The SMC5/6 complex recruits the PAF1 complex to facilitate DNA double‐strand break repair in Arabidopsis

Abstract

DNA double‐strand breaks (DSBs) are one of the most toxic forms of DNA damage, which threatens genome stability. Homologous recombination is an error‐free DSB repair pathway, in which the evolutionarily conserved SMC5/6 complex (SMC5/6) plays essential roles. The PAF1 complex (PAF1C) is well known to regulate transcription. Here we show that SMC5/6 recruits PAF1C to facilitate DSB repair in plants. In a genetic screen for DNA damage response mutants (DDRMs), we found that the Arabidopsis ddrm4 mutant is hypersensitive to DSB‐inducing agents and is defective in homologous recombination. DDRM4 encodes PAF1, a core subunit of PAF1C. Further biochemical and genetic studies reveal that SMC5/6 recruits PAF1C to DSB sites, where PAF1C further recruits the E2 ubiquitin‐conjugating enzymes UBC1/2, which interact with the E3 ubiquitin ligases HUB1/2 to mediate the monoubiquitination of histone H2B at DSBs. These results implicate SMC5/6‐PAF1C‐UBC1/2‐HUB1/2 as a new axis for DSB repair through homologous recombination, revealing a new mechanism of SMC5/6 and uncovering a novel function of PAF1C.

A screen for DNA damage response mutants reveals an unexpected cooperation of SMC5/6 and PAF1 complexes facilitating E2/E3 enzyme recruitment for local histone H2B ubiquitination.

Introduction

Maintenance of genome stability is essential for all organisms to survive and reproduce. However, genome stability is often threatened by endogenous and exogenous factors, resulting in various types of DNA lesions, among which DNA double‐strand break (DSB) is one of the most toxic forms (Friedberg et al, 2004; Schubert, 2021). If DSB is not repaired properly, it will cause genome instability and affect DNA replication and gene transcription (Jackson & Bartek, 2009). To deal with DSB, organisms have evolved elaborate DNA damage response (DDR) mechanisms including cell cycle arrest, transcriptional reprogramming, DNA repair, and cell death. DSB is recognized by the MRE11‐RAD50‐NBS1 (MRN) complex, which recruits a central protein kinase ATAXIA‐TELANGIECTASIA MUTATED (ATM) (Hu et al, 2016). ATM phosphorylates itself and many other substrates including the histone variant H2AX. The phosphorylated H2AX (γ‐H2AX) accumulates at DSB sites to initiate a cascade of downstream events. Homologous recombination (HR) and nonhomologous end‐joining (NHEJ) are the two major repair pathways for DSB (Brandsma & Gent, 2012). NHEJ is a relatively fast but error‐prone process that occurs in all the cell cycle phases, while HR is a relatively slow but error‐free repair pathway restricted to the late S and G2 phases.

The structural maintenance of chromosomes (SMC) complexes, including cohesin, condensin, and the SMC5/6 complex, are essential regulators of chromosome architecture and organization in cells (Uhlmann, 2016). Compared with cohesin and condensin, SMC5/6 is less well understood. Nevertheless, accumulating evidence suggested that SMC5/6 plays multiple roles in DDR (Aragón, 2018). On one hand, SMC5/6 promotes HR by mediating the cohesion of sister chromatids and facilitating the resolution of DNA intermediates formed during repair (Chen et al, 2009; Stephan et al, 2011; McAleenan et al, 2012; Wu et al, 2012). On the other hand, SMC5/6 inhibits HR by negatively regulating repair proteins such as RAD51, DMC1, RAD51D, BRCA2, SWS1, ATR, and RAD17 (Potts et al, 2006; Durrant et al, 2007; Wang et al, 2010; Song et al, 2011; Yan et al, 2013; Chen et al, 2021b). Furthermore, we have shown that SMC5/6 also negatively regulates cell cycle progression by inhibiting E2F transcription factors (Wang et al, 2018). Recent studies in Arabidopsis revealed that the recruitment of SMC5/6 is dependent on the IDN2–CDC5–ADA2b module (Lai et al, 2018; Jiang et al, 2019, 2022).

The polymerase‐associated factor 1 (PAF1) complex (PAF1C) is evolutionarily conserved in all eukaryotes (Cao & Ma, 2011). It was originally identified in budding yeast (Wade et al, 1996). The yeast PAF1C contains five subunits, Paf1, Ctr9, Cdc73, Rtf1, and Leo1 (Shi et al, 1997; Mueller & Jaehning, 2002), whose orthologs in Arabidopsis are PAF1, VIP6, CDC73, VIP5, and VIP4, respectively (Zhang & Van Nocker, 2002; He et al, 2004; Oh et al, 2004; Yu & Michaels, 2010). The most well‐known function of PAF1C is transcriptional regulation by directly interacting with RNA polymerase II (RNA Pol II) (Jaehning, 2010). PAF1C recruits E2 ubiquitin‐conjugating enzymes and E3 ubiquitin ligases to mediate the monoubiquitination of H2B (H2Bub), which is required for di‐ or tri‐methylation of H3K4, H3K36, and H3K79 (Krogan et al, 2003; Wood et al, 2003; Zhang et al, 2013). Due to its key roles in transcription and chromatin modification, PAF1C was reported to regulate multiple important biological processes (Francette et al, 2021; van den Heuvel et al, 2021). For example, in Arabidopsis, PAF1C was reported to inhibit flowering (Li et al, 2019; Nasim et al, 2022). However, the roles of PAF1C in DSB repair are still unknown.

H2Bub can directly regulate chromatin status by changing nucleosome stability, nucleosome disintegration, and high‐order chromatin structure (Fuchs & Oren, 2014), thereby is involved in multiple molecular and biological processes such as DNA replication (Trujillo & Osley, 2012), transcription (Zhu et al, 2005; Pavri et al, 2006; Chaudhary et al, 2007; Laribee et al, 2007; Kim & Roeder, 2009) and DNA repair. In human cells, RNF20/RNF40 is recruited to the DNA damage sites and is phosphorylated by ATM to promote H2Bub (Moyal et al, 2011). This process prevents excessive excision of DSB terminals and allows the recruitment of RAD51 by promoting the release of nucleosomes (Nakamura et al, 2011). Bre1, the yeast homologous protein of RNF20/RNF40, plays similar roles (Zheng et al, 2018). It remains to be determined whether HUB1 and HUB2 (two Arabidopsis homologs of RNF20 and RNF40) are involved in DNA damage repair.

In this study, we found that PAF1C is required for DSB repair in Arabidopsis. Mechanistically, SMC5/6 recruits PAF1C to DSB sites, where PAF1C recruits UBC1/2 and HUB1/2 to promote H2Bub, facilitating DSB repair. Our study thus reveals a new mechanism underlying DSB repair and uncovers the relationship between SMC5/6 and PAF1C for the first time.

