Histone H2A monoubiquitination marks are targeted to specific sites by cohesin subunits in Arabidopsis

Histone H2A monoubiquitination (H2Aub1) functions as a conserved posttranslational modification in eukaryotes to maintain gene expression and guarantee cellular identity. Arabidopsis H2Aub1 is catalyzed by the core components AtRING1s and AtBMI1s of polycomb repressive complex 1 (PRC1). Because PRC1 components lack known DNA binding domains, it is unclear how H2Aub1 is established at specific genomic locations. Here, we show that the Arabidopsis cohesin subunits AtSYN4 and AtSCC3 interact with each other, and AtSCC3 binds to AtBMI1s. H2Aub1 levels are reduced in atsyn4 mutant or AtSCC3 artificial microRNA knockdown plants. ChIP-seq assays indicate that most binding events of AtSYN4 and AtSCC3 are associated with H2Aub1 along the genome where transcription is activated independently of H3K27me3. Finally, we show that AtSYN4 binds directly to the G-box motif and directs H2Aub1 to these sites. Our study thus reveals a mechanism for cohesin-mediated recruitment of AtBMI1s to specific genomic loci to mediate H2Aub1.

The nucleosome is the basic building block of chromatin and is composed of 146 base pairs (bp) of DNA wrapped around a histone octamer (H2A-H2B and H3-H4 dimers). The arrays of nucleosomes are connected by a shorter stretch of linker DNA and linker histone H1 1,2 . Histone tails (mostly within N-terminals) undergo various posttranslational modifications (PTMs), such as methylation, acetylation, phosphorylation, ubiquitination, and SUMOylation 3 . These PTMs impact genome functions by altering chromatin structure which plays critical roles in DNA-dependent processes, such as DNA replication, transcription, and damage repair 3 .
Similar to mammalian systems, the prevailing model in which H3K27me3 deposition by PRC2 leads to the recruitment of PRC1 and subsequent deposition of H2Aub1 has once been accepted for the recruitment of Polycomb group (PcG) complexes to specific genomic loci in Arabidopsis 7,8 . This model explained that H2Aub1 is usually associated with the H3K27me3 mark and plays a repressive role in gene transcription 5,9 . However, recent studies have shown that the functions of PRC1/2 are complex 7 . H2Aub1 is essential for H3K27me3 and PRC1mediated transcriptional regulation in Marchantia polymorpha 10 . PRC1 activity is required for the H3K27me3 modification of seed maturation genes in Arabidopsis 11 . Moreover, Arabidopsis H3K27me3 enrichment covers gene bodies whereas H2Aub1 is enriched at the regions surrounding the transcription start site (TSS) 12,13 . The ubiquitinationindependent repression by PRC1 plays an important role in determining neuronal fate in mammals 14 . Nevertheless, PRC1 can associate with polycomb response elements (PREs) even without PRC2 participation in Drosophila 15,16 . EMBRYONIC FLOWER 1 (EMF1) and LIKE HETEROCHROMATIN PROTEIN 1 (LHP1) were once thought to be PRC1 subunits but were recently identified as PRC2-associated proteins 7,17 . EARLY BOLTING IN SHORT DAYS (EBS) and its homolog SHORT LIFE (SHL) were also defined as PRC2-associated proteins 18 , and PRC-associated coiled-coil protein plays a role in the maintenance of rice shoot apical meristem activity by regulating H3K27me3 19 . As PRC1 contains no DNA-binding domain, the molecular mechanism for recruiting PRC1 to the specific genomic loci remains unclear.
