• 2018 NAS Awards Nominations Now Open
  • Sign-up for PNAS eTOC Alerts

Epigenetic transitions leading to heritable, RNA-mediated de novo silencing in Arabidopsis thaliana

  1. David C. Baulcombe1
  1. Department of Plant Sciences, University of Cambridge, Cambridge CB2 3EA, United Kingdom

  1. Edited by Steven E. Jacobsen, University of California, Los Angeles, CA, and approved November 18, 2014 (received for review July 10, 2014)

Significance

Using virus-induced gene silencing (VIGS) in wild-type and mutant Arabidopsis, we characterize a novel mechanism associated with the de novo establishment of heritable epigenetic marks in plants. Once established by this novel mechanism, the epigenetic mark is then reinforced by the previously characterized PolIV pathway of RNA-directed DNA methylation. A similar transition from the novel mechanism to the PolIV pathway is likely to explain many epigenetic phenomena in which RNA-directed DNA methylation is established de novo, including transposon silencing and paramutation. A practical benefit of our work is the identification of a mutant plant genotype in which the maintenance mechanism of epigenetic VIGS is reinforced. This genotype would aid the use of epigenetic VIGS for dissection of gene structure and function.

 
Next Section

Abstract

In plants, RNA-directed DNA methylation (RdDM), a mechanism where epigenetic modifiers are guided to target loci by small RNAs, plays a major role in silencing of transposable elements (TEs) to maintain genome integrity. So far, two RdDM pathways have been identified: RNA Polymerase IV (PolIV)-RdDM and RNA-dependent RNA Polymerase 6 (RDR6)-RdDM. PolIV-RdDM involves a self-reinforcing feedback mechanism that maintains TE silencing, but cannot explain how epigenetic silencing is first initiated. A function of RDR6-RdDM is to reestablish epigenetic silencing of active TEs, but it is unknown if this pathway can induce DNA methylation at naïve, non-TE loci. To investigate de novo establishment of RdDM, we have used virus-induced gene silencing (VIGS) of an active FLOWERING WAGENINGEN epiallele. Using genetic mutants we show that unlike PolIV-RdDM, but like RDR6-RdDM, establishment of VIGS-mediated RdDM requires PolV and DRM2 but not Dicer like-3 and other PolIV pathway components. DNA methylation in VIGS is likely initiated by a process guided by virus-derived small (s) RNAs that are 21/22-nt in length and reinforced or maintained by 24-nt sRNAs. We demonstrate that VIGS-RdDM as a tool for gene silencing can be enhanced by use of mutant plants with increased production of 24-nt sRNAs to reinforce the level of RdDM.

Methylation of cytosine (C) residues of DNA is a stable and heritable modification that mediates epigenetic control in eukaryotic genomes. In plants this modification occurs at CG, CHG, and CHH sequences (where H can be C, A, or T) and involves RNA-directed DNA methylation (RdDM) pathways. The DNA methyltransferases in these pathways are guided to target loci by ribonucleoproteins in which a small (s)RNA is the specificity determinant (1, 2).

There are two mechanisms to maintain DNA methylation. In RdDM pathways acting at predominantly transposable elements (TEs), SAWADEE HOMEODOMAIN HOMOLOG (SHH)1 binds and recruits RNA Polymerase IV (PolIV) to transcribe methylated DNA (3, 4). The PolIV transcripts are then made double-stranded by RNA-dependent RNA polymerase (RDR)2 and cleaved by Dicer-like 3 (DCL3) into 24-nt sRNAs that are loaded into and guide AGONAUTE (AGO)4 to complementary scaffold RNAs transcribed from the same locus by RNA polymerase V (PolV). The sRNA-bound AGO and the PolV transcript interact and recruit the de novo methyltransferase DRM2 that modifies C residues of the DNA strand acting as the template for PolV (5).

Maintenance of CG and CHG methylation during DNA replication occurs via MET1, the plant homolog of the mammalian DNA methyltransferase DNMT1, and the plant-specific CHROMOMETHYLASE 3 (CMT3) that works in concert with the SU(VAR)3-9 HOMOLOG (SUVH) SET domain histone methyltransferse KRYPTONITE. Both MET1 and CMT3 act independently of sRNAs but maintenance of CHH methylation within euchromatin is dependent on the continuous operation of the sRNA establishment mechanism, resulting in a self-reinforcing loop (1). This PolIV-RdDM maintenance step is stabilized by SHH1 and the SUVH2/9 SET domain proteins, which bind methylated DNA and continue to recruit PolIV and PolV, respectively to RdDM sites (3, 4, 6, 7). Maintenance of CHH methylation within heterochromatin involves CMT2 and relies on repressive chromatin modifications, independent of sRNAs (8, 9).

The detailed analyses of PolIV-RdDM provide a good explanation of how methylated Cs in an asymmetric context (CHH) can be maintained through feedback mechanisms, but they do not explain how DNA methylation at naïve loci can be first established. A second RdDM pathway has been proposed in which the sRNAs are 21/22-nt rather than 24-nt in length and PolIV is not involved (10⇓–12). This process is referred to as RDR6-RdDM to reflect the identity of the required RDR ortholog in Arabidopsis, and it was hypothesized that, in connection with the silencing of TEs and DNA methylation at transacting-siRNA (tasiRNA) loci, posttranscriptional gene silencing (PTGS) of PolII-derived transcripts is the trigger. However, the analysis of this mechanism did not conclusively demonstrate a role in RNA-mediated de novo silencing, and the establishment and maintenance phases of silencing at these loci cannot be easily distinguished.

