A virus-encoded protein suppresses methylation of the viral genome in the Cajal body through its interaction with AGO4

In plants, establishment ofde novoDNA methylation is regulated by the RNA-directed DNA methylation (RdDM) pathway. RdDM machinery is known to concentrate in the Cajal body, but the biological significance of this localization has remained elusive. Here, we show that the anti-viral methylation of theTomato yellow leaf curl virus(TYLCV) genome occurs specifically in the Cajal body ofNicotiana benthamianacells. Methylation of the viral genome is countered by a virus-encoded protein, V2, which interacts with the central RdDM component AGO4, interfering with its binding to the viral DNA; Cajal body localization of the V2-AGO4 interaction is required for the viral protein to exert this function. Taken together, our results draw a long sought-after functional connection between RdDM, the Cajal body, and anti-viral DNA methylation, paving the way for a deeper understanding of DNA methylation and anti-viral defences in plants.


INTRODUCTION 20
DNA methylation in cytosine residues is a conserved epigenetic mark essential for protecting the 21 eukaryotic genome against invading nucleic acids, namely viruses and transposable elements. In 22 plants, establishment of de novo DNA methylation is believed to be regulated by the RNA-directed 23 DNA methylation (RdDM) pathway. The canonical RdDM pathway requires the activity of two 24 plant-specific RNA polymerase II-related enzymes, Pol IV and Pol V, and leads to cytosine 25 methylation in a sequence-specific manner. In brief, the current understanding of RdDM is as 26 follows: Pol IV generates RNA transcripts subsequently converted to double-stranded RNA 27 (dsRNA) by RDR2 (Haag et al., 2012, Law et al., 2011, and then diced into 24-nt siRNAs by 28 DCL3 (Xie et al., 2004); the resulting 24-nt siRNAs are loaded into AGO4 (Zilberman et al., 2003), 29 which is guided to scaffold RNA molecules generated by Pol V via sequence complementarity 30 and recruits the de novo methyl transferase DRM2 (Böhmdorfer et al., 2014al., , Chan et al., 2005 Gao et al., 2010al., , Zhong et al., 2014, which in turn catalyses methylation of adjacent DNA 32 sequences. RdDM generally creates a chromatin environment refractive to gene expression. Part 33 of the RdDM machinery, including AGO4, was found to concentrate in the Cajal body, a 34 subnuclear compartment that is the site of maturation of ribonucleoprotein complexes (Li et al., 35 2006). This observation led to the suggestion that Cajal bodies might be a centre for the assembly 36 of an AGO4/NRPE1/siRNA complex, which would facilitate RdDM at target loci (Li et al., 2006). 37 However, the biological relevance of the Cajal body localization of RdDM components, and of this 38 compartment on DNA methylation, has so far not been demonstrated. 39 Geminiviruses are a family of plant viruses with circular single-stranded (ss) DNA genomes 40 infecting multiple crops and causing dramatic yield losses worldwide. The geminiviral genome 41 replicates in the nucleus of the infected cell by using the host DNA replication machinery, made 42 available following the viral re-programming of the cell cycle (reviewed in 43 V2 does not hamper production or loading of vsiRNA but interferes with AGO4 binding to 145 the viral genome 146 The canonical function of AGO4 in the RdDM pathway requires loading of siRNA and association 147 to Pol V, and results in the recruitment of DRM2 to the target loci and the subsequent methylation 148 of the adjacent DNA (Matzke et al., 2015, Matzke & Mosher, 2014. Through physical interaction, 149 V2 could affect AGO4 function on the