Virus-Mediated Targeted DNA Methylation Illuminates the Dynamics of Methylation in an Endogenous Plant Gene

DNA methylation maintains genome stability and regulates gene expression in plants. RNA-directed DNA methylation (RdDM) is critical for appropriate methylation. However, no efficient tools are available for the investigation of the functions of specific DNA methylation. In this study, the cucumber mosaic virus vector was used for targeted DNA methylation. Methylation was rapidly induced but gradually decreased from the 3′ end of the target endogenous sequence in Nicotiana benthamiana, suggesting a mechanism to protect against the ectopic introduction of DNA methylation. Increasing 24-nt siRNAs blocked this reduction in methylation by down-regulating DCL2 and DCL4. RdDM relies on the sequence identity between RNA and genomic DNA; however, this identity does not appear to be the sole determinant for efficient DNA methylation. The current findings provide new insight into the regulation of DNA methylation and promote additional effort to develop efficient targeted DNA methylation in plants.

DNA methylation for transposon control and gene regulation at specific loci has been investigated in previous studies with mutants that show a global change in DNA methylation. Caution is needed in concluding that any observed change directly reflects the regulation at a target region. One strategy to overcome this problem is targeted DNA methylation/demethylation. Targeted methylation of a foreign gene has been successfully achieved by expressing homologous RNAs with a target sequence from a transgene [2][3][4][5][6]. Targeted methylation of an endogenous gene was reported in petunia [4], 2 of 17 maize [7], potato [8], rice [6,9], and A. thaliana [10,11], though silencing of the endogenous gene was less efficient as compared to an exogenous gene [6,12]. Zinc-finger and CRISPR technologies used for gene editing were successfully applied to targeted DNA methylation. Johnson et al. demonstrated that Zinc-finger-fused SUVH9 enabled targeted DNA methylation in A. thaliana FWA gene [13]. Recent studies also showed that several zinc-finger-fused RdDM components could induce targeted DNA methylation [14] and CRISPR/CAS-based recruitment of tobacco DRM could also achieve targeted DNA methylation in A. thaliana FWA gene [15]. For targeted demethylation, the zinc-finger-fused ten-eleven translocation 1 (TET1) catalytic domain could induce DNA demethylation at the FWA gene and CACTA1 transposon in A. thaliana [16].
Both expression of homologous RNAs against target regions and utilization of effector proteins for gene editing are powerful tools, and these approaches require transgenic plants that are laborious and time-consuming to develop. Further, transgenic technology is only useful if experimental systems are available to create transgenic plants (e.g., gene delivery into plant cells and regeneration of plants). Plant virus vector technology can overcome these problems since it does not require transgenic plants. Targeted methylation of a foreign gene by a plant virus vector was achieved by expressing RNA identical to the target sequence [17][18][19][20]. Similar methylation of endogenous genes has also been achieved in petunia [21,22], tomato [21], and A. thaliana [23]. However, as observed in transgene expression, targeted DNA methylation using a plant virus vector against endogenous genes was less efficient compared to methylation of exogenous genes. Further, success has only been achieved for limited genes in a few plant species. More efficient targeted DNA methylation by plant virus vectors will require a deeper understanding of the characteristics of targeted DNA methylation. At present, such understanding is limited.
In this study, characteristics of targeted DNA methylation against an endogenous gene using a cucumber mosaic virus (CMV) vector in Nicotiana benthamiana was investigated. Previous successful studies used virus vectors to target regions that are natively regulated by DNA methylation in tomato LeSPL-CNR [21] and A. thaliana FWA [23], but this study targeted the promoter region in the phytoene desaturase (PDS) gene, which does not appear to be under native regulation by DNA methylation. This strategy was adopted to avoid intrinsic regulation of DNA methylation that would complicate the interpretation of outcomes. The tetraploid plant, N. benthamiana [24], was chosen as a model because more than 70% of flowering plants are polyploids and epigenetic control plays a critical role in genome evolution [25]. CMV is a positive-stranded RNA virus consisting of three RNA segments, RNA1-3. CMV infects a variety of plants, including more than 1200 species in over 100 families [26]. A CMV vector based on the CMV-Y strain has been developed for gene silencing in plants, including N. benthamiana [20,27], Glycine max [28], Capsicum annuum [29], and for efficient gene expression in N. benthamiana [30,31]. Targeted DNA methylation by the CMV vector was achieved against a foreign gene in N. benthamiana [20] and also against endogenous genes in petunia and tomato [21].
