miRNAs trigger widespread epigenetically activated siRNAs from transposons in Arabidopsis

Abstract

In plants, post-transcriptional gene silencing (PTGS) is mediated by DICER-LIKE 1 (DCL1)-dependent microRNAs (miRNAs), which also trigger 21-nucleotide secondary short interfering RNAs (siRNAs) via RNA-DEPENDENT RNA POLYMERASE 6 (RDR6), DCL4 and ARGONAUTE 1 (AGO1), whereas transcriptional gene silencing (TGS) of transposons is mediated by 24-nucleotide heterochromatic (het)siRNAs, RDR2, DCL3 and AGO4 (ref. 4). Transposons can also give rise to abundant 21-nucleotide 'epigenetically activated' small interfering RNAs (easiRNAs) in DECREASED DNA METHYLATION 1 (ddm1) and DNA METHYLTRANSFERASE 1 (met1) mutants, as well as in the vegetative nucleus of pollen grains and in dedifferentiated plant cell cultures. Here we show that easiRNAs in Arabidopsis thaliana resemble secondary siRNAs, in that thousands of transposon transcripts are specifically targeted by more than 50 miRNAs for cleavage and processing by RDR6. Loss of RDR6, DCL4 or DCL1 in a ddm1 background results in loss of 21-nucleotide easiRNAs and severe infertility, but 24-nucleotide hetsiRNAs are partially restored, supporting an antagonistic relationship between PTGS and TGS. Thus miRNA-directed easiRNA biogenesis is a latent mechanism that specifically targets transposon transcripts, but only when they are epigenetically reactivated during reprogramming of the germ line. This ancient recognition mechanism may have been retained both by transposons to evade long-term heterochromatic silencing and by their hosts for genome defence.

Figure 1. miRNAs trigger RDR6-dependent 21-nt epigenetically activated (ea)siRNA biogenesis from reactivated transposons

Whole-genome representation illustrating miRNAs triggering widespread easiRNA biogenesis at transposons in Arabidopsis. Outermost to innermost tracks depicting: miRNA abundance in ddm1-2 (Histogram); Arabidopsis Chromosome I – V, pericentromeric region (black) (Ideogram); Gene and transposon frequency, low density (red), high density (blue) (Heat-map); Retrotransposon derived 21-nt sRNAs (dark red = unique, light red = multiple mapping) and DNA transposon-derived 21-nt sRNAs (dark blue = unique, light blue = multiple mapping) 21-nt siRNAs in order of Col-0, ddm1-2, ddm1-2 rdr6-15, and ddm1-2 dcl1-11 (Histogram); miRNAs targeting transposons (Connectors); miR859a (Chr I, red), miR390a (Chr II, blue), miR172d (Chr III, red), eamiR2 ATHILAIV (Chr IV, grey) and miR172e (Chr V, blue). Transposons are post-transcriptionally targeted by 50 known miRNAs and newly discovered eamiRNAs (Table 1) giving rise to abundant RDR6-dependent 21-nt easiRNAs at transposons in Arabidopsis. (Refer to Extended Data Fig. 1; Supplementary Table 1; 3).

Figure 2. miRNA cleavage at ATHILA ORF1 and ATCOPIA43

(a, b) 21-nt easiRNAs at ATHILA ORF1 (AT2G10280), in comparison to non-easiRNA generating ATCOPIA43 (AT1G36040) in track order Col-0 (i), ddm1-2 (ii), rdr6-15 (iii), ddm1-2 rdr6-15 (iv) and ddm1-2 dcl1-11 (v). (c, d) Parallel Analysis of RNA Ends (PARE) at ATHILA ORF1 and ATCOPIA43 in track order Col-0 (i), ddm1-2 (ii) and ddm1-2 rdr6-15 (iv). Individual reads, sense = red, antisense = blue. (e, f) 5′ RLM RACE-PCR products (arrows; black bars in a, b) from ATHILA ORF1 and ATCOPIA43 corresponding to cleaved mRNA fractionated by gel electrophoresis. Inset: miRNA alignment with 5′ RLM RACE-PCR cleavage sites, and frequencies of cleavage products indicated as fractions.

