The Arabidopsis active demethylase ROS1 cis-regulates immune-responsive genes by pruning DNA methylation at promoter-regulatory regions

Active DNA demethylation has emerged as an important regulatory process of plant and mammalian immunity. However, very little is known about the mechanisms by which active demethylation controls transcriptional immune reprogramming and disease resistance. Here, we first show that the Arabidopsis active demethylase ROS1 promotes basal resistance towards Pseudomonas syringae by antagonizing RNA-directed DNA methylation (RdDM). Furthermore, we found that ROS1 facilitates the flagellin-triggered induction of the disease resistance gene RMG1 by limiting RdDM at the 3’ boundary of a remnant RC/Helitron transposable element (TE) embedded in its promoter. We further identify flagellin-responsive ROS1 putative primary targets, and show that at a subset of promoters, ROS1 erases methylation at discrete regions exhibiting WRKY transcription factors (TFs) binding. In particular, we demonstrate that ROS1 removes methylation at the orphan immune receptor RLP43 promoter, to ensure DNA binding of WRKY TFs. Finally, we show that ROS1-directed demethylation of the RMG1 and RLP43 promoters is causal for both flagellin responsiveness of these genes and for basal resistance. Overall, these findings significantly advance our understanding of how active demethylases shape transcriptional immune reprogramming to enable antibacterial resistance. HIGHLIGHTS The TNL RMG1 that is regulated by ROS1 positively regulates basal resistance towards Pto DC3000 ROS1 regulates the flg22-triggered differential expression of more than 2000 thousands genes, among which 10% are demethylated by ROS1 ROS1 facilitates the transcriptional activation of a subset of flg22-induced genes by antagonizing RdDM at discrete promoter regions WRKY transcription factors (TFs) bind to the demethylated promoter regions of a subset of flg22-induced ROS1 targets The hypermethylation at the RLP43 promoter, caused by the lack of ROS1, repels DNA binding of two PAMP-responsive WRKY TFs Specific hypermethylation at the ROS1-targeted promoter regions of RMG1 and RLP43 is causal for their silencing as well as for disease susceptibility against Pto DC3000


46
Plants are permanently exposed to microbes including pathogens and rely on a potent immune response to control infections. The first layer of the plant immune system involves the recognition of Pathogen-might ensure the DNA/chromatin binding of TFs during pathogen infection and/or elicitation, although 126 this hypothesis has never been tested experimentally.

127
Here, we characterize Arabidopsis demethylase ROS1 in the context of antibacterial immunity. We first 128 demonstrate that ROS1 positively regulates basal resistance against Pto DC3000 by antagonizing RdDM 129 activity. Consistent with this observation, we found that the TNL RESISTANCE METHYLATED GENE 130 1 (RMG1) contributes to basal resistance against Pto DC3000, and that ROS1-directed suppression of Figure 1. The enhanced Pto DC3000 disease susceptibility observed in ros1-infected mutants is mainly dependent on DCL2-and/or DCL3 functions. A. Increased Pto DC3000 vascular propagation in the two independent ros1 mutant alleles, ros1-3 and ros1-4. Secondary veins of five-week-old Col-0, ros1-3 and ros1-4 mutants were inoculated with a virulent GFP-tagged Pto DC3000 strain (Pto DC3000-GFP) at 10 7 cfu ml -1 using the toothpick inoculation method. Inoculation was done on 6 secondary veins per leaf and 2 sites of inoculation per vein. The number of Pto DC3000-GFP spreading events from the wound inoculation sites was quantified after 3 days under UV light using a macrozoom. When the bacteria propagated away from any of the 12 inoculation sites, it was indexed as propagation with a possibility of maximum 18 propagations per leaf (see Material & Methods). The values from three independent experiments were considered for the comparative analysis. Statistical significance was assessed using a one-way ANOVA test and Tukey's multiple comparisons test. B. Representative pictures of the GFP fluorescence observed at the whole leaf level on the plants presented in A. C. Enhanced Pto DC3000 apoplastic growth in the two independent ros1 mutant alleles, ros1-3 and ros1-4. Fiveweek-old plants of Col-0, ros1-3 and ros1-4 mutants were dip-inoculated with Pto DC3000-GFP at 5 x 10 7 cfu ml -1 . Bacterial titers were monitored at 3 days post-inoculation (dpi). Three leaves out of four plants per line and from three independent experiments were considered for the comparative analysis. Statistical significance was assessed using a one-way ANOVA test and Tukey's multiple comparisons test. D. Increased bacterial propagation in the vein observed in ros1-3 is rescued in the ros1dcl23 triple mutant. Secondary veins of five-week-old Col-0, ros1-3 and the triple ros1dcl23 mutants were inoculated as in A. and the results analysed as in A. E. Representative pictures of the GFP fluorescence observed at the whole leaf level on the plants presented in D. F. Enhanced Pto DC3000 apoplastic growth in ros1 is partially rescued in the ros1dcl23 background. Five-weekold plants of Col-0, ros1-3 and ros1dcl23 were inoculated as in C. and the results were analysed as in C.