To further investigate the de novo establishment of DNA methylation in plants we infected transgenic plants expressing a GFP transgene with an RNA virus carrying the promoter sequence of the reporter gene (13). On infection, the plant’s antiviral defense mechanism produced sRNAs against the virus and the promoter through virus-induced gene silencing (VIGS) (14). The reporter gene was silenced in the infected plant because of increased DNA methylation at its promoter that was maintained in later generations when the virus was no longer present. It was presumed that, like heritable silencing of TEs, the RdDM pathway was responsible for the establishment of VIGS-RdDM. However, these experiments in Nicotiana benthamiana could not test for involvement of pathways that had been previously characterized in Arabidopsis.

Using VIGS in Arabidopsis, we show that sRNAs can initiate heritable DNA methylation and transcriptional gene silencing (TGS) at an endogenous locus. The virus-derived sRNAs responsible for establishing DNA methylation are likely to be 21/22-nt in length and establishment occurs without canonical PolIV-RdDM, but requires PolV and DRM2. An earlier initiation stage of RNA-mediated de novo silencing may occur independently of DNA methylation. Furthermore, mutant plants producing highly abundant 24-nt sRNAs exhibit reinforced maintenance of RdDM. From our results with VIGS-RdDM, we propose three steps for de novo, heritable DNA methylation at endogenous loci in plants—initiation, establishment, and maintenance—which can occur with or without the presence of 24-nt sRNAs.

Previous SectionNext Section

Results

VIGS of FLOWERING WAGENINGEN.

To test the involvement of the PolIV-RdDM pathway in the establishment of DNA methylation, we set up a VIGS system in Arabidopsis thaliana. Our strategy was to first characterize VIGS in wild-type plants and then to test RNA silencing mutants for effects on epigenetic silencing of an endogenous target locus. We used tobacco rattle virus (TRV) as the viral vector (15, 16) and an active FLOWERING WAGENINGEN (FWA) epiallele, in the fwa-d epimutant (17), as the target locus. Most of the analyses described here involved plants carrying this epiallele, although for some comparative analyses the plants were wild-type Columbia (Col-0) carrying a naturally silenced allele of FWA (18). We refer to the two different epigenotypes as Col-0(FWAC) and Col-0(FWACme), respectively, to reflect the status of C methylation at FWA.

One TRV construct (TRV:FWAtr) included the direct tandem repeats from the FWA promoter and transcribed sequence extending to the second intron within the 5′UTR, a region previously shown to be required for silencing of FWA (17, 19). This construct was designed to test epigenetic silencing mechanisms. A second construct, to investigate PTGS, carried part of the FWA coding sequence (TRV:FWAcds) (Fig. S1 A and B).

In the infected plants (V0 generation) the level of FWA sRNAs correlated with the levels of viral RNA (Fig. 1 A and B) and, with TRV:FWAcds, the FWA mRNA was correspondingly less abundant than in control plants. Some repression of FWA mRNA was evident in plants infected with TRV:FWAtr, but the level of silencing did not correlate with the level of virus and by 45 d postinfection (dpi) it returned to levels in mock-infected plants (Fig. 1C and Fig. S1 C and D). From these data we conclude that effective posttranscriptional but not epigenetic silencing of FWA could be triggered by these TRV VIGS constructs. However, the TRV:FWAcds-infected plants did not flower early (Fig. S1E), as would be expected from the reduced expression of FWA. This lack of phenotype is most likely because the silencing was too weak or because early flowering was repressed by FWA that had been produced before virus inoculation (20).

Fig. 1.
View larger version:
Fig. 1.

Establishing VIGS of FWA. (A, Upper) Northern blot with TRV RNA species (Fig. S1B) and (Lower) sRNA Northern blot with FWAtr sRNAs in TRV:FWAtr-, TRV (T)-, or mock (M)-infected Col-0(FWAC) plants, 14 dpi. FDH and U6 were probed as loading controls, respectively. TRV:FWAtr is marked with a red asterisk. The black asterisks represent the 20-nt and 30-nt sRNA markers. For the sRNA Northern blot analysis, TRV:FWAtr-infected samples 7–12 were run on a separate gel as indicated by the white separating line between samples 6 and 7. (B, Upper) RT-PCR of TRV and TRV:FWAcds in TRV:FWAcds-, TRV (T)-, or mock (M)-infected Col-0(FWAC) plants, 14 dpi. FDH was assayed as an internal control. (Lower) sRNA Northern blot with FWAcds sRNAs in the same plants. U6 was probed as a loading control. The black asterisks represent the 20-nt and 30-nt sRNA markers. (C) FWA expression in TRV:FWAtr- (tr), TRV:FWAcds- (cds), TRV- (T), or mock (M)-infected Col-0(FWAC) plants, 14 dpi and 45 dpi. Each sample is represented by a black diamond and the average of all samples per treatment is represented by the red horizontal line. A two-tailed Student t test suggested FWA expression was repressed by TRV:FWAtr (*P < 0.05) and TRV:FWAcds (**P < 0.01), 14dpi. (D) Proportion of early- (black), intermediate- (dark gray), and late- (light gray) flowering V1 progeny from TRV:FWAtr-, TRV (T)-, or mock (M)-infected Col-0(FWAC) plants (A), compared with Col-0(FWACme) and Col-0(FWAC). Lines with a ratio of <1 (early):3 (late) are marked with a black asterisk.