The cohesin complex is highly conserved in eukaryotes and consists of SISTER CHROMATID COHESION 3 (SCC3), the α-kleisin family protein, and structural maintenance of chromosome (SMC) proteins SMC1 and SMC3. Within cohesin, SMC1 and SMC3 proteins form a long-armed V-shaped heterodimer, and the α-kleisin subunit bridges the SMC dimer, forming a ring-like structure. SCC3 interacts with the αkleisin subunit, and stabilizes the ring-like structure 20 . In Arabidopsis, homozygous T-DNA knockout mutants of AtSMC1, AtSMC3, and AtSCC3 show developmental defects in embryo and endosperm, underlying early developmental arrest in the seed [21][22][23][24] . The Arabidopsis genome encodes single copies of AtSMC1, AtSMC3, and AtSCC3, but four α-kleisin genes (AtSYN1-4) 25,26 . AtSYN1 is a meiosis-specific cohesin subunit, and the homozygous atsyn1 T-DNA insertion line is sterile in male and female gametophytes 25,27,28 . AtSYN3 is indispensable for meiotic recombination, and the null allele of AtSYN3 shows gametophyte lethality 26,29 . AtSYN2 and AtSYN4 play roles in DNA doublestrand break (DSB) repair 30,31 . After recruitment to chromatin, cohesin is established and maintained in the genome by the activity of an acetyltransferase (establishment of cohesin 1/Chromosome transmission fidelity 7, Eco1/Ctf7). Most of the homozygous atctf7 mutant seeds show embryo arrest during their early development. Only a small number of atctf7 plants can survive, but exhibit major defects in vegetative growth and development, and are completely sterile 32 . Initially, cohesin was found to function as a 'molecular glue' to cohere sister chromatids. In addition to the well-illustrated function of cohesin in mitosis and meiosis 33 , growing evidence has indicated a pivotal role of cohesin in modulating the three-dimensional (3D) genome organization and transcriptional regulation 34 . In animals, cohesin and CCCTCbinding factor (CTCF) are essential for forming topologically associating domains. The dynamic interaction of cohesin with its chromatin binding site builds a bridge between enhancers/insulators and promoters, thereby regulating gene expression 34 . In humans, the cohesin-NIPBL complex mediates chromatin loop extrusion by a 'swing and clamp' mechanism 35 . In Drosophila, PRC1 subunits and cohesin subunits SMC1/3 can be co-purified 36 . However, a functional link between cohesin and PRC1 has not been established, and the potential coordination of cohesin and H2Aub1 remains to be illustrated.
In this work, we show that AtSYN4 interacts with AtSCC3 which binds to AtBMIs. In addition, the cohesin subunits AtSYN4 and AtSCC3 are functionally related to catalytic subunits of PRC1, AtBMI1A/B/C. Chromatin immunoprecipitation followed by sequencing (ChIP-seq) assays indicate that most binding events of AtSYN4 and AtSCC3 are associated with H2Aub1 along the genome. Moreover, AtSYN4 directly binds to the genomic loci enriched with the G-box motif of DNA and associates with AtBMI1A/B/C through AtSCC3, thus affecting H2Aub1 levels at these sites. Together, these results reveal the role of cohesin subunits in the recruitment of AtBMIs to certain genomic loci to mediate H2Aub1 in these specific locations.
We have not detected a direct interaction between AtSYN4 and AtB-MI1A/B/C through luciferase complementation assay. Given that AtSCC3 can interact with AtSYN4 and AtBMIs (Fig. 1a, b and Supplementary Fig. 1b), we speculated that AtSCC3 might act as a bridge that can connect AtSYN4 with AtBMI1s. To test this hypothesis, we performed an in vivo Co-IP assay. When AtBMI1A-YFP, AtSYN4-FLAG, and AtSCC3-FLAG were co-expressed in tobacco leaves, AtSCC3-FLAG and AtSYN4-FLAG could be pulled down by AtBMI1A-YFP. In contrast, when AtBMI1A-YFP and AtSYN4-FLAG were co-expressed in tobacco leaves without AtSCC3-FLAG, AtSYN4-FLAG could not be pulled down by AtBMI1A-YFP. Similar results were obtained from AtBMI1B-YFP and AtBMI1C-YFP ( Fig. 1c; Supplementary Fig. 1c). AtBMI1A/B/C can bind AtRING1A or AtRING1B 37 . Therefore, we examined the potential interactions between AtSCC3 or AtSYN4 and AtRING1A or AtRING1B. We found that AtRING1A interacts with AtSCC3 and AtSYN4, and the interaction between AtRING1A and AtSCC3 is much stronger than that between AtRING1A and AtSYN4 (Supplementary Fig. 1d).