Results

The ddrm4‐1 mutant is hypersensitive to DNA‐damaging agents

To identify new regulators of DSB repair in plants, we performed a genetic screen for DNA Damage Response Mutants (DDRM) in Arabidopsis. Plants were grown vertically on medium containing the DSB‐inducing agents such as bleomycin (BLE) or camptothecin (CPT). The plants with shorter or longer roots than wild type (Col‐0) were considered ddrm mutants. Previously, we have reported the identification of ddrm1 (Wang et al, 2022) and ddrm6 (Li et al, 2021). Here we reported the ddrm4 mutant. As shown in Fig 1A and B, the root length of ddrm4‐1 was similar to that of Col‐0 in the control condition, but was significantly shorter than that of Col‐0 in the presence of BLE, suggesting that ddrm4‐1 is hypersensitive to BLE. In addition, we also found that ddrm4‐1 is hypersensitive to CPT (Appendix Fig S1A and B).

Figure 1. PAF1 is required for homologous recombination.

• A
  Pictures of plants treated with BLE. Plants were grown vertically on 1/2 MS medium with or without 4 μM BLE for 9 days. Scale bar = 1 cm.
• B
  The relative root length of the indicated plants. The data are represented as means ± SD (n = 10 plants) relative to the values of Col‐0 under the control condition. The statistical significance was determined using two‐way ANOVA analysis. ****P < 0.0001. The experiments were repeated three times with similar results.
• C
  The genomic structure of PAF1. The T‐DNA insertion sites in ddrm4‐1 and ddrm4‐2 and the genotyping primers are shown. The exons are shown as dark green boxes, and the introns are shown as black lines. The UTRs are shown as light green boxes. ATG and TGA indicate the start and stop codons, respectively.
• D
  Gene ontology (GO) analysis of genes upregulated in ddrm4‐1 compared with Col‐0. The top 10 significantly enriched GO terms are shown. Three biological replicates were used to identify the differentially expressed genes.
• E, F
  Comet assays. Representative pictures of comet assays (E) and the percentages of DNA in the comet tails (F) are shown. Scale bars = 100 μm. The data are represented as means ± SD (n > 200 technical replicates). The statistical significance was determined using two‐tailed Student's t‐test. ****P < 0.0001. The experiments were repeated three times with similar results.
• G
  Schematic representation of HR reporter system. The reporter line contains an I‐SceI restriction site within the GUS gene as well as a nearby donor sequence (U). A single DSB is induced when the reporter line is crossed with the I‐SceI trigger line. When the DSB is repaired through HR, the functional GUS is restored.
• H, I
  Representative GUS staining images of cotyledons (H) and quantification of the relative HR efficiency (I). The reporter line and trigger line in either ddrm4‐1 or Col‐0 background were crossed and the F1 seedlings were used for GUS staining analysis. Scale bar = 1 mm. The number of blue sectors was scored. The data are presented as means ± SD (n > 132 cotyledons) relative to the values of Col‐0. The statistical significance was determined using two‐tailed Student's t‐test. ****P < 0.0001.

Source data are available online for this figure.

DDRM4 encodes PAF1

The ddrm4‐1 mutant was identified in the Arabidopsis TRANSPLANTA collection (Coego et al, 2014). To determine the T‐DNA insertion site in ddrm4‐1, we carried out high‐efficiency thermal asymmetric interlaced PCR (hiTAIL‐PCR) analysis followed by sequencing (Appendix Fig S2A and B). We found that the T‐DNA was inserted into the first exon of the PAF1 (AT1G79730) gene (Fig 1C), which was confirmed through genotyping analysis (Appendix Fig S2C). The quantitative reverse transcription PCR (RT‐qPCR) analysis showed that the transcription level of DDRM4 was dramatically reduced in ddrm4‐1 (Appendix Fig S2D). To confirm that PAF1 is the DDRM4 gene, we crossed ddrm4‐1 with another paf1 mutant ddrm4‐2 (Fig 1C). Similar to ddrm4‐1, the ddrm4‐2 mutant and all the resulting F1 seedlings were hypersensitive to BLE (Fig EV1A and B), suggesting that ddrm4‐1 is allelic to ddrm4‐2.

Figure EV1. The ddrm4 mutants are hypersensitive to BLE.

• A–D
  (A, C) Pictures of plants treated with BLE. Plants were grown vertically on 1/2 MS medium with or without 4 μM BLE for 7 days (A) or 9 days (C). Bar = 1 cm. 4–1 × 4–2 F1, the F1 seedlings derived from the cross between ddrm4‐1 and ddrm4‐2. COM, the complementation lines expressing PAF1 driven by the native promoter in ddrm4‐1. (B and D) The relative root length of the indicated plants. The data are represented as means ± SD (n = 10 plants) relative to the values of Col‐0 under the control condition. The statistical significance was determined using two‐way ANOVA analysis. ns, not significant; ****P < 0.0001. The experiments were repeated three times with similar results.

To further confirm that PAF1 is DDRM4, we performed the genetic complementation test by transforming the genomic sequence of PAF1 driven by its native promoter (pPAF1:PAF1) into ddrm4‐1. The root length of the complementation lines (COM) was similar to those of Col‐0 in the presence of BLE (Fig EV1C and D), indicating that PAF1 complements the ddrm4‐1 mutant.

The PAF1 complex is required for DSB repair

PAF1 is a core subunit of PAF1C, which contains the other five subunits VIP3, VIP4, VIP5, VIP6, and CDC73 (Antosz et al, 2017). To determine whether the hypersensitivity of ddrm4 to DNA damage is attributed to the specific function of PAF1 or the general function of PAF1C, we examined the phenotypes of the mutants defective in the other five subunits. The mutants of VIP4 and VIP6 were generated by CRISPR‐Cas9 technology (Fig EV2A) and the others were T‐DNA insertion mutants. All the mutants displayed early flowering phenotypes (Fig EV2B), indicating that they are loss‐of‐function mutants. Similar to ddrm4, all these mutants showed hypersensitivity to BLE (Fig EV2C–L), suggesting that the whole PAF1C is involved in DSB repair.

Figure EV2. The PAF1C mutants are hypersensitive to BLE.

• A
  The indels in the vip4‐c1 and vip6‐c1 mutants generated using CRISPR/Cas9 technology. The letters in red indicate the inserted nucleotides. The lowercase letters indicate intron. The uppercase letters indicate exon.
• B
  The early flowering phenotypes of the PAF1C mutants. Plants were grown in soil for 3 weeks. Scale bar = 1 cm.
• C–L
  (C, E, G, I, K) Pictures of plants treated with BLE. Plants were grown vertically on 1/2 MS medium with or without 4 μM BLE (C, G) or 2 μM BLE (E, I, K) for 8 days. Scale bar = 1 cm. (D, F, H, J, L) The relative root length of the indicated plants. The data are represented as means ± SD (n = 10 technical replicates) relative to the values of Col‐0 under the control condition. The statistical significance between different genotypes was determined using two‐way ANOVA analysis. ****P < 0.0001. The experiments were repeated three times with similar results.

