIAL2 was able to interact with MORC6 in a Yeast Two-Hybrid assay (Fig. 3d). We also confirmed this interaction by an in vivo co-immunoprecipitation assay, observing that MORC6-FLAG was able to interact with PIAL2-Myc (Fig. 3e). While there could certainly be other important interactions, these results suggest that the MOM1 complex likely recruits MORC6 in part via a physical interaction between PIAL2 and MORC6. MORC6 then triggers FWA methylation via its interaction with the RdDM machinery as previously reported 40 .
To investigate if the MOM1 complex also recruits the MORC6 protein at other loci, ChIP-seq was performed with Myc-tagged MORC6 in the backgrounds of Col-0, morc6-3 mutant, mom1-3 mutant and pial1/2 double mutant. MORC6 ChIP-seq signal over Pol V peaks was strongly decreased in the mom1-3 and pial1/2 mutant backgrounds compared to that in the backgrounds of Col-0 and morc6-3 mutant (Supplementary Fig. 6a and b), while the MORC6-Myc protein expression levels were not decreased in the mom1-3 and pial1/2 mutant backgrounds ( Supplementary Fig. 6c). At the same time, there was still residue MORC6 ChIP-seq signal over Pol V peaks in the mom1-3 and pial1/2 mutant backgrounds ( Supplementary Fig. 6a, b). These data suggest that MORC6 is recruited to RdDM sites by the MOM1 complex as well as other mechanisms.

The MOM1 complex facilitates the process of transgene silencing
Several previous screens identified MOM1 as a key component in the maintenance of the silenced state of the transgene reporters used in the screen 3,5,6 . RdDM is involved in the maintenance of DNA methylation, but also in the initial establishment of methylation. For example, studies have shown that when an extra copy of the FWA gene is introduced into Arabidopsis plants via Agrobacterium-mediated transformation, it is very efficiently methylated and silenced in the wild type background. However, this methylation and silencing is blocked in RdDM mutants, leading to overexpression and a late flowering phenotype 14,28,47 . Interestingly, the silencing of FWA transgenes was previously shown to be less efficient in the morc mutants 40 . Since the MOM1 complex is closely linked with the RdDM machinery and MORC6, we suspected that the MOM1 complex may also facilitate the efficient establishment of transgene silencing. To test this, the FWA transgene was transformed into Col-0 plants (wild type) and the mutant background of nrpe1-11, mom1-3, pial1/2, mom2-22, aipp3-1 and phd1-2. As expected 40 , the T1 transgenic plants in the nrpe1-11 background flowered much later (mean leaf number: 33.81) compared to those in the wild type background (mean leaf number: 15.91) ( Fig. 4a and Supplementary Fig. 7a). We found that T1 plants containing the FWA transgene in mom1-3 or pial1/2 mutant backgrounds also flowered later than in those in the Col-0 background, with a mean leaf number of 27.55 (mom1) and 31.98 (pial1/2) ( Fig. 4a and Supplementary Fig. 7a). We examined four late flowering T1 plants in each of the mom1-3 and pial1/2 mutant backgrounds and observed that, consistent with their late flowering phenotype, FWA mRNA levels were higher than in the wild type background (Fig. 4b upper panel). The unmethylated FWA promoter DNA fraction, as detected with McrBC digestion assay, was also higher in these T1 plants compared to wild type, suggesting that efficient establishment of DNA methylation on the FWA transgene was impaired in mom1-3 and pial1/2 mutants (Fig. 4b lower panel).
Although a small number of T1 FWA transgenic plants with a late flowering time was also observed in the mom2-2, aipp3-1 and phd1-2 backgrounds, the average flowering time of these T1 plants was not  significantly later than that of the T1 plants in the wild type background ( Fig. 4a and Supplementary Fig. 7a). In fact, the FWA transgene T1 population in the aipp3-1 background flowered significantly earlier than in wild type ( Supplementary Fig. 7a), likely due to the fact that the aipp3-1 mutant itself flowers earlier than wild type plants (Supplementary Fig. 7b), as previously reported 46 . These data suggests that MOM2, AIPP3 and PHD1 contribute minimally to efficient silencing of the FWA transgene, whereas MOM1 and PIAL1/2 contribute significantly.
In strong RdDM mutants such as nrpe1, the FWA transgene stays unmethylated and all of the T2 offspring plants with the FWA transgene show a late flowering phenotype 40 (Fig. 4c). These data suggests that instead of completely blocking FWA transgene silencing as in strong RdDM mutants like nrpe1, mutation of MOM1 or PIAL1/2 reduces the efficiency of FWA transgene silencing, similar to what was previously observed for mutation of MORC genes 40 .