AtSCC3 and AtSYN4 functionally correlate with AtBMI1A/B/C Arabidopsis has a single copy of the AtSCC3 gene and its homozygous full loss-of-function alleles are embryonic lethal 24 . Therefore, we knocked down AtSCC3 by using artificial microRNAs (amiRNAs) with the miR319a precursor as a backbone. We obtained two RNA interference (RNAi) lines, SCC3RNAi-1 and SCC3RNAi-2, with the transcript level of AtSCC3 decreasing by approximately 1/3 and 2/3, respectively ( Supplementary Fig. 2a). In addition to AtSCC3 knockdown lines, we also used the atctf7 mutant, in which the cohesin complex cannot be stably maintained on DNA 38 , and atctf7 seedlings display severe dwarf and major defects in vegetative growth and development (Supplementary Fig. 2b).
Atsyn4, SCC3RNAi-1, SCC3RNAi-2, and atbmi1a/b mutant plants all exhibit phenotypes of delayed flowering and serrated rosette leaves ( Supplementary Fig. 2c-e). The phenocopies among atsyn4, SCC3RNAi-1, SCC3RNAi-2, and atbmi1a/b mutant lines and the physical interactions among AtSCC3, AtSYN4, and AtBMIs implicated the potential functional connections among these proteins. We then performed RNA sequencing (RNA-seq) to quantify transcripts in 7-day-old seedlings of Col-0, atsyn4, SCC3RNAi-2, and atctf7. The quality-control metrics of RNA-seq data are shown in Supplementary Fig. 3a, b. We found that the differentially expressed genes (DEGs) between SCC3RNAi-2 and atbmi1a/b/c 12 significantly overlapped and were positively correlated (Fig. 2a Supplementary Fig. 4c). These results implicated that AtSYN4 plays a significant role in the AtBMI-mediated pathway, and AtSYN2 might be involved in this process to some degree upon the complete depletion of AtSYN4. Comparison of atctf7 and atbmi1a/b/c transcriptome data showed that genes with the same trend of disordered expression are also apparently intersected, and positively correlated (Fig. 2e, f, Supplementary Data 2, 5).
Given that AtSCC3 interacts with AtSYN4 and AtBMI1A/B/C which are associated with H2Aub1, we examined the global H2Aub1 levels in atsyn4, SCC3RNAi-1, SCC3RNAi-2, and atctf7 mutants. Western blots showed that the global levels of H2Aub1 in SCC3RNAi-2 and atctf7 mutant plants are much lower than that in Col-0 ( Supplementary  Fig. 5a). Consistent with the weaker phenotypes of atsyn4 and SCC3RNAi-1 lines than SCC3RNAi-2 line (Supplementary Figs. 2e and 5b), a milder decrease of the global H2Aub1 level in atsyn4 and SCC3RNAi-1 lines than SCC3RNAi-2 was detected (Supplementary Fig. 5a).

AtSYN4/AtSCC3 and H2Aub1 tend to flank TSS closely
To address genomic relationships among AtSYN4, AtSCC3, and H2Aub1, we first detected the genome-wide occupancy of AtSYN4 and AtSCC3 using ChIP-seq. The quality-control metrics of ChIP-seq experiments are shown in Supplementary Fig. 6a We then genome-widely compared AtSCC3 and AtSYN4 targeted genes to H2Aub1 occupancy 12 . By visualizing on Integrative Genomics Viewer (IGV), we observed that AtSYN4 and AtSCC3 peaks tend to locate slightly upstream of the TSS, and the peaks of H2Aub1 locate slightly downstream of the TSS (Fig. 3a). In addition, we found that a large proportion of genes (~76%, 1203 of 1591) bound by AtSCC3 showed significant enrichment of AtSYN4. About 53% (4696 of 8907) and 42% (665 of 1591) of genes occupied by AtSYN4 and AtSCC3, respectively, are H2Aub1-marked genes (Fig. 3b). We then compared AtSYN4, AtSCC3, and H2Aub1 enrichment patterns across the ±5 kb flanking TSS. The heatmap and signal density plots across the ±5 kb flanking TSS suggested that AtSYN4 colocalizes largely with AtSCC3 along the genome, and the AtSCC3/AtSYN4 peaks locate at~100 bp upstream of the TSS (Fig. 3c, d, Supplementary Data 6, 7), and H2Aub1 ChIP-seq peaks are enriched at~250 bp downstream of the TSS (Fig. 3c, d).