The ddrm4 mutant is defective in DSB repair

Given that PAF1C mainly functions in transcriptional regulation, we speculated that the expression of DDR‐related genes in the ddrm4‐1 mutant may be reduced. To test this possibility, we performed transcriptome analysis using RNA sequencing technology (RNA‐Seq). Compared with Col‐0, 609 genes were upregulated and 259 genes were downregulated in ddrm4‐1 (|Log₂FoldChange| > Log₂1.5, P < 0.05, Table EV1). Unexpectedly, we found that many genes related to DDR were upregulated rather than downregulated in ddrm4‐1. Consistently, gene ontology (GO) analysis revealed that the DDR‐related GO terms (e.g. DNA repair, cellular response to DNA damage stimulus) were significantly enriched in the upregulated genes but not in the downregulated genes (Fig 1D, Table EV2 and EV3). These data indicated that PAF1C may regulate DSB repair through other functions in addition to transcriptional regulation.

The enhanced expression of DDR‐related genes suggested that DDR is activated in ddrm4 under normal growth conditions. Previously, we have reported that the sni1 mutant accumulates more DNA damage, which activates DDR to induce the expression of DDR‐related genes under normal growth conditions (Yan et al, 2013). Therefore, we speculated that ddrm4 accumulated more DNA damage than Col‐0. To test this, we performed comet assays. As expected, the DNA in the tails was significantly higher in the ddrm4‐1 mutant than in Col‐0 (Fig 1E and F).

The accumulation of DNA damage may be due to defects in DSB repair. Since HR is one of the major DSB repair mechanisms, we tested the HR efficiency of ddrm4‐1 using a previously established HR reporter system including the DU.GUS reporter line and the I‐SceI trigger line (Orel et al, 2003). The DU.GUS reporter line contains a unique recognition site of the endonuclease I‐SceI between two nonfunctional GUS fragments as well as a nearby GUS fragment as a donor sequence (Fig 1G). When the reporter line is crossed with the trigger line, I‐SceI induces DSB, which can be repaired through HR and thus produces a functional GUS gene, resulting in blue sectors after GUS staining. We found that the HR efficiency of ddrm4‐1 is about half of Col‐0 (Fig 1H and I). These results suggested that PAF1 is required for HR.

PAF1 is recruited to DSB sites

Many proteins (e.g., RAD51 and BRCA1) involved in HR are recruited to DSB sites and form discrete foci. To test whether PAF1 is recruited to DSB sites, we first examined the localization of CFP‐PAF1 transiently expressed in Nicotiana benthamiana and Arabidopsis protoplasts. Confocal microscopy analysis revealed that most CFP‐PAF1 were evenly distributed in the nucleus in the control condition. However, after BLE treatment, most of CFP‐PAF1 formed foci (Fig 2A–D). It is well known that the H2A variant H2AX is phosphorylated when DSB occurs and forms foci. Arabidopsis encodes two H2AX, called H2AXA and H2AXB. When CFP‐PAF1 was coexpressed with H2AXA‐YFP in Arabidopsis protoplasts, the BLE‐induced CFP‐PAF1 foci colocalized with H2AXA‐YFP foci (Appendix Fig S3). To further investigated the localization of PAF1, we transformed 35S:CFP‐PAF1 in ddrm4‐1. The root lengths of the 35S:CFP‐PAF1/ddrm4‐1 transgenic lines were similar to that of Col‐0 (Appendix Fig S4), suggesting that CFP‐PAF1 is functional. Then we examined the CFP‐PAF1 localization in the root cells. We found that after BLE treatment, CFP‐PAF1 formed foci in about 80% of cells (Fig 2E and F).

Figure 2. PAF1 is recruited to DSB sites.

• A–F
  (A, C, E) Representative images of CFP‐PAF1 expressed in N. benthamiana leaves (A), Arabidopsis protoplasts (C), and transgenic Arabidopsis (E). The N. benthamiana leaves and the transgenic seedlings were treated with 40 μM BLE for 4 h before imaging. The Arabidopsis protoplasts were treated with 20 μM BLE for 2 h before imaging. Scale bars in (A, C) = 2.5 μm. Bars in (E) = 10 μm. (B, D, F) The percentage of cells with CFP‐PAF1 foci in N. benthamiana leaves (B), Arabidopsis protoplasts (D), and transgenic Arabidopsis (F). The data are represented as means ± SEM (n = 3 biological replicates). For each sample, more than 35 cells were scored. The statistical significance was determined using two‐tailed Student's t‐test. ***P < 0.001. ****P < 0.0001.
• G
  Schematic representation of the SSDIS vector. GR, glucocorticoid receptor. I‐SceI, restriction endonuclease I‐SceI. 35S, 35S promoter. HYGR, hygromycin‐resistance gene. The recognition sequence of I‐SceI was named as DSB site. The position downstream of the DSB site and the PCR fragment in (K) were indicated (0–4.5 kb).
• H, I
  Representative images (H) and percentage of cells with CFP‐PAF1 foci in the root of SSDIS/CFP‐PAF1 transgenic Arabidopsis. Plants were mock treated or treated with 100 μM DEX for 4 h before imaging. Scale bars = 10 μm. The red arrowheads indicate CFP‐PAF1 foci. The data in (I) are represented as means ± SEM (n = 3 biological replicates). For each sample, more than 49 cells were scored. The statistical significance was determined using two‐tailed Student's t‐test. ***P < 0.001.
• J
  eChIP‐seq assays. The SSDIS/CFP‐PAF1 transgenic Arabidopsis was mock‐treated or treated with 100 μM DEX for 4 h. The eChIP assays were carried out using an anti‐GFP antibody. Both the immunoprecipitated DNA (eChIP) and the input DNA were subjected to next‐generation sequencing analysis. The experiments were repeated twice times with similar results.
• K
  eChIP‐qPCR assays. The immunoprecipitated DNA and the input DNA were subjected to qPCR analysis. The ratios of eChIP and input are shown. UBQ5 serves as a negative control. The positions of the PCR fragment of 0.5–4.5 kb were shown in (I). The data are represented as means ± SD (n = 3 technical replicates) relative to the values of Mock. The statistical significance was determined using two‐tailed Student's t‐test. ns, not significant; **P < 0.01; ***P < 0.001. The experiments were repeated three times with similar results.

Source data are available online for this figure.

To further confirm that PAF1 is recruited to DSB sites, we generated a site‐specific DSB inducing system (SSDIS) construct (Fig 2G). This construct contains the sequence encoding the endonuclease I‐SceI fused with glucocorticoid receptor (GR) driven by 35S promoter, after which there is a single I‐SceI recognition site. The fusion of GR to I‐SceI allows us to control the nucleo‐cytoplasmic localization of I‐SceI by using the hormone dexamethasone (DEX). In the absence of DEX, the GR‐I‐SceI fusion protein is sequestered in the cytoplasm. After DEX treatment, the GR‐I‐SceI fusion protein translocates into the nucleus, where it cleaves its recognition site and produces a site‐specific DSB. The SSDIS construct was transformed into the 35S:CFP‐PAF1 transgenic plants. One of the transgenic lines was used for further characterization. Through whole genome sequencing, we found that the SSDIS was inserted after the nucleotide 24944088 on Chromosome 1 (Appendix Fig S5A), which was further confirmed through genotyping analysis (Appendix Fig S5B). We found that in the absence of DEX, CFP‐PAF1 was evenly distributed in the nucleus. But after DEX treatment, CFP‐PAF1 formed foci in about 30% of root cells (Fig 2H and I).