MOM1 influences DNA methylation and chromatin accessibility at some RdDM sites
The strong co-localization of the MOM1 complex with RdDM sites suggests that the MOM1 complex might facilitate the endogenous function of the RdDM machinery. To test this hypothesis, we performed Whole Genome Bisulfite Sequencing (WGBS) in mom1-3, pial1/ 2, phd1-2, phd1-3, aipp3-1, and mom2-2 and analyzed these together with previously published WGBS data from the morc6-3 23 and morc1/2/ 4/5/6/7 hextuple (morchex) 39 mutants, followed by analysis using the High-Confidence Differentially Methylated Regions (hcDMRs) pipeline 7 . We observed 120 hypo CHH hcDMRs in mom1-3 (shared by two replicates), and 93 hypo CHH hcDMRs in the pial1/2 double mutant. Over these hypo CHH hcDMRs, Pol V ChIP-seq signal was enriched, and the CHH methylation level were strongly decreased in the nrpe1 mutant, suggesting that these hypo CHH hcDMRs in mom1-3 and pial1/2 mutants are RdDM sites (Supplementary Fig. 8a). Similarly, the hypo CHG hcDMRs in mom1-3 and pial/2 mutants are also mainly RdDM sites (Supplementary Fig. 8b). On the contrary, the majority of hypo CG hcDMRs in mom1-3 and pial1/2 mutants were barely enriched for Pol V ChIP-seq signal, were devoid of CHH and CHG methylation in Col-0, and were located in genes and in chromosome arms, suggesting that they are likely sites of gene body methylation ( Supplementary  Fig. 8c-e). In addition, only a very small proportion of hypo CG hcDMRs in mom1-3 and in pial1/2 double mutant overlapped (Supplementary Fig. 8f), suggesting that the majority of these hypo CG hcDMRs are unlikely due to the function of the MOM1 complex. It's likely that these hypo CG hcDMRs are accumulated random natural variations in CG methylation. Overall, these data suggests that the MOM1 complex helps maintain DNA methylation at some RdDM sites.
We next focused on CHH methylation to compare the effects of MOM1 complex and MORC mutants on RdDM sites. The hypo CHH hcDMRs in mom1-3 and pial1/2 notably overlapped with those of morc6 and morchex at RdDM sites (520 DMRs in morchex) 39 (Fig. 5a, b, Supplementary Fig. 9, Supplementary Data 3). This is consistent with an earlier analysis that showed a strong overlap of mom1 hypomethylated DMRs with those of the morchex mutant 7 . In addition, a small number of hypo CHH hcDMRs were detected in mom2-2 (n = 23) and aipp3-1 (n = 13), which also showed some overlap with those of the morchex mutant. Neither of the phd1 mutant alleles tested showed any hypo CHH hcDMRs (Supplementary Data 3). To further explore the CG variation CHG variation CHH variation  functions of MOM1 complex components at these sites, we performed RNA-seq in Col-0, morc6-3, morchex and mutants of the MOM1 complex components. We observed that the expression level of the genomic regions within 1 kb of the 520 CHH hypo-DMR regions previously found in the morchex mutant were slightly upregulated in mom1-3, pial1/2, morc6-3 and morchex mutants, but not in phd1-2, aipp3-1, pial1-2, pial2-1, or mom2-2 mutants (Fig. 5c), showing that MOM1/PIAL1/PIAL2, along with MORCs, are required for the maintenance of CHH methylation and gene silencing at a small subset of RdDM sites, while AIPP3, PHD1, and MOM2 seem to play little role in this process. Since MOM1-ZF is able to trigger early flowering in the suvh2/9 fwa background, it is possible that endogenously, the MOM1 complex is also able to recruit RdDM machinery in the absence of SUVH2/9. To test this hypothesis, we performed WGBS of the suvh2/9 double mutant together with mom1-3 and the Col-0 control. Consistent with a previous report 27 , DNA methylation in the CHH context was lost in the suvh2/9 double mutant over the majority of hypo CHH hcDMRs in the nrpe1 mutant ( Supplementary Fig. 10a), suggesting that the recruitment of the RdDM pathway by SUVH2/9 plays a predominant role at endogenous RdDM sites. Interestingly, 46 out of the 120 hypo CHH hcDMRs of the mom1-3 mutant were not identified as hypo CHH hcDMRs in the suvh2/9 double mutant. The CHH DNA methylation over these sites was largely preserved in the suvh2/9 double mutant background ( Supplementary Fig. 10b, c), suggesting that similar to the MOM1-ZF result, MOM1 is still able to trigger RdDM at these sites without SUVH2/9. At the same time, over the other 74 mom1-3 hypo CHH hcDMRs (also identified as hypo CHH hcDMRs in suvh2/9), CHH DNA methylation was strongly decreased in the mom1-3 mutant and in the suvh2/9 double mutant ( Supplementary Fig. 10d), suggesting that MOM1 and SUVH2/9 are both required for RdDM function at these loci.
We also performed ATAC-seq and detected 342 regions with increased ATAC-seq signal in the mom1-3 mutant compared to Col-0 (Fig. 5d). As expected, these regions were enriched for MOM1 ChIP-seq signal (Fig. 5d). We also found that Pol V Chip-seq signal was highly enriched in these regions (Fig. 5d, e), suggesting that the MOM1 complex reduces chromatin accessibility at a subset of RdDM sites. Consistently, DNA methylation levels in CG, CHG and CHH contexts were decreased over the majority of these regions ( Supplementary  Fig. 11). Together, these results suggest that the MOM1 complex contributes to the endogenous function of the RdDM machinery, facilitating the maintenance of DNA methylation and a more closed chromatin state at some RdDM sites.