The lack of epigenetic VIGS with TRV:FWAtr could indicate that the target promoter sequence was refractory to silencing or that epigenetic silencing was restricted to one allele and was masked by the expressed second allele. To test the second scenario we investigated the flowering time of progeny of TRV:FWAtr-infected plants (V1 generation) in the expectation that some plants might have inherited two silenced alleles of FWA, as in transgene-induced FWA silencing (17, 19). Many of the V1 plants flowered late, as did the Col-0(FWAC) controls. However, in the progeny of TRV:FWAtr- but not TRV- or TRV:FWAcds-infected plants, there were individuals that flowered as early as Col-0(FWACme) plants with fully silenced FWA (Fig. 1D and Fig. S2 A and B). The proportion of early flowering plants did not follow a simple 1:3 ratio, as would be expected if heterozygous epialleles in the V0 plant were stably inherited: it was either more or less than 1:3. We conclude that FWA silencing affected both alleles but that effects were not seen in the infected plant because they were initially weak and became progressively stronger during the transition between generations. Similar results were obtained with VIGS of the unsilenced fwa-2 epimutant (18) in the Ler background (Fig. S2C).

The progressive silencing of FWA was not a result of persistence of TRV:FWAtr because we failed to detect virus in ∼200 tested early flowering progeny using a PCR test (SI Materials and Methods). It was likely therefore that silencing was initiated in the V0 plant without affecting expression of FWA. The silencing of this gene would then have been established more completely in the transition between generations, or in the V1 plants. The proportion of early-flowering V1 progeny correlated positively with the level of virus infection and FWAtr sRNAs in the parent plant, consistent with initiation of epigenetic silencing in the infected plants (compare Fig. 1A with Fig. 1D).

The early-flowering time in the V1 progeny of infected plants was associated with changes in DNA methylation at the TRV:FWAtr target site (FWAtr). The FWAtr DNA in leaf or floral tissue from V0 plants was not methylated (Fig. S1F), but in the V1 progeny it was hypermethylated at levels similar to those in Col-0(FWACme) plants (Fig. 2A). Hypermethylation was established in all C sequence contexts (Fig. 2B) and, in the earliest flowering V1 plants, CHH methylation was 10–20% higher than wild-type Col-0(FWACme) with naturally silenced FWA. In V1 plants with an intermediate-flowering time the degree of methylation at FWA was 20–40% less than in early-flowering plants, although the proportion of methylated C residues in the CHH context was 20–50%: two to five times more than in Col-0(FWACme) (Fig. 2B and Fig. S2D). Most of this intermediate DNA methylation was at the 5′ end of the TRV:FWAtr target region (Fig. S2D), suggesting that VIGS-RdDM was established at the 5′ end of the tandem repeats and that it then spread in the 5′→3′ direction.

Fig. 2.
View larger version:
Fig. 2.

VIGS-RdDM of FWA induces early flowering in V1 progeny plants. (A) DNA methylation (McrBC-qPCR) at FWAtr in V1 progeny of Col-0(FWAC) plants infected with TRV:FWAtr, TRV, or mock conditions compared with Col-0(FWACme) and Col-0(FWAC) plants. DNA from early- and late-flowering progeny from TRV:FWAtr-infected lines 6 and 12 (Fig. 1A) were assayed. Each independent sample is represented by a black diamond and the average of all samples per treatment is represented by the red horizontal line. (B) Bisulfite sequencing analysis of DNA methylation at FWAtr in early- and late-flowering V1 progeny plants from A. Col-0(FWACme) and Col-0(FWAC) were included as a comparison. The position of C residues along the FWAtr region (black arrows) is represented by the ticks on the x axis and the context of methylation is represented by the different colors: CG is in black, CHG is in blue, and CHH is in red.

Flowering time and DNA methylation at FWA cosegregate in a Mendelian ratio in an F2 population produced from a cross between plants with FWA in a silent (FWACme) and active (FWAC) state (18). In contrast, consistent with a progressive increase in FWA silencing between generations, the V2 progeny of intermediate-flowering V1 plants all gained DNA methylation at FWA and flowered at the same time as Col-0(FWACme) (Fig. S3 A and B). This progressive silencing, together with the complete transmission of the FWA silenced and hypermethylated phenotype to the V2 generation from the early-flowering V1 plants (Fig. S3), indicates that TRV:FWAtr infection leads to transgenerational epigenetic silencing of the FWA promoter sequence, as with the original TRV:35S system (13). However, unlike the TRV:35S system, we could not detect the epigenetic silencing in the infected plants (Fig. S1F). To explain the FWA silencing in the V1 generation we propose that establishment of the epigenetic mark involved cryptic epigenetic changes to the FWAtr sequence in the infected plant. In the following sections we first describe a genetic analysis of FWA promoter VIGS and, second, a further analysis of establishment.

Separate RdDM Pathways for CG and CHH Methylation.

To investigate the genetic requirements for FWA promoter VIGS, we infected FWAC plants that carried mutations in various genes known to act in PolIV-RdDM (Table S1). Inoculation was with TRV:FWAtr, TRV, or mock conditions (Fig. S4A) and, as with the Col-0(FWAC), there was no change in flowering time of the V0 plants in any of the mutant backgrounds (Fig. S4B).

In the V1 progeny of infected poliv(FWAC), rdr2(FWAC), dcl3(FWAC), and ago4(FWAC) genotypes, as with Col-0(FWAC), there were early- and intermediate-flowering plants (Fig. 3A and Fig. S5) that gained DNA methylation at FWAtr (Fig. 3B and Fig. S6A). However, the RdDM in all instances with these mutants was predominantly in a CG context and, in the intermediate plants, it was restricted to the 5′ end of the tandem repeat (Fig. S6B). In contrast, none of the V1 progeny of polv(FWAC)- or drm1/2(FWAC)-infected plants were early flowering. From these data we conclude that there are at least two mechanisms involved in RdDM. First, there is a PolIV-, RDR2-, DCL3-, and AGO4-independent process leading to RdDM of C residues in a CG or CHG context. This process is clearly sufficient for silencing of FWA. A second process is dependent on PolIV, RDR2, DCL3, AGO4, PolV, and DRM2, and it affects CHH methylation.

