Abstract
Transcriptional silencing is an essential mechanism for controlling the expression of genes, transgenes and heterochromatic repeats through specific epigenetic marks on chromatin that are maintained during DNA replication. In Arabidopsis, silenced transgenes and heterochromatic sequences are typically associated with high levels of DNA methylation, while silenced genes are enriched in H3K27me3. Reactivation of these loci is often correlated with decreased levels of these repressive epigenetic marks. Here, we report that the DNA helicase REGULATOR OF TELOMERE ELONGATION 1 (RTEL1) is required for transcriptional silencing. RTEL1 deficiency causes upregulation of many genes enriched in H3K27me3 accompanied by a moderate decrease in this mark, but no loss of DNA methylation at reactivated heterochromatic loci. Instead, heterochromatin exhibits DNA hypermethylation and increased H3K27me3 in rtel1. We further find that loss of RTEL1 suppresses the release of heterochromatin silencing caused by the absence of the MOM1 silencing factor. RTEL1 is conserved among eukaryotes and plays a key role in resolving DNA secondary structures during DNA replication. Inducing such aberrant DNA structures using DNA cross-linking agents also results in a loss of transcriptional silencing. These findings uncover unappreciated roles for RTEL1 in transcriptional silencing and in stabilizing DNA methylation and H3K27me3 patterns.
Introduction
In eukaryotic cells, DNA is packaged with histone proteins to form a highly compacted and organized structure known as chromatin. Chromatin exists in two different main states: euchromatin, which is mainly composed of protein coding genes and sparsely condensed, and heterochromatin, which shows a high density of transposable elements and repetitive sequences and a higher level of condensation. By influencing DNA accessibility, chromatin structure largely influences various DNA-related biological processes, including transcription, DNA replication and DNA repair. The chromatin structure is also related to the presence of epigenetic marks, including various post-translational modifications of histone proteins and DNA methylation.
In Arabidopsis thaliana, DNA methylation of cytosine residues occurs in three sequence contexts: the symmetric CG and CHG contexts and the asymmetric CHH context (where H is any base but G) (Law & Jacobsen, 2010). Typically, heterochromatin sequences are transcriptionally silent and associated with high levels of DNA methylation at all three cytosine sequence contexts. In contrast, most constitutively active genes in euchromatin are either unmethylated or DNA methylated in their coding sequence specifically at CG sites (Tran et al, 2005; Zhang et al, 2006; Lister et al, 2008). On the other hand, genes with spatiotemporal expression are typically silenced in certain tissues or at certain developmental stages by histone H3 lysine 27 trimethylation (H3K27me3) deposited by Polycomb group (PcG) proteins (Bieluszewski et al, 2021; Xiao & Wagner, 2015).
DNA methylation is established de novo by the RNA-directed DNA methylation (RdDM) pathway and then maintained by specific pathways to each cytosine sequence context. CG sites are the most frequently and highly methylated, with maintenance dependent on METHYLTRANSFERASE 1 (MET1) that acts during DNA replication. CHROMOMETHYLASE 3 (CMT3) is the primary methyltransferase for CHG methylation maintenance. Methylation at CHH sites in heterochromatin is mainly maintained by CHROMOMETHYLASE 2 (CMT2), while DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2) acting in the RdDM pathway methylates CHH positions in shorter, less heterochromatic sequences (Du et al, 2015; Law & Jacobsen, 2010)L. In the poorly accessible heterochromatin environment, DNA methylation depends on Decrease in DNA Methylation 1 (DDM1) to overcome the linker histone H1 barrier and allow access to DNA for MET1, CMT2 and CMT3 (Zemach et al, 2013; Vongs et al, 1993).
Disruption of DNA methylation pathways, particularly CG methylation, leads to the reactivation of numerous transposable elements and some protein coding genes (Lister et al, 2008; Zhang et al, 2006; Mathieu et al, 2007). However, DNA methylation is not the sole mechanism which the cell relies on to secure transcriptional gene silencing. There is growing evidence that DNA methylation-independent pathways participate in robust silencing. Several key factors involved in these pathways have been identified, including MORPHEUS’ MOLECULE 1 (MOM1), a protein that mediates silencing of various pericentromeric sequences (Amedeo et al, 2000; Steimer et al, 2000; Vaillant et al, 2006; Yokthongwattana et al, 2010), and members of the conserved Microrchidia (MORC) ATPase family, AtMORC1 and AtMORC6, which regulate silencing of several, mostly heterochromatic, TEs (Moissiard et al, 2012). The precise mechanisms by which these actors control silencing are not yet fully understood. There is evidence that MOM1 and AtMORC6 act in largely distinct pathways in controlling silencing of pericentromeric TEs, although MOM1 appears to recruit the RdDM machinery via MORC6 at some RdDM sites (Moissiard et al, 2014; Li et al, 2023). Transcriptional reactivation in mom1 mutants requires the MED14 subunit of the MEDIATOR complex (Bourguet et al, 2018). More recently, we found that the MAINTENANCE OF MERISTEMS (MAIN) and MAIN-LIKE1 (MAIL1) proteins, which contain a transposon-related plant mobile domain, also control silencing of a set of pericentromeric TEs independently of DNA methylation through a molecular pathway distinct from those involving MOM1 and AtMORC6 (Ikeda et al, 2017).
During DNA replication, both heterochromatin and euchromatin are disrupted ahead of the replication fork and are then reassembled into their original epigenetic states behind the fork (Probst et al, 2009). Similarly, chromatin structure must be modulated during DNA repair to allow repair factors access to the DNA molecule, and subsequently restored to its original state to maintain transcriptional activity of the underlying sequences. Stalling of the replication fork leads to epigenetic instability and loss of silencing, suggesting that DNA unwinding, replication and chromatin re-formation must be tightly coordinated (Sarkies & Sale, 2012)L. Mutations in several factors involved in DNA replication and/or repair have been found to affect silencing, indicating a tight link between these processes. These factors include, among others, the DNA polymerases POLα (Liu et al, 2010), POLδ (Iglesias et al, 2015), POLL (Bourguet et al, 2020; Yin et al, 2009), REPLICATION PROTEIN 2 (RPA2) (Elmayan et al, 2005)L, the nuclear protein BRUSHY1 (BRU1) (Takeda et al, 2004), the flap endonuclease FEN1 (Zhang et al, 2016) and the chromatin assembly factor CAF-1 (Schönrock et al, 2006). Mutants with defects in these factors exhibit increased sensitivity to DNA damaging agents, indicating the accumulation of DNA damage and/or constitutive replication stress. Notably, release of silencing in these mutants is not associated with a loss of DNA methylation.