Figure 3. miR845b targets AtGP1 promoting easiRNA biogenesis

(a) Transgenic line over-expressing AtGP1 containing the 20-nt miR845b target region (marked miR845) and novel overlapping 21-nt easiRNAs (star). Individual reads, sense = red, antisense = blue. (b) Transgenic line over-expressing AtGP1 that did not contain the 20-nt miR845b predicted target region, with no 21-nt easiRNAs overlapping this site. (c) Frequencies of 5′ RLM RACE-PCR miR845b AtGP1 cleavage products indicated as fractions.

Figure 4. Transposons targeted by miRNA, gain hetsiRNA when the easiRNA pathway is lost

(a) Overlap of transposons targeted by hetsiRNAs in Col-0, ddm1-2 and ddm1-2 rdr6-15. (b) Overlap of transposons targeted by miRNAs, that have lost hetsiRNAs in ddm1-2, gained hetsiRNAs in ddm1-2 rdr6-15, and those whose transcripts undergo productive miRNA cleavage resulting in easiRNA biogenesis in ddm1-2. 21-nt easiRNAs, 24-nt hetsiRNAs and PARE degradome at easiRNA-generating (c) ATHILA (AT3G32118) in comparison to non-easiRNA generating (d) ATENSPM1 (AT4G02314). Individual reads, sense = red, antisense = blue in track order Col-0 (i), ddm1-2 (ii) rdr6-15 (iii), and ddm1-2 rdr6-15 (iv).

Extended Data Figure 1. 21-nt easiRNAs originate from transposons in ddm1 and are miRNA and RDR6-dependent

(a) 18-nt to 26-nt small RNA abundance in Col-0, ddm1-2, and ddm1-2 rdr6-15. Normalized reads per million (RPM). (b) 18-nt to 26-nt small RNA abundance in Col-0, ddm1-2, and ddm1-2 dcl1-11. Normalized reads per million (RPM).

Extended Data Figure 2. miRNA target genes and transposons that do not promote tasiRNA nor easiRNA, respectively, have degradation covering the entire region

Read pattern distribution of 21-nt unique reads (represented as a histogram of read density (grey bars)) and PARE signatures at (a) TAS2 (AT2G39681), (b) RCC (AT3G02300), SEP2 (AT3G02310), (c) MEE58 (AT4G13940), and transposons (d) ATENSPM6 (AT2G06720), (e) ATLINE1_4 (AT2G15540), (f) ATCOPIA43 (AT3G0410), in track order Col-0, ddm1-2, rdr6-15, ddm1-2 rdr6-15 and ddm1-2 dcl1-11.

Extended Data Figure 3. Relationship between DNA methylation, easiRNA and hetsiRNA at transposons that miRNA are predicted to target

DNA methylation, by CG (red), CHG (blue) and CHH+/− context (green) cytosine context (scale 1 = methylated cytosine, 0 = unmethylated cytosine) at (a) ATCOPIA93 (AT5G17125), (b) ATMU5 (AT4G08680) DNA transposon and the surrounding region, and (c) ATHILA6A (AT4TE15030) retrotransposon and the surrounding region, in Col-0, ddm1-2, rdr6-15 and ddm1-2 rdr6-15 (i). 21-nt and 24-nt siRNAs, represented as a histogram of read density (grey bars) in track order Col-0, ddm1-2, rdr6-15 and ddm1-2 rdr6-15 (ii). Key for sRNA reads, unique (U) mapping to one location of the genome, and multiple (M) mapping to more than one location of the genome. PARE read density in Col-0 and ddm1-2 (iii).

Extended Data Figure 4. Epigenetically activated (ea)miRNA immature precursor sequence and predicted structure

Epigenetically activated (ea)miRNAs immature precursor sequences have methylated cytosines (*) in Col-0, that are unmethylated in ddm1-2. Mature miRNA are underlined, and the putative stem-loop structures of the precursors are illustrated.

Extended Data Figure 5. 24-nt hetsiRNAs at transposons in Col-0, are lost in ddm1, and gained in ddm1 rdr6 and ddm1 dcl1

24-nt hetsiRNAs by transposon class in Col-0, ddm1-2, rdr6-15, ddm1-2 rdr6-15, and ddm1-2 dcl1-11. Normalized reads per million (RPM).