Next, we analyzed the genes with their H2Aub1 levels downregulated by AtBMI mutations. We found that the genes with decreased H2Aub1 levels in the atbmi1a/b/c triple mutant overlap significantly with AtSYN4-or AtSCC3-targeted genes in Col-0 ( Fig. 4a), in contrast to the rare gene sharing between AtSYN4/AtSCC3 and H3K27me3 targets (Fig. 4b). In addition, H2Aub1 peaks were found closer to AtSYN4 or AtSCC3 binding sites than regions associated with the H3K27me3 signature (Fig. 4c). To examine the transcriptional states of AtSYN4 and AtSCC3 target genes, we quantified the relative transcript levels of their target genes by RNA-seq on 7-day-old wild-type seedlings. We found that the transcript levels of AtSYN4 or AtSCC3 targeted genes are significantly higher than genes nontargeted by AtSYN4 or AtSCC3, respectively (Fig. 4d), indicating that genes associated with AtSYN4/AtSCC3 are more transcriptionally active 12 .
Based on combined AtSYN4/AtSCC3 ChIP-seq and atsyn4/ AtSCC3RNAi-1 transcriptome analysis, we found that the DEGs identified in atsyn4 and AtSCC3RNAi-1 are significantly related to AtSYN4 and AtSCC3 target genes marked by H2Aub1, respectively (Supplementary Fig. 8a and 8b). Gene Ontology (GO) analysis indicated that the representative categories based on biological function include responses to wounding, chitin, and water deprivation ( Supplementary  Fig. 8c, 8d).
AtSYN4 and AtSCC3 affect H2Aub1 levels neighboring G-boxcontaining DNA How AtBMI1A/B/C finds the target sites and mediates H2Aub1 in a DNA sequence-specific manner is not clear. We speculated that AtSYN4 and AtSCC3 could facilitate AtBMI1A/B/C binding to a specific DNA sequences. To test the hypothesis, we identified the G-box motif (CACGTG) in the co-occupied loci of AtSYN4, AtSCC3, and H2Aub1 (Fig. 5a).
The IGV screenshots of the selected genes showed that AtSYN4 peaks closely locate the G-box sites and H2Aub1 peaks are adjacent to these G-box sites (Fig. 5a). In addition, more than half of the genes (~58%, 1211 of 2106) with the G-box bounded by AtSYN4 are marked by H2Aub1 (Fig. 5b), and about 35% (564 of 1591) of the genes with the G-box bounded by AtSCC3 are H2Aub1-marked. Moreover, we plotted AtSYN4, AtSCC3, and H2Aub1 enrichment patterns across the ±5 kb flanking TSS of genes targeted by AtSYN4 with the G-box element (Fig. 5c, d). The heatmap and signal density plots centered on TSS suggested that the AtSYN4 peaks with the G-box element had similar chromatin signatures with AtSCC3 and H2Aub1 across the TSS (Figs. 3c, d and 5c, d).