To confirm the binding of CFP‐PAF1 to DSB, we performed enhanced chromatin immunoprecipitation (eChIP), followed by sequencing analysis (eChIP‐seq). We focused our analysis on the sequencing reads within the SSDIS region. Since there were two identical 35S promoter sequences, their corresponding reads were not used for alignment. Compared to the input sample, the reads in the eChIP sample were much higher, especially at the region 2.5 kb downstream of DSB site (Fig 2J; Appendix Fig S5C), suggesting that CFP‐PAF1 can bind these regions. These data were confirmed through eChIP‐qPCR analysis (Fig 2K). It is of note that in the mock‐treated sample, there was also some enrichment. This may be because some leaky GR‐I‐SceI can translocate to the nucleus in the absence of DEX and generate DSB, which recruited CFP‐PAF1. An alternative explanation is that CFP‐PAF1 can bind to these regions to regulate transcription. Nevertheless, we found that the peaks were higher in the DEX‐treated sample than the mock‐treated sample, suggesting that at least some of the CFP‐PAF1 binding is due to the presence of DSB.

PAF1 interacts with SMC5/6

Next, we sought to investigate how PAF1 is recruited to DSB sites. One possibility is that other DSB‐binding proteins can interact with and recruit PAF1. In a previous study, many PAF1‐interacting proteins were identified (Antosz et al, 2017). Among them, SMC5 and SMC6B attracted our attention because the SMC5/6 complex plays important roles in DSB repair (Aragón, 2018). To confirm their interactions, we first performed yeast two‐hybrid (Y2H) assays, which revealed that both SMC5 and SMC6B interact with PAF1 (Fig 3A and B). The SMC5 and SMC6 proteins contain head domain, coiled‐coil domain, and hinge domain, among which the hinge domains mediate the interaction between SMC5 and SMC6 (Alt et al, 2017). Since PAF1 interact with both SMC5 and SMC6B, it is possible that PAF1 interacts with their hinge domains. To test this, we performed Y2H assays using truncated forms of SMC5 and SMC6B. We found that the hinge domains of SMC5 and SMC6B can interact with PAF1, and the truncated forms without hinge domains cannot interact with PAF1, suggesting that the hinge domains are necessary and sufficient for the interaction between PAF1 and SMC5/6 (Fig EV3A and B). To test whether they interact in vivo, we carried out bimolecular fluorescence complementation (BiFC) assays (Figs 3C and D, and EV3C and D), split luciferase assays (Figs 3E and F, and EV3E and F), and coimmunoprecipitation (CoIP) assays (Fig EV3G and H). All these assays suggested that PAF1 interacts with SMC5 and SMC6B. To test whether PAF1 interacts with SMC5 and SMC6B directly or indirectly, we performed pull‐down assays using purified recombinant proteins (Figs 3G and H, and EV3I and J), which suggested that their interactions are direct. Therefore, PAF1 directly interacts with SMC5 and SMC6B both in vitro and in vivo.

Figure 3. PAF1 interacts with SMC5 and SMC6B.

• A, B
  Yeast two‐hybrid assays. PAF1 was fused with the activation domain (AD). SMC5 and SMC6B were fused with the DNA‐binding domain (BD). DDO, double dropout (SD/−Trp/−Leu) medium. QDO, quadruple dropout (SD/−Trp/−Leu/‐His/−Ade) medium.
• C, D
  BiFC assays. The proteins were fused to either the C‐ or N‐terminal half of YFP (cYFP or nYFP) and were transiently expressed in N. benthamiana. The YFP fluoresce detected by confocal microscopy indicates interaction. Scale bars = 10 μm.
• E, F
  Split luciferase assays. The proteins were fused to either the C‐ or N‐terminal half of luciferase (cLUC or nLUC) and were transiently expressed in N. benthamiana. The luminescence detected by a CCD camera indicates interaction.
• G, H
  In vitro pull‐down assays. The dextrin beads coupled with MBP, MBP‐SMC5, or MBP‐SMC6B were incubated with His‐PAF1, respectively. After washing, the beads were subjected to western blotting analysis using anti‐His or anti‐MBP antibodies.
• I, J
  CFP‐PAF1 foci colocalize with SMC5‐YFP and SMC6B‐YFP foci in Arabidopsis protoplasts. The Arabidopsis protoplasts were treated with 20 μM BLE for 2 h before imaging. Scale bars = 2.5 μm. All experiments were repeated at least three times with similar results.

Source data are available online for this figure.

Figure EV3. PAF1 interacts with the hinge domains of SMC5 and SMC6B.

• A, B
  Yeast two‐hybrid assays. The domains of SMC5 and SMC6B are shown in different colors. For the truncated proteins, colored boxes represent retained domains and black lines represent deleted domains. 5, SMC5; 6B, SMC6B; H, head; C, coil; BD, DNA binding domain; AD, activation domain; DDO, double dropout (SD/−Trp/−Leu) medium; QDO, quadruple dropout (SD/−Trp/−Leu/‐His/−Ade) medium.
• C, D
  BiFC assays. The proteins were fused to either the C‐ or N‐terminal half of YFP (cYFP or nYFP) and were transiently expressed in N. benthamiana. The YFP fluoresce detected by confocal microscopy indicates interaction. Scale bars = 10 μm.
• E, F
  Split luciferase assays. The proteins were fused to either the C‐ or N‐terminal half of luciferase (cLUC or nLUC) and were transiently expressed in N. benthamiana. The luminescence detected by a CCD camera indicates interaction.
• G, H
  CoIP assays. PAF1‐FLAG was coexpressed with Hinge_5‐YFP, Hinge_6B‐YFP, or GFP in N. benthamiana leaves. The proteins precipitated by anti‐GFP beads were subjected to western blotting analysis using anti‐FLAG or anti‐GFP antibodies.
• I, J
  In vitro pull‐down assays. The dextrin beads coupled with MBP‐tagged proteins were incubated with His‐PAF1. After washing, the beads were subjected to western blotting analysis using anti‐His and anti‐MBP antibodies. All experiments were repeated at least three times with similar results.

Previous studies have shown that SMC5/6 can form foci after BLE treatment (Lai et al, 2018). Therefore, we wanted to know whether PAF1 foci colocalize with the SMC5/6 foci. CFP‐PAF1 was coexpressed with SMC5‐YFP or SMC6B‐YFP in N. benthamiana (Appendix Fig S6) and in Arabidopsis protoplast (Fig 3I and J). We found that the foci of CFP‐PAF1 and SMC5‐YFP or SMC6B‐YFP colocalized very well in the nucleus after BLE treatment. In addition, we found that the other subunits of PAF1C (VIP3, VIP4, VIP5, VIP6, and CDC73) can also form DNA damage‐induced foci, which colocalized with CFP‐PAF1 and SMC5‐YFP (Fig EV4A–E). These results indicated that PAF1C and SMC5/6 colocalize to DSB sites.