viral genome in different ways, for example by impairing 150 loading of viral siRNA (vsiRNA) onto this protein or by displacing endogenous interactors, such 151 as Pol V or DRM2. In order to shed light on the molecular mechanism underlying the V2-mediated 152 interference of AGO4-dependent methylation of the viral genome, we tested binding of AGO4 to 153 the viral DNA in the presence or absence of V2 in local infections with TYLCV WT and V2 null 154 mutant, respectively, by Chromatin immunoprecipitation (ChIP). As shown in Figure 5A, 3xFLAG-155 NbAGO4 could bind both the IR and the V2-encoding region of the viral genome in the absence 156 of V2 (TYLCV-V2null), but the signal decreased to background levels when V2 was present 157 (TYLCV). Therefore, AGO4 has the capacity to bind the viral DNA molecule, but this binding is 158 impaired by the virus-encoded V2 protein. AGO4 binding in the TYLCV V2 null mutant hence 159 correlates with the detected increase in viral DNA methylation ( Figure 4A). 160 Several viral silencing suppressors encoded by different viruses have been shown to inhibit 161 formation of AGO/sRNA complexes (e. g. Burgyán et al., 2011, Rawlings et al., 2011, Schott et al., 162 2012. To test whether this strategy is also employed by V2, we immunoprecipitated 3xFLAG-163 NbAGO4 co-expressed with WT or V2 null mutant TYLCV in local infection assays in N. 164 benthamiana, and visualized AGO4-bound vsiRNA by sRNA northern blotting. While infected 165 samples contained both 21-and 24-nt vsiRNA, and the occurrence and accumulation of these 166 sRNA species was not affected by the presence of virus-encoded V2, mostly 24-nt vsiRNA co-167 immunoprecipitated with AGO4 ( Figure 5B). Interestingly, a higher amount of vsiRNA associated 168 to AGO4 in the samples infected with the WT virus ( Figure 5B). Taken together, these results 169 demonstrate that V2 does not affect the production or accumulation of vsiRNA, nor does it hamper 170 loading of these vsiRNA molecules into AGO4, but interferes with binding of this protein to the 171 viral genome in order to suppress DNA methylation and promote virulence. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint In Arabidopsis, AGO4 has been shown to co-localize with its interactor NRPE1 (NRPD1b), a 176 subunit of Pol V, in the Cajal body, which was then suggested to be a center for the assembly of 177 AGO4/NRPE1/siRNA complexes, enabling RdDM at target loci (Li et al., 2008, Li et al., 2006. 178 However, the functional significance of this subnuclear localization has so far remained elusive. 179 Interestingly, both V2-GFP and the different RFP-AGO4 orthologues from N. benthamiana and 180 tomato co-localize in a distinct subnuclear compartment, identified as the Cajal body by the 181 accumulation of the nucleolus and Cajal body marker fibrillarin (Barneche et al., 2000), upon 182  which is based on visualization and hence provides spatial information, unveiled that, strikingly, 189 the association between these two proteins occurs mostly or exclusively in the Cajal body, where 190 V2 homotypic interactions also occur ( Figure 6B) . 191 In order to evaluate the relevance of the Cajal body localization of the V2-AGO4 interaction, we 192 took advantage of the fact that a GFP-V2 fusion protein, as opposed to the previously mentioned The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint  The plant DNA viruses geminiviruses and pararetroviruses are both targets and suppressors of 215 DNA methylation; this possibly extends to the third family of plant DNA viruses, nanoviruses, 216 although experimental evidence is lacking (reviewed in Poogin, 2013;Pumplin and Voinnet, 2013). 