The MOM1 complex has endogenous function divergent from the RdDM machinery
Previous studies have shown that the mom1 mutants show derepression of pericentromeric heterochromatin regions, while the targets of the RdDM machinery tends to locate in euchromatic regions of the chromosome arms 5,35,48,49 . Consistent with these differences, we observed that ChIP-seq signals of MOM1, MORCs, and to a lesser extent PIAL2 were more highly enriched on transposable elements (TEs) located in pericentromeric regions as compared to TEs located in the chromosome armsthe opposite pattern to that of Pol V ChIP-seq 33 (Fig. 6a). From our RNA-seq, mom1 and pial1/2 mutants also showed transcriptional upregulation mainly in pericentromeric regions, while up-regulated TEs in the nrpe1-11 mutant were located more broadly over the chromosomes including both pericentromeric regions and the euchromatic arms ( Supplementary Fig. 12a). Consistent with previous reports 6 , morc6-3 and morchex mutants also displayed derepression of pericentromeric regions (Supplementary Fig. 12a). Upregulated differentially expressed TEs (DE-TEs) in the morc6-3 and morchex mutants showed a prominent overlap with those of the mom1-2, mom1-3, and pial1/2 mutants (Supplementary Fig. 12b). Meanwhile, many upregulated TEs in the mom1-3 mutant are not derepressed or only mildly derepressed in the morc6-3 and morchex mutants, suggesting that the functions of the MOM1 complex and the MORC proteins do not fully overlap. The phd1, aipp3, and mom2 mutants on the other hand showed little change in expression at these sites (Supplementary Fig. 12a, b), suggesting that these factors are less important for this silencing function.
We also discovered a set of MOM1 ChIP-seq peaks that did not overlap with DNA methylation (Supplementary Fig. 12c). We initially discovered these by performing unsupervised clustering of MOM1 ChIP-seq data with Pol V ChIP-seq data 33 , and identified a group of MOM1 unique peaks not colocalizing with Pol V sites (Fig. 6b). We named the MOM1 and Pol V co-bound peaks as Cluster 1 peaks and the MOM1 unique peaks as Cluster 2 peaks (Fig. 6b). As expected, the Cluster 1 peaks were DNA methylated in all sequence contexts, while DNA methylation levels over Cluster 2 peaks were very low (Supplementary Fig. 12c). Other components of the MOM1 complex, such as the PIAL2, AIPP3 and to a lesser extent, PHD1 were also enriched at cluster 2 peaks (Fig. 6b). In addition, MORC4 40 , MORC6 and MORC7 40 co-localized with MOM1 at both Cluster 1 and 2 peaks (Fig. 6b). Interestingly, we found that the Cluster 2 peaks were enriched for active histone marks H3K4me3 and H3Ac 50 , as well as accessible chromatin indicated by ATAC-seq signal (Fig. 6c). This observation is consistent with a recent study reporting that MORC7 protein binds to active chromatin regions devoid of RdDM 40 . While H3K4me3 tends to peak after the Transcription Start Site (TSS), the MOM1 ChIP-seq signal tended to peak around the TSS of the genes near Cluster 2 peaks, similar to the ATAC-seq signal (Fig. 6d, e). The function of the MOM1 complex at these non-DNA methylated sites is currently unknown.
Overall, the ChIP-seq data suggests that while MOM1 and PIAL2 show strong localization to RdDM sites, they and the MORC proteins are more enriched in pericentromeric regions compared to the RdDM machinery. In addition, they are also present at unique active chromatin sites. The recruitment mechanism and the endogenous function of the MOM1 complex binding at the active chromatin sites need to be further investigated.