Figure EV4. The PAF1C foci colocalize with SMC5 foci.

• A–E
  The indicated proteins were coexpressed in N. benthamiana leaves. The leaves were treated with 40 μM BLE for 2 h before imaging. Scale bars = 2.5 μm. The experiments were repeated three times with similar results.

PAF1C functions downstream of SMC5/6

Next, we wanted to know the relationship between PAF1C and SMC5/6. To test the effect of PAF1C on the SMC5/6 foci, we transiently expressed SMC5‐YFP or SMC6B‐YFP in the protoplasts of the ddrm4‐1 mutant. We found that the SMC5‐YFP or SMC6B‐YFP foci were similar in ddrm4‐1 and Col‐0 (Appendix Fig S7), suggesting that the SMC5 and SMC6B foci are independent of PAF1. Then, we tested the effect of SMC5/6 on the PAF1 foci. Since the SMC5 knock‐out mutant is lethal, we used RNAi strategy to knock down the expression of SMC5 by transforming the RNAi construct of SMC5 (SMC5‐Ri) into the Arabidopsis protoplasts. The expression of SMC5 was confirmed through RT‐qPCR analysis (Appendix Fig S8). We found that the BLE‐induced CFP‐PAF1 foci were significantly reduced when SMC5 was knocked down (Fig 4A and B). To investigate the effect of SMC6 on the PAF1 foci, we transiently expressed CFP‐PAF1 in the smc6b protoplasts. Consistently, the CFP‐PAF1 foci were dramatically reduced in smc6b compared with Col‐0 (Fig 4C and D).

Figure 4. PAF1 functions downstream of SMC5/6.

• A–F
  The PAF1 foci are dependent on SMC5 and SMC6B. (A) CFP‐PAF1 was transfected alone or cotransfected with the RNAi construct of SMC5 (SMC5‐Ri) into Col‐0 protoplasts. (C) CFP‐PAF1 was transfected into the protoplasts of Col‐0 or smc6b, respectively. The protoplasts were treated with 20 μM BLE for 2 h before imaging. Scale bars = 2.5 μm. (E) The 35S:CFP‐PAF1/Col‐0 or 35S:CFP‐PAF1/smc6b transgenic seedlings were treated with 40 μM BLE for 4 h before imaging. Scale bars = 5 μm. (B, D, F) The percentage of cells with foci. The data were represented as means ± SEM (n = 3 biological replicates). For each sample, at least 32 cells were scored. The statistical significance was determined using two‐way ANOVA analysis. **P < 0.01; ***P < 0.001.
• G, H
  eChIP‐qPCR assays. (G) The SMC5‐Ri construct was transfected into the protoplasts of SSDIS/CFP‐PAF1 transgenic plants. (H) The 35S:CFP‐PAF1 construct was transfected into the protoplasts of SSDIS/smc6b or SSDIS/Col‐0 transgenic plants. The protoplasts were treated with 100 μM DEX for 4 h to induce DSBs. The eChIP assays were carried out using an anti‐GFP antibody. The immunoprecipitated DNA (eChIP) and the input DNA were subjected to qPCR analysis. The ratios of eChIP and input are shown. UBQ5 serves as a negative control. The data are represented as means ± SD (n = 3 technical replicates). The statistical significance was determined using two‐tailed Student's t‐test. ns, not significant; ****P < 0.0001.
• I–L
  (I, K) Pictures of plants treated with BLE. Plants were grown vertically on 1/2 MS medium with or without 4 μM BLE (I) or 2 μM BLE (K) for 8 days. Scale bar = 1 cm. (J, L) The relative root length of the indicated plants. The data are represented as means ± SD (n = 10 plants) relative to the values of Col‐0 under the control condition. The statistical significance was determined using two‐way ANOVA analysis. ns, not significant; ****P < 0.0001. The experiments were repeated three times with similar results.

Source data are available online for this figure.

To further confirm this result, we generated 35S:CFP‐PAF1/smc6b by crossing 35S:CFP‐PAF1/ddrm4‐1 with smc6b. As expected, the BLE‐induced CFP‐PAF1 foci were significantly reduced in smc6b (Fig 4E and F). To investigate whether SMC5/6 is required for the recruitment of PAF1C at DSB sites, we transfected SMC5‐Ri in the protoplasts of SSDIS/35S:CFP‐PAF1 and performed eChIP‐qPCR assays. The enrichment of CFP‐PAF1 was significantly decreased when SMC5 was knocked down (Fig 4G). Similarly, the enrichment of CFP‐PAF1 was significantly decreased in smc6b (Fig 4H). These results indicated that SMC5/6 is required for the recruitment of PAF1C to DSB sites.

To test the relationship between PAF1C and SMC5/6 genetically, we generated the smc6b ddrm4‐1 double mutant. Compared with ddrm4‐1, smc6b was significantly more sensitive to BLE (Fig 4I and J). Notably, the sensitivity of smc6b ddrm4‐1 double mutant to BLE was similar to that of smc6b (Fig 4I and J), indicating that PAF1 and SMC6B function in the same pathway. More interestingly, we found that overexpression of CFP‐PAF1 partially suppressed the hypersensitivity of smc6b to BLE (Fig 4K and L), suggesting that PAF1C functions downstream of SMC5/6.

PAF1C recruits the E2 ubiquitin conjugases UBC1/2 to DSB sites

The next question we wanted to address is how PAF1C regulates DSB repair. Previous studies have shown that PAF1C regulates transcription by promoting the monoubiquitination of H2B (H2Bub). In addition, H2Bub was shown to be required for DSB repair (Moyal et al, 2011; Nakamura et al, 2011; Zheng et al, 2018; Liu et al, 2021). Based on these findings, we hypothesized that PAF1C facilitates DSB repair by promoting H2Bub. Recent studies in Saccharomyces cerevisiae revealed that Cdc73 and Rtf1, two subunits of yeast PAF1C, interact with the E2 ubiquitin conjugase Rad6, which functions together with the E3 ubiquitin ligase Bre1 to mediate H2Bub (Van Oss et al, 2016; Chen et al, 2021a). Therefore, we tested whether the Arabidopsis CDC73 and VIP5 (the homolog of Rtf1) interact with UBC1 and UBC2 (two homologs of Rad6). In the split luciferase assays, CDC73 and VIP5 interacted with both UBC1 and UBC2 (Fig 5A and B; Appendix Fig S9). Since UBC1 and UBC2 share 99% of sequence identity, we used UBC2 in the subsequent experiments. BiFC assays (Fig 5C and D), CoIP assays (Fig 5E and F), and pull‐down assays (Fig 5G and H) suggested that UBC2 directly interacts with CDC73 and VIP5 in vitro and in vivo.

Figure 5. CDC73 and VIP5 interact with UBC2.