In this study, we show that CMV vector-induced DNA methylation in the PDS gene's promoter region is dynamically changed in N. benthamiana. The vector induces DNA methylation within two weeks at all cytosine contexts, but the methylation level gradually decreased from the 3 end of the targeted region. Increasing 24-nt siRNAs instead of 21-nt and 22-nt siRNAs by down-regulating DCL2 and DCL4 expression leads to enhanced durability of the high DNA methylation state. Targeted DNA methylation is also induced in the region that contains mismatch sequences to expressed RNAs, suggesting that sequence identity may not be the sole determinant of the efficiency of DNA methylation.

CMV Vector Induces DNA Methylation Specifically in the Targeted Promoter Sequences of the PDS Gene
The phytoene desaturase (PDS) gene was chosen as a model for characterizing virusinduced targeted DNA methylation. Two copies of the PDS gene (designated as NbPDSa and NbPDSb in this study) are found in N. benthamiana draft genome sequences [32].

CMV Vector Induces DNA Methylation Specifically in the Targeted Promoter Sequences of the PDS Gene
The phytoene desaturase (PDS) gene was chosen as a model for characterizing virusinduced targeted DNA methylation. Two copies of the PDS gene (designated as NbPDSa and NbPDSb in this study) are found in N. benthamiana draft genome sequences [32]. The region, −1225 to −1 of NbPDSb, had promoter activity via transient expression of β-glucuronidase (GUS)-fused construct using agroinfiltration in N. benthamiana leaf tissues (Figure 1A,B). The DNA methylation inducers for the promoter region using the CMV vector are provided ( Figure 1C,D) [20,21]. The bisulfite sequencing indicated that the methylation level around the target region (−957 to −565) was quite low in both plants without CMV infection and plants inoculated with vector control expressing a partial sequence of GFP at 21 days after inoculation (dpi) (Figure 2A). The expression of the T2s inducer RNA significantly induced methylation at all cytosine contexts (CG, CHG, and CHH) throughout the target region ( Figure 2A). The average levels of CG, CHG, and CHH methylation in the target region were approximately 45%, 81%, and 62%, and no clear differences in DNA methylation levels were observed between constructs designed in a coding strand (T2s) and an opposite strand (T2as) (Figure 2A,B). When T3s inducer (5' half of the analyzed area) was expressed, DNA methylation was specifically induced in the targeted region (Figure 2A), and there were no clear differences in DNA methylation levels between constructs based on a coding (T3s) and an opposite strand (T3as) (Figure 2A  The DNA methylation inducers for the promoter region using the CMV vector are provided ( Figure 1C,D) [20,21]. The bisulfite sequencing indicated that the methylation level around the target region (−957 to −565) was quite low in both plants without CMV infection and plants inoculated with vector control expressing a partial sequence of GFP at 21 days after inoculation (dpi) (Figure 2A). The expression of the T2s inducer RNA significantly induced methylation at all cytosine contexts (CG, CHG, and CHH) throughout the target region ( Figure 2A). The average levels of CG, CHG, and CHH methylation in the target region were approximately 45%, 81%, and 62%, and no clear differences in DNA methylation levels were observed between constructs designed in a coding strand (T2s) and an opposite strand (T2as) (Figure 2A,B). When T3s inducer (5 half of the analyzed area) was expressed, DNA methylation was specifically induced in the targeted region (Figure 2A), and there were no clear differences in DNA methylation levels between constructs based on a coding (T3s) and an opposite strand (T3as) (Figure 2A,B).
Twenty-one-nt, 22-nt, and 24-nt small RNAs play an important role in RdDM. Almost no small RNAs (18-50 nt) accumulated around the region (−957 to −565) in healthy plants ( Figure 2C,D). In contrast, various sizes of small RNAs, including 21-, 22-, and 24-nt, accumulated in the targeted region in T3s-expressing plants at 21 dpi ( Figure 2C,D). The order of accumulation levels was 21-nt > 22-nt > 24-nt. Secondary siRNAs appeared not to be generated in both upstream and downstream regions ( Figure 2C), which is consistent with no spread of DNA methylation from the target region to the downstream region ( Figure 2A). mRNA levels of the NbPDS gene by real-time PCR show that total mRNA levels of NbPDSa and NbPDSb decreased in the presence of inducers, and no clear difference between constructs was seen based on coding and an opposite strand ( Figure 2E).  Figure 2C,D). The order of accumulation levels was 21-nt > 22-nt > 24-nt. Secondary siRNAs appeared not to be generated in both upstream and downstream regions ( Figure 2C), which is consistent with no spread of DNA methylation from the target region to the downstream region ( Figure 2A). mRNA levels of the NbPDS gene by real-time PCR show that total mRNA levels of NbPDSa and NbPDSb decreased in the presence of inducers, and no clear difference between constructs was seen based on coding and an opposite strand ( Figure 2E).