24 nt siRNAs is abolished. Interestingly, we found that both the enhanced vascular propagation and 153 apoplastic growth of Pto DC3000-GFP detected in the ros1-3 mutant were reduced in ros1dcl23 154 mutants, and almost comparable to the phenotypes of Col-0-infected plants ( Figure 1D-F). These data 155 indicate that DCL2 and/or DCL3 are mainly responsible for the enhanced susceptible phenotypes 156 observed in ros1-3-infected mutants. They also suggest that some defence genes are likely 157 hypermethylated and silenced in the ros1-3 mutants through the action of, at least in part, DCL2-and/or 158 DCL3-dependent siRNAs.

160
The ROS1 target RMG1 is a functional disease resistance gene that contributes to basal resistance 161 towards Pto DC3000 162 ROS1 has previously been shown to ensure a proper flg22-triggered induction of RMG1, an orphan TNL 163 that is demethylated by ROS1 in its promoter (Yu et al., 2013). As a result, a strong reduction in flg22-  The TNL gene RMG1 contributes to apoplastic and vascular resistance against Pto DC3000 and its flg22-triggered induction is negatively regulated by RdDM in the absence of ROS1. A. RMG1 mRNA levels in Col-0 and the two rmg1 mutant alleles, namely rmg1-1 and rmg1-2, were monitored by RT-qPCR at 6h after syringe-infiltration of mock (water) or 1 µM of flg22 peptide. B. RMG1 positively regulates vascular resistance towards Pto DC3000. Secondary veins of five-week-old Col-0, ros1-3, rmg1-1 and rmg1-2 plants were inoculated with Pto DC3000-GFP at 10 7 cfu ml -1 using the toothpick inoculation method. Inoculation was done on 6 secondary veins per leaf and 2 sites of inoculation per vein. The number of Pto DC3000-GFP spreading events from the wound inoculation sites was quantified after 3 days under UV light using a macrozoom. When the Pto DC3000-GFP propagated away from any of the 12 inoculation sites, it was indexed as propagation with a possibility of maximum 18 propagations per leaf. The values from three independent experiments were considered for the comparative analysis. Statistical significance was assessed using a one-way ANOVA test. C. RMG1 positively regulates apoplastic resistance towards Pto DC3000. Five-week-old Col-0, ros1-3, rmg1-1 and rmg1-2 plants were dip-inoculated with Pto DC3000-GFP at 5 x 10 7 cfu ml -1 . Bacterial titers were monitored at 3 days post-infection (dpi). Three leaves out of four plants per line and from three independent experiments were considered for the comparative analysis. Statistical significance was assessed using a one-way ANOVA test. D. Flg22-triggered induction of RMG1 is compromised in ros1-3-elicited mutant and correlates with an increased DNA methylation and siRNA levels at the remnant RC/Helitron TE AtREP11, and particularly at its 3' boundary. IGV snapshots showing mRNA levels (mRNA-seq) after syringe infiltration of mock (water) or 1 µM of flg22 peptide for Col-0 and ros1-3, and cytosine DNA methylation levels (Bs-Seq) and siRNA levels (sRNA-seq) in 5week-old untreated rosette leaves of Col-0 and ros1-3, at the RMG1 locus. The differentially methylated region (DMR) is highlighted by the dotted box. E. Levels of different siRNA species at the 3' boundary of AtREP11. IGV snapshots representing the levels of methylation (BS-seq), total siRNA species (siRNA-seq) and different size of siRNA species (21nt, 22nt, 23nt and 24nt siRNAs) in 5-week-old rosette leaves of Col-0, ros1-3 and ros1dcl23. F. The flg22-triggered induction of RMG1 is fully restored in ros1dcl23-elicited triple mutants. RT-qPCR analysis depicting RMG1 and FRK1 mRNA levels in Col-0, ros1-3, dcl23 and ros1dcl23 5-week-old rosette leaves treated with either mock (water) or 