Here, we identify a role in silencing for REGULATOR OF TELOMERE ELONGATION 1 (RTEL1) encoding a DNA helicase protein linked to DNA repair and replication. We show that RTEL1 is required for the silencing of both a transgene and a set of endogenous loci independently of a loss of DNA methylation, and through a pathway likely related to its replication function. Additionally, we show that RTEL1 is necessary from preventing a gain of H3K27me3 and DNA methylation in heterochromatin regions. Upregulated genes in rtel1 show a moderate loss of H3K27me3, while heterochromatin regions gain H3K27me3 and DNA methylation. Moreover, our study indicates that RTEL1 is required for the release of heterochromatin silencing in the absence of MOM1, and for viably in pol2a and mail1 mutant backgrounds.
Results
Mutations in RTEL1 release transcriptional silencing
The Arabidopsis L5 line carries transcriptionally silent tandem repeats of a ß-glucuronidase (GUS) transgene under control of the CaMV 35S promoter (Morel et al, 2000). We and others previously showed that silencing of the L5-GUS transgene is sensitive to mutations in a wide range of transcriptional silencing regulators (Bourguet et al, 2018; Ikeda et al, 2017; Bourguet et al, 2020; Elmayan et al, 2005; Takeda et al, 2004; Morel et al, 2000). We conducted a genetic screen on an ethyl methanesulfonate (EMS) mutagenized L5 population for mutants that reactivate the L5-GUS transgene independently of a loss of DNA methylation. We used GUS histochemical staining assay on isolated leaves and methylation-sensitive restriction enzyme digestion followed by PCR to select M2 plants with destabilized L5-GUS silencing and no loss of DNA methylation at the 5’ end of the transgene GUS coding region. We identified six mutants, for which silencing release of the L5-GUS transgene followed a 1/3 ratio in the F2 progeny of backcrosses with L5, indicating that the causative mutations were single-locus nuclear recessive. Further mapping-by-sequencing of candidate mutations using outcross F2 populations identified new mutant alleles of RPA2, MOM1, POLa2, and two mutant alleles of POL2A already described (Figure S1) (Yin et al, 2009; Bourguet et al, 2020), validating our screening strategy as mutations in these genes are known to affect silencing without associated loss of DNA methylation (Vaillant et al, 2006; Liu et al, 2010; Bourguet et al, 2020; Yin et al, 2009; Elmayan et al, 2005). In the last mutant, p14 12-1, we identified a C to T substitution in AT1G79950 (REGULATOR OF TELOMERE LENGTH 1, RTEL1), causing a premature stop codon at the amino acid position 712 (Figure 1a-c, S2a), most likely leading to an abortion of translation. Complementation test between p14 12-1 and the independent rtel1-1 T-DNA insertion mutant line (Recker et al, 2014), rescue of p14 12-1 defects with a RTEL1 genomic copy, and introducing L5-GUS in the rtel1-1 background confirmed that mutation of RTEL1 was responsible for the observed silencing defects (Figure 1, S2b-d). Hence, p14 12-1 was renamed rtel1-3. In addition to the L5-GUS transgene, silencing of the Athila retrotransposon-derivates TSI repeats was destabilized to some extent in rtel1 mutants, and expression of the silenced 45S rRNA variant VAR1 was activated in rtel1-3 (Figure 1d).
Loss of rtel1 upregulates DNA damage response genes and genes enriched in H3K27me3
To define how mutations in RTEL1 affect transcription across the genome, we conducted RNA sequencing (RNA-seq) on rtel1-1 and rtel1-3 mutant seedlings and compared their transcriptomes to that of WT (L5) plants. We identified 346 and 781 differentially expressed genes (DEGs) in rtel1-1 and rtel1-3, respectively. The large majority of DEGs in rtel1-1 (75%) and rtel1-3 (90%) were upregulated, indicating that loss of RTEL1 primarily leads to the upregulation of transcripts (Figure 2a). Notably, the upregulated transcripts were mostly derived from protein coding genes (PCGs), although 16 and five TEs were also upregulated in rtel1-3 and rtel1-1, respectively. We identified 687 PCGs upregulated in rtel1-3 and 261 in rtel1-1, with about 76% of the rtel1-1 upregulated PCGs being also upregulated in rtel1-3 (Figure 2b). Thus, rtel1-3 has a stronger impact on transcription compared to rtel1-1. These results indicate that RTEL1 is required for proper expression of several hundreds of genes and silencing of a subset of heterochromatic repeats.
Previous studies in Arabidopsis have shown that RTEL1 is required for proper DNA replication and DNA repair, and its loss causes activation of DNA repair genes (Dorn et al, 2019; Recker et al, 2014). Accordingly, Gene Ontology (GO) analysis indicated that genes involved in response to DNA damage and DNA damage repair were overrepresented among PCGs upregulated in both rtel1-1 and rtel1-3 (Figure 2c, Table S1). This suggests that rtel1 mutants undergo constitutive replicative stress, which likely contributes to the upregulation of a subset of PCGs. Interestingly, rtel1-upregulated PCGs were also enriched in stress responsive genes (Figure 2c, Table S1). Stress-responsive genes are enriched in the histone variant H2A.Z, and the majority of H2A.Z nucleosomes associate with the H3K27me3 repressive histone modification in Arabidopsis (Carter et al, 2018; Coleman-Derr & Zilberman, 2012). We determined genome-wide H3K27me3 patterns in wild-type and rtel1-3 seedlings using chromatin immunoprecipitation followed by sequencing (ChIP-seq). More than 65% of rtel1-3 upregulated PCGs were associated with high levels of H3K27me3 in the wild type (Figure 2d, e). Notably, rtel1-upregulated PCGs associated with H3K27me3 were enriched in GO-terms related to stress response, whereas GO-terms involved in DNA damage response were only enriched in rtel1-upregulated PCGs that were not marked by H3K27me3 (Table S2-5). Consistent with H3K27me3 repressing their transcription, rtel1-3-upregulated PCGs enriched in H3K27me3 had significantly enhanced transcript accumulation in double mutants for the CURLY LEAF and SWINGER H3K27me3 methyltransferases, as compared to rtel1-upregulated PCGs not marked by H3K27me3 (Figure 2f). Additionally, we found that loss of RTEL1 was associated with a moderate but reproducible reduction in H3K27me3 at upregulated PCGs associated with this mark in the WT (Figure 2d). Together, these findings suggest that both constitutive replicative stress and perturbed H3K27me3-mediated silencing likely contribute to PCG upregulation in rtel1 through at least partly independent pathways.