We then employed yeast one-hybrid (Y1H) assays to study the specific recognition between AtSYN4/AtSCC3 and the selected three DNA sequences containing the G-box in the AtSYN4, AtSCC3, and H2Aub1 co-occupied loci. The results showed that AtSYN4 could bind the G-box containing DNA directly, blocked by the mutations in the G-box (Fig. 6a). We found that the AtSCC3-occupied DNA in this G-box containing loci decreases in the atsyn4 mutant (Fig. 6b). We further performed ChIP-qPCRs and confirmed that AtSYN4, AtSCC3, and H2Aub1 occupy DNA fragments containing the G-box motif (Fig. 6c). In addition, we found that H2Aub1 levels decrease at these G-boxcontaining loci in atsyn4 and SCC3RNAi-2 plants (Fig. 6d). Altogether, these findings supported a model in which AtSYN4 directly binds to the G-box containing DNA, and AtSCC3 functions as a linker between AtSYN4 and AtBEMI1A/B/C, which in turn recruits AtBMI1A/B/C to this G-box containing loci to mediate H2Aub1 on nucleosomes at these specific genomic sites (Fig. 7).

Discussion
Developmental transitions and cell fate decisions are key processes of the Arabidopsis lifecycle and require specific and stable remodeling of gene expression patterns. PcG proteins are crucial epigenetic regulators of developmental transitions and cell fate decisions. PRC1 and PRC2 play a role in seed embryonic trait determinacy. Numerous embryo-characteristic genes are ectopically expressed in various PRC1 and PRC2 mutant seedlings, as reflected at a morphological level by indeterminate growth, resulting in the formation of callus-like structures 5,37,39,40 . Besides, PRC1 and PRC2 also function in shoot stem cell fate determinacy 5,9,41 . Similar to AtBMI1s 5,9,41 , we found AtSYN4 or AtSCC3 depletion can prolong the vegetative growth period, characterized by late flowering, implicating that cohesin-mediated H2Aub1 might play a role in the transition from vegetative growth to reproductive flowering in Arabidopsis.
H2Aub1 in Arabidopsis is catalyzed by PRC1 and acts as a pivotal regulator of chromatin-associated processes, including DNA replication, damage repair, and gene expression 15,42 . The traditional view is that H2Aub1 starts with H3K27me3 modification by PRC2. Then, PRC1 binds/reads H3K27me3 via its subunits and further catalyzes H2Aub1, which plays a repressive role in transcriptional regulation. Recent studies have come up with different models. For example, H2Aub1 may lead to the recruitment of PRC2. In other words, PRC1 works upstream of PRC2 11,43 . H2Aub1 marks are widespread in the Arabidopsis genome, not only colocalizing with H3K27me3 but also occupying a set of transcriptionally active genes devoid of H3K27me3 12,13 . However, how PRC1 is recruited to loci in an H3K27me3-independent manner and mediates downstream functions are not answered.
In this study, we found that cohesin subunits AtSCC3 and AtSYN4 are involved in H2Aub1 through their direct or indirect interaction with AtBMI1A/B/C, the E3 ligases in PRC1 for H2Aub1. By direct interaction between AtSCC3 and AtSYN4 which binds to genomic loci containing consensus G-box motifs, AtSCC3 recruits AtBMIs to these genomic loci to monoubiquitinate H2A. AtSCC3 and AtBMIs might be more dynamic than AtSYN4 at these sites as more than half of G-box-containing genes bounded by AtSYN4 are marked by H2Aub1, and fewer G-box-containing genes bounded by AtSCC3 are marked by H2Aub1 (Fig. 5b), supporting a 'work and go' model for these proteins, especially for AtSCC3 and AtBMI1A/B/C. Downregulation of AtSCC3 by RNAi or AtSYN4 mutation causes the global reduction of the H2Aub1 level. In addition, cohesinassociated H2Aub1 is preferentially deposited to transcriptionally active genes. Our results thus largely revealed the mechanism for H2Aub1 targeting and gene regulation through the cohesin-PRC1 module.
Arabidopsis genome encodes four α-kleisin genes, AtSYN1-4 25,26 . The functions of AtSYN1 (the orthologue of yeast Rec8) and AtSYN3 seem to be meiosis-specific [25][26][27][28][29] . In this study, a specific role of AtSYN4 in H2Aub1 is reported. AtSYN2 might be involved in the AtSYN4-mediated process to some degree upon the complete deletion of AtSYN4. The functions of Arabidopsis α-kleisin-like proteins AtSYN1-4 might be cell cycle, cell type, tissue, or development stage-specific; therefore, the functional divergence among AtSYN1-4 is of interest to be fully investigated in the future.