Discussion
Due to the lack of major change in DNA methylation status in derepressed transgenes and endogenous TEs in the mom1 mutant, MOM1 function has long been considered as independent of DNA methylation or downstream of DNA methylation. In our study, we observed a close link between the MOM1 complex and the RdDM machinery. By tethering the MOM1 complex with ZF in the fwa mutant, heritable DNA methylation was established at the FWA promoter, suggesting that the RdDM machinery was recruited as a result. Consistent with this, silencing and methylation of FWA were blocked in mutants of the DDR complex, as well as the nrpe1 and drm1/2 mutants, but not in the suvh2/ 9 and nrpd1 mutants. Thus, the recruitment of the de novo DNA methyltransferase DRM2 by the MOM1 complex requires the Pol V arm of the RdDM pathway. Previous MORC6-ZF tethering experiments resulted in similar results 43 , i.e., the DDR complex and the downstream Pol V arm was required for silencing of FWA. In addition, we found that mutation of MORC6 blocked FWA silencing mediated by ZF fusion to MOM1 complex components, suggesting that the MOM1 complex recruits the RdDM machinery via MORC6. This was also consistent with our observed physical interaction between PIAL2 of the MOM1 complex and MORC6. However, these observations do not exclude the possibility that physical interactions might also exist between MOM1 complex components and other components of the RdDM machinery.
We also found that MOM1 and PIAL1/2 are required for the efficiency of the establishment of methylation and silencing of FWA transgenes. Compared to RdDM mutants that completely block DNA methylation and silencing of FWA transgenes, the mom1 and pial1/2 mutants only showed a reduced efficiency of silencing, similar to what was observed in the morchex mutant. How the MOM1 complex performs this function is unclear. The MOM1 complex might facilitate the initial loading of the RdDM machinery onto the FWA transgene, or it might allow for greater retention of the loaded RdDM machinery for more efficient DNA methylation and silencing, as has been suggested for the MORCs 40 . Consistent with the connections between MOM1 and RdDM revealed by ZF tethering results and FWA transgene silencing results, our ChIP-seq data showed that the MOM1 complex highly colocalized with RdDM sites in the genome. Our analysis of WGBS data also showed that MOM1 and PIAL1/2 were required to maintain CHH methylation at a small subset of RdDM sites, which notably overlap with CHH hypoDMR sites in the morchex mutants. A previous study also reported a similar observation with WGBS data from a different mom1 mutant allele (mom1-2) 7 . Thus, aside from the previous findings that that transgene and TE silencing are released in the mom1 mutant background without major DNA methylation changes 3,5,6 , the MOM1 complex 7 , together with the MORC proteins, are also required for the maintenance of DNA methylation at a small subset of RdDM sites. It seems likely that this would be mechanistically related to the role of both MOM1 and MORCs in the establishment of FWA transgene silencing, and it is intriguing to speculate that this might reflect an ancient role of these proteins in the initial establishment of methylation and silencing of novel invading transposable elements. We also found that MORC6 ChIP-seq signal at RdDM sites was strongly decreased in the mom1 and pial1/2 mutants, suggesting that MORC6 is loaded onto endogenous RdDM sites by the MOM1 complex, together with other mechanisms 29,37,51 . Based on these observations, we propose a model (Fig. 7) for how the MOM1 complex influences the RdDM machinery as follows. The MOM1 complex is first loaded onto RdDM target sites through an unknown mechanism to facilitate the binding of the MORC6 protein. MORC6 would then enhance the recruitment the Pol V arm of the RdDM machinery to methylate target DNA, by topologically entrapping the DNA as well as directly interacting with RdDM components, thus serving as a tethering factor 29,37,51,52 . From previous studies 4,5 and data from this study, it seems clear that the MOM1 complex has at least two functions in epigenome regulation, a role in the establishment and maintenance of RdDM, and a role in the maintenance of silencing of TEs in pericentromeric regions. Furthermore, it appears that these two functions are mechanistically distinct. For example, comparison of DE-TEs and DEgenes in the nrpe1 and mom1 mutants in previous studies 5,35 indicates that the majority of their endogenous targets do not overlap. Thus, for instance, the loss of RdDM function in the nrpe1 mutant does not impair the silencing function of MOM1 at the majority of its TE targets. It's possible that the localization of MOM1 at RdDM is only needed for the silencing of the relatively small number of shared TE targets between MOM1 and RdDM. In future studies, it will be interesting to investigate the relationship between the two functions of the MOM1 complex, and identify the MOM1 complex component(s) or protein domain(s) that might be required for only one of the functions.
In addition to the localization at RdDM sites, we identified a unique set of MOM1 peaks which are enriched with active chromatin marks. This is reminiscent of an earlier study reporting that MOM1 regulates transcription in intermediate heterochromatin, which is associated with both active and repressive histone marks 48 . Interestingly, the MOM1 complex and MORCs seem to behave similarly in binding active chromatin, as MORC7 was also reported to bind active chromatin devoid of RdDM 53 , and MORCs are colocalized at these MOM1 unique peaks. The mechanism of recruiting the MOM1 complex to these unique peaks and the function of MOM1 at these active chromatin sites is unknown.
In summary, our results uncover the function for the MOM1 complex in the efficiency of both the establishment and maintenance of RNA-directed DNA methylation and gene silencing, and point to a potential function at some unmethylated euchromatic regions, suggesting that MOM1 plays multifaceted roles in epigenome regulation.