RTEL1 prevents heterochromatin DNA hypermethylation
The QUA-QUINE STARCH (QQS) gene is among the few PCGs that are downregulated in rtel1 mutants. QQS is located in the pericentromeric heterochromatin of chromosome 3, and its expression is negatively correlated with the DNA methylation level of repeated sequences located within the 5’ end of the gene (Silveira et al, 2013; Rigal et al, 2016). This led us to speculate that RTEL1 deficiency may alter DNA methylation patterns at some endogenous sequences. We investigated this possibility by determining genome-wide DNA methylation profiles in WT and rtel1-3 plants using bisulfite sequencing (BS-seq). This showed that DNA methylation levels of the 5’ end of QQS were indeed increased in rtel1-3 in correlation with transcriptional downregulation of the gene (Figure S3a). More surprisingly, we found that CHG methylation levels were significantly increased at the L5-GUS transgene in rtel1-3, despite transcriptional reactivation (Figure S3b).
As expected from the general anticorrelation between DNA methylation and H3K27me3 (Mathieu et al, 2005; Deleris et al, 2012; Weinhofer et al, 2010), rtel1-3 upregulated PCGs, which are enriched in H3K27me3, showed low DNA methylation levels in the WT, and these remained unaffected in rtel1-3 (Figure S3c). We plotted average DNA methylation levels along chromosomes and found no overt change in CG DNA methylation in rtel1-3 mutants (Figure 3a). However, we observed that DNA methylation levels of pericentromeric regions were noticeably increased at CHG, and to a lesser extent, CHH sites (Figure 3a). Consistently, average CHG, and to a lesser extent, CHH methylation levels were increased over TEs in rtel1-3 compared to WT (Figure S3d). All three CHG sequence subcontexts (CAG, CTG and CCG) were equally hypermethylated in rtel1-3 (Figure S3d). Separating TEs according to their genomic localization showed that non-CG hypermethylation in rtel1-3 was much more pronounced at pericentromeric TEs than at TEs located on chromosome arms (Figure 3b). We also determined differentially methylated regions (DMRs) in rtel1-3 relative to WT plants and found that rtel1-3 mostly induced CHG hypermethylation, which was largely clustered at pericentromeric TE-rich heterochromatin (Figure 3c, d). Arabidopsis heterochromatin is associated with the H2A.W histone variant and enriched in H3K27me1 (Yelagandula et al, 2014; Jacob et al, 2009). We found that the gain in DNA methylation in rtel1-3 specifically occurred at regions associated with these heterochromatin features and was particularly marked at CHG sites, although heterochromatic CG and CHH sites did show some degree of hypermethylation (Figure 3e). However, this increase in DNA methylation levels was not associated with the upregulation of DNA methyltransferases genes (including CMT3) or the downregulation of DNA demethylase genes (Figure S3e). These results demonstrate that RTEL1 is required for maintaining proper patterns of heterochromatin DNA methylation, particularly at CHG sites.
Higher order chromatin organization appears unaltered in rtel1
In 4’,6-diamidino-2-phenylindole (DAPI) stained WT nuclei, heterochromatin appears as brightly stained foci called chromocenters, which contain most of the highly repetitive DNA in the genome, including ribosomal DNA (rDNA) and satellite repeats. The nuclei of WT and rtel1-3 plants contained a similar number of chromocenters and an equivalent relative heterochromatic fraction (Figure 4a, b). Previous studies have shown that rtel1-1 mutants display a reduced number of 45S rDNA repeats (Röhrig et al, 2016; Dorn et al, 2019), however, we were unable to detect variation in the amount of 45S rDNA repeats in rtel1-3 (Figure S4). To further investigate a potential impact of rtel1-3 on higher-order chromatin structure, we profiled chromatin accessibility in WT and rtel1-3 seedlings using Assay for Transposase-Accessible Chromatin followed by high-throughput sequencing (ATAC-seq). We found that rtel1-3 did not affect overall chromatin accessibility profiles (Figure 4c, d). A closer examination of rtel1 upregulated PCGs revealed that increased chromatin accessibility was not required for, nor a necessary consequence of, the transcriptional upregulation of these genes. Rather, these showed a moderate decrease in accessibility around their transcription start site (TSS), possibly reflecting active loading of the transcription machinery at these genes in the mutant (Figure 4d). In conclusion, our results suggest that higher-order chromatin organization remains largely intact in rtel1-3.
RTEL1 influences gene expression largely independently of POL2A and MAIL1
Two mutant alleles of the POL2A gene encoding the catalytic subunit of polymerase epsilon were recovered from our screen (Figure S1). We previously reported that pol2a mutations releases heterochromatin silencing and leads to increased levels of DNA methylation, particularly at CHG sites (Bourguet et al, 2020)L. However, in contrast to rtel1-3, TEs located outside pericentromeric heterochromatin along chromosome arms also exhibit marked CHG hypermethylation in pol2a mutants (Figure 3b, S5a)L (Bourguet et al, 2020). Part of the CHG hypermethylation in pol2a is accounted by increased expression of CMT3 (Bourguet et al, 2020)L; however, CMT3 transcript accumulation was not up-regulated in rtel1-3 (Figure S3e). Additionally, heterochromatin gains CG methylation in rtel1-3, but not in pol2a-12 (Bourguet et al, 2020). We sought to analyze epistasis between pol2a and rtel1 mutations. However, we were unable to isolate rtel1 pol2a double mutants in the progeny of self-fertilized rtel1-3 pol2a-12 double heterozygotes or either sesquimutants, suggesting a lethal genetic interaction between rtel1-3 and pol2a-12. Comparing the transcriptomes of pol2a-12 and rtel1-3 single mutants revealed little overlap of upregulated PCGs (Figure S5b). Although perturbed H3K27me3-mediated silencing likely contributes to transcriptional upregulation of PCGs in both pol2a and rtel1 (see Figure 2d-f) (Bourguet et al, 2020), the sets of H3K27me3-marked PCGs upregulated in each mutant were mostly distinct (Figure S5b), and rtel1-upregulated PCGs displayed much higher average H3K27me3 levels than pol2a-upregulated genes (Figure S5c). Altogether, our data suggest that RTEL1 and POL2A function in at least partly distinct pathways to regulate gene expression and restrain DNA methylation.
We previously showed that MAIL1 regulates the expression of numerous genes and ensures silencing of a set of pericentromeric TEs in addition to the L5-GUS transgene independently of DNA methylation (Ikeda et al, 2017). Although mail1 mutants display severe developmental defects that do not resemble rtel1-3 mutant phenotype, TEs show increased levels of non-CG methylation in both mutants (Figure S5d) (Nicolau et al, 2020; Bourguet et al, 2020). We sought to test the genetic interaction between rtel1 and mail1 mutations in non-CG DNA hypermethylation by crossing rtel1-3 with the mail1-1 knockout mutant. However, we were unable to obtain any rtel1-3 mail1-1 double mutant plants in the progeny of double heterozygotes or in the progeny of rtel1-3 MAIL1/mail1-1 sesquimutant, indicating that RTEL1 and MAIL1 are synergistically required for viability. Transcriptome comparison of rtel1-3 and mail1-1 single mutants highlighted distinct sets of upregulated PCGs with limited overlap regardless of their association with H3K27me3 (Figure S5e). We conclude that RTEL1 and MAIL1 regulate gene expression through largely non-overlapping molecular pathways.