PREs are a series of DNA sequences that interact with PcG proteins 44 . PREs can be recognized by specific transcription factors to recruit PRC1 and subsequently cooperate with additional proteins through DNA-protein or protein-protein interactions [45][46][47] . Interestingly, PREs can exist long distances (thousands of base pairs) before the TSS 45,48-50 . If and how PRC1 plays a role in long-range regulation remains a mystery. Although the role of cohesin in 3D genome organization, including chromatin loop formation, was revealed in mammalian systems, our model raised a possibility regarding the role of PREs in the long-distance regulation of H2Aub1 through cohesin, which might mediate long-distance chromatin interaction and recruit catalytic subunits in PRC. In addition, we found that AtSYN4 and AtSCC3 are mostly located within promoter regions, typically the core promoter regions immediately upstream of the TSS, and that there is a shift of~350 bp between H2Aub1 and these two cohesin subunits. This may be related to the occupancy of nucleosomes around the TSS as ã 140 bp nucleosome-free promoter region exists upstream of the TSS, and the sites with the highest occupancy of nucleosomes are located 100 bp downstream of the TSS 51,52 . Cohesin mediates promoter-enhancer or promoter-insulator interactions in animals 34 . Considering the conservation of cohesin structure and function, it is interesting to study whether plant cohesin mediates long-range chromatin interactions and long-distance H2Aub1. However, the plant has no homolog of the CTCF factor. It is still an open question if there are CTCF-like factors in plants that play a role in 3D chromatin organization together with cohesin, similar to animal systems.

Plant materials and growth conditions
Arabidopsis thaliana (ecotype Col-0), T-DNA insertion mutants atsyn2 (SALK_044851), atsyn4 (SALK_076116), atctf7 (SALK_059500) were 4  Density Distance to SCC3 SCC SC SC SCC SCC C S S SC C CC S C C S pea p p p p k(bp k(bp k(bp k(bp k(bp bp bp k(bp k(bp k(bp (bp b k(bp p bp b bp ( p ( ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) Density Distance to SYN4 peak(bp) Significance was tested using a hypergeometric test. c Density plots displaying the distances of H2Aub1 and H3K27me3 to the center of AtSYN4 peak (left) or AtSCC3 peak (right). The dotted line represents the average distance to AtSYN4 or AtSCC3 peak. d Comparisons of the transcription levels between AtSYN4-or AtSCC3-targeted genes and not targeted genes. The p-value was calculated by the one-sided Wilcoxon test, p-value: "***" < 0.001, "**" < 0.01, "ns" not significant. Data are mean ± SD of three biological repeats. The properties of the box plots are defined as follows; minima: lower whisker = smallest observation greater than or equal to lower hinge−1.5*IQR (IQR = interquartile range: the difference between the 75th and 25th percentiles), box lower hinge = 25% quantile; box middle = median, 50% quantile; box upper hinge = 75% quantile; maxima: upper whisker = largest observation less than or equal to upper hinge + 1.5*IQR. The transcription levels of AtSYN4-or AtSCC3-targeted genes and not targeted genes are provided as a Source Data file.

Yeast two-hybrid assay
Yeast two-hybrid interaction assays were performed according to the Yeast Maker Yeast Transformation System 2 User Manual (Clontech). The coding sequences of AtSCC3 and AtBMI1A were subcloned into pGBKT7 or pGADT7, respectively. The constructs were then cotransformed into yeast (AH109). The yeast cells containing the bait and the prey constructs were grown on selective plates (SD-Leu-Trp-His-Ala and SD-Leu-Trp) for analysis. The results were tested after 3-7 days of growth at 30˚C. The primers used are listed in Supplementary Data 8. Values came from three biological replicates.