Fig. 3.
View larger version:
Fig. 3.

VIGS-RdDM of FWA requires PolV and DRM2 and is enhanced in a dcl2/4 mutant. (A) Proportion of early- (black), intermediate- (dark gray), and late- (light gray) flowering V1 progeny from Col-0(FWAC), poliv(FWAC), rdr2(FWAC), dcl3(FWAC), ago4(FWAC), polv(FWAC), drm1/2(FWAC), dcl2/4(FWAC), and dcl2/3/4(FWAC) mutants infected with TRV:FWAtr (blue diamond), TRV (red diamond), or mock (orange diamond) conditions. (B) DNA methylation (McrBC-qPCR) at FWAtr in early- (E) and late- (L) flowering V1 progeny from two independent lines of poliv(FWAC), dcl3(FWAC), dcl2/4(FWAC), and dcl2/3/4(FWAC) mutants. Each independent sample is represented by a black diamond and the average of all samples per treatment is represented by the red horizontal line. (C) Proportion of different size classes of sRNAs that map to FWAtr in Col-0(FWAC) and dcl2/4(FWAC) TRV:FWAtr-infected V0 plants. The proportion of 21-nt (red), 22-nt (orange), 23-nt (light blue), and 24-nt (dark blue) sRNA reads was determined from the actual read count, corrected for multiple mapping. Three independent samples per genotype were assessed (Fig. S4 A and C).

In addition to DCL3, Arabidopsis encodes DCL1, -2, and -4 that generate 21-nt miRNAs, 22-nt sRNAs, and 21-nt sRNAs, respectively (21) and, in a separate assay for FWA silencing, there was a redundant requirement of DCL proteins (22). To further explore the DCL requirement in VIGS-RdDM, we assessed the dcl2(FWAC), dcl4(FWAC), dcl2/4(FWAC), and dcl2/3/4(FWAC) mutants in our assay. The viral load in dcl2/4(FWAC) and dcl2/3/4(FWAC) V0 plants was much greater than Col-0(FWAC) plants (Fig. S4C), consistent with the role of DCL2 and DCL4 acting in the antiviral RNA silencing pathway (23). The proportion of early-flowering V1 progeny from TRV:FWAtr-infected dcl2(FWAC) and dcl4(FWAC) plants was similar to Col-0(FWAC), but there were many more early-flowering V1 progeny from dcl2/4(FWAC) plants (Fig. 3A and Fig. S5). For some lines almost 100% of V1 progeny flowered early and, correspondingly, there was a high degree of DNA methylation at FWAtr in these plants (Fig. 3B and Fig. S6A). This FWAtr hypermethylation in dcl2/4(FWAC)-infected plants was more biased to the CG context than with the VIGS of Col-0(FWAC) plants.

In the TRV:FWAtr-infected dcl2/4(FWAC) V0 plants, unlike the wild-type plants, repression of FWA correlated with DNA methylation at FWAtr (10–50%) (Fig. S4 D–G). In a minority of plants, FWA was repressed independent of DNA methylation, suggesting PTGS of FWA may precede TGS, or an alternative repressive epigenetic mark could be present at FWA before initiation of DNA methylation. The pattern of DNA methylation established in the dcl2/4(FWAC) V0-infected plants was in all sequence contexts, with increased levels of CHH methylation presumably because of the high level of 24-nt sRNAs in these plants (Fig. S4A). However, deep sequencing of sRNAs from TRV:FWAtr-infected dcl2/4(FWAC) V0 plants indicated that ∼22–35% of sRNAs mapping to the FWAtr target region are 21/22-nt in length (compared with 90% in wild-type) (Fig. 3C and Fig. S7), consistent with the possibility that 21/22-nt sRNAs play a role in the establishment of DNA methylation in this system.

In contrast, the RdDM and FWA silencing in the V1 progeny of infected dcl2/3/4(FWAC) plants was not enhanced. TRV:FWAtr infection triggered early or intermediate flowering in a minority of the V1 generation plants (Fig. 3A and Fig. S5) and, overall, the level of DNA methylation, predominantly in the CG context, was less than in the V1 progeny of Col-0(FWAC) plants (Fig. 3B and Fig. S6A). This finding indicates that the greater silencing of FWA in dcl2/4(FWAC) V1 plants was dependent on DCL3, and consistent with this interpretation there were abundant 24-nt FWAtr sRNAs in the infected dcl2/4(FWAC) plants (Fig. 3C and Figs. S4A and S7). In dcl2/3/4(FWAC) V0 plants there were only a low abundance of 21/22-nt FWAtr sRNAs (Fig. S4A), despite comparable levels of viral load in infected dcl2/4(FWAC) and dcl2/3/4(FWAC) mutants (Fig. S4C). However, because FWA silencing was not eliminated in the V1 progeny of dcl2/3/4(FWAC) plants, the epigenetic silencing of FWA must have been initiated and established in the absence of DCL3 as with the V1 progeny of the dcl3(FWAC) mutant.

Establishment of FWA Epigenetic Silencing.

To further investigate the establishment step in VIGS-RdDM of FWA we made reciprocal crosses between mock-infected and TRV:FWAtr-infected Col-0(FWAC) plants (Fig. S8A) and analyzed the flowering time of the F1 progeny. Most of these F1 plants flowered late, indicating that FWA was not silenced. However, F1 progeny from 15 of 61 lines tested had an intermediate flowering time phenotype (irrespective of the direction of the cross) and had ∼50% DNA methylation at FWAtr (Fig. 4 and Fig. S8 B and C). Furthermore, the flowering-time behavior of self-fertilized F2 populations from the intermediate F1 plants segregated in a 1:2:1 (early:intermediate:late) ratio (Fig. S8D). We conclude from these data that the methylated allele in F1 plants was inherited from the TRV:FWAtr-infected parent, consistent with the semidominant mutation of FWA (18), and that this silent state was not transmitted to the allele from the mock-infected plant. The initiation of FWA silencing must have occurred at the DNA or chromatin level in the V0 plant and these data rule out that a transacting initiating factor, for example an RNA, is transmitted from the infected plant into its progeny.