RTEL1 stabilizes H3K27me3 level over heterochromatin
We also recovered a mutant allele of MOM1 from our screen for mutants that release silencing without loss of DNA methylation (Figure S1). MOM1 was the first identified protein required for transcriptional silencing independent of DNA methylation in Arabidopsis (Amedeo et al, 2000), and disruption of MOM1 releases silencing of the L5-GUS transgene and of many pericentromeric sequences (Amedeo et al, 2000; Elmayan et al, 2005; Vaillant et al, 2006). We obtained rtel1-3 mom1-2 double mutants, which were phenotypically indistinguishable from rtel1-3 single mutants (Figure 5a). Histochemical staining for GUS activity revealed that L5-GUS silencing was more strongly released in mom1 than in rtel1 and rtel1 mom1 (Figure 5b), suggesting a genetic interaction between rtel1 and mom1. Similarly, release of silencing of the TSI repeats was significantly stronger in mom1-2 than in rtel1-3, and rtel1 mom1 double mutants exhibited intermediate levels of TSI transcript accumulation (Figure 5c), indicating that rtel1 restrains mom1-induced reactivation at these loci. To further investigate epistasis between rtel1 and mom1 mutations, we compared silencing defects in mom1 and mom1 rtel1 using RNA-seq. We identified 72 reactivated TEs in mom1, the majority of which showed a sharp reduction in transcript levels in mom1-2 rtel1-3 (Figure 5d), suggesting that RTEL1 is required for silencing release in mom1-2. We previously reported that TE transcription in mom1-2 requires MEDIATOR14 (MED14) (Bourguet et al, 2018). However, MED14 transcript levels were similar in WT and rtel1 plants (Figure S6), suggesting that the requirement for RTEL1 and MED14 were independent. We hypothesized that RTEL1 dependency for mom1-induced silencing release might rely on changes in chromatin structure and/or patterns of epigenetic marks at these loci in the absence of RTEL1. Chromatin accessibility of TEs silenced by MOM1 was not altered in rtel1 (Figure S7). However, these TEs showed marked hypermethylation of CHG sites and, more surprisingly, an increased association with H3K27me3 in the absence of RTEL1 (Figure 5e, f). Thus, increased levels of these two repressive marks possibly contribute to limiting mom1-induced activation of these endogenous loci in the rtel1 mom1 background.
Since MOM1-regulated TEs are mostly located in heterochromatin, we further analyzed changes in H3K27me3 over heterochromatic sequences in rtel1-3. We found that, on average, TEs located in chromosome arms or in pericentromeres exhibited equally low association with H3K27me3 compared to PCGs in the WT (Figure 6a). Loss of RTEL1 was associated with increased levels of H3K27me3 over TEs located in pericentromeres, while TEs in chromosome arms and PCGs remained largely unaffected (Figure 6a), suggesting that heterochromatin specifically gain H3K27me3 in rtel1. To strengthen this conclusion, we determined H3K27me3 changes in rtel1 in relation to H2A.W and H3K27me1. Genomic regions depleted in these heterochromatin marks showed no change in H3K27me3, whereas enrichment in H2A.W and H3K27me1 positively correlated with increased association with H3K27me3 in rtel1 (Figure 6b). Thus, our data indicate that in the absence of RTEL1, heterochromatin regions of the genome specifically gain H3K27me3.
Accumulation of DNA cross-links destabilizes silencing
It has been shown previously that RTEL1 functions in the repair events of DNA crosslinks (Hu et al, 2015; Recker et al, 2014). We used mitomycin C (MMC) and cis-platin (CP) crosslinking agents to assess whether accumulation of DNA crosslinks impacts transcriptional silencing. MMC mainly generates inter-strand cross-links on DNA, whereas CP preferentially forms intra-strand cross-links. Histochemical staining showed that applying CP or MMC to WT (L5) plants reactivated L5-GUS expression (Figure 7a, b). In these plants, L5-GUS expression appeared more pronounced in (non-replicating embryonic tissues) cotyledons, with MMC treated plants additionally showing GUS reactivation in hypocotyl and adult root regions, a pattern similar to that observed in untreated rtel1 mutants (Figure 7a). Drug treatment did not enhance GUS reactivation in rtel1-3 (Figure 7a, b). RNA-seq revealed that about half of the PCGs upregulated upon MMC treatment in the WT were also upregulated in untreated rtel1-3 plants (Figure 7c). Conversely, PCGs differentially expressed in rtel1-3 in the absence of drug were on average similarly differently expressed when WT plants were exposed to MMC (Figure 7d). Globally, transcriptional changes associated with rtel1-3 mutation and those induced by treating WT plants with MMC showed positive correlation (Figure S8). These data show that inducing DNA crosslinks leads to release of silencing and suggest that accumulation of these DNA lesions may be, at least partly, at the origin of transcriptional defects in rtel1 mutants.
In Arabidopsis, the FANCJB helicase, closely related to RTEL1, is also involved in the repair of inter-strand crosslinks. However, both helicases were shown to act in independent pathways of inter-strand crosslinks repair (Dorn et al, 2019). We combined rte1-3 mutation with the fancjb-1 mutation by crossing and obtained rtel1-3 fancjb-1 double mutants containing the L5-GUS transgene. The fancjb-1 mutants were indistinguishable from their WT siblings; however, rtel1-3 fancjb-1 plants exhibited a more severe developmental phenotype compared to rtel1-3and first generation double mutants were complete sterile (Figure 7e, Figure S9a). We found that L5-GUS expression remained silenced in fancjb-1, whereas L5-GUS silencing was further released in rtel1-3 fancjb-1 double mutants compared to rtel1-3 single mutants (Figure 7f). Similarly, silencing release of the Athila retrotransposon-derivates repeats (TSI) was enhanced in rtel1-3 fancjb-1 compared to rtel1-3 (Figure S9b). Thus, in the absence of RTEL1, FANCJ appears to restrain silencing release. Taken together, these results indicate that RTEL1-mediated repair of inter-strand crosslinks has a prominent role in silencing, and that FANCJB backs up RTEL1 to promote efficient silencing.
Discussion
In this study, we uncover unappreciated roles for the DNA helicase RTEL1 in transcriptional silencing and in controlling DNA methylation and H3K27me3 patterns. RTEL1 deficiency releases transcriptional silencing and is associated with changes in H3K27me3 levels characterized by a reduction at up-regulated genes and an increase over heterochromatin. In addition, DNA methylation levels increase in heterochromatin, particularly at CHG sites.