1 µM of flg22 for 6h. The mRNA levels are relative to the level of UBQ transcripts. Statistical significance of flg22 treatment on expression was assessed using a two-way ANOVA test and a Sidak's multiple comparisons test. maintain a low basal expression of this gene in untreated conditions (Yu et al., 2013). By contrast, sequencing approach (Yu et al., 2013). These results were confirmed by using a whole-genome bisulfite 182 sequencing (BS-seq) approach in rosette leaves of untreated Col-0 and ros1-3 mutants ( Figure 2D). We 183 found that the hyper Differentially Methylated Region (hyperDMR) detected at the AtREP11 repeat, 184 was particularly pronounced at the 3' boundary of this remnant RC/Helitron TE in the ros1-3 mutant, 185 while this region was unmethylated in Col-0 plants ( Figure 2D). This result is consistent with a role of  Figure 2DE). All these siRNA species were no longer produced in ros1dcl23 mutants, a molecular effect 193 which was associated with an absence of methylation at this ROS1-targeted region ( Figure 2E).

194
Collectively, these data indicate that DCL2 and/or DCL3-dependent siRNAs direct RdDM 195 predominantly at the 3' boundary of AtREP11 in the ros1-3 mutant. They also suggest that, in Col-0 196 plants, ROS1 counteracts the biogenesis of DCL2-and DCL3-dependent siRNAs at this RMG1 197 promoter region to restrict methylation spreading at the 3' boundary of AtREP11. Col-0, ros1-3, dcl2-1 dcl3-1 (dcl23) and ros1dcl23 plants with either mock or flg22 for 6 hours and further monitored the mRNA accumulation of RMG1. We found a compromised induction of RMG1 in 205 the ros1-3-elicited mutant, which was not observed for Flg22-induced Receptor-like Kinase 1 (FRK1), 206 which is not hypermethylated in ros1-3 mutants and thus served as a control ( Figure 2F; Figure S2; 207 Figure S3). By contrast, a full restoration of flg22-triggered inducibility of RMG1 was observed in the 208 ros1dcl23-elicited mutant ( Figure 2F; Figure S3). These data indicate that RdDM of AtREP11 and its 3' 209 boundary, which is specifically detected in ros1-3 mutants, negatively regulates the flg22 inducibility 210 of RMG1. They also suggest that ROS1-directed demethylation of this RMG1 promoter region is crucial   Figure 3B; Figure S4). We found that the 227 overall increase of methylation levels at the 102 less-induced genes and at the whole 2943 ROS1 targets, 228 in ros1-3 compared to Col-0 plants, was almost exclusively observed in 2 Kb upstream sequence regions 229 ( Figure 3C). An increase in siRNA accumulation was also detected at both the promoters of the flg22- A. Proportion of flg22-responsive genes that are regulated by ROS1. One hundred and two flg22-responsive genes that are "less-induced" and 115 flg22-responsive genes that are "less-repressed" in ros1-3-elicited mutant exhibit hypermethylated DMRs (hyperDMRs). Venn diagram representing the overlap of genes presenting hyper-DMRs in the ros1 mutant (in orange) with genes presenting a compromised induction (in red) or repression (in green) in ros1-3 compared to Col-0 treated with mock (water) or 1 µM of flg22 for 6h. B. Heat map representing the relative expression of the 102 less-induced genes in Col-0 and ros1-3 treated with mock (water) or 1 µM of flg22 for 6h. Merged data from two independent biological replicates are presented. C. Increased global DNA methylation levels observed in untreated ros1-3 compared to Col-0 at the whole set of genes exhibiting hyperDMRs (left panel) and at the 102 less-induced genes (right panel) are restored in the ros1dcl23 triple mutant. Heatmap representing global DNA methylation levels within regions comprising 2Kb upstream of