Extended Data Figure 6. Overlap of TEs that undergo easiRNA biogenesis, hetsiRNA loss and miRNAs targeting

Individual transposons were grouped depending on small RNA abundance in each genotype. (a) TEs that lose 24-nt hetsiRNAs in ddm1-2, gain 21-nt easiRNAs in ddm1-2 overlap with those that gain 24-nt hetsiRNAs in ddm1-2 rdr6-15. (b) TEs that are targeted and cleaved by miRNAs overlap with those that gain 21-nt easiRNAs in ddm1-2 and 24-nt hetsiRNAs in ddm1-2 rdr6-15. (c) TEs that are targeted and cleaved by two or more miRNAs overlap with those that gain 21-nt easiRNAs in ddm1-2, and those that gain 24-nt hetsiRNAs in ddm1-5 rdr6-15. (d) TEs that are predicted to be targeted by miRNAs, but without supporting PARE cleavage data, also overlap with those that gain 21-nt easiRNAs in ddm1-2 and 24-nt hetsiRNAs in ddm1-2 rdr6-15.

Extended Data Figure 7. Loss of methylation at transposons in ddm1 is partially restored in ddm1 rdr6

(a) Transposon methylation in Col-0, ddm1-2, rdr6-15 and ddm1-2 rdr6-15 replicates. Scale 1 = methylated cytosine, 0 = unmethylated cytosine. Total DNA methylation at transposons, grouped by superfamily in (b) Col-0, (c) ddm1-2, (d) rdr6-15 and (e, f) ddm1-2 rdr6-15 replicates. Key (1) LTR Retrotransposons ATGYPSY, (2) LTR Retrotransposons ATCOPIA, (3) NonLTR Retrotransposons ATLINE, (4) nonLTR Retrotransposons TSCL, (5) TIR DNA Transposons MuDR, (6) non-TIR DNA Transposons MuDR, (7) DNA Transposons EnSpM, (8) DNA Transposons Helitron and (9) Other Repeats. Total methylation by total converted cytosine to thymine and non-converted cytosine counts (at least 10 reads per cytosine). Scale 0 = unmethylated, 1 = methylated. Boxplots indicate median, range and standard deviations (box). DNA methylation, by CG (red), CHG (blue) and CHH+/− context (Green) at (g) ATHILA ORF1 (AT2G10280) and (h) ATCOPIA43 (AT1G36040). Track order Col-0 (i), ddm1-2 (ii), rdr6-15 (iii) and ddm1-2 rdr6-15 (iv).

Extended Data Figure 8. Transposon transcript abundance in ddm1-2 and ddm1 rdr6

(a, b, c) Ath1 Affymetrix microarray expression (log2 signal intensity) in Col-0 in comparison to ddm1-2, rdr6-15 and ddm1-2 rdr6-15. (d) TEs upregulated in ddm1-2 were not further upregulated in ddm1-2 rdr6-15. Key: red = transposons, black = genes.

Extended Data Figure 9. miRNA-directed easiRNA biogenesis from activated transposons

When TEs are epigenetically activated, through the loss of DNA methylation and/or heterochromatin, transposon mRNA transcripts become preferentially targeted by miRNAs (DCL1-dependent) bound by AGO1. Productive cleavage of transposon transcripts engages RDR6 and DCL4, which generate 21-nt epigenetically-activated (ea)siRNAs from transposon open reading frames, in a post-transcriptional gene silencing (PTGS) mechanism, that are then loaded into AGO1, and thus, prevents engagement of RDR2 and RdDM. This antagonism accounts for the retention of miRNA binding sites by transposons, to evade long-term heritable silencing, elicited by DNA methylation via RDR2. This model also accounts for the retention of the miRNA-directed mechanism by the host organism, in order to generate easiRNAs to silence TEs when they are epigenetically reprogrammed in the germline.

Extended Data Figure 10. Arabidopsis miRNAs target transposons

Known Arabidopsis miRNAs, and novel epigenetically-activated (ea)miRNAs that arise in ddm1-2, are predicted to target transposon transcripts and confirmed to cleave transposon transcripts by PARE (Supplementary Table 3; Supplementary Methods). eamiRNAs, some known to be developmentally regulated, and TE-derived eamiRNAs that target specific transposon families. Transposons are identified by EVRY TE identifier (Supplementary Table 2, for further annotation refer to The Arabidopsis Information (TAIR10) annotation ORF ID). Transposon transcripts giving rise to 21-nt easiRNAs (bold); those that are targeted by multiple miRNA (*); and, those miRNAs that target multiple transposons of the same family (italics) are highlighted.