Fig. 4.
View larger version:
Fig. 4.

VIGS-RdDM at FWA is established before or during gametogenesis. (A) Proportion of early- (black), intermediate- (dark gray), and late- (light gray) flowering F1 progeny from reciprocal crosses between mock- and TRV:FWAtr-infected Col-0(FWAC) plants, compared with self-fertilized Col-0(FWACme), Col-0(FWAC), mock-, and TRV:FWAtr-infected Col-0(FWAC) plants. (B) DNA methylation (McrBC-qPCR) at FWAtr in intermediate- (I) and late- (L) flowering F1 progeny from reciprocal crosses between mock- and TRV:FWAtr-infected Col-0(FWAC) plants (A), compared with self-fertilized control plants. Each independent sample is represented by a black diamond and the average of all samples per treatment is represented by the red horizontal line.

The early-flowering phenotype was an infrequent characteristic in V1 progeny of Col-0(FWAC)-infected plants (185 of 1,061; progeny from 27 TRV:FWAtr-infected lines presented here). However, the FWA DNA of all early-flowering V1 plants was either fully methylated or became fully methylated in the V2 generation (Fig. S3), suggesting that both gametes had undergone an epigenetic conversion in the V0 plant that would be manifested as full DNA methylation in the V1 or V2 plants. It must be, therefore, that the infrequent epigenetic conversion in the infected plant affects either both FWA alleles or none. To test this hypothesis, we harvested individual siliques from Col-0(FWAC) V0 plants infected with TRV:FWAtr (Fig. S8A) and measured the flowering time of the individual progeny. We found that most individual siliques produced 100% late-flowering progeny (Fig. S8E), consistent with the low frequency of early-flowering behavior transmitted to the V1 progeny. However, there were some rare siliques with 50–100% of seed that grew into early-flowering plants. The concentration of FWA silencing in a few siliques further supports that an epigenetic factor is inherited through both gametes within certain flowers of the infected plant.

To further investigate when silencing of FWA is initiated, we assessed the degree of DNA methylation at FWAtr in a population of pollen cells from V0 plants. In Col-0(FWACme) plants, the FWA tandem repeat is demethylated at certain C residues within the vegetative nucleus (VN) of pollen because of the action of the DNA glycosylase DEMETER (DME), but full FWAtr DNA methylation is maintained in the sperm cells (SC) (24, 25). Therefore, on average 66% of DNA from pollen of a Col-0(FWACme) plant will be fully methylated at FWAtr. We predict if FWA silencing is initiated in the gametes or earlier, a proportion of pollen SCs from TRV:FWAtr-infected Col-0(FWAC) plants will be fully methylated like Col-0(FWACme).

To test this hypothesis, we bisulfite-sequenced pollen DNA from mock- and TRV:FWAtr-infected Col-0(FWAC) (Fig. S8A) and dcl2/4(FWAC) (Fig. S4D) plants and compared this to Col-0(FWACme) and dcl2/4(FWACme) plants (Fig. 5 and Fig. S9 A and B). Approximately 20% of clones representing Col-0(FWACme) and dcl2/4(FWACme) DNA lacked C methylation at residues that are demethylated in the VN (24). This value is consistent with bisulfite sequencing clones representing DNA from SC and the VN, although the proportion of unmethylated DNA is lower than the predicted 33%. The deviation is likely a result of preferential amplification of methylated DNA in the PCR.

Fig. 5.
View larger version:
Fig. 5.

FWAtr is methylated in pollen of TRV:FWAtr-infected plants. Bisulfite sequencing analysis of DNA methylation at FWAtr in pollen from mock- and TRV:FWAtr-infected (A) Col-0(FWAC) and (B) dcl2/4(FWAC) plants. Data for two independent TRV:FWAtr-infected lines per genotype are presented. Col-0(FWACme) and dcl2/4(FWACme) plants were assayed as a comparison. The position of C residues along the FWAtr region is represented by the ticks on the x axis and the context of methylation is represented by the different colors: CG is in black, CHG is in blue, and CHH is in red. The four CG residues that are demethylated in the VN by DME (24, 25) are represented by an asterisk.

No clones of pollen DNA from Col-0(FWAC) and dcl2/4(FWAC) mock-infected plants contained methylated Cs at FWAtr. In contrast, 10–25% of clones representing pollen DNA from Col-0(FWAC) TRV:FWAtr-infected plants were methylated in all sequence contexts at the FWA target (Fig. 5A and Fig. S9A). Because of this small sample size, we are unable to differentiate DNA from the SC and VN with confidence, but the presence of fully methylated clones indicates that DNA from SC is represented. Consistent with this interpretation is the correlation of DNA methylation in pollen of TRV:FWAtr-infected Col-0(FWAC) plants with the proportion of early-flowering progeny (Fig. S9C). This correlation was most striking in samples from dcl2/4(FWAC)-infected plants in which 80–85% of pollen DNA was methylated (Fig. 5B and Fig. S9B), corresponding to highly abundant early-flowering V1 progeny (Fig. 3A and Fig. S9D). From these results we conclude that VIGS-RdDM of FWA in TRV:FWAtr-infected plants takes place either at or before gametogenesis.

Previous SectionNext Section

Discussion

Epigenetic Transitions in de Novo Silencing.