IP-MS and cross-linking IP-MS
MS/MS database searching was performed using MaxQuant (1.6.10.43) against newest Arabidopsis thaliana proteome database [http://www.uniprot.org]. Analysis of raw data was obtained from the LC-MS runs using MaxQuant with the integrated Andromeda peptide search engine using default setting with enabled LFQ normalization. Data sets were filtered at a 1% FDR at both the PSM and protein levels. The MaxQuant peptide intensity and MS/MS counts were used for all peptide quantitation. For Fig. 1c, fold of change of MS/MS counts and P value of MOM1-FLAG lines crosslinking IP-MS compared to crosslinking IP-MS of Col-0 control were calculated by LIMMA 58 (v3.52.4).

Chromatin immunoprecipitation sequencing
For chromatin immunoprecipitation sequencing (ChIP-seq), 15 mL of unopened flower buds were collected for each ChIP and flash-frozen in liquid nitrogen. The flower tissue was ground to fine powder with Retsch homogenizer in liquid nitrogen and resuspended in nuclei extraction buffer (50 mM HEPES pH 8.0, 1 M sucrose, 5 mM KCl, 5 mM MgCl 2 , 0.6% Triton X-100, 0.4 mM PMSF, 5 mM benzamidine, cOmplete EDTA-free Protease Inhibitor Cocktail (Roche), 50 µM MG132). For transgenic lines of MOM1-Myc in mom1-7 and PIAL2-Myc in pial2-1, EGS was first added to resuspended lysate to 1.5 mM and the tissue lysate was incubated at room temperature for 10 min with rotation. Then the lysate was supplemented with formaldehyde at 1% and rotated at room temperature for another 10 min followed by adding glycine to stop crosslinking. For ChIP of all other proteins, crosslinking was performed by directly supplementing formaldehyde to 1% without adding EGS, then rotated at room temperature for 10 min followed by adding glycine to stop crosslinking. The crosslinked nuclei were isolated, lysed with Nuclei Lysis Buffer and diluted with ChIP Dilution Buffer 40 . Then the lysate was sonicated for 22 min (30 s on/30 s off) with Bioruptor Plus (Diagenode). After centrifugation, antibody for FLAG epitope (M2 monoclonal antibody, Sigma F1804, 10 µL per ChIP added at a final dilution of 1:400) or for Myc epitope (Cell Signaling, 71D10, 20 µL per ChIP added at a final dilution of 1:200) were added to the supernatant and incubated at 4°C overnight with rotation. Then, Protein A and Protein G Dynabeads (Invitrogen) were added and incubated at 4°C for 2 hours with rotation. After that, the beads were washed and eluted, and the eluted chromatin was reverse-crosslinked by adding 20 µL 5 M NaCl and incubated at 65°C overnight followed by treatment of Proteinase K (Invitrogen) for 4 hours at 45°C. DNA was purified and precipitated with 3 M Sodium Acetate, GlycoBlue (Invitrogen) and ethanol at −20°C overnight. After centrifugation, the precipitated DNA was washed with ice cold 70% ethanol, air dried and dissolved in 120 µL of H 2 O. ChIP-seq libraries were prepared with Ovation Ultra Low System V2 kit (NuGEN), and sequenced on Illumina NovaSeq 6000 or HiSeq 4000 instruments.
For unsupervised clustering of Pol V and MOM1 peaks (Fig. 6b), RPKM of Pol V 33 , MOM1 and corresponding control ChIP-seqs over merged peaks of Pol V and MOM1 were calculated with custom scripts. Then, log 2 (PolV RPKM/control RPKM) and log 2 (MOM1 RPKM /control RPKM) were calculated and used for unsupervised clustering with the ConcensusClusterPlus R package (v1.60.0) 63 . For analysis of ChIP signal over TEs located in euchromatic arms versus TEs located in pericentromeric regions (Fig. 6a), the pericentromeric regions were defined by Bourguet et al. 64 .