Firefly luciferase (LUC) complementation imaging assay
For LUC complementation imaging assays 53 , AtSYN4, AtSCC3, AtBMI1A/ B/C, and AtRING1A were fused to the N-or C-terminal fragment of LUC (N and C), respectively. The primers used are listed in Supplementary Data 8. The fused plasmids were introduced into Agrobacterium tumefaciens strain GV3101 by electroporation and then incubated in LB (with 50 mg/L kanamycin and 25 mg/L gentamycin) plate medium at 30°C for 48-72 h to OD 600 = 0.8. Then mix the corresponding Agrobacterium tumefaciens strains equally and then co-infiltrated into tobacco (N. benthamiana) leaves using an injection syringe. 48 h later, the infiltrated leaves were injected with 100 mM luciferin (Sango, dissolved in water), and the luciferase signals were detected by the PMCapture software (Version 1.00) of a Chemiluminescence Imaging System (Tanon 5500, Shanghai, China). Values came from three biological replicates.

Constructs and transgenic plants
The artificial microRNA was used to direct gene silencing 54 . For the AtSCC3RNAi vector construct, a 404-bp PCR product corresponding to the pre-miRNA stem-loop sequence of the Arabidopsis miR gene miR319a was amplified from genomic DNA and cloned into the vector pCambia1300. The vector was then used as a template for subsequent PCR amplification and replacement of the endogenous miR319a/ miR319a* with amiRNA/amiRNA* 55 . The targets of AtSCC3 were designed by a web-based tool (http://wmd.weigelworld.org). The primers are listed in Supplementary Data 8. The full length of AtSYN4 genomic DNA including the upstream regulatory sequence was subcloned into EcoRI/SalI-treated vector pCambia1300-N1-YFP to generate pSYN4-SYN4-YFP construct. The atsyn4 mutant was transformed with GV3101 harboring pSYN4-SYN4-YFP to generate pSYN4-SYN4-YFP/atsyn4. Primers used were shown in Supplementary Data 8.
The coding sequences of AtSCC3 were subcloned into EcoRI/ SalI-treated vector pCambia1300-35S-N1-YFP to generate Pro35-S:AtSCC3-YFP. Arabidopsis Col-0 plants were transformed with GV3101 harboring Pro35S:AtSCC3-YFP to generate the AtSCC3-YFP/Col-0 overexpression lines. Primers used were shown in Supplementary Data 8. All constructs were confirmed by sequencing and introduced into Agrobacterium tumefaciens strain GV3101 by electroporation. Transgenic Arabidopsis plants were generated by floral dipping 56 .
The extracted chromatin was then treated with 0.4 N H 2 SO 4 overnight, and the proteins were precipitated with trichloroacetic acid (TCA) in a final concentration of 20% (w/v). The precipitated proteins were then washed three times with cold acetone, air-dried, suspended, and boiled for 10 min. Proteins were separated in 15% SDS-PAGE gels. Western blots were performed with anti-H2Aub1 (Cell Signaling 8240S, 1:2000 dilution) and anti-H3 (Sigma H0164, 1:3000 dilution) antibodies. Blotting signals were detected by the PMCapture software (Version 1.00) of a Chemiluminescence Imaging System (Tanon 5500, Shanghai, China).

ChIP-seq and ChIP-qPCR assay
Arabidopsis Col-0 and pSYN4-SYN4-YFP/atsyn4 plants were used for ChIP-seq. Plants were grown at 21°C on MS agar plates supplemented with 1.5% sucrose and 0.8% agar. For ChIP assay 57 , 10 g fresh 7-day-old whole seedlings were cross-linked in the crosslink buffer (0.4 M sucrose, 10 mM Tris-HCl (pH 8.0), 1 mM PMSF, 1 mM EDTA, 1% formaldehyde) for 3 × 5 min using vacuum infiltration and the reaction was terminated in 2 M glycine. The extracted nuclei were immunoprecipitated with anti-CSCC3 (ABclonal WG-02026D), anti-GFP (Abcam ab290) antibodies (5 μg/2 g plant sample), and incubated at 4°C overnight. After reverse cross-linking, DNA was extracted by phenol-chloroform method and sheared to an average size of 300 bp using a sonicator (Bioruptor, Diagenode). Sequencing library was constructed according to the standard Illumina protocol. For ChIP-seq, two immunoprecipitations from independent biological replicates were processed for next-generation sequencing.