In Arabidopsis, mouse, human cells, and C. elegans, RTEL1 plays an essential and conserved DNA function in facilitating replication, which ensures genome stability and cell cycle progression (Dorn et al, 2019; Hu et al, 2015; Recker et al, 2014; Röhrig et al, 2016; Vannier et al, 2013; Uringa et al, 2012; Barber et al, 2008). RTEL1 associates with PCNA via a PIP box domain in its C terminus that is conserved in human, mouse and Arabidopsis, and accumulates at sites of replication stress, suggesting that RTEL1 could target stalled replisomes and facilitate replication fork progression by resolving DNA secondary structures (Hu et al, 2015; Vannier et al, 2013). We previously reported that inhibiting replication fork progression by depleting the cellular dNTP pool through the supply of HU induces loss of silencing (Bourguet et al, 2020). RTEL1 deficiency is associated with prolonged S-phase duration, likely due to the accumulation of cross-linked DNA (Hu et al, 2015), and we show here that treatment with the cross-linking agents, which also likely interferes with replication fork progression, also results in release of silencing. These findings support that replication stress can trigger transcription of normally silent regions. Interestingly, we found that exposing rtel1 plants to cross-linking agents did not further enhance the release of silencing, suggesting that rtel1 mutation and cross-linking agents may act through a common pathway to cause silencing loss. This supports the hypothesis that RTEL1-mediated repair of inter-strand crosslinks is essential for stabilizing transcriptional silencing. FANCJB, one of the 5’-3’ DNA helicases, has been reported to play a role in repairing inter-strand crosslinks in Arabidopsis independently of RTEL1 (Dorn et al, 2019). We find that the combined deficiency of both helicases results in developmental defects and complete sterility. This synthetic sterility appears to be evolutionarily conserved as C. elegans double mutants for rtel1 and the FANCJ homolog dog-1 fail to develop embryos due to germline proliferation defects, which were assigned to replicative stress (Barber et al, 2008). Our data suggest that in the absence of RTEL1, FANCJB-mediated repair of inter-strand crosslinks serves a backup function restraining loss of silencing. FANCJ was shown to function alongside RTEL1 to overcome covalent DNA-protein crosslinks that constitute bulky roadblocks to the advancing DNA replication machinery (Sparks et al, 2019; Yaneva et al, 2023); however, inter-strand DNA crosslinks constitute different barriers to replisome progression. Repair of covalent links between DNA strands is a complex multi-step process with several functional levels including inter-strand crosslink detection and unhooking by DNA incisions that generate a double strand break, monoadduct bypass and removing, and finally recombinational repair of double strand break, which eventually restore replication fork progression. In vitro, purified human RTEL1 mediates dismantling of displacement loops (D loops) that are formed as recombination intermediates (Barber et al, 2008), and this anti-recombinase activity of RTEL1, which allows replication fork progression, is conserved in Arabidopsis (Recker et al, 2014; Hu et al, 2015; Olivier et al, 2018). Thus, RTEL1 deficiency may results in the accumulation of incompletely processed and toxic joint DNA molecules during repair of inter-strand crosslink-induced DNA damage, which perturbs DNA replication. FANCJ plays a crucial function in branch-migration, which suppresses these intermediate DNA structures (Awate et al, 2020). Additionally, FANCJ can destabilize the RAD51 nucleoprotein filament associated with single-stranded DNA and inhibit the strand exchange reaction to complete homologous recombinational repair (Sommers et al, 2009). Therefore, we hypothesize that, to compensate the lack of RTEL1 and the presence of toxic recombination intermediates, FANCJB supports inter-strand crosslinks repair by inhibiting the strand exchange reaction, thereby alleviating replicative stress and silencing release.
How may replication stress destabilize silencing? Transcriptionally upregulated loci in rtel1 are typically associated with either high levels of DNA methylation (e.g., L5-GUS transgene, TSI repeats) or enriched in H3K27me3 (PCGs). The propagation of these repressive epigenetic marks is tightly linked to DNA replication to ensure prompt restoration of the epigenetic landscape following doubling of the DNA amount (Jiang & Berger, 2017; Xu et al, 2010; Law & Jacobsen, 2010). As the replication fork progresses, newly synthesized DNA strands acquire their DNA methylation pattern by copying the preexisting DNA methylation information from the template strand. Parental histones are displaced and recycled with their modifications and re-deposited at their pre-replicative positions. New, unmodified histones are deposited, resulting in a twofold dilution of parental histone modification levels. These new histones are then modified to restore the parental level of histone modifications (Alabert et al, 2015; Reverón-Gómez et al, 2018). In yeast and animal cells, parental histone recycling is facilitated by the tethering of parental histones to the DNA polymerases alpha and epsilon, and disruption of this interaction leads to transcriptional defects (Bellelli et al, 2020; Evrin et al, 2018; Li et al, 2020). In HeLa cells, replication stress induced by HU treatment leads to local loss of histone (H3.1) recycling and severely impairs the spatial distribution of parental histones in the surrounding regions, which may likely contribute to epigenomic instability (Clément et al, 2018). We propose that, upon RTEL1 deficiency, defective repair of inter-strand crosslink-induced DNA damage creates local DNA perturbation at the replication fork that impair recycling of parental histones and delays epigenome restoration following DNA replication. This may provide a short time-window with locally reduced levels of H3K27me3 or DNA methylation, allowing transcription factor binding and transcription initiation in typically silent genomic regions. The gain in H3K27me3 over heterochromatin in rtel1 might reflect relocation of this mark from chromosome arms towards this late replicating genome compartment, which appears less affected by replication stress-induced histone recycling defects in HeLa cells (Clément et al, 2018). RTEL1 deficiency is associated with increased levels of DNA methylation in heterochromatin, most prominently at CHG sites, which occurs independently of transcriptional upregulation of the DNA methyltransferases. We previously reported that replication stress induced by HU treatment and mutations in other DNA replication-related genes, such as POL2A, RPA2A, and POLD2, also lead to increased CHG methylation levels in heterochromatin (Bourguet et al, 2020). Hence, heterochromatin CHG hypermethylation appears to be a molecular hallmark of replication stress. Recent studies in yeast have shown that fork stalling results in local chromatin alterations associated with higher levels of the H3K9me3 heterochromatin mark. This locally compacts chromatin and avoids fork collapse (Feng et al, 2019). In Arabidopsis, H3K9me2 serves as the functional equivalent of H3K9me3. Considering the tight interconnection between H3K9me2 and CHG methylation maintenance, a similar mechanism could contribute to the increased CHG methylation observed under replication stress. Methylation of CHG sites mostly occurs in heterochromatin, and because its maintenance is not as efficient as that of CG methylation (Niederhuth et al, 2016), it may explain why heterochromatin hypermethylation is skewed toward CHG sites in rtel1 and other mutants that exhibit constitutive replication stress. Heterochromatin is difficult to replicate, regularly causing replication fork stalling even in the WT (Zeman & Cimprich, 2014), which could contribute to maintaining higher CHG methylation levels in these genomic regions. Increased levels of CHG methylation (and H3K27me3) in heterochromatin likely mitigate transcriptional de-repression of the underlying sequences under replication stress. This is consistent with the impact of RTEL1 deficiency on mom1-induced transcription of heterochromatic TEs, and with the strong upregulation of TEs we previously observed in pol2a cmt3 double mutants (Bourguet et al, 2020).