Gradual Reduction of DNA Methylation from the 3' End of the Targeted Region
A time-course analysis of DNA methylation using bisulfite sequencing indicated that DNA methylation in all cytosine contexts began to be induced from 6 dpi in both inoculated ( Figure 3A,C) and upper non-inoculated leaves ( Figure 3B,D) infected with CMV carrying T3s inducer. This timing corresponds to the time when the virus was significantly The percent of cytosine methylation in each context was calculated in the targeted region. The black, blue, red, and gray bars indicate CG, CHG, CHH, and total methylation, respectively. (C) IGV snapshots of 21-nt, 22-nt, and 24-nt small RNAs mapped in the region analyzed by bisulfite sequencing (−957 to −565). Small RNA sequencing used the total RNAs from the leaves where the T3s inducer (−938 to −739) was expressed from the CMV vector at 21 days after inoculation. The y-axes indicate raw reads on a log scale, normalized by total mapped reads. (D) Length distribution of 18-50-nt small RNAs in the targeted region. (E) Total mRNA levels of NbPDSa and NbPDSb were analyzed in the upper non-inoculated leaves infected with CMV/NbPDSb-T2s, -T2as, -T3s, -T3as, and -asGFP (vector control) at 21 days after inoculation. The expression levels were normalized relative to NbEF1α. The error bars indicate the standard deviation of four biological replicates. Statistical analysis used the Dunnett's method. Data from vector control (asGFP) was used as a control for statistical analysis. **, p < 0.01.

Gradual Reduction of DNA Methylation from the 3 End of the Targeted Region
A time-course analysis of DNA methylation using bisulfite sequencing indicated that DNA methylation in all cytosine contexts began to be induced from 6 dpi in both inoculated ( Figure 3A,C) and upper non-inoculated leaves ( Figure 3B,D) infected with CMV carrying T3s inducer. This timing corresponds to the time when the virus was significantly accumulated ( Figure 3E,F). No spread of DNA methylation to the downstream region was observed at any time points in both inoculated and upper non-inoculated leaves. Unexpectedly, we found that DNA methylation level gradually decreased in all cytosine contexts, especially at the 3 region of the target region in the NbPDSb promoter in plants inoculated with CMV carrying T2s inducer ( Figure 4). observed at any time points in both inoculated and upper non-inoculated leaves. Unexpectedly, we found that DNA methylation level gradually decreased in all cytosine contexts, especially at the 3' region of the target region in the NbPDSb promoter in plants inoculated with CMV carrying T2s inducer ( Figure 4).   As 24-nt siRNAs are important for maintaining DNA methylation [23], increasing the amount of 24-nt siRNAs was expected to maintain the elevated DNA methylation levels in our system. We used DCL2 and DCL4-double knock-down (d2d4) N. benthamiana transgenic plants [34], where increases in 24-nt siRNAs are expected by reducing the transcription of 21-nt and 22-nt siRNAs that are produced by DCL4 and DCL2, respectively [35]. To confirm the increase in 24-nt siRNAs generated from inducers expressed from CMV vector in d2d4 plants, we analyzed the pattern of small RNAs produced from the T2s inducer by small RNA sequencing. The results in 21 dpi showed that most siRNAs produced from the T2s inducer in d2d4 plants were 24-nt siRNAs, which accounted for about 74.3 % of total 21-nt, 22-nt, and 24-nt siRNAs, in contrast to about 4.9 % in wild-type plants (Supplementary Figure S1). CMV/NbPDSb-T2s and CMV/NbPDSb-T3s were inoculated onto wild-type and d2d4 plants, and methylation levels were assessed. At 35dpi, DNA methylation level at NbPDSb was dramatically reduced in wild-type plants but remained high in d2d4 plants ( Figure 5A,B). Small RNA sequencing indicated that amounts of 24-nt siRNAs dramatically increased in d2d4 plants in the low-methylation region observed in wild-type plants at 35 dpi ( Figure 5C,D; Supplementary Figure S2). Virus accumulation in d2d4 plants is not consistently higher compared to wild-type plants at 35 dpi ( Figure 5E). These results suggest that low accumulation of 24-nt siRNAs leads to a reduction in DNA methylation levels in wild-type plants, and enhanced accumulation of 24-nt siRNAs suppresses this reduction in d2d4 plants. Virus accumulation