Our analysis of FWA promoter VIGS was based on the expectation that the canonical PolIV-RdDM pathway would be involved in establishing heritable silencing in the infected plants. In subsequent generations we anticipated that, as with 35S promoter VIGS in N. benthamiana, persistence of heritable epigenetic marks would be independent of RNA and dependent on the maintenance methyltransferase, MET1 (13). However, our data are not consistent with that prediction. The data indicate instead that heritable silencing of FWA involves a complex sequence of epigenetic transitions in which the canonical PolIV-RdDM pathway is involved but not essential. We propose below that the same sequence of epigenetic transitions applies generally when epigenetic marks in plants are established de novo by an RNA-based mechanism.

Our conclusions are based primarily on two key findings. First there is the unexpected observation that FWA promoter VIGS in wild-type plants is largely independent of key components of PolIV-RdDM, including PolIV, RDR2, DCL3, and AGO4 (Fig. 3 and Fig. S6). This finding prompts a radical shift from previous views about RdDM in which most of these proteins were viewed as necessary for establishment of DNA methylation (1, 2).

The second key finding with FWA promoter VIGS, unlike 35S promoter VIGS, was that the epigenetic silencing could be established without detectable effect in the V0 plant and is progressively reinforced over two or three generations. The progression was associated eventually with DNA methylation in CG, CHG, and CHH contexts that spread from the 5′ to the 3′ part of the promoter relative to the direction of transcription (Fig. S2D). Consistent with this observation, sRNAs mapped primarily to the 5′ end of FWAtr in Col-0(FWAC) and dcl2/4(FWAC) TRV:FWAtr-infected V0 plants (Fig. S7).

We discuss below how the slow progression of silencing in the FWA promoter VIGS system allowed resolution of the epigenetic transitions, including establishment and partially redundant systems for maintenance that is either dependent or independent of RNA. The transitions were less apparent with 35S promoter VIGS because it segued rapidly from establishment in the infected plant into RNA-independent maintenance in the progeny.

Initiation and Establishment of Epigenetic Silencing.

Mutation of PolV function led to complete loss of VIGS-RdDM (Fig. 3A and Figs. S5 and S6A) and it is therefore likely that this protein acts in both establishment and RNA-mediated maintenance of silencing. In other RdDM silencing systems, PolV has been implicated in the production of scaffold RNAs that are the binding site of AGO-bound sRNAs that direct DNA methylation to the adjacent chromatin (1, 2). It is anticipated that PolV has a scaffold RNA role in FWA promoter VIGS, as proposed for these other RdDM systems. The most abundant sRNAs in the TRV:FWAtr-infected plants are the 21/22-nt size class generated by DCL2, and DCL4 (23) (Figs. 1A and 3C and Figs. S4A and S7), and it is probable, as in the RDR6-RdDM system (12), that these RNAs account for the primary RdDM mediated by VIGS.

Mutation of the de novo methyltransferase DRM2 leads to complete loss of VIGS-RdDM (Fig. 3A and Figs. S5 and S6A), suggesting that DNA methylation is an early epigenetic mark in the establishment mechanism. DNA methylation at FWAtr could be detected in the pollen of plants undergoing VIGS-RdDM, indicating that DNA methylation is established early (Fig. 5 and Fig. S8). However, we could not detect this FWA promoter methylation in the vegetative tissue of Col-0(FWAC) V0 plants (Fig. S1F), and it remains possible that another process—initiation—precedes establishment of DNA methylation in the gametes. Such an alternative process could involve histone modifications (1) or it could involve persistent RNAs, like those involved in “recovery” from virus infection (26). Recovery is a long-lived and sequence-specific immunity to secondary infection by plant viruses including TRV that is mediated by RNA silencing.

Maintenance of Epigenetic Silencing.

Our data indicate that, following establishment of FWA promoter VIGS, the epigenetic marks are maintained in the V1 and subsequent progeny by two partially redundant maintenance mechanisms. An RNA-independent maintenance relies on the recognition of hemimethylated Cs in the symmetrical context of newly replicated DNA motifs by the DNA methyltransferases MET1 and CMT3. In contrast, the RNA-dependent maintenance is sequence motif-independent. It involves canonical PolIV-RdDM in which methyl DNA-binding proteins recruit the 24-nt sRNA biogenesis proteins to the genomic site of primary RdDM. The 24-nt sRNAs would then continue to direct the DNA methyltransferases to the unmethylated strand of newly replicated DNA. DCL3 is a key protein in this process because it generates the 24-nt sRNAs.

The dcl3 phenotype is consistent with this interpretation because the mutation did not affect establishment but it did reduce the level of CHH methylation (Fig. 3 and Fig. S6). That there was no effect of dcl3 on the frequency of early-flowering progeny is probably because the overall level of FWAtr DNA methylation in the dcl3 mutant was similar to that observed for the wild-type, although it was more biased to CG and CHG methylation. We propose that initiation and establishment of FWA silencing in dcl3 would have been as in wild-type plants but that maintenance was only via the RNA-independent mechanism.

Previous SectionNext Section

Conclusion

Our proposed model for FWA promoter VIGS involves initiation and establishment mechanisms likely mediated by 21/22-nt sRNAs followed by two maintenance mechanisms that are either dependent on 24-nt sRNAs or independent of RNA (Fig. S10). This model can account for the high level of DNA methylation at FWAtr in all contexts in the V1 progeny of dcl2/4 plants and the reduction of this strong silencing in dcl2/3/4 V1 progeny to levels of CG methylation that were lower than in the dcl3 mutant (Fig. 3B and Fig. S6). To explain these effects, we propose that initiation and establishment of FWA silencing in the dcl2/4 and dcl2/3/4 mutants is at a lower level than in wild-type and dcl3 because the 21/22-nt initiator sRNAs would be produced, presumably, by DCL1 alone rather than DCL1, -2, and -4 in wild-type plants. The low level of primary RdDM in the dcl2/4 plants would be compensated for by the RNA-dependent maintenance mechanism that would be supercharged by the very abundant 24-nt sRNAs (Fig. 3C and Figs. S4A and S7).