217 The independent evolution of viral suppressors of DNA methylation argues for an anti-viral effect 218 of this epigenetic modification. Indeed, seminal experiments by Brough et al. (1992) andErmak 219 et al. (1993) demonstrated that methylation of the geminivirus genome interferes with its 220 replication in transformed protoplasts, likely due to a dual effect on viral gene expression and 221 function of the replication complex. More specifically, RdDM seems to play a prominent role in 222 plant defence against geminiviruses, since RdDM mutants or silenced plants display increased 223 susceptibility to geminivirus infection (Raja et al., 2008, Zhong et al., 2017, and DNA methylation 224 and repressive histone marks typical of RdDM are deposited on the viral genome (Castillo-225 Gonzalez et al., 2015, Ceniceros-Ojeda et al., 2016, Coursey et al., 2018, Dogar et al., 2006, 226 Jackel et al., 2016, Kushwaha et al., 2017, Wang et al., 2018. However, the subnuclear 227 distribution of these anti-viral methylation events is currently unknown. 228 AGO4 is a central component of the canonical RdDM pathway, and as such an obvious target for 229 viral inhibition. However, AGO4 also affects susceptibility to RNA viruses and viroids, and is 230 targeted by proteins encoded by RNA viruses, which raises the idea that either RdDM on the host 231 genome plays a role in modulating plant-virus interactions broadly, or AGO4 has an anti-viral role 232 beyond RdDM (Brosseau et al., 2016, Ma et al., 2015, Minoia et al., 2014. Supporting the latter, 233 AGO4-dependent defences against a potexvirus are independent of other RdDM components 234 and do not require nuclear localization of AGO4 (Brosseau et al., 2016).
hypothesize that the viral protein might mask a surface required for AGO4 recruitment to the viral 241 DNA through the association with an endogenous interactor (e. g. NRPE1;Li et al., 2006), or 242 interfere with the complementarity-based pairing to the nascent Pol V transcript. Our results 243 indicate that AGO4-dependent methylation of viral DNA occurs quickly in the absence of V2 244 ( Figure 4A,B). Nevertheless, AGO4 silencing still has a detectable, if minor, positive impact on 245 the accumulation of the WT virus, which correlates with decreased viral DNA methylation ( Figure  246 3E,F; Figure 4C), suggesting that the V2-mediated suppression of AGO4 function is not complete. 247 On the other hand, WT levels of viral DNA methylation are not restored in the V2 mutant upon 248 AGO4 silencing, which raises the idea that AGO4 might not be the only methylation-related target 249 of V2. In agreement with this, V2 has been shown to bind HDA6 and interfere with its promotion 250 of MET1-dependent methylation of the viral DNA (Wang et al., 2018). The finding that the physical association between TYLCV V2 and AGO4 takes places in a specific 262 nuclear body, the Cajal body, and has an impact on the methyl-state of the viral population in the 263 cell, suggests that all or most viral DNA molecules must localize in this subnuclear compartment 264 at some point of the viral cycle. Supporting this notion, the activity of V2 as a suppressor of 265 methylation of the viral genome requires the Cajal body localization of its interaction with AGO4 266 ( Figure 7A The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint The Cajal body has also been connected to systemic infection of plant RNA viruses, and proteins 274 encoded by RNA viruses can bind coilin, which impacts plant-virus interactions (Kim et al., 2007a, 275 Kim et al., 2007b, Semashko et al., 2012, Shaw et al., 2014, although these effects are likely 276 independent on DNA methylation.

