A two-step process for epigenetic inheritance in Arabidopsis

Abstract

In Arabidopsis, multisubunit RNA polymerases IV and V orchestrate RNA-directed DNA methylation (RdDM) and transcriptional silencing, but what identifies the loci to be silenced is unclear. We show that heritable silent locus identity at a specific subset of RdDM targets requires HISTONE DEACETYLASE 6 (HDA6) acting upstream of Pol IV recruitment and siRNA biogenesis. At these loci, epigenetic memory conferring silent locus identity is erased in hda6 mutants such that restoration of HDA6 activity cannot restore siRNA biogenesis or silencing. Silent locus identity is similarly lost in mutants for the cytosine maintenance methyltransferase, MET1. By contrast, pol IV or pol V mutants disrupt silencing without erasing silent locus identity, allowing restoration of Pol IV or Pol V function to restore silencing. Collectively, these observations indicate that silent locus specification and silencing are separable steps that together account for epigenetic inheritance of the silenced state.

Figure 1. Functional relationships among HDA6, Pol IV and Pol V in transcriptional silencing

(A) Hierarchical clustering of 99 transposable element families based on polyA+ RNA-seq read counts. The single mutants hda6-7, nrpd1-3 (pol IV) or nrpe1-11 (pol V), and the double mutants hda6 pol IV and hda6 pol V are compared to wild-type (Col-0). Expression levels are normalized relative to the total number of mapped reads per genotype. (B) Derepressed transposable element (TE) Superfamilies. Families were considered “derepressed” if at least 40 polyA+ RNA-seq reads mapped to a family in the given mutant and fewer than 20 reads were detected in wild-type Col-0. (C) Venn diagrams summarizing locus derepression in single and double mutants of hda6, pol IV and pol V. Derepressed loci were determined using NOISeq (Tarazona et al., 2012) by comparing polyA+ RNA-seq read counts in the mutant relative to Col-0. A minimum 4-fold increase in the mutant relative to wild-type, and p value < 0.01 were set as thresholds. (D) Relative abundance of genes versus transposons among Group E and Group DL loci detected in hda6 pol IV (left) or hda6 pol V (right) double mutants. (E) RT-PCR analyses of Group DL and Group E loci in wild-type (Col-0) and the indicated mutant backgrounds. Actin serves as a loading control. Pol II* denotes nrpb2-3, a mutant allele of the Pol II second-largest subunit that affects RdDM at some loci (Zheng et al., 2009). Controls omitting reverse transcriptase (−RT) were performed for all loci; no products were obtained, as shown for the Actin control (See also Figure S1).

Figure 2. Molecular and phenotypic consequences of “double-locked” silencing involving HDA6 and Pol IV/V

(A) AtGP5 retrotransposon RNA-seq data. Long terminal repeats (LTRs) flank the transposable element body (see diagram). Vertical black bars denote polyA+ RNA read numbers (log₂ units), normalized to total mapped reads, in wild-type Col-0, homozygous hda6-7, pol IV and pol V mutants and the double mutants, hda6 pol IV and hda6 pol V. Numbers at right are RPKM (reads per kilobase per million) log-ratios, comparing expression in each mutant to Col-0. (B) Phenotypes of 21-day-old plants: wild-type Col-0, or the indicated mutants (see also Figure S2). (C) Diagram of the SDC locus (At2g17690), showing upstream tandem repeats (purple arrows), the Pol II transcription start site (+1), and PolyA+ RNA-seq data plotted on a log₂ scale (black). PCR amplicons for chromatin immunoprecipitation (ChIP) and bisulfite sequencing (black and red arrows, respectively) are indicated. (D) Bisulfite sequencing analysis of DNA methylation upstream of +1 (red amplicon, panel C). Data from triple mutant drm1 drm2 cmt3 serves as a hypomethylated control (Henderson and Jacobsen, 2008). (E) ChIP analysis of Histone 3 Lysine 9/14 acetylation at three positions in the SDC locus. Error bars show the propagated standard error of the mean from three replicates.

Figure 3. Epistasis of HDA6 and Pol IV/V at a Group E locus

(A) Diagram of the ERT9 gene (At5g24240) and adjacent annotated loci, At5g24250 and At5g24260. Arrows mark PCR amplicons used for RT-PCR, bisulfite sequencing or ChIP analyses. Data tracks show polyA+ RNA read counts (log₂ units) in wild-type Col-0, hda6-7, pol IV, pol V and the double mutants hda6 pol IV and hda6 pol V. Numbers to the left are log-ratios, comparing expression in each mutant to Col-0. Grey shading represents RNA-seq reads mapped to the 3′-terminal exons of At5g24260. (B) RT-PCR verification of ERT9 gene derepression in the mutant hda6-7 (intervals b, c, d), and lack of detectable At5g24250 expression (interval a). (C) ChIP analysis of Histone 3 Lysine 9/14 acetylation at the ERT9 promoter region. Error bars show the propagated standard error of the mean from three replicates. (D) Bisulfite sequencing analysis of DNA methylation in the interval −131 to +190 relative to the ERT9 transcription start site (+1) (see also Figure S3).