RNA sequencing
For RNA-seq experiments, twelve-day old seedlings grown on half MS medium (Murashige and Skoog Basal Medium) were collected and flash-frozen in liquid nitrogen. RNA was extracted with Direct-zol RNA MiniPrep kit (Zymo Research) and 1 µg of total RNA was used to prepare RNA-seq libraries with TruSeq Stranded mRNA kit (Illumina), and the libraries were sequenced on Illumina NovaSeq 6000 instruments.
The raw reads of RNA-seq were aligned to the TAIR10 reference genome with bowtie2. Rsem-calculate-expression (v1.3.1) from RSEM 65 with default settings was used to calculate expression levels. DEGs and DE-TEs were calculated with "run_DE_analysis.pl" from Trinity version 2.8.5 66 and log 2 FC ≥ 1 and FDR < 0.05 were used as the cut off. RNA-seq track files in bigWig format were generated using bamCoverage of deeptools (v3.1.3) with RPKM normalization.

Whole genome bisulfite sequencing
Rosette leaves of about one-month-old Arabidopsis Col-0 wild type, phd1-2, phd1-3, mom2-2, aipp3-1, fwa plants and ZF transgenic lines (MOM1-ZF, MOM2-ZF, PIAL1-ZF, PIAL2-ZF and PHD1-ZF) T2 plants with early flowering phenotype were collected for DNA extraction using DNeasy Plant Mini Kit (QIAGEN). 500 ng DNA was sheared with Covaris S2 (Covaris) into around 200 bp at 4°C. The DNA fragments were used to perform end repair reaction using the Kapa Hyper Prep kit (Roche), and together with Illumina TruSeq DNA sgl Index Set A/B (Illumina) to perform adapter ligation. The ligation products were purified with AMPure beads (Beckman Coulter), and then converted with EpiTect Bisulfite kit (QIAGEN). The converted ligation products were used as templates, together with the primers from the Kapa Hyper Prep kit (Roche) and MyTaq Master mix (Bioline) to perform PCR. The PCR products were purified with AMPure beads (Beckman Coulter) and sequenced by Illumina NovaSeq 6000 instrument.
The WGBS data analysis was performed by aligning the raw reads to both strands of the TAIR10 reference genome using BSMAP (v.2.74) 67 , allowing up to 2 mismatches and 1 best hit. Reads with more than 3 consecutive methylated CHH sites were removed, and the methylation level was calculated with the ratio of C/(C + T). For Fig. 2d, the methylation levels at 1 kb flanking regions of ZF off target sites 43 in MOM1-ZF, MOM2-ZF, PIAL1-ZF, PIAL2-ZF and PHD1-ZF were subtracted by the methylation level of fwa and plotted with R package pheatmap (v1.0.12).