Our study reveals that POL2A and RTEL1 are synergistically required for Arabidopsis viability. This finding is consistent with a recent report indicating that removal of both RTEL1 and DNA polymerase epsilon leads to loss of viability in worms and vertebrates (Bellelli et al, 2020), indicating that this synthetic lethal interaction is conserved across different kingdoms. Notably, the combined loss of RTEL1 and DNA polymerase epsilon confers extensive genome instability and cessation of DNA replication in both worms and mouse cells (Bellelli et al, 2020). We also find that rtel1 is synthetic lethal when combined with mail1, which is known to cause DNA damage accumulation (de Luxán-Hernández et al, 2020). This suggests that RTEL1 function is essential to DNA replication in the presence of DNA damage. Together, our results highlight the importance of RTEL1 and its interaction with other DNA replication factors in maintaining epigenome integrity and cell viability.
Materials and Methods
Plant material
Plants were grown in soil in long-day conditions (16 h light, 8 h dark) at 23 °C with 60% relative humidity. The rtel1-3 mutant allele reported in this study was isolated from a population of EMS mutagenized L5 plants that we previously described (Ikeda et al, 2017)L. The rtel1-3 mutant was backcrossed twice to the parental L5 line before analysis. The rtel1-1 (SALK_113285), pol2a-12 (Bourguet et al, 2020), mail1-1 (GK-840E05), mom1-2 (SAIL_610_G01) and fancjb-1 (SALK_101493) mutant lines used in this study were all in a L5 genetic background. Primers used in this study are listed in Table S6.
Histochemical staining
Whole seedlings or rosette leaves were vaccum infiltrated twice 5 min with 3 ml of X-Gluc staining solution (50 mM NaHPO4 pH 7; 10mM EDTA; 0.2% Triton-X-100; 0.04% X-Gluc) and incubated 24 h at 37°C in the dark. Chlorophyll was subsequently cleared with repeated washes in ethanol at room temperature.
Mutagenesis and mapping
The mutant L5 population used in this study was obtained by mutagenesis as described in Ikeda et al., 2017. To screen for mutants impaired in transcriptional gene silencing, one leaf per M2 plant was dissected and histochemically stained as previously described. DNA of plant selected at this stage was extracted from rosette leaves using Wizard Genomic DNA Purification kit (Promega) and digested with methylation-sensitive restriction enzyme (MspI or CfoI, NEB) according to the manufacturer’s instructions. GUS 5’ extremity and FWA gene (control region without MspI/CfoI restriction site) was amplified, and methylation levels were estimated by measuring the amplification ratio between the transgene and the control gene to select mutants that reactivate L5-GUS transgene independently of a loss of DNA methylation. We performed mapping-by-sequencing using genomic DNA from a pool of 50-100 individual F2 mutant plants segregating from a p14 12-1 x Ler cross. Libraries were prepared using the TruSeq Nano DNA Sample Preparation kit (Illumina) and sequenced on a NextSeq 500 instrument (Illumina) at Fasteris S.A. (Geneva, Switzerland) to generate ∼ 100M 150-bp paired-end reads. Reads were mapped on the Arabidopsis TAIR10 genome using BWA (Li & Durbin, 2009) with the default parameters. Single-nucleotide polymorphisms calling was performed using SAMtools mpileup followed by BCFtools view with the -bvgN option. Three candidate genes with EMS-induced non-synonymous mutations were identified in the mapping interval on chromosome 1. T-DNA mutant lines for the three candidate genes were analyzed for release of gene silencing and allelic tests were performed, allowing identification of RTEL1.
Complementation of rtel1-3
A genomic fragment containing the RTEL1 gene with its promoter region (1 kb upstream the ATG codon) was PCR amplified from Col-0 genomic DNA and cloned into the pDONR221 vector (Invitrogen) by BP recombination. Following sequence verification, this fragment was introduced in the GATEWAY destination vector pB7FWG2,0 Δp35S by LR recombination. The resulting pRTEL1::RTEL1 construct was introduced into rtel1-3 plants using Agrobacterium-mediated floral-dip transformation.
Drug Treatments
The seeds were surface-sterilized with 0.05% SDS, 70% ethanol for 5 min, rinsed 5 min with 95% ethanol, air dried and evenly spread on sterile 0.8% agar, 1% sucrose and 1X Murashige and Skoog (M0255; Duchefa Biochemie) medium supplemented with or without the indicated concentrations of MMC (M0503; Sigma-Aldrich) or CP (TEVA Pharma). Whole seedlings were harvest at 14-day-old for analysis.
Transcript analysis
Total RNA was extracted from immature inflorescences (stages 1-12) or aerial parts of 16-day-old seedlings using NucleoSpin RNA Plus kit (Macherey-Nagel) following the manufacturer’s recommendations. Gel-based RT-PCR analyses were performed using the One-Step RT-PCR kit (Qiagen) in a final volume of 10 µL starting from 0.03 µg of DNase-treated (RQ1 Rnase-free Dnase supplemented with Rnasin Ribonuclease Inhibitor; Promega) total RNA. For RT-qPCR analyses, 1 µg of Dnase-treated total RNA was reverse transcribed using the M-MLV Reverse Transcriptase (Promega) and Random Hexadeoxynucleotides (Promega) in a final volume of 25 µL. Two microliters of cDNA was used for subsequent amplification using SensiFAST SYBR No-Rox Kit (Bioline) on an Eco™ Real-Time PCR System (Ilumina), following a program of 5 min at 95°C, 40 cycles of 10 s at 95°C, and 30 s at 60°C, in a final reaction volume of 10 µL. Amplification of ACTIN2 gene transcripts was used as a reference for normalization and data were analyzed according to the 2-ΔΔCt method (Livak & Schmittgen, 2001).