ChIP-seq data processing
Low-quality and adapter sequences were trimmed from the reads using cutadapt (v1.18) with parameters: -a AGATCGGAAGAGC -A AGATCGGAAGAGC-trim-n -m 50 -q 20, 20. Then, the reads were mapped to the Arabidopsis thaliana TAIR10 genome using Bowtie2 (v2.3.4.3) with parameters: -N 0--no-discordant --no-mixed --nounal. SAMTools (v1.9) was used to transfer the mapping results from SAM format to position-sorted SAM format. Next, the duplicated reads were removed by markdup from sambamba (v0.6.8). The BAM files were then converted to BigWig files using bam-Coverage from the deepTools suite (v3.1.3) with parameters: -normalizeUsing RPKM -binSize 25. The Integrative Genomics Viewer (IGV) was used to show the signal of AtSYN4/AtSCC3 enrichment in a certain genomic region in a track view. Peaks were called using MACS2 (v 2.1.1) callpeak with default parameters. Motif calling was performed using Homer. Peak locations and annotated genes of histone modifications (H3K27me3 and H2Aub1) were obtained from Zhou et al. 12 .

Yeast one-hybrid assay
Yeast one-hybrid assays were performed according to the Matchmaker Yeast One-Hybrid System User Manual (PT1031-1 Clontech). Coding sequences of AtSCC3 and AtSYN4 were PCR-amplified and subcloned into pGADT7 (Clontech). The sequence1, sequence2, and sequence3 were 50 bp G-box (CACGTG) counting DNA selected from the promoter region of AT1G64385, AT3G23440, and AT4G16380, respectively. Mutant 1, mutant 2, and mutant 3 are mutations with the three sequences of G-box (CACGTG) in the promoters mutated into GGGGGG. The sequences were generated by primer annealing from primers synthesized, then subcloned into EcoRI/SacI-treated vector pHISi-1. Resultant constructs were transformed into yeast strain YM4271. Yeast transformants were grown on synthetic dropout (-Leu/-His) medium containing 80 mM 3-AT for 3 days and observed. Primers used are listed in Supplementary Data 8.

RNA extraction and qRT-PCR analysis
The Arabidopsis Col-0, SCC3RNAi-2, atsyn4, atctf7, and atsyn2atsyn4 mutants were used for RNA-seq. Plants were grown at 21°C in long-day conditions (16 h light and 8 h dark cycles) on Murashige & Skoog (MS) medium supplemented with 1.5% sucrose and 0.8% agar. The total RNAs were extracted from 7-day-old Arabidopsis seedlings using Trizol reagent (TIANGEN), and the libraries were constructed according to a standard protocol (Illumina). All of the data were from three biological replicates.
For each sample, RNA (about 2 µg) was used as a template for reverse transcription using ReverTra Ace qPCR RT Master Mix with gDNA Remove Kit (Toyobo) according to the manufacturer's instructions. qRT-PCRs were conducted in a total volume of 20 μL containing 10 μL SYBR Premix Ex-Taq, 0.2 μg cDNA, primers (0.2 mM), and 8.3 μL double distilled water. ACTIN2 (At3g18780) was used for data normalization. Primers used for qRT-PCRs were listed in Supplementary Data 8. Values came from three biological replicates each with three technical repeats. qRT-RCR data was collected by Bio-Rad CFX96.

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

Data availability
The AtSYN4-ChIP, AtSCC3-ChIP, Input-ChIP, and RNA-seq datasets have been deposited to NCBI under accession PRJNA681034. Source data are provided with this paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.