This model could also explain the transition from PTGS to TGS of the Evade retroelement (27), the RDR6-RdDM pathway in which 21/22-nt sRNAs direct DNA methylation at active TEs (12), the 21-nt sRNAs from VN of pollen having a role in epigenetic modification of DNA in the SC (28), and with secondary epigenetically activated siRNAs (easiRNAs) and tasiRNAs that guide RdDM at active TEs and TAS genes, respectively (11, 29). The paradigm could also be reconciled with the finding that PolIV is apparently required for de novo RdDM of an FWA transgene in plants with an endogenous FWACme (30). We propose that, independent of PolIV, there is very low-level silencing of this transgene similar to the establishment of silencing in our V0 plants (Fig. 5A). Transgenes are often prone to spontaneous silencing because they produce aberrant RNA. However, in the presence of PolIV, 24-nt sRNAs would maintain and reinforce the silencing of this transgene. Some of these 24nt sRNAs may have been produced at FWACme and acted in trans at the transgene locus.

The interpretation that 24-nt sRNAs are not sufficient for establishment is also consistent with the similar levels of 24-nt sRNAs complementary to FWAtr in plants with active and silent FWA, despite the different levels of DNA methylation at FWAtr (19). Similarly, in paramutation, there is evidence that 24-nt sRNAs are not sufficient to transmit silencing from one allele to another (31). It is therefore possible that the requirement for 21/22-nt sRNAs for establishment and the switch to 24-nt sRNAs for maintenance is a general feature of de novo RdDM in plants (12).

An attraction of our model is that it requires only a single process to explain diverse initiation of epigenetic silencing in many systems. However, we have not yet been able to carry out a rigorous test of the model because no mutant devoid of 21/22-nt viral sRNAs is available. In the absence of such a test we concede that a second mechanism for establishment of RdDM involving 24nt sRNAs remains possible.

In addition to mechanistic insights, our data also suggest an enhanced strategy for heritable, epigenetic VIGS of endogenous loci. The target locus should be in a dcl2/4 mutant to establish gene silencing and there should be a functional DCL3 to ensure maximal reinforcement through the RNA-dependent maintenance mechanism. To boost the system further, a target with a high number of C residues in the CG context will ensure efficiency of the RNA-independent maintenance mechanism. This strategy may have application if epigenetic silencing would be useful for crop plant improvement.

Previous SectionNext Section

Materials and Methods

Plant Methods.

Wild-type [Col-0(FWACme)], the fwa-d epimutant (17) [Col-0(FWAC)] and various RNA silencing mutants in the Col-0 background (Table S1) were grown using standard plant growth methods. See SI Materials and Methods.

Viral Inoculations.

Previously described TRV VIGS vectors (15, 16) were modified to contain the FWAtr or FWAcds DNA sequences. Virus replication was carried out according to ref. 15 and young leaves of 3-wk-old Arabidopsis plants were mechanically inoculated. See SI Materials and Methods.

Nucleic Acid Analyses.

Standard protocols for RNA and DNA extraction were performed. See SI Materials and Methods for subsequent techniques and Table S2 for oligonucleotides.

Previous SectionNext Section

Acknowledgments

We thank Sebastian Müller for bioinformatic support; Shuoya Tang and Mel Steer for technical assistance; and Ian Henderson, Tim Hore, Jake Harris, and Quentin Gouil for critical reading prior to submission. Work in the D.C.B. laboratory is supported by The Gatsby Charitable Foundation, the EU FP7 Collaborative Project Grant AENEAS, and the European Research Council Advanced Investigator Grant ERC-2013-AdG 340642. D.C.B. is the Royal Society Edward Penley Abraham Research Professor.

Previous SectionNext Section

Footnotes

  • ↵1To whom correspondence should be addressed. Email: dcb40{at}cam.ac.uk.

Freely available online through the PNAS open access option.