pattern showed that the virus level reached a peak at 10 dpi and declined thereafter ( Figure 5F), suggesting that amounts of 24-nt siRNAs generated from inducers decline in a similar fashion, which leads to a reduction of DNA methylation. Consistent with this notion, elimination of DNA methylation was not observed throughout the region in three transgenic plants (T0 generation) that constitutively express an inverted repeat sequence of the NbPDSb promoter along with the accumulation of 24-nt siRNAs ( Figure 6). Notably, the amount of 25-50-nt small RNAs, the size of P4RNA suggested to be important for RdDM [36], increased in low methylation regions in d2d4 plants ( Figure 5C; Supplementary Figure S2). This observation suggests that increased 25-50-nt small RNAs enhance de novo methylation or accumulate as precursors for 24-nt RNAs. As 24-nt siRNAs are important for maintaining DNA methylation [23], increasing the amount of 24-nt siRNAs was expected to maintain the elevated DNA methylation levels in our system. We used DCL2 and DCL4-double knock-down (d2d4) N. benthamiana transgenic plants [34], where increases in 24-nt siRNAs are expected by reducing the transcription of 21-nt and 22-nt siRNAs that are produced by DCL4 and DCL2, respectively [35]. To confirm the increase in 24-nt siRNAs generated from inducers expressed from CMV vector in d2d4 plants, we analyzed the pattern of small RNAs produced from the T2s inducer by small RNA sequencing. The results in 21 dpi showed that most siRNAs produced from the T2s inducer in d2d4 plants were 24-nt siRNAs, which accounted for about 74.3% of total 21-nt, 22-nt, and 24-nt siRNAs, in contrast to about 4.9% in wild-type plants (Supplementary Figure S1). CMV/NbPDSb-T2s and CMV/NbPDSb-T3s were inoculated onto wild-type and d2d4 plants, and methylation levels were assessed. At 35dpi, DNA methylation level at NbPDSb was dramatically reduced in wild-type plants but remained high in d2d4 plants ( Figure 5A Figure S2). Virus accumulation in d2d4 plants is not consistently higher compared to wild-type plants at 35 dpi ( Figure 5E). These results suggest that low accumulation of 24-nt siRNAs leads to a reduction in DNA methylation levels in wild-type plants, and enhanced accumulation of 24-nt siRNAs suppresses this reduction in d2d4 plants. Virus accumulation pattern showed that the virus level reached a peak at 10 dpi and declined thereafter ( Figure 5F), suggesting that amounts of 24-nt siRNAs generated from inducers decline in a similar fashion, which leads to a reduction of DNA methylation. Consistent with this notion, elimination of DNA methylation was not observed throughout the region in three transgenic plants (T0 generation) that constitutively express an inverted repeat sequence of the NbPDSb promoter along with the accumulation of 24-nt siRNAs ( Figure 6). Notably, the amount of 25-50-nt small RNAs, the size of P4RNA suggested to be important for RdDM [36], increased in low methylation regions in d2d4 plants ( Figure 5C; Supplementary Figure S2). This observation suggests that increased 25-50-nt small RNAs enhance de novo methylation or accumulate as precursors for 24-nt RNAs.

Sequence Identity May Not Be the Sole Determinant of Efficiency of Silencing and Ind of DNA Methylation of PDS Genes
RdDM induction is homology-dependent, and sequence identity between in and target is a critical factor in determining the induction efficiency of DNA methy We investigated whether sequence identity affects the induction of DNA methylatio silencing the target genes using the closely related two PDS genes (NbPDSa and Nb found in the N. benthamiana genome. Real-time PCR analysis indicated that basal e

Sequence Identity May Not Be the Sole Determinant of Efficiency of Silencing and Induction of DNA Methylation of PDS Genes
RdDM induction is homology-dependent, and sequence identity between inducer and target is a critical factor in determining the induction efficiency of DNA methylation. We investigated whether sequence identity affects the induction of DNA methylation and silencing the target genes using the closely related two PDS genes (NbPDSa and NbPDSb) found in the N. benthamiana genome. Real-time PCR analysis indicated that basal expression levels of NbPDSa were about 4.3× higher than the expression of NbPDSb (Supplementary Figure S3).