• A, B
  Split luciferase assays. The proteins were fused to either the C‐ or N‐terminal half of luciferase (cLUC or nLUC) and were transiently expressed in N. benthamiana. The luminescence detected by a CCD camera indicates interaction.
• C, D
  BiFC assays. The proteins were fused to either the C‐ or N‐terminal half of YFP (cYFP or nYFP) and were transiently expressed in N. benthamiana. The YFP fluoresce detected by confocal microscopy indicates interaction. Scale bars = 10 μm.
• E, F
  CoIP assays. The fusion proteins were coexpressed in N. benthamiana leaves. The proteins precipitated by anti‐GFP beads were subjected to western blotting analysis using anti‐FLAG or anti‐GFP antibodies.
• G, H
  In vitro pull‐down assays. The dextrin beads coupled with MBP or MBP‐UBC2 were incubated with His‐CDC73 or His‐VIP5. After washing, the beads were subjected to western blotting analysis using anti‐His or anti‐MBP antibodies.
• I, J
  The UBC2 foci are dependent on VIP5. UBC2‐mCherry and/or VIP5‐YFP were transfected into the vip5 protoplasts. The protoplasts were treated with 20 μM BLE for 2 h before imaging. Scale bars = 2.5 μm. The data in (J) were represented as means ± SEM (n = 3 biological replicates). For each sample, at least 30 cells were scored. The statistical significance was determined using two‐way ANOVA analysis. ****P < 0.0001. All experiments were repeated at least three times with similar results.

Source data are available online for this figure.

The interaction between UBC1/2 and VIP5 or CDC73 suggested that PAF1C may recruit UBC1/2 to DSB sites through VIP5 or CDC73. First, we examined VIP5 foci in Arabidopsis protoplasts and found that VIP5‐YFP formed foci after BLE treatment. Similar to PAF1 (Fig 4A–D), the VIP5‐YFP foci significantly decreased in smc6b or SMC5‐Ri, suggesting that VIP5 foci are also dependent on SMC5/6 (Appendix Fig S10). Next, we transiently expressed UBC2‐mCherry in the protoplasts of vip5 (Fig 5I and J). In the absence of VIP5, UBC2‐mCherry could not form foci after BLE treatment. However, when VIP5‐YFP was coexpressed, UBC2‐mCherry formed foci, which colocalized with VIP5‐YFP foci. These data suggested that PAF1C is required for the recruitment of UBC1/2 to DSB sites.

UBC1/2 recruit the E3 ubiquitin ligases HUB1/2 to DSB sites

Previously, it was reported that UBC1 and UBC2 interact with the E3 ubiquitin ligase HUB1 and HUB2 (two homologs of Bre1) in yeast‐two hybrid assays (Cao et al, 2008). To further confirm their interactions, we performed split luciferase assays (Fig 6A and B) and pull‐down assays (Fig 6C and D), which revealed that UBC1 and UBC2 directly interact with HUB1 and HUB2 both in vitro and in vivo.

Figure 6. HUB1 and HUB2 interact with UBC2.

• A, B
  Split luciferase assay. The proteins were fused to either the C‐ or N‐terminal half of luciferase (cLUC or nLUC) and were transiently expressed in N. benthamiana. The luminescence detected by a CCD camera indicates interaction. The experiments were repeated three times with similar results.
• C, D
  In vitro pull‐down assay. The dextrin beads coupled with MBP, MBP‐HUB1, or MBP‐HUB2 were incubated with His‐UBC2. After washing, the beads were subjected to western blotting analysis using anti‐His or anti‐MBP antibodies. The experiments were repeated three times with similar results.
• E–J
  The HUB1/2 foci are dependent on UBC2. VIP5‐CFP, HUB1/2‐YFP, and/or UBC2‐mCherry were transfected into the ubc1 ubc2‐c1 protoplasts. The protoplasts were treated with 20 μM BLE for 2 h before imaging. Bars = 2.5 μm. The data in (F, G, I, J) were represented as means ± SEM (n = 3 biological replicates). For each sample, at least 60 cells were scored. The statistical significance was determined using two‐way ANOVA analysis. ns, not significant; **P < 0.01.

Source data are available online for this figure.

To test whether UBC1/2 can recruit HUB1/2 to DSB sites, we first generated the ubc1 ubc2 double mutant using CRISPR/Cas9 technology because UBC1 and UBC2 are functionally redundant (Cao et al, 2008). Sequence analysis of ubc1 ubc2‐c1 and ubc1 ubc2‐c2 revealed that there were some indels in both UBC1 and UBC2 (Appendix Fig S11A). Similar to the PAF1C mutants, the ubc1 ubc2 mutants showed early flowering phenotype (Appendix Fig S11B). Next, we transiently expressed VIP5‐CFP and HUB1/2‐YFP in the protoplasts of ubc1 ubc2‐c1. As shown in Fig 6E–J, in the absence of UBC1/2, while most of VIP5‐CFP could form foci after BLE treatment, HUB1/2‐YFP could not. However, when UBC2‐mCherry was coexpressed, HUB1/2‐YFP formed foci, which colocalized with VIP5‐CFP and UBC2‐mCherry foci. These data suggested that UBC1/2 are required for the recruitment of HUB1/2 to DSB sites.

To test whether the recruitment of HUB1/2 is dependent on PAF1C, we transiently expressed HUB1/2‐YFP and UBC2‐mCherry in the protoplasts of vip5 (Appendix Fig S12). In the absence of VIP5, HUB1/2 and UBC2 could not form foci after the treatment of BLE. When VIP5‐CFP was coexpressed, HUB1/2 and UBC2 formed foci and their foci were colocalized with VIP5 foci, suggesting that the recruitment of HUB1/2 to DSB sites is dependent on VIP5.

H2Bub is required for DSB repair

The recruitment of UBC1/2 and HUB1/2 to DSB sites suggested that they may regulate DSB repair. Therefore, we treated ubc1 ubc2‐c, hub1‐5, and hub2‐2 mutants with BLE. As expected, the root lengths of these mutants were significantly shorter than that of Col‐0 (Fig 7A–F), suggesting that UBC1/2 and HUB1/2 are required for DSB repair. To test their genetic relationship with PAF1C, we generated the hub1‐5 ddrm4‐1 double mutant. We found that the root length of hub1‐5 ddrm4‐1 was not significantly different from that of ddrm4‐1 (Fig 7G and H), suggesting that PAF1 and HUB1 function in the same pathway.

Figure 7. H2Bub is required for DSB repair.

• A, C, E, G
  Pictures of plants treated with BLE. Plants were grown vertically on 1/2 MS medium with or without 6 μM BLE (A) or 4 μM BLE (C, E, G) for 8 days. Scale bar = 1 cm.
• B, D, F, H
  The relative root length of the indicated plants. The data are represented as means ± SD (n = 10 plants) relative to the values of Col‐0 under the control condition. The statistical significance was determined using two‐way ANOVA analysis. ns, not significant; ***P < 0.001; ****P < 0.0001. The experiments were repeated three times with similar results.
• I, J
  eChIP‐qPCR assays. (I) The SSDIS/Col‐0 transgenic Arabidopsis was mock‐treated or treated with 100 μM DEX for 4 h. (J) The SSDIS/Col‐0 and SSDIS/ddrm4‐1 transgenic Arabidopsis were treated with 100 μM DEX for 4 h. The eChIP assays were carried out using an anti‐H2Bub antibody. The immunoprecipitated DNA (eChIP) and the input DNA were subjected to qPCR analysis. The ratios of eChIP and input are shown. UBQ5 serves as a negative control. The data are represented as means ± SD (n = 3 technical replicates). (I) The data are relative to the values of Mock. (J) The data are relative to the values of UBQ5 in Col‐0. The statistical significance was determined using two‐tailed Student's t‐test. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. The experiments were repeated at least twice with similar results.
• K
  A proposed working model.