mRNA sequencing
Total RNA was extracted from 14-or 16-day-old seedlings with RNAzol (Euromedex) and then treated with the RQ1 RNase-free DNase (Promega) supplemented with RNasin Ribonuclease Inhibitor (Promega), cleaned up either with phenol-chloroform and precipitated with ethanol or with RNA Clean & Concentrator -5 kit (Zymo Research). Two biological replicates were collected for each genotype. Libraries were prepared using the TruSeq Stranded mRNA Library Prep kit (Illumina) and sequenced on a HiSeq 4000 instrument (Illumina) at Fasteris S.A. (Geneva, Switzerland) to generate ∼ 26–35M 50-bp single-end reads. Reads were aligned to the TAIR10 Arabidopsis genome plus the L5-GUS transgene with STAR version 2.7.9a (Dobin et al, 2013) retaining multi-mapped reads mapping up to 10 positions. Subsequent read counting on TEs and PCGs was performed with featureCounts version 2.0.3 (Liao et al, 2014) using the Araport11 genome annotation. Normalization and differential analyses were done using DESeq2 version 1.26.0 (Love et al, 2014) with default parameters. Only loci with absolute fold-change ≥ 1 and Benjamini-Hochberg adjusted P values < 0.05 were considered differentially expressed. We also re-analyzed publicly available data for mail1 (ERR1593753, ERR1593754, ERR1593761, ERR1593762 (Ikeda et al, 2017)), pol2a-12 (ERR4194076 -ERR4194079 (Bourguet et al, 2020)) and clf swn (SRR1931614, SRR1931620, SRR2500947, SRR2962411 (Wang et al, 2016)).
Gene Ontology analysis
GO enrichment analysis was performed using the PANTHER Overrepresentation Test (version 17.0; http://pantherdb.org/ (Mi et al, 2019; Ashburner et al, 2000; The Gene Ontology Consortium et al, 2021)). Statistical significance was calculated by Fisher’s exact test and adjusted using the Bonferroni correction for multiple testing (multiplication of the single-test P value by the number of independent tests to obtain an expected error rate) and P values < 0.05 were deemed statistically significant. Fold enrichment corresponds to the ratio of upregulated PCGs numbers annotated in the ontology term over all genes annotated in this term and was considered overrepresented if it was greater than 1.
Bisulfite sequencing (BS-seq)
Genomic DNA was extracted from 16-day-old seedlings with the Wizard Genomic DNA Purification Kit (Promega), following the manufacturer’s instructions. Sodium bisulfite conversion, library preparation, and sequencing on a HiseqXten were performed at the Beijing Genomics Institute (Hong Kong), producing ∼ 26–35M 150-bp paired-end reads. Data analysis was done using the Methylpy analysis pipeline v1.4.6 (Schultz et al, 2015). Reads were filtered to remove PCR duplicates using Picard v2.23.3 and mapped to TAIR10 using Bowtie2 v2.4.1 (Langmead & Salzberg, 2012). Only cytosines with a minimum coverage of 6 reads were retained. To calculate average methylation levels at specific regions, we used the add-methylation-level function from Methylpy (Schultz et al, 2015). TEs located in euchromatic arms and TEs located in pericentromeric regions were previously defined (Bourguet et al, 2021). We also re-analyzed publicly available data for mail1 (ERR1593765 -ERR1593768 (Ikeda et al, 2017)) and pol2a-12 (ERR4194158, ERR4194163 (Bourguet et al, 2020)).
Nuclei isolation and microscopy
Rosette leaves from 3-week-old plants were fixed for one hour in 4% formaldehyde in 1X PBS and subsequently chopped on a glass slide. Tissues were covered with a coverslip, manually squashed and frozen in liquid nitrogen before removing the coverslip. Slides were washed twice in 1X PBS and DNA was stained with DAPI in Vectashield mounting medium (Vector Laboratories). Nuclei were visualized using a Zeiss Axio Imager Z.1 epifluorescence microscope (Carl Zeiss AG) with a PL Apochromat 100X/1.40 oil objective and images were captured with a Zeiss AxioCam MRm camera using the Zeiss ZEN software. The relative heterochromatin fraction was computed for each nucleus by calculating the ratio of the sum of chromocenter volume (area x intensity) over that of the entire nucleus using the ImageJ software.
Chromatin immunoprecipitation followed by sequencing (ChIP-seq)
ChIP-seq was performed starting from 2-week-old seedlings grown in soil. Briefly, 1 g. of aerial tissues were fixed in 1% formaldehyde-containing MC buffer (Yelagandula et al, 2017). Fixed tissues were nitrogen frozen and ground to fine powder by vortexing in the presence of seven ceramic beads without thawing the extracts. Nuclei were isolated as previously described (Villar & Köhler, 2010) and resuspended in 1ml of 20mM Tris pH8, 2mM EDTA, 0.1% SDS 0.1%, 1X complete protease inhibitor cocktail (Roche). Chromatin was then sonicated in a Covaris S220 with the following settings: treatment time 5 min, acoustic duty factor 20%, PIP 170W, Cycles per burst 200 and max temperature 8°C. Following centrifugation at 13000rpm 4°C for 10 min., about half of the chromatin extract was used for immunoprecipitation with 5µg of H3K27me3 antibody (C15410069 Diagenode). IP, washes, and reverse crosslink were carried out as described in (Yelagandula et al, 2017). Reverse crosslinked DNA was purified using Zymo ChIP DNA clean and concentrator columns following the manufacturer’s instructions. Library preparation from the eluted DNA and PE100 sequencing on a DNBseq sequencing platform was performed at the Beijing Genomics institute (Hong Kong). Two biological replicates were performed for each genotype.
Reads were trimmed and filtered for quality and adapter contamination using Trim Galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), and aligned to the TAIR10 genome plus the L5-GUS transgene using bowtie2. Multi-mapping reads were discarded, and probable PCR duplicates were removed using the Picard Tools suite. Sample tracks and metaplots over genomic regions were obtained using deeptools bamCoverage (-normalizeUsing RPGC).
ATAC-seq
One hundred mg of 14-day-old seedlings were ground and nuclei were isolated with 4°C Buffer (0,25M Sucrose, 10mM Tris-HCl, 10mM MgCL2, 1% Triton, 5mM ß-mercaptoethanol) containing proteinase inhibitor cocktail (Roche) and filtered in 63 µm. Nuclei were resuspended in 1X TD Buffer (Illumina FC-121-1030) and 2.5 µl of Tn5 Transposase (Illumina FC-121-1030) were added. Tagmentation reaction was performed at 37°C for 30min, and DNA was purified using a Qiagen MinElute Kit. DNA libraries were amplified for a total of 8 cycles. The libraries were then subjected to 1L×L75Lbp high-throughput sequencing using a NextSeq 500 instrument (Illumina). Reads were mapped to the Arabidopsis TAIR10 reference genome using STAR version 2.7.2b (Dobin et al, 2013). Peaks were called with MACS2 v2.2 (Zhang et al, 2008). Data for generating metaplots was obtained with DeepTools v3.3.2 (Ramírez et al, 2016).