NbPDSb-based inducers were used in the above experiments, and each inducer has mismatch sequences with NbPDSa: T2s and T3s have 84.6% and 87.2% nucleotide identities with corresponding regions of NbPDSa, respectively (Supplementary Figure S4). We found that both T2s and T3s efficiently induced DNA methylation in the homologous region of NbPDSa (Figure 7A,B). Further, the DNA methylation level of NbPDSa was lower than that of NbPDSb in all cytosine contexts ( Figure 7A,B). Unexpectedly, regardless of the lower DNA methylation level in NbPDSa, both NbPDSb-based T2s and T3s reduce NbPDSa expression more efficiently compared to NbPDSb expression ( Figure 7C). Also, the time-course analysis indicated that NbPDSa mRNA was constantly reduced compared to NbPDSb mRNA in both inoculated (Supplementary Figure S5A) and upper non-inoculated leaves (Supplementary Figure S5B). These results indicated that NbPDSb-based inducer reduced NbPDSa mRNA more efficiently than NbPDSb mRNA independent of DNA methylation level.  Figure 7C). Also, the time-course analysis indicated that NbPDSa mRNA was constantly reduced compared to NbPDSb mRNA in both inoculated (Supplementary Figure S5A) and upper non-inoculated leaves (Supplementary Figure S5B). These results indicated that NbPDSb-based inducer reduced NbPDSa mRNA more efficiently than NbPDSb mRNA independent of DNA methylation level.   Figure 2B). The black, blue, red, and gray bars indicate CG, CHG, CHH, and total methylation, respectively. (C) mRNA levels of NbPDSa and NbPDSb were separately quantified by real-time PCR in the upper non-inoculated leaves with CMV/NbPDSb-T2s, -T3s, -asGFP (vector control) at 21 days after inoculation. The expression level of vector control was set to 1.0 for NbPDSa and NbPDSb. The expression levels were normalized relative to NbEF1α. The error bars indicate the standard deviation of four biological replicates. Statistical analyses used the Dunnett's method. Data from vector control (asGFP) was used as a control for statistical analysis. **, p < 0.01.
The reciprocal experiment for inducing DNA methylation by the NbPDSa-based inducer (T30s) revealed, by bisulfite sequencing, that NbPDSa-based T30s efficiently induced DNA methylation against the NbPDSb sequence ( Figure 8A,B). The level of DNA methylation in NbPDSb sequences induced by the NbPDSa-based T30s was higher than that induced by the NbPDSb-based T3s in all cytosine contexts. Further, T30s' expression reduced both NbPDSa and NbPDSb mRNA levels more effectively compared to T3s ( Figure 8C), though no significant difference in virus accumulation was seen ( Figure 8D). NbPDSa-based T30s is an effective DNA methylation inducer for NbPDSb regardless of the presence of mismatches, suggesting that sequence identity alone does not determine the efficiency of induction of DNA methylation. The reciprocal experiment for inducing DNA methylation by the NbPDSa-based inducer (T30s) revealed, by bisulfite sequencing, that NbPDSa-based T30s efficiently induced DNA methylation against the NbPDSb sequence ( Figure 8A,B). The level of DNA methylation in NbPDSb sequences induced by the NbPDSa-based T30s was higher than that induced by the NbPDSb-based T3s in all cytosine contexts. Further, T30s' expression reduced both NbPDSa and NbPDSb mRNA levels more effectively compared to T3s (Figure 8C), though no significant difference in virus accumulation was seen ( Figure 8D). NbPDSa-based T30s is an effective DNA methylation inducer for NbPDSb regardless of the presence of mismatches, suggesting that sequence identity alone does not determine the efficiency of induction of DNA methylation.

Discussion
Virus-mediated targeted DNA methylation has significant advantages over transgenic technology, but is to date, poorly characterized. In this work, we reveal rapid induction and gradual reduction of DNA methylation in the promoter sequence of the PDS gene using a CMV vector in N. benthamiana. To the best of our knowledge, this report is the first to capture dynamic changes in DNA methylation in an endogenous genomic region in plants.