Agrobacterium-mediated transient gene expression in N. benthamiana 313
All binary plasmids were transformed into Agrobacterium tumefaciens strain GV3101, with the 314 exception of pBINTRA6, which was transformed into A. tumefaciens strain C58c1. A. tumefaciens 315 clones carrying the constructs of interest were liquid-cultured in LB with appropriate antibiotics at 316 28℃ overnight. Bacterial cultures were then centrifuged at 4,000 g for 10 min and resuspended 317 in the infiltration buffer (10 mM MgCl2, 10 mM MES pH 5.6, 150 μM acetosyringone) and adjusted 318 to an OD600 = 0.5. Next, the bacterial suspensions were incubated in the buffer at room 319 temperature and in the dark for 2-4 hours and then infiltrated 3-4-week-old N. benthamiana plants. 320 For co-expression experiments, the different agrobacterium suspensions were mixed at 1:1 ratio 321 before infiltration. 322

Quantitative PCR (qPCR) and Reverse Transcription PCR (RT-qPCR) 376
To determine viral accumulation, total DNA was extracted from N. benthamiana leaves (from 377 BioRad CFX96 real-time system as described previously (Wang et al., 2017b). Total RNA was 387 extracted from the leaves of tomato plants mock-inoculated or infected with TYLCV at 3 weeks 388 post-inoculation (wpi). SlActin was used as reference gene (Exposito-Rodriguez et al., 2008). 389 Similarly, RT-qPCR was performed on RNA extracted from tomato to detect the expression of 390  al., 2008). 400

Chromatin immunoprecipitation (ChIP) assay 401
The agrobacterium clone carrying the binary vector to express 3xFLAG-NbAGO4-1 was co-402 infiltrated with those carrying the TYLCV or TYLCV-V2null infectious clones in N. benthamiana 403 leaves, and tissues were collected at 2 dpi. Chromatin immunoprecipitation (ChIP) assays were  Table 3; the primers for Small RNA (sRNA) extraction and northern blot was performed as described (Yang et al., 2015). 433 Briefly, sRNAs were purified from total extracts or AGO4 immunoprecipitates and subjected to 434 northern blot analysis. For each sample, sRNAs were separated on a 17% polyacrylamide gel, 435 which was electrotransferred to a Hybond N+ membrane (GE Lifesciences). Membranes were 436 cross-linked, incubated for 2 hours at 65°C, and hybridized overnight at 38°C with 32 P-labeled 437 probes for the intergenic region (IR) of the viral genome amplified by PCR (Fw: 438 TCCTCTTTAGAGAGAGAACAATTGGGA, Rv: ACAACGAAATCCGTGAACAG) or 439 oligonucleotides in PerfectHyb buffer (Sigma). Washed membranes were exposed to X-ray films   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint (B) Pairwise identity and genetic distance matrix among AtAGO4, NbAGO4 and SlAGO4 proteins.
(C) 3xHA-SlAGO4a, 3xHA-SlAGO4b, and 3xHA-SlAGO4d specifically interact with V2-GFP in co-immunoprecipitation (co-IP) assays upon transient expression in N. benthamiana. Free GFP was used as negative control. CBB, Coomassie brilliant blue staining. Three independent biological replicates were performed with similar results.
(D) SlAGO4a and SlAGO4b interact with V2 in split-luciferase assays. V2-N-luc and C-luc-SlAGO4a/b were transiently co-expressed in N. benthamiana; C-luc-SlWRKY75 is used as negative control. The luciferase bioluminescence from at least three independent leaves per experiment was imaged two days after infiltration. The average bioluminescence, measured in counts per second (cps), as well as an image of a representative leaf are shown. Values represent the mean of three independent biological replicates; error bars indicate SEM.
Asterisks indicate a statistically significant difference (according to a Student's t-test, ***: P<0.001.) compared to the negative control.

Figure 2
SlAGO4c . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint infection assays with TYLCV WT or V2 null mutant (TYLCV-V2null) in AGO4-silenced (TRV-NbAGO4) or control (TRV-EV) N. benthamiana plants at 3 weeks post-inoculation (wpi), as detected by bisulfite sequencing. Samples come from the same plants used in Figure 3F.
The original single-base resolution bisulfite sequencing data are shown in Figure  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint negative control), or cYFP-NbAGO4, cYFP-SlAGO4, or cYFP-V2 (as a positive control) in N.
benthamiana leaves. CFP-Fibrillarin was used as a nucleolus and Cajal body marker.
Confocal images were taken at two days after infiltration. Yellow fluorescence indicates a positive interaction. Arrowheads indicate the position of the Cajal body. Bar, 5μm. This experiment was repeated more than three times with similar results.
. CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint   . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint

Supplementary table 4. Values of IR methylation in independent experiments and
replicates.

Percentage of methylated cytosines in the intergenic region (IR) of TYLCV in systemic infection assays with TYLCV
or the V2 null mutant in Nbcoilin-silenced or control plants ( Figure 7D). . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/811091 doi: bioRxiv preprint