Source data are available online for this figure.

Since UBC1/2 and HUB1/2 are known to mediate H2Bub (Liu et al, 2007; Cao et al, 2008), the hypersensitivities of the ubc1 ubc2, hub1, hub2 mutants to BLE suggested that H2Bub is required for DSB repair in Arabidopsis. To test this possibility, we first examine whether H2Bub is enriched at DSB sites by performing eChIP‐qPCR analysis using an anti‐H2Bub antibody in the SSDIS transgenic lines. Compared with the mock treatment, H2Bub was significantly enriched in the SSDIS regions after DEX treatment (Fig 7I). Interestingly, this enrichment was reduced in the ddrm4‐1 mutant (Fig 7J), suggesting that the enrichment of H2Bub at DSB sites is dependent on PAF1C.

Discussion

Based on the above data, we proposed a simplified working model to illustrate how PAF1C regulates DSB repair (Fig 7K). In the absence of DSBs, SMC5/6 binds to DNA evenly. When DSBs occur, SMC5/6 is recruited to the DSB sites, where SMC5/6 recruits PAF1C. Then, the VIP5 and CDC73 subunits of PAF1C recruit the E2 ubiquitin conjugases UBC1/2, which further recruit the E3 ubiquitin ligases HUB1/2 to promote H2Bub, facilitating DSB repair through HR.

The roles of PAF1C in transcription regulation have been well‐studied (Moniaux et al, 2006; Jaehning, 2010; Antosz et al, 2017; Vos et al, 2018; Nasim et al, 2022). Here we found that PAF1C plays a critical role in DSB repair. Since many DDR genes were upregulated in the paf1 mutant (Fig 1D, Tables EV1 and EV2), it is less likely that PAF1C contributes to DSB repair by regulating transcription of DDR genes albeit we cannot completely rule out this possibility. In support, we found that PAF1C can be recruited to DSB sites by SMC5/6 and promotes HUB1/2‐mediated H2Bub, which has been shown to contribute to DSB repair in animals (Moyal et al, 2011; Nakamura et al, 2011; Zheng et al, 2018). Therefore, the role of PAF1C in DSB repair may represent a transcription‐independent function of PAF1C.

Since PAF1 is highly conserved in eukaryotes, it is likely the role of PAF1 in DSB repair is also conserved in other organisms. Previously, it was reported that PAF1 is highly expressed in cancers including pancreatic cancer (Vaz et al, 2016) and lung cancer (Zhi et al, 2015) and is related to tumor metastasis (Nimmakayala et al, 2018), drug resistance (Vaz et al, 2014), and poor prognosis (Zhi et al, 2015). Therefore, PAF1 is considered a potential proto‐oncogene and an attractive target for cancer therapy (Moniaux et al, 2006; Karmakar et al, 2018). Since DNA damage is closely associated with cancers, it is possible that the role of PAF1 in DNA damage repair also accounts for its relationship with cancers, which may open a new research direction to study the roles of PAF1 in cancer development.

Previous studies have revealed several mechanisms of SMC5/6 in HR (Potts et al, 2006; Durrant et al, 2007; Stephan et al, 2011; Yan et al, 2013; Chen et al, 2021b). Our finding that PAF1C functions downstream of SMC5/6 further enhances our understanding of SMC5/6 in HR. Recently, we and others have reported that SMC5/6 is also important for meiosis and the SMC5/6 mutants are partially sterile (Díaz et al, 2019; Zelkowski et al, 2019; Chen et al, 2021b). Interestingly, the paf1 mutant is also partially sterile, indicating that the interaction of SMC5/6 and PAF1C may also occur in meiotic cells.

The roles of H2Bub in DNA repair have been reported in multiple organisms (Moyal et al, 2011; Nakamura et al, 2011; Zheng et al, 2018). These studies suggested that H2Bub promotes DSB repair by facilitating the recruitment of repair proteins and histone eviction at the damage sites, which is likely applied to plants. Despite its importance in DNA damage, how H2Bub is activated in response to DNA damage is not very clear. A recent study revealed that RPA recruits the E3 ubiquitin ligase Bre1 to mediate H2Bub in yeast (Liu et al, 2021). Our study suggested that SMC5/6 and PAF1C mediate the recruitment of the E3 ubiquitin ligase HUB1/2, representing a new mechanism for H2Bub activation.

Both SMC5/6 and PAF1C are fundamental protein complexes in eukaryotic cells. Our study revealed their relationship for the first time. Given the conservation of SMC5/6 and PAF1C, the interplay between these two complexes may be a general mechanism in all eukaryotes, which is worthwhile to study, especially in humans.

Materials and Methods

Plant materials and growth conditions

All Arabidopsis mutants used in this study are in the Columbia‐0 (Col‐0) background. The TRANSPLANTA collection (N2101415) was obtained from the Nottingham Arabidopsis Stock Centre. The smc6b‐2 (SALK_124719) and ddrm4‐2 (SALK_046605) mutants were obtained from the Arabidopsis Biological Resource Center. The vip3‐7 (SALK_139885), cdc73‐1 (SALK_150644), hub1‐5 (SALK_044415), and hub2‐2 (SALK_071289) mutants were obtained from Arashare (https://www.arashare.cn/). The vip5‐2 (SALK_062223) mutant was described previously (Lu et al, 2017). The vip4‐c1 and vip6‐c1 mutants were generated using CRISPR/CAS9 technology (Wang et al, 2015). All the transgenic plants were generated using the floral‐dip method (Clough & Bent, 1998). Seeds were sterilized with 2% PPM (Plant Cell Technology), stratified at 4°C in the dark for 2 days and then plated on half‐strength (1/2) Murashige and Skoog (MS) medium containing 1% sucrose and 0.35% phytagel. Plants were grown under long‐day conditions (16 h of light and 8 h of dark) in a growth chamber at 22°C.

Mutant screening

The TRANSPLANTA collection (Coego et al, 2014) was used to screen for ddrms. The seeds from each line were grown vertically on 1/2 MS medium containing 4 μM BLE for 7–9 days. The plants with shorter or longer roots than Col‐0 were considered as ddrms.

Identification of insertion site through hiTAIL‐PCR

To identify the T‐DNA insertion site(s) in ddrm4‐1, the hiTAIL‐PCR analysis was performed as described previously (Liu & Chen, 2007). After three rounds of PCR, the PCR product was purified and subjected to DNA sequencing. The primers are listed in Table EV4.

Vector construction

All vectors were constructed using the Lighting Cloning kit (BDIT0014, Biodragon Immunotechnology, China). The primers used for cloning were listed in Table EV4.