Previous work showed that DNA methylation is not induced in the targeted region of the FWA gene in A. thaliana infected with a TRV-carrying target fragment of FWA [23]. The lack of DNA methylation in the FWA gene might be due to intrinsic epigenetic regulation of the FWA gene in A. thaliana. The PDS gene in N. benthamiana is apparently not regulated by DNA methylation, and DNA methylation in this gene would not be influenced by intrinsic regulation. Even so, T2s-induced DNA methylation gradually declined (Figure 4). Small RNA sequencing showed a positive correlation between levels of 24-nt siRNA and DNA methylation. A model from a previous study may provide an explanation [23]. Where initiation and establishment of RdDM are mediated by 21-nt and 22-nt siRNAs, and induced DNA methylation is reinforced and maintained by 24-nt siRNAs: (1) inducer dsRNA generated from the CMV vector are cleaved into 21-nt and 22-nt siRNAs by DCL4 and DCL2, respectively (small amounts of 24-nt siRNAs are also generated by DCL3); (2) 21/22-nt siRNAs are incorporated into ARGONAUTE proteins, and cytosine is methylated via the Pol V pathway; (3) once methylation is established, canonical Pol IV-mediated RdDM machinery is recruited to a target region, and cytosine methylation is reinforced and maintained through 24-nt siRNAs. At 35 dpi in the 3 low methylation region of wild-type plants ( Figure 5A,B), production of 21/22-nt and 24-nt siRNAs was low due to reduced virus accumulation for efficient initiation/establishment and reinforcement/maintenance of DNA methylation, respectively ( Figure 5C,D). In d2d4 plants, high levels of 24-nt siRNAs might maintain DNA methylation ( Figure 5C,D). Notably, 25-50-nt small RNAs may be involved in maintaining DNA methylation in d2d4 plants. Previous reports show that Pol IV generates 25-50-nt small RNAs called P4RNA in A. thaliana [36][37][38][39]. Yang et al. found that DNA methylation at many loci do not require DCLs and proposed that P4RNA itself is a trigger for DNA methylation at the initiation phase. P4RNA may also function in establishment/maintenance, independent of 24-nt siRNAs [36]. As the amount of 25-50-nt RNAs was also positively correlated with the level of DNA methylation ( Figure 5C), these classes of RNAs might also contribute to the maintenance of high methylation in d2d4 plants. The 25-50-nt RNA might provide evidence that canonical RdDM is newly initiated after the DNA is methylated by a virus-mediated Pol IV-independent non-canonical RdDM pathway.
The reduction of DNA methylation was biased at the 3 end of the targeted region of the PDS promoter sequence (Figures 4 and 5A). In the A. thaliana FWA gene, DNA methylation was introduced in progenies from plants infected with TRV, but it was not always introduced throughout the target region. Some plants showed methylation only at the 5 end region near the transcription start site [33]. A recent study demonstrates that artificial site-specific transcriptional activation reduces promoter methylation of the FWA gene in A. thaliana [15]. These reports suggest that transcription by polymerase II perturbs the introduction of DNA methylation near the transcriptional start site (Figure 9). This suggestion is not inconsistent with genome-wide analyses that show low DNA methylation levels in this region [40]. RNA-seq data available in the Sol Genomics Network database [32] indicate an accumulation of the PDS gene transcripts with a 5 end within T2s and T3s-targeted regions in N. benthamiana. Thus, some transcription could start within the target region and reduce DNA methylation in wild-type plants, and this reduction might be countered by maintenance mechanisms involving highly accumulated 24-nt siRNAs and/or 25-50-nt RNAs in d2d4 plants. Importantly, demethylases, such as ROS1, might also be involved in the removal of DNA methylation (Figure 9), which is consistent with previous work showing CMV-mediated targeted methylation against a 35S promoter. This methylation was enhanced by down-regulating ROS1 expression in N. benthamiana [41]. Thus, strong maintenance activity of RdDM is required to introduce stable DNA methylation around transcriptional start sites. targeted regions in N. benthamiana. Thus, some transcription could start within the target region and reduce DNA methylation in wild-type plants, and this reduction might be countered by maintenance mechanisms involving highly accumulated 24-nt siRNAs and/or 25-50-nt RNAs in d2d4 plants. Importantly, demethylases, such as ROS1, might also be involved