BS-PCR-seq
Rosette leaves of about one-month-old plants were collected and subject to DNA extraction with CTAB method followed by bisulfite DNA conversion using the EpiTect Bisulfite kit (QIAGEN) kit. Three regions of the FWA gene were amplified from the converted DNA with Pfu Turbo Cx (Agilent): Region 1 (chr4: 13038143-13038272), Region 2 (chr4: 13038356-13038499) and Region3 (chr4: 13038568-13038695). Primers used are listed in Supplementary Data 4. Libraries were prepared with the purified PCR product by the Kapa DNA Hyper Kit (Roche) together with TruSeq DNA UD indexes for Illumina (Illumina) and were sequenced on Illumina iSeq 100 or HiSeq 4000 instruments.
BS-PCR-seq data was analyzed by aligning the raw reads to both strands of the TAIR10 reference genome with BSMAP (v.2.90) 67 allowing up to 2 mismatches and 1 best hit. After quality filtering, the methylation level of cytosines was calculated as the ratio of C/(C + T), and customized R scripts were used to plot methylation data over the FWA region 1-3.

ATAC-seq
Fresh unopened flower buds of about one-month-old Col-0 and mom1-3 mutant plants were collected for nuclei extraction and ATAC-seq, with two replicates for each genotype. We collected nuclei from unopened flower buds 33 , which were used for ATAC-seq 69 . Unopened flower buds were collected for extraction of nuclei as follows. About 5 grams of unopened flower buds was collected and immediately transferred into ice-cold grinding buffer (300 mM sucrose, 20 mM Tris pH 8, 5 mM MgCl 2 , 5 mM KCl, 0.2% Triton X-100, 5 mM β-mercaptoethanol, and 35% glycerol). The samples were ground with Omni International General Laboratory Homogenizer on ice and then filtered through a two-layer Miracloth and a 40-µm nylon mesh Cell Strainer (Fisher). Samples were spin filtered for 10 min at 3,000 g, the supernatant was discarded, and the pellet was resuspended with 25 mL of grinding buffer using a Dounce homogenizer. The wash step was performed twice in total, and nuclei were resuspended in 0.5 mL of freezing buffer (50 mM Tris pH 8, 5 mM MgCl 2 , 20% glycerol, and 5 mM β-mercaptoethanol). Nuclei were subjected to a transposition reaction with Tn5 (Illumina). For the transposition reaction, 25 µL of 2× DMF (66 mM Tris-acetate pH 7.8, 132 mM K-Acetate, 20 mM Mg-Acetate, and 32% DMF) was mixed with 2.5 µL Tn5 and 22.5 µL nuclei suspension at 37°C for 30 min. Transposed DNA fragments were purified with ChIP DNA Clean & Concentrator Kit (Zymo). Libraries were prepared with Phusion High-Fidelity DNA Polymerase (NEB) in a system containing 12.5 µL 2x Phusion, 1.25 µL 10 mM Ad1 primer, 1.25 µL 10 mM Ad2 primer, 4 µL ddH2O, and 6 µL purified transposed DNA fragments. The ATAC-seq libraries were sequenced on HiSeq 4000 platform (Illumina).
For ATAC-seq data analysis, raw reads were adaptor-trimmed with trim_galore and mapped to the TAIR10 reference genome with Bowtie2 59 (-X 2000 -m 1). After removing duplicate reads and reads mapped to chloroplast and mitochondrial, ATAC-Seq open chromatin peaks of each replicate were called using MACS2 with parameters "-p 0.01-nomodel-shift −100-extsize 200". Consensus peaks between replicates were identified with bedtools (version 2.26.0) intersect and differential accessible peaks were called with the R packge edgeR 70 (version 3.30.0). Merged bigwig file of the two replicates were used for heatmap and metaplot.

RT-qPCR
Rossette leaves of about one-month-old plants were collected for RNA extraction with Zymo Direct-Zol RNA miniprep Kit (Zymo Research). A total of 1 µg of RNA were used for cDNA synthesis with iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed with iQ SYBR Green Supermix (Bio-Rad) and primers for qPCR were listed in Supplementary Data 4.

McrBC assay
Genomic DNA extracted with the CTAB method were treated with RNase A (Qiagen) and diluted to about 100 ng/µL. 10 µL of diluted DNA were used for McrBC digestion (NEB, 4 h at 37°C) or mock digestion (the same volume of H 2 O instead of McrBC enzyme was added with all other components the same in the reaction, was also kept for 4 h at 37°C). Relative undigested FWA promoter quantity (McrBC treated/ H 2 O treated) was determined with qPCR and primers used were listed in Supplementary Data 4.