the transcription start site (TSS), the gene body, and 2 Kb downstream of the transcription end site (TES) of the whole set of genes exhibiting hyperDMRs in ros1-3 versus Col-0 (left panel) and of the 102 lessinduced genes (right panel). These heatmaps were generated from BS-seq datasets obtained from 5-week-old rosette leaves of Col-0, ros1-3 and ros1dcl23 mutants. D. Increased 24 nt siRNA levels in ros1-3 at the whole set of genes exhibiting hyperDMRs in ros1-3 versus Col-0 (left panel) and at the 102 less-induced genes (right panel) are restored in the ros1dcl23 triple mutant. Heatmap representing 24-nt siRNA levels within regions comprising 2Kb upstream of the transcription start site (TSS), the gene body, and 2 Kb downstream of the transcription end site (TES) of the whole set of genes exhibiting hyperDMRs in ros1-3 versus Col-0 (left panel) and of the 102 less-induced genes (right panel). These heatmaps were generated from sRNA-seq datasets obtained from 5-week-old rosette leaves of untreated Col-0, ros1-3 and ros1dcl23 mutants. Average of the two replicates are represented. E. Methylation levels at RBA1 are partially restored in ros1dcl23 whereas RLP43 retains methylation levels similar to methylation levels observed in the single ros1-3 mutant. IGV snapshots showing siRNA levels and methylation levels at the DMRs of RBA1 and RLP43 in Col-0, ros1-3 and ros1dcl23. F. IGV snapshots representing mRNA-seq data in Col-0 and ros1-3 mutant in mock-and flg22-treated conditions as well as BS-seq and sRNA-seq of untreated Col-0 and ros1-3 plants. G. RBA1 gene induction is partially restored whereas RLP43 remains in a repressed state in ros1dcl23. RT-qPCR analyses from 5-week-old rosette leaves of Col-0, ros1-3, dcl23 and ros1dcl23 treated with either mock (water) or 1 µM of flg22 for 6h. The mRNA levels are relative to the level of UBQ transcripts. Statistical significance of flg22 treatment on expression was assessed using a two-way ANOVA test and a Sidak's multiple comparisons test.
induced and overall ROS1 targets in ros1-3 versus Col-0 plants ( Figure 3D; Figure S5). By contrast, the 231 overall enhanced DNA methylation and siRNA levels observed in ros1-3 mutants at these loci was 232 almost restored to Col-0 levels in ros1dcl23 mutants (Figure 3C/D; Figure S5). The latter data suggest 233 that ROS1 directs promoter demethylation mostly by antagonizing DCL2 and/or DCL3 functions.  Figure 3E/G). The flg22-triggered induction of RBA1 was restored in ros1dcl23 to levels 249 similar to dcl23 mutants, but not to the same extent as in elicited Col-0 plants ( Figure 3G). Furthermore, 250 this effect was associated with a partial decrease in RBA1 promoter methylation at the RBA1 hyperDMR, 251 in ros1dcl23 compared to ros1-3 mutants ( Figure 3E). By contrast, RLP43 remained fully silenced in 252 the ros1dcl23-treated mutants, as in elicited ros1-3 mutants, which is consistent with comparable levels 253 of RLP43 promoter methylation in both untreated mutants ( Figure 3E). It is noteworthy that the 254 unaltered RLP43 hyperDMR detected in the ros1dcl23 mutants, was associated with a moderate 255 remaining accumulation of siRNAs, which was shifted from 23-24 nt siRNAs in ros1-3 to 21 nt siRNAs 256 in ros1dcl23 mutants ( Figure 3E; Figure S6). It was also accompanied by the presence of longer RNAs (25 to 30 nt), which exhibit a preference for having an A at their 5'ends and that overlap with siRNAs 258 produced from the hyperDMR, as observed for typical P4RNA species ( Figure

. Several WRKY transcription factors bind to the demethylated region of the RLP43 promoter, which contains a functional and flg22-responsive W-box cis-element.