in the removal of DNA methylation (Figure 9), which is consistent with previous work showing CMV-mediated targeted methylation against a 35S promoter. This methylation was enhanced by down-regulating ROS1 expression in N. benthamiana [41]. Thus, strong maintenance activity of RdDM is required to introduce stable DNA methylation around transcriptional start sites. Figure 9. Model for induction and reduction of DNA methylation by the CMV vector in the N. benthamiana PDS gene. The CMV vector produces dsRNAs against a target sequence, which are processed into 21-nt, 22-nt, and 24-nt siRNAs by DCL4, DCL2, and DCL3 proteins, respectively. At the early stages of infection (until 21dpi) in wild-type plants, DNA methylation is initiated and established through highly accumulated 21-/22-siRNAs (step1) and maintained via 24-nt siRNAs (step2). In d2d4 plants, though the level of 21-nt and 22-nt siRNAs are low, highly accumulated 24-nt siRNAs effectively maintain high DNA methylation states. At later stages of infection (after 21dpi) in wild-type plants, the introduced DNA methylation is gradually eliminated, presumably due to the reduction in accumulated small RNAs. In contrast, in d2d4 plants, highly accumulated 24-nt siRNAs maintain a high methylation state. The reduction of DNA methylation may be caused by competition between RdDM machinery and polymerase II transcription. Alternatively, but not exclusively, DNA demethylase may be preferentially recruited to the 3′ end of the target region by an unknown mechanism and erase DNA methylation because the DNA methylation appears to be erased from the 3' end.
The expression of the NbPDSb-based inducer efficiently introduced DNA methylation at the homologous sequence of NbPDSa, and vice versa ( Figures 7A,B and 8A,B). . Model for induction and reduction of DNA methylation by the CMV vector in the N. benthamiana PDS gene. The CMV vector produces dsRNAs against a target sequence, which are processed into 21-nt, 22-nt, and 24-nt siRNAs by DCL4, DCL2, and DCL3 proteins, respectively. At the early stages of infection (until 21dpi) in wild-type plants, DNA methylation is initiated and established through highly accumulated 21-/22-siRNAs (step1) and maintained via 24-nt siRNAs (step2). In d2d4 plants, though the level of 21-nt and 22-nt siRNAs are low, highly accumulated 24-nt siR-NAs effectively maintain high DNA methylation states. At later stages of infection (after 21dpi) in wild-type plants, the introduced DNA methylation is gradually eliminated, presumably due to the reduction in accumulated small RNAs. In contrast, in d2d4 plants, highly accumulated 24-nt siRNAs maintain a high methylation state. The reduction of DNA methylation may be caused by competition between RdDM machinery and polymerase II transcription. Alternatively, but not exclusively, DNA demethylase may be preferentially recruited to the 3 end of the target region by an unknown mechanism and erase DNA methylation because the DNA methylation appears to be erased from the 3 end.
The expression of the NbPDSb-based inducer efficiently introduced DNA methylation at the homologous sequence of NbPDSa, and vice versa ( Figure 7A,B and Figure 8A,B). Many small RNAs generated from NbPDSaor NbPDSb-based inducers are imperfectly matched with NbPDSb or NbPDSa sequence, respectively. Thus, perfectly matched siR-NAs might direct DNA methylation, and some mismatches might also share this activity. Contrary to our notion that methylation is more effectively induced by perfectly matched inducers than inducers that have mismatches, the expression of NbPDSa-based the T30s inducer introduced DNA methylation in the homologous region of NbPDSb more efficiently than NbPDSb-based T3s ( Figure 8B). These results suggest that the sequence identity of inducers is not the sole determinant of DNA methylation efficiency. Future detailed analyses of the efficiency of small RNA generation from inducers, sequence composition, and siRNAs responsible for methylation of specific sites will identify factors that determine efficiency and specificity. Further, regardless of similar or higher levels of DNA methylation in NbPDSb than in NbPDSa (Figures 7B and 8B), the degree of reduction in NbPDSb mRNA level was smaller than that for NbPDSa mRNA (Figures 7C and 8C). One explanation might be that the degree of reduction was small because of lower basal expression levels of NbPDSb (Supplementary Figure S3). Alternatively, regulatory regions required for transcriptional suppression might differ between NbPDSa and NbPDSb.