Figure 4. HDA6 facilitates Pol IV-dependent siRNA biogenesis

(A) RNA-seq profiles at three Group E targets. Data tracks show polyA+ RNA reads (black vertical bars) and small RNA reads (red and yellow vertical bars; yellow denotes positions where siRNA and poly A+ reads overlap) in wild-type (Col-0), hda6-7, nrpd1-3 (pol IV), and the double mutant, hda6 pol IV. Read counts (log₂ units) were normalized by total mapped reads. Analogous data for Group DL loci are in Figure S4. (B) Scatter plot showing the relative dependence of 24 nt siRNA production on Pol IV and HDA6, genome-wide. Each data point represents a TAIR10 gene locus, with the x-axis indicating the log-ratio of small RNA read density in Col-0 relative to hda6-7, and the y-axis indicating the log-ratio of small RNA read density in Col-0 relative to pol IV. Group E loci are marked in red.

Figure 5. HDA6 rescue fails to restore silent locus identity at Group E loci

(A) Agrobacterium-mediated transformation of the hda6-6 null mutant with a transgene expressing HDA6 fused to a FLAG epitope (HDA6-F). HDA6 was detected by immunoblotting using an antibody raised against a C-terminal peptide. (B) RT-PCR assays for Group DL and Group E locus expression in Col-0 or hda6-6, nrpd1 (pol IV) or nrpe1 (pol V) mutants, or in lines in which these mutants were rescued by corresponding full-length, FLAG-tagged transgenes (HDA6-F, NRPD1-F, or NRPE1-F). Actin serves as a loading control. Controls omitting reverse transcriptase (−RT) produced no amplification products, as shown for Actin (see also Figure S5). (C) Quantitative droplet digital RT-PCR assays of SDC (Group DL) and ERT7 (Group E) expression. Transcript levels are reported relative to ACT2. Controls omitting RT yielded almost no positive droplets. (D) H3K9/14 acetylation measured by chromatin immunoprecipitation (ChIP). Percentage enrichment was determined by real-time PCR analysis of immunoprecipitated DNA compared to input chromatin. Error bars show the propagated standard error of the mean for three replicates. (E) Chop-PCR assay assessing DNA methylation by digesting genomic DNA with methylation-sensitive restriction endonucleases prior to PCR amplification using flanking primers. The endonucleases report on cytosine methylation in different sequence contexts: HpaII (CG), MspI (CHG), HaeIII (CHH) or AluI (CHH), where H is any base other than G. See also Figures S5 and S6.

Figure 6. Epigenetic memory loss results in failure to produce Pol IV-dependent siRNAs

(A) Abundance of 24 nt siRNAs mapping to chromosomes 3 and 5 in wild-type (Col-0), hda6-6 and HDA6-FLAG rescued hda6-6 plants. Small RNA clusters are plotted using a 200-kb sliding window stepped in 100-kb increments. Black squares mark the estimated positions of centromeres. (B) Relative abundance of siRNAs of different size classes mapping to Group DL loci (left panel) and a browser view of siRNA abundance and distribution at the soloLTR locus (right panel). (C) Relative abundance of siRNAs of different size classes mapping to Group E loci (left panel) and a browser view of siRNA profiles at ERT7 (right panel). Supporting polyA+ RNA-seq data and gene annotations are in Figures 4A and S4. (D) Boxplot analysis of 21 nt miRNAs, revealing no significant overall change in miRNA abundance in hda6 or hda6 HDA6-F rescue lines relative to wild-type Col-0. (E) Boxplot analysis of the effects of HDA6 mutation, and HDA6 rescue, on 24 nt siRNAs that are either SHH1-independent or SHH1-dependent (as defined in Law et. al. 2013). All read counts are normalized to total mapped read numbers. Asterisks above boxplots in panel E indicate significant reduction relative to Col-0 (p < 0.002, Wilcoxon rank-sum test). See also Figure S7.

Figure 7. HDA6, MET1, SHH1 and Pol IV collaborate in gene silencing

(A) 24 nt siRNAs at Group E loci are reduced in met1 mutants. Red vertical bars depict the abundance of siRNA reads (log₂ units) detected in Col-0 or met1-3 null mutants in the datasets of Lister et al. 2008, normalized relative to the number of mapped reads. (B) Boxplot analysis of the MET1-dependence of SHH1-dependent siRNA clusters (*indicates significant reduction; p < 0.002, Wilcoxon rank-sum test). (C) Erasure of silent locus identity occurs in met1 mutants, as in hda6 mutants. The histograms show the expression levels of the Group E loci, ERT7 and ERT9 in wild-type Col-0 (+/+), homozygous mutants (−/−) for hda6-7 (a null mutant), met1-7 (a null mutant) or met1-1 (a partial loss-of-function, or hypomorphic, mutant), or in F1 heterozygous progeny (+/−) that result from outcrossing the homozygous mutants to a wild-type plant. Quantitative droplet digital RT-PCR was used to quantify ERT7 and ERT9 expression levels (on a Log₁₀ scale) relative to an ACT2 control. (D) Two-step model for epigenetic inheritance in which silent locus identity at Group E loci is heritable and is a prerequisite for, but separable from, silencing by RNA-directed DNA methylation. Following replication of fully CG-methylated DNA, hemi-methylated DNA is assembled into nucleosomes by chromatin assembly factors that use acetylated histones as substrates. De-acetylation of these histones by HDA6 may be a prerequisite for CG methylation maintenance by MET1. HDA6-dependent CG maintenance methylation perpetuates the heritable epigenetic memory accounting for silent locus identity. Subsequent recruitment of Pol IV, assisted by SHH1 or other chromatin binding partners, brings about silencing of the loci by RdDM in each new generation.