Flowering time measurement
Total true leaf numbers (sum of rosette leaf number and cauline leaf number) after bolting of the plants were used as measurement of flowering time. Plants with less than 20 true leaf number were considered as early flowering. Detailed flowering times as raw leaf count for each plant are listed in Source Data of Figs. 2a, 3a, 4a, c, Supplementary Fig. 2a, Supplementary Fig. 2c, Supplementary Fig. 3c, Supplementary Fig. 4a-d, and Supplementary Fig. 7b. The numbers of independent plants (n) scored for each population and detailed statistics of flowering time comparison between different populations are listed in Supplementary Data 5.

Yeast two-hybrid (Y2H)
The cDNA sequences of PIAL1, PIAL2, MOM2, MORC6, and MOM1 CMM2 domain (aa1660-aa1860) 5 were first cloned into gateway entry vectors followed by LR reaction with pGBKT7-GW (Addgene 61703) and pGADT7-GW (Addgene 61702) destination vectors. Pairs of plasmid DNA for the desired protein interaction to be tested were cotransformed into the yeast strain AH109. Combinations of the empty pGBKT7-GW or pGADT7-GW vectors and the plasmids of desired proteins were used for transformation of yeast cells to test for selfactivation. Transformed yeast cells were plated on synthetic dropout medium without Trp and Leu (SD-TL) and incubated for 2-3 days to allow for the growth of positive colonies carrying both plasmids. Three yeast colonies of each tested protein interaction pairs were picked and mixed in 150 μL 1×TE solution, and 3 μL of the 1×TE solution with the yeast cells were blotted on synthetic dropout medium without Trp, Leu, and His (SD-TLH) and with 5 mM 3-amino-1,2,4-triazole (3AT) to inhibit background growth. Growth of yeast on SD-TLH with 5 mM 3AT medium after 2-3 days of incubation indicates the interaction between the GAL4-AD fusion protein and the GAL4-BD fusion protein.

Co-immunoprecipitation
A total of 2 grams of 2-week-old seedling tissue were collected from MORC6-FLAG X PIAL2-Myc F1 generation and PIAL2-Myc transgenic plants and ground into fine powder in liquid nitrogen. The tissue powder was resuspended with 10 mL IP buffer, and incubated for 20 min at 4°C. Then the lysate was centrifuged and filtered with Miracloth twice. A total of 30 μL of anti-FLAG M2 Affinity Gel (Millipore) was added to the supernatant and incubated for 2 h at 4°C. Then, the anti-FLAG beads were washed with IP buffer for 5 times, and eluted with 40 µL elution buffer (IP buffer with 100 µg/mL 3xFLAG peptide). The eluted protein was used for western blot. Anti-Myc/c-Myc antibody (9E10) HRP (Santa cruz Biotechnology sc-40 HRP, 1:3000 dilution) and monoclonal ANTI-FLAG M2 HRP (Sigma-Aldrich A8592, 1:7500 dilution) were used for western blot.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The high-throughput sequencing data generated in this study have been deposited in the National Center for Biotechnology information Gene Expression Omnibus database under accession code GSE221679. The mass spectrometry proteomics data generated in this study have been deposited in the ProteomeXchange Consortium via the PRIDE 71 partner repository under accession code PXD039991. The TAIR10 reference genome used in this study are available at The Arabidopsis Information Resource website [https://www.arabidopsis.org/ download/index-auto.jsp%3Fdir%3D%252Fdownload_files%252FGenes %252FTAIR10_genome_release]. The Col-0 DNA methylation data used in this study for screenshots in Figs. 1b, 3c, 5e, 6e and Supplementary  Fig. 6b are available in the National Center for Biotechnology information Gene Expression Omnibus database under accession code GSE124746. The DNA methylation data of the nrpe1 mutant and corresponding Col-0 control plants used in this study for Supplementary  Fig. 8a-c and Supplementary Fig. 10b-d are available in the National Center for Biotechnology information Gene Expression Omnibus database under accession code GSE39901. The Col-0 DNA methylation data used in this study for Supplementary Fig. 12c are available in the National Center for Biotechnology information Gene Expression Omnibus database under accession code GSE54677. The Col-0 DNA methylation data used in this study for Supplementary Fig. 3a are available in the National Center for Biotechnology information Gene Expression Omnibus database under accession code GSE80302 . Source data are provided as a Source Data file. Source data are provided with this paper.