Data availability
High throughput sequencing data generated in this study has been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE197614. All other data supporting the findings of this study are available within the manuscript and its Supplementary Files or are available from the corresponding author upon request.
Supporting information
Supplementary Figure 1: Mutant alleles of known silencing factors isolated in this screen. (a) Representation of the genes with the mutation and the equivalent amino acid change in red. The numbers in brackets correspond to the position of the mutated nucleotide relative to the transcription start site. (b) Photos of the histochemical staining for GUS activity in leaves of the isolated mutants and WT plant (L5) as control.
Supplementary Figure 2: A mutation in RTEL1 is responsible for silencing defects in p14 12-1. (a) Mapping of p14 12-1 mutation by mapping-by-sequencing Distribution of Landsberg (Ler) Single Nucleotide Polymorphism (SNP) frequency along the five Arabidopsis chromosomes in the F2 Ler x p14 12-1 progeny. The dashed red box marks the region depleted in Ler SNP frequency that carries the causative mutation. (b) Photos of 3-week-old plants (top) and histochemical staining for GUS activity (bottom) of the p14 12-1 mutant and complementation test between rtel1-1 and p14 12-1. (c) Representative images of histochemical staining for GUS activity in leaves from plants of the indicated genotypes. Three independent T1 lines were selected for rtel1-3::RTEL1 complemented genotype. (d) RT-qPCR analysis of rtel1-3 upregulated transcripts from inflorescences of indicated genotyped. Transcript levels are normalized to ACT2 and further normalized to rtel1-3 mutant. Sample means are shown with error bars representing standard error of the mean across three technical replicates.
Supplementary Figure 3: DNA methylation profiles in rtel1-3 mutant. (a) Genome browser tracks showing DNA methylation enrichment (in all three cytosine contexts) and mRNA profiles at QQS locus in WT and rtel1-3. Local hypermethylation is represented in subtracting WT control WGBS-seq to rtel1-3. Two replicates are shown for mRNA-seq data. (b) CG, CHG and CHH DNA methylation levels over L5-GUS transgene measured by WGBS-seq in WT and rtel1-3 plants. Error bars represent standard error of the mean across two biological replicates. (c) Metaplots showing average methylation rates in CG, CHG and CHH contexts found in rtel1-3 upregulated PCGs (left) or in all PCGs of the Arabidopsis genome (right). PCGs were scaled to 2 kb and sequences 1 kb upstream their transcription start site (TSS) or downstream their transcription end site (TES) were included. Average methylation levels per 100 bp windows were plotted. (d) CG, CHG, CHH (left) and CHG subcontexts (CAG, CCG and CTG, right) DNA methylation levels over all TEs in WT and rtel1-3. TEs were aligned as in Figure 3b. (e) Transcript accumulation in reads per kilobase per million mapped reads (RPKM) at the indicated genes for WT and rtel1-3. Error bars represent standard error of the mean across two biological replicates.
Supplementary Figure 4: Analysis of rDNA copy number in rtel1 mutants. The 45S rDNA copy number was determined by quantitative PCR in rtel1-1 and rtel1-3 mutants, in relation to WT (L5 background) plants (n=3). Sample means are shown with error bars representing standard error of the mean across two biological replicates.
Supplementary Figure 5: RTEL1 function is mostly independent of POL2A and MAIL1 (a) Metaplots showing average DNA methylation levels in all three cytosine contexts at TEs located on chromosome arms (left) or in pericentromeric heterochromatin (right) in pol2a-12 mutant. TEs were aligned as in Figure 3b. (b) Venn diagram showing the overlap between rtel1-3, pol2a-12 upregulated PCGs and all H3K27me3-marked genes. (c) H3K27me3 ChIP signal over rtel1-3, pol2a-12 upregulated PCGs and all PCGs of the Arabidopsis genome. PCGs were scaled and plotted as in Figure 2d. (d) Metaplots showing TE methylation in mail1-1 mutant represented as in a. (e) Venn diagram showing the overlap between rtel1-3, mail1-1 upregulated PCGs and all H3K27me3-marked genes. Methylome and transcriptome of pol2a-12 and mail1-1 mutants were re-analyzed from published data (Bourguet et al, 2020; Ikeda et al, 2017).
Supplementary Figure 6: MED14 transcript levels in rtel1 mutants. Genome browser tracks showing mRNA profiles at MED14 (AT3G04740) locus in WT, rtel1-1 and rtel1-3. Two replicates are shown.
Supplementary Figure 7: Chromatin accessibility at mom1-2 upregulated TEs. ATAC-seq depth over mom1-2 up-regulated TEs in WT, rtel1-1 and rtel1-3 single mutants. TEs were scaled to 6 kb and sequences 3 kb upstream or downstream their 5’ and 3’ ends respectively were included. Average RPGC signal over 200 bp bins is plotted.
Supplementary Figure 8: Gene expression changes in MMC-treated WT plants and untreated rtel1-3 plants. Scatter plot showing correlation of transcriptional changes of all PCGs in rtel-3 [log2(rtel1-3-NT/L5-NT)] and in MMC-treated WT plants [log2(WT-MMC/WT-NT)]. The colored bin represents the count of PCGs from low (purple) to high (yellow). Linear regression line (red) and R2 (p < 2.2×10-16) are indicated.
Supplementary Figure 9: Sterility and silencing release observed in rtel1-3 fancjb-1 double mutants. (a) Inflorescence (top) and individual silique (bottom) pictures of the indicated genotype plants. WT (L5 background) plants and single mutants show normal siliques while transformed flowers in rtel1-3 fancjb-1 double mutants did not self-fertilize and silique elongation did not occur indicating no developing seeds. (b) RT-PCR analysis of transcripts from endogenous repeats, TSI (Transcriptionally Silent Information) and 106B. Each line represents biological replicates. Amplification of UBQ1 (UBIQUITIN EXTENSION PROTEIN 1) was used as a loading control. For each target, PCR in the absence of reverse transcription (no RT) was performed to ensure any DNA contamination.
Acknowledgments
We thank Hervé Vaucheret for the kind gift of the L5 line. Work in the Benhamed laboratory was supported by the Institut Universitaire de France (IUF). Y.H. was supported by China Scholar Council fellowships (201806690005). This work was supported by CNRS, Inserm, Université Clermont Auvergne, a Young Researcher grant from the Auvergne Regional Council (to O.M.), and a grant from the European Union under the Fonds Européen de Développement Régional (FEDER, project CELCIUS, to O.M.). A.H. was supported by a PhD studentship from the Ministère de l’éducation nationale, de l’enseignement supérieur et de la recherche. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