A. A subgroup of the ROS1 targets exhibit an over-representation of WRKY DNA binding at the promoter regions corresponding the hyperDMRs that were retrieved in ros1-3. GAT analysis performed on publicly available DAPsequencing data (O'Malley et al., 2016) identified WRKY transcription factors with significant enrichment in the regions corresponding to the hyperDMRs observed in ros1-3 for the 14 stringent ROS1 primary targets. B. Several WRKY transcription factors specifically bind at the region that is demethylated by ROS1 in the RLP43 promoter. Snapshots representing, from top to bottom, mRNA-seq (Rep1), BS-seq (Rep1) and DAP-seq data at the RLP43 locus. To better appreciate the overlap between the promoter region of RLP43 subjected to ROS1directed demethylation and the region where WRKY transcription factors bind to DNA, a zoom in is depicted at the level of the hyperDMR (box on the right panel).    . The artificial siRNA-directed targeting of remethylation at the RMG1 and RLP43 promoters impairs the flg22-triggered inducibility of these genes and enhances disease susceptibility towards Pto DC3000. A. Scheme depicting the chimeric inverted repeat (IR) construct designed to simultaneously direct DNA remethylation at the RMG1 and RLP43 promoter regions. The IR-RMG1p/RLP43p contains sequences corresponding to the DNA sequence regions of RMG1 (blue) and RLP43 (orange) promoters that are demethylated by ROS1 in Col-0 and hypermethylated in ros1 mutants. This inverted repeat transgene is driven by the constitutive 35S promoter, hence hypothesized to constitutively produce two populations of siRNA species designed to force remethylation of the RMG1 and RLP43 promoter regions that are normally demethylated by ROS1 in Col-0. B. The RMG1 and RLP43 promoters exhibit hypermethylated in IR-RMG1p/RLP43p lines such as in ros1-3 mutants. Genomic DNAs from Col-0, ros1-3 and two independent IR-RMG1p/RLP43p lines (two biological replicates per line) were digested using McrBC and further analysed by qPCR. Ratio between digested DNA and undigested DNA was quantified to assess the proportion of methylation. C. The flg22-triggered induction of RMG1 and RLP43 is impaired in the two independent IR-RMG1p/RLP43p lines. Five-week old rosette leaves of Col-0, ros1-3 and two independent IR-RMG1p/RLP43p lines were syringeinfiltrated with either mock (water) or 1 µM of flg22 for 6h, and the mRNA levels of RMG1 and RLP43 were monitored by RT-qPCR analyses. The mRNA levels are relative to the level of UBQ transcripts. Statistical significance of flg22 treatment on expression was assessed using a two-way ANOVA test and a Sidak's multiple comparisons test. D. The flg22-triggered induction of RBA1 and FRK1 is not affected in the two independent IR-RMG1p/RLP43p lines. Five-week old rosette leaves of Col-0, ros1-3 and two independent IR-RMG1p/RLP43p lines were syringeinfiltrated with either mock (water) or 1 µM of flg22 for 6h, and the mRNA levels of RBA1 and FRK1 were monitored by RT-qPCR analyses. The mRNA levels are relative to the level of UBQ transcripts. Statistical significance of flg22 treatment on expression was assessed using a two-way ANOVA test and a Sidak's multiple comparisons test. E. IR-RMG1p/RLP43p lines exhibit increased Pto DC3000 titer. Five-week-old plants of Col-0, ros1-3 and two independent IR-RMG1p/RLP43p lines were dip-inoculated with Pto DC3000-GFP at 5 x 10 7 cfu ml -1 . Bacterial titers were monitored at 3 days post-infection (dpi). Three leaves out of four plants per line and from three independent experiments were considered for the comparative analysis. Statistical significance was assessed using a one-way ANOVA test and Tukey's multiple comparisons test.