Previous Section
 

References

  1. ↵
    1. Law JA,
    2. Jacobsen SE
    (2010) Establishing, maintaining and modifying DNA methylation patterns in plants and animals. Nat Rev Genet 11(3):204–220
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  2. ↵
    1. Matzke MA,
    2. Mosher RA
    (2014) RNA-directed DNA methylation: An epigenetic pathway of increasing complexity. Nat Rev Genet 15(6):394–408
    .
    CrossRefMedlineGoogle Scholar
  3. ↵
    1. Law JA, et al.
    (2013) Polymerase IV occupancy at RNA-directed DNA methylation sites requires SHH1. Nature 498(7454):385–389
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  4. ↵
    1. Zhang H, et al.
    (2013) DTF1 is a core component of RNA-directed DNA methylation and may assist in the recruitment of Pol IV. Proc Natl Acad Sci USA 110(20):8290–8295
    .
    Abstract/FREE Full Text
  5. ↵
    1. Zhong X, et al.
    (2014) Molecular mechanism of action of plant DRM de novo DNA methyltransferases. Cell 157(5):1050–1060
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  6. ↵
    1. Johnson LM, et al.
    (2014) SRA- and SET-domain–containing proteins link RNA polymerase V occupancy to DNA methylation. Nature 507(7490):124–128
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  7. ↵
    1. Liu Z-W, et al.
    (2014) The SET domain proteins SUVH2 and SUVH9 are required for Pol V occupancy at RNA-directed DNA methylation loci. PLoS Genet 10(1):e1003948
    .
    CrossRefMedlineGoogle Scholar
  8. ↵
    1. Zemach A, et al.
    (2013) The Arabidopsis nucleosome remodeler DDM1 allows DNA methyltransferases to access H1-containing heterochromatin. Cell 153(1):193–205
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  9. ↵
    1. Stroud H, et al.
    (2014) Non-CG methylation patterns shape the epigenetic landscape in Arabidopsis. Nat Struct Mol Biol 21(1):64–72
    .
    MedlineGoogle Scholar
  10. ↵
    1. Pontier D, et al.
    (2012) NERD, a plant-specific GW protein, defines an additional RNAi-dependent chromatin-based pathway in Arabidopsis. Mol Cell 48(1):121–132
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  11. ↵
    1. Wu L,
    2. Mao L,
    3. Qi Y
    (2012) Roles of Dicer-like and argonaute proteins in TAS-derived small interfering RNA-triggered DNA methylation. Plant Physiol 160(2):990–999
    .
    Abstract/FREE Full Text
  12. ↵
    1. Nuthikattu S, et al.
    (2013) The initiation of epigenetic silencing of active transposable elements is triggered by RDR6 and 21-22 nucleotide small interfering RNAs. Plant Physiol 162(1):116–131
    .
    Abstract/FREE Full Text
  13. ↵
    1. Jones L,
    2. Ratcliff F,
    3. Baulcombe DC
    (2001) RNA-directed transcriptional gene silencing in plants can be inherited independently of the RNA trigger and requires Met1 for maintenance. Curr Biol 11(10):747–757
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  14. ↵
    1. Lu R,
    2. Martin-Hernandez AM,
    3. Peart JR,
    4. Malcuit I,
    5. Baulcombe DC
    (2003) Virus-induced gene silencing in plants. Methods 30(4):296–303
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  15. ↵
    1. Ratcliff F,
    2. Martin-Hernandez AM,
    3. Baulcombe DC
    (2001) Tobacco rattle virus as a vector for analysis of gene function by silencing. Plant J 25(2):237–245
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  16. ↵
    1. Liu Y,
    2. Schiff M,
    3. Dinesh-Kumar SP
    (2002) Virus-induced gene silencing in tomato. Plant J 31(6):777–786
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  17. ↵
    1. Kinoshita Y, et al.
    (2007) Control of FWA gene silencing in Arabidopsis thaliana by SINE-related direct repeats. Plant J 49(1):38–45
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  18. ↵
    1. Soppe WJJ, et al.
    (2000) The late flowering phenotype of fwa mutants is caused by gain-of-function epigenetic alleles of a homeodomain gene. Mol Cell 6(4):791–802
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  19. ↵
    1. Chan SW-L,
    2. Zhang X,
    3. Bernatavichute YV,
    4. Jacobsen SE
    (2006) Two-step recruitment of RNA-directed DNA methylation to tandem repeats. PLoS Biol 4(11):e363
    .
    CrossRefMedlineGoogle Scholar
  20. ↵
    1. Ikeda Y,
    2. Kobayashi Y,
    3. Yamaguchi A,
    4. Abe M,
    5. Araki T
    (2007) Molecular basis of late-flowering phenotype caused by dominant epi-alleles of the FWA locus in Arabidopsis. Plant Cell Physiol 48(2):205–220
    .
    Abstract/FREE Full Text
  21. ↵
    1. Bologna NG,
    2. Voinnet O
    (2014) The diversity, biogenesis, and activities of endogenous silencing small RNAs in Arabidopsis. Annu Rev Plant Biol 65:473–503
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  22. ↵
    1. Henderson IR, et al.
    (2006) Dissecting Arabidopsis thaliana DICER function in small RNA processing, gene silencing and DNA methylation patterning. Nat Genet 38(6):721–725
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  23. ↵
    1. Deleris A, et al.
    (2006) Hierarchical action and inhibition of plant Dicer-like proteins in antiviral defense. Science 313(5783):68–71
    .
    Abstract/FREE Full Text
  24. ↵
    1. Schoft VK, et al.
    (2011) Function of the DEMETER DNA glycosylase in the Arabidopsis thaliana male gametophyte. Proc Natl Acad Sci USA 108(19):8042–8047
    .
    Abstract/FREE Full Text
  25. ↵
    1. Calarco JP, et al.
    (2012) Reprogramming of DNA methylation in pollen guides epigenetic inheritance via small RNA. Cell 151(1):194–205
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  26. ↵
    1. MacDiarmid R
    (2005) RNA silencing in productive virus infections. Annu Rev Phytopathol 43:523–544
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  27. ↵
    1. Marí-Ordóñez A, et al.
    (2013) Reconstructing de novo silencing of an active plant retrotransposon. Nat Genet 45(9):1029–1039
    .
    CrossRefMedlineGoogle Scholar
  28. ↵
    1. Slotkin RK, et al.
    (2009) Epigenetic reprogramming and small RNA silencing of transposable elements in pollen. Cell 136(3):461–472
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  29. ↵
    1. Creasey KM, et al.
    (2014) miRNAs trigger widespread epigenetically activated siRNAs from transposons in Arabidopsis. Nature 508(7496):411–415
    .
    CrossRefMedlineWeb of ScienceGoogle Scholar
  30. ↵
    1. Chan SW-L, et al.
    (2004) RNA silencing genes control de novo DNA methylation. Science 303(5662):1336
    .
    FREE Full Text
  31. ↵
    1. Hollick JB
    (2012) Paramutation: A trans-homolog interaction affecting heritable gene regulation. Curr Opin Plant Biol 15(5):536–543
    .
    CrossRefMedlineGoogle Scholar

Online Impact