RNA extraction, RT‐qPCR, and RNA‐sequencing

The total RNA was extracted using RNA Isolation Reagent (BS259A, Biosharp). The cDNA was synthesized using First‐Strand cDNA Synthesis Kit (R312‐01, Vazyme). The qPCR was performed using CFX384 Touch™ Fluorescent Quantitative PCR Detection System (Bio‐Rad). UBQ5 was used as the reference gene. For RNA‐sequencing, 8‐day‐old seedlings of Col‐0 and ddrm4‐1 grown vertically on 1/2 MS medium were used. The sequencing library was constructed using VAHTS^® Universal V6 RNA‐seq Library Prep Kit for Illumina (NR604, Vazyme) and was sequenced using HiSeq4000 (Illumina). Raw reads were processed and aligned to the Arabidopsis genome TAIR 10 (https://www.arabidopsis.org) using Hisat2, version 2.5.1b. The transcript assembly and read quantification were performed by StringTie. Genes with over 20 reads were filtered and processed using DESeq2 to identify the differentially expressed genes (P < 0.05, |Log₂FoldChange| > Log₂1.5). Gene Ontology (GO) analysis was carried out using PANTHER Classification System database (http://pantherdb.org/).

Comet assays

Comet assays were performed as described previously (Speit & Rothfuss, 2012). The leaves from 3‐week‐old Col‐0 and ddrm4‐1 were used. The nuclei were stained with 20 μg/ml ethidium bromide for 10 min, and fluorescence was captured using confocal laser scanning microscopy (TCS SP8, Leica). The images were analyzed using the CASP Comet Assay Software (Końca et al, 2003).

HR efficiency assays

HR efficiency assays were performed as described previously (Roth et al, 2012; Wang et al, 2022). The reporter line and trigger line were respectively introduced into the ddrm4‐1 mutant through crosses.

BLE treatments and fluorescence imaging

For the fluorescent proteins transiently expressed in the leaves of N. benthamiana, the leaves were incubated with 40 μM BLE for 2 h before imaging. For the fluorescent proteins transiently expressed in Arabidopsis protoplasts, BLE was added to the culture solution to a final concentration of 20 μM and incubated for 2 h before imaging. For the fluorescent proteins in transgenic Arabidopsis, the seedlings grown vertically on 1/2 MS medium were transferred to 1/2 MS medium containing 40 μM BLE and continued to grow for 4 h before imaging. The fluorescence was captured using confocal laser scanning microscopy (TCS SP8, Leica). To avoid fluorescence interference, the sequential scan mode was used.

eChIP assays

The eChIP assays were performed as previously described with some modifications (Zhao et al, 2020). The seedlings or protoplasts were cross‐linked with 1% formaldehyde. The tissues were lysed in 300 μl Buffer S (50 mM HEPES‐KOH [pH7.5], 150 mM NaCl, 1 mM EDTA, 1% Triton X‐100, 0.1% sodium deoxycholate, 1% SDS, 1 mM PMSF). The homogenate was mixed with 1.2 ml Buffer F (50 mM HEPES‐KOH [pH 7.5], 150 mM NaCl, 1 mM EDTA, 1% Triton X‐100, 0.1% sodium deoxycholate, 1 mM PMSF), followed by sonication using the Bioruptor Plus sonication system (Diagenode, Belgium). The lysates were centrifuged at 20,000 g for 10 min at 4°C, and the supernatant was transferred to a new tube containing protein G beads (C600022‐0001, BBI) coupled with an anti‐GFP antibody (11814460001, Roche) or an anti‐H2Bub antibody (MM‐0029‐P, MediMabs). After incubating for 4 h, the beads were subsequently washed with low salt buffer (50 mM HEPES‐KOH pH 8.0, 150 mM NaCl, 1 mM EDTA pH 8.0, 1% Triton X‐100, 0.1% sodium deoxycholate, 0.1% SDS), high salt buffer (50 mM HEPES‐KOH pH 8.0, 350 mM NaCl, 1 mM EDTA pH 8.0, 1% Triton X‐100, 0.1% sodium deoxycholate, 0.1% SDS), ChIP buffer (10 mM Tris–HCl pH 8.0, 250 mM LiCl, 0.5% NP‐40, 1 mM EDTA, 0.1% sodium deoxycholate), and TE Buffer (10 mM Tris–HCl pH 8.0, 1 mM EDTA). The immunoprecipitated protein‐DNA complexes were eluted from beads by adding freshly prepared ChIP Elution buffer (1% SDS, 0.1 M NaHCO₃). The purified DNA samples were subjected to qPCR analysis or DNA sequencing. The library was constructed using NGS Fast DNA Library Prep Set for Illumina (CW2585, CWBIO) and was sequenced using HiSeq4000 (Illumina). The sequencing data were analyzed as follows: the low‐quality reads were excluded after quality control by fastqc, and adapters were removed by fastp; High‐quality reads were mapped to the Arabidopsis reference genome (TAIR 10) using Bowtie2 (version 2.3.3) with parameter “‐‐chunkmbs 320 ‐m 1 ‐‐best”; Duplicate reads were removed by Picard; MACS2 was used to detect peaks with parameter “‐‐fraglen 147 ‐‐pval‐thresh 0.00001 ‐‐make‐signal”. The eChIP‐Seq data was normalized using RPKM (Reads Per Kilobase per Million mapped reads).

Protein interaction analysis

The Y2H assays, BiFC assays, split luciferase assays, CoIP assays, and pull‐down assays were performed as described previously (Yan et al, 2013; Wang et al, 2018, 2021; Pan et al, 2021). For Y2H assays, the corresponding constructs were cotransformed into yeast strain AH109. For CoIP assays, the fusion proteins were transiently expressed in N. benthamiana. The anti‐GFP (Nanoantibody) Beads (AE079, ABclonal) were used for immunoprecipitation. For in vitro pull‐down assays, all recombinant proteins were expressed in E. coli BL21 (DE3). The MBP‐tagged proteins coupled to dextrin beads (SA026005, Smart‐lifesciences) were used to pull‐down His‐tagged proteins. For BiFC assays, the fusion proteins were coexpressed in N. benthamiana. The YFP fluorescence was examined using confocal laser scanning microscopy (TCS SP8, Leica). For split luciferase assays, the fusion proteins were coexpressed in N. benthamiana. The luminescences were captured using Lumazone Imaging System equipped with 2048B CCD camera (Roper).

Author contributions

Cunliang Li: Formal analysis; validation; investigation; visualization; project administration; writing—review and editing. Yuyu Guo: Formal analysis; validation; investigation; visualization; writing—original draft; writing—review and editing. Lili Wang: Funding acquisition; writing—review and editing. Shunping Yan: Conceptualization; formal analysis; supervision; funding acquisition; project administration; writing—review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Table EV4

PDF+

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

The EMBO Journal (2023) 42: e112756

Data availability

The raw data of RNA‐Seq and eChIP‐Seq were deposited in BioProject with the dataset identifier PRJNA883821 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA883821).