In this work, targeted DNA methylation induced by a CMV vector was shown to dynamically change the methylation of an endogenous gene promoter. Additionally, the introduction of DNA methylation against homeologs in N. benthamiana was characterized. The findings contribute to the development of efficient targeted DNA methylation and the understanding of epigenetic regulation of polyploid genomes. Targeted DNA methylation by a CMV vector will be a powerful tool to elucidate readily and rapidly functions of specific methylation in a biological context and create plants with altered expression of target genes. The latter development would allow the expression of desirable agronomical traits without changing DNA sequences.

Plant Growth
N. benthamiana was grown in soil or hydroponically in a nutrient solution (Otsuka hydroponic composition, Otsuka Chemical Co., Ltd., Osaka, Japan) at 23 • C-25 • C and a 16 h light/8h dark cycle as described earlier [42].

Preparation of Plasmids
For promoter analysis, a genomic fragment of −1225 to −1 from the translational start site of NbPDSb was amplified by PCR and exchanged with the 35S promoter upstream of the GUS gene in pBE2113 (pBE2113/NbPDSp::GUS). For targeted DNA methylation with the CMV vector, T2, T3, T30, and asGFP fragments were amplified by PCR and cloned into CMV2 A1 [20] at StuI and MluI sites using primers described in Supplementary Table S1. The "s" is added as a suffix of each inducer indicates strands coding sense transcripts of the NbPDS gene, and "as" is a complementary strand of "s." For quantification of mRNA levels of NbPDSa and NbPDSb, fused NbPDSa and NbPDSb fragments containing amplified region by real-time PCR, and the fragment was cloned into pCR4 (pCR4/NbPDSab-st) according to manufacturer's instructions. Fragments containing the amplified region of NbEF1α by real-time PCR were also cloned into the pCR4 vector (pCR4/NbEF1α-st) as an endogenous reference. Primer information is described in Supplementary Table S1.

Virus Inoculation
CMV inoculation was conducted following Kanazawa et al. [21]. CMV RNAs 1-3 were obtained by in vitro transcription using linearized plasmids containing cDNAs of Basal expression levels of NbPDSa and NbPDSb were quantified by real-time PCR using standard plasmids, pCR4/NbPDSab-st for NbPDSa and NbPDSb, pCR4/NbEF1α-st for NbEF1α. NbPDSa and NbPDSb expression levels were normalized to expression levels of NbEF1α.

Conclusions
In this study, we demonstrated that CMV vector-induced DNA methylation in the promoter region of the PDS gene was dynamically changed, and strikingly, the methylation level gradually decreased from the 3 end of the targeted region, which are blocked by increasing 24-nt siRNAs by down-regulating the DCL2 and DCL4 expressions in N. benthamiana. We also showed that DNA methylation was induced in the region that contains mismatched sequences to expressed RNAs, suggesting that sequence identity may not be the sole determinant of the efficiency of DNA methylation. These findings contribute to the development of efficient targeted DNA methylation and understanding of epigenetic regulation of polyploid genomes.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/ijms22084125/s1, Figure S1: Small RNA profiles in wild-type and d2d4 plants infected with CMV vector carrying inducer for DNA methylation; Figure S2: Small RNA profiles in wildtype and d2d4 plants infected with CMV vector carrying inducer for DNA methylation.; Figure S3: Small RNA profiles in wild-type and d2d4 plants infected with CMV vector carrying inducer for DNA methylation.; Figure S4: Sequence alignment of nucleotide sequences of region upstream of translational start.; Figure S5: Time-course analysis of NbPDSa and NbPDSb expressions in plants infected with CMV vector carrying inducer for DNA methylation.; Table S1: Primers used in this study.; Table S2: Primers used in this study.  Data Availability Statement: The GenBank/ENA/DDBJ accession numbers for coding sequences of NbPDSa (Niben101Scf01283g02002.1) and NbPDSb (Niben101Scf14708g00023.1) are LC543532 and LC543533, respectively. Accession numbers for genomic regions of NbPDSa (−3260 to +88 from translational start site), NbPDSb (−3322 to +94) are LC543534 and LC543535, respectively. Sequencing data have been deposited in the DDBJ Sequenced Read Archive under the accession number DRA010371. These sequences and sequencing data are available at the DNA Data Bank of Japan (DDBJ) (https://www.ddbj.nig.ac.jp/ddbj/index-e.html).