revealed levels of siRNA accumulation and of methylation at these promoter regions that were 341 comparable to the ones achieved in ros1-3 mutants ( Figures 6B; Figure S12), thereby validating our 342 experimental strategy. We next challenged two independent IR-RMG1p/RLP43p transgenic lines with 343 flg22 and monitored the levels of RMG1 and RLP43 mRNAs at 6 hours post-treatment. Significantly, 344 we found that the induction of these genes was strongly altered in the two IR-RMG1p/RLP43p 345 independent transgenic lines, almost to the same extent as in the ros1-elicited mutant background 346 ( Figure 6C; Figure S12C). By contrast, the flg22-triggered induction of RBA1 and FRK1 was 347 comparable in Col-0 and IR-RMG1p/RLP43p transgenic lines ( Figure 6D; Figure

387
We also performed a detailed analysis of the flg22-induced ROS1 putative primary targets. Most of 388 these genes exhibit hypermethylation at their promoters, an effect which was almost abolished in 389 ros1dcl23 mutants ( Figure 3CD). These data suggest that ROS1 antagonizes RdDM at these promoters, 390 presumably to facilitate their transcriptional activation during plant defence signalling. Consistent with 391 this hypothesis, we found that the impaired induction of RMG1, RBA1 and of the phosphoglycerate 392 mutase genes observed in ros1-elicited plants was either fully or partially restored in ros1dcl23-elicited mutants (Figures 2, 3). This was however not the case of RLP43, which remained fully silenced in 394 ros1dcl23-elicited mutants, a phenotype that was associated with a moderate gain of 21 nt siRNAs at 395 the RLP43 hyperDMR in ros1dcl23 mutants (Figure 3; Figure S6). This observation suggests that DCL4 396 and/or DCL1 might, at least in part, compensate for the lack of DCL2 and DCL3 in the ros1 mutants to Interestingly, our work demonstrates that the dominance of ROS1 over RdDM at two immune-408 responsive gene promoters can be alleviated by artificially producing siRNAs against these genomic 409 regions ( Figure 6). In particular, we showed that the stable expression of an inverted repeat transgene 410 bearing sequence homologies to the ROS1 targeted regions of RMG1 and RLP43 promoters, resulted in 411 their hypermethylation along with a compromised flg22-triggered induction of these genes ( Figure 6). 412 Furthermore, we found that these transgenic lines exhibited an enhanced susceptibility towards Pto 413 DC3000, indicating that the silencing of these genes is sufficient to dampen basal resistance ( Figure 6).

414
We are anticipating that this approach will be extensively used in the future to assess the regulatory     For this analysis, we focused on protein-coding genes having a hyper-DMR "within 2Kb" (i.e. the region 684 covering the gene-body plus 2Kb upstream and downstream). Using this method, we selected 2944 685 protein coding genes. This list was then crossed with the list of "less-induced genes", which allowed us 686 to retrieve 102 candidates for further analyses.

706
The size of the circle represents the log2-fold-change of GAT analysis (a higher log2-fold-change, 707 corresponding to a bigger circle, means the TF binds more often on this particular promoter than in 1000 708 simulations of randomly assigned genomic intervals of the same size). ahead of publication; Robert Fischer for providing the ros1-3 and ros1-4 alleles; Florence Jay for its 714 contribution to the selection of the ros1-3 dcl2-1 dcl3-1 mutants, and all the members of the Navarro 715 lab for their inputs, discussions and critical reading of the manuscript. This work was supported by a