Cotranscriptional Role for Arabidopsis DICER-LIKE 4 in Transcription Termination

See allHide authors and affiliations

Science  30 Mar 2012:
Vol. 335, Issue 6076, pp. 1621-1623
DOI: 10.1126/science.1214402


Transcription termination is emerging as an important component of gene regulation necessary to partition the genome and minimize transcriptional interference. We have discovered a role for the Arabidopsis RNA silencing enzyme DICER-LIKE 4 (DCL4) in transcription termination of an endogenous Arabidopsis gene, FCA. DCL4 directly associates with FCA chromatin in the 3′ region and promotes cleavage of the nascent transcript in a domain downstream of the canonical polyA site. In a dcl4 mutant, the resulting transcriptional read-through triggers an RNA interference–mediated gene silencing of a transgene containing the same 3′ region. We conclude that DCL4 promotes transcription termination of the Arabidopsis FCA gene, reducing the amount of aberrant RNA produced from the locus.

In eukaryotes, termination of RNA polymerase II (Pol II) activity is highly coordinated with RNA 3′ processing and cotranscriptional degradation of nascent RNA (1). Cleavage of the nascent transcript at the polyadenylation site produces a entry point for an exoribonuclease, which degrades the Pol II–associated RNA and displaces the Pol II from the template DNA (2, 3). In cases where RNA 3′ processing factors fail to cleave at the polyadenylation site, other endoribonucleases supply the exoribonuclease entry point further downstream (4, 5). Dicer proteins are one class of endoribonuclease identified through their production of small RNAs in the size range of 21 to 24 nucleotides (nt). These are often derived from RNA hairpin structures or long double-stranded RNAs (dsRNAs) from repeated sequences and multicopy transposons (6). In Arabidopsis, four different DICER-LIKE proteins (DCL1 to 4) have been associated with gene-silencing mechanisms via microRNA, small interfering RNA (siRNA), and trans-acting siRNA (tasiRNA) action (7). However, it is becoming clear that Dicer-dependent mechanisms are also important for regulation of endogenous single-copy genes (8, 9).

The nuclear-localized RNA binding protein FCA induces chromatin silencing of the Arabidopsis flowering-time regulator FLOWERING LOCUS C (FLC). FCA alters poly (A) site usage of FLC antisense transcripts, which, in turn, induces H3K4me2 demethylase activity and reduced FLC expression (10, 11). In a suppressor screen (using a genotype overexpressing FCA), we identified a mutant (sof59) that increases levels of FLC (Fig. 1A) through reduction of expression of the 35S::FCAγ transgene (Fig. 1B) (12). We found the causative mutation to be a premature stop codon in DCL4 (Fig. 1C). Complementation of the mutant with a wild-type (WT) DCL4 transgene fully restored expression of 35S::FCAγ (Fig. 1B). We named this mutant allele of DCL4 in Ler (Landsberg erecta) background as dcl4-15.

Fig. 1

Identification of DCL4 as a regulator of FCA expression. (A) FLC::LUC signal in sof59 and its parental line C2. (B) Expression of 35S::FCAγ transgene is reduced in sof59 and is rescued by reintroduction of DCL4. Total seedling RNA hybridized with single-stranded antisense probe from 35S::FCAγ. The asterisk indicates read-through transcripts. Blots reprobed with β-TUBULIN (β-TUB) are shown as loading control. (C) Schematic representation of DCL4 and the position of the mutation.

In Arabidopsis, DCL4 is required for the biogenesis of tasiRNAs. tasiRNA generation involves RNA DEPENDENT RNA POLYMERASE 6 (RDR6) and SUPPRESSOR OF GENE SILENCING 3 (SGS3) (13, 14). Expression of the 35S::FCAγ transgene was not affected by rdr6 or sgs3 mutations (fig. S1), arguing that reduced expression of the 35S::FCAγ transgene in dcl4 was not due to loss of tasiRNA or related developmental defects.

The requirement of DCL4 to maintain expression of the 35S::FCAγ transgene runs counter to its role in gene silencing. The 35S::FCAγ transgene is composed of the strong 35S promoter, FCAγ cDNA, and ~420 base pairs (bp) of the 3′ region from the endogenous FCA gene as the 3′ termination region (fig. S2) (15). We detected polyadenylated read-through transcripts from this transgene (Fig. 1B); they either terminated in the vector sequence 861 or 947 nt downstream of the major canonical FCA polyadenylation site or were spliced and terminated a further 1000 nt downstream, within the terminator sequence of the convergently transcribed adjacent kanamycin-resistance transgene NPTII (fig. S2). These data suggested that the FCA 3′ sequences terminate transcription inefficiently.

We also observed transcriptional read-through for the endogenous FCA gene in multiple dcl4 mutant alleles: dcl4-15, dcl4-2t, dcl4-3, and dcl4-4, but not in the weak allele dcl4-6 (Fig. 2B and figs. S3 to S5) (16), and this was associated with increased Pol II occupancy in this region (Fig. 2C). Transcription termination within Schizosaccharomyces pombe centromeric repeats also depends on Dicer function (17). To further confirm a direct role for DCL4 in transcription termination, we performed nuclear run-on experiments. This analysis showed a clear increase in transcriptional read-through into the region downstream of the FCA polyadenylation site in dcl4 compared with the wild type (Fig. 2D). The read-through transcripts overlap with the adjacent convergently transcribed gene At4g16270 (fig. S3). This overlapping convergent transcription did not seem to trigger a sense/antisense pairing and destabilization mechanism as quantitative strand-specific reverse transcription polymerase chain reaction (qRT-PCR) showed that expression of At4g16270 did not change in dcl4 (fig. S3). In addition, no 21-nt siRNA could be detected using high-sensitivity Northern analysis. These data suggest that DCL4 functions to repress accumulation of endogenous FCA read-through transcripts in a strand-specific manner.

Fig. 2

Transcriptional read-through of endogenous FCA and association of DCL4 with the FCA 3′ region. (A) Location of PCR fragments in the 3′ region of FCA assayed in (B) to (E). pA indicates an FCA polyadenylation site; numbers under each fragment indicate the nucleotide distance from FCA pA. (B) Semi–qRT-PCR over rt9 analyzing FCA sense read-through RNA normalized to UBC (ubiquitin conjugating enzyme At5G25760). Data are normalized to the Ler wild type. (C) Pol II ChIP enrichment (IP/input) over rt3 normalized to that of actin (At5g09810). (D) Transcripts from nuclear run-on experiments over the rt3 region normalized to the L region. (E) DCL4 ChIP assayed by qPCR over the FCA gene body (L and J2) and 3′ region (rt3, rt5, rt7, rt6). Enrichment (IP/input) is normalized to that of STM (At1g62360). Values in (B) to (E) are means ± SEM; n = 3 biological replicates.

Using chromatin immunoprecipitation (ChIP) analysis, we found that DCL4 associates directly (an enrichment of ~threefold) with endogenous FCA chromatin in a region 600 to 1200 bp downstream of the polyadenylation site (Fig. 2E and fig. S6). DCL4 was not detected in the body of the FCA gene (Fig. 2E and fig. S6) or in the 3′ region of FLC or ACTIN (fig. S6), and transcriptional read-through at these loci was not affected by dcl4 (fig. S7). Thus, DCL4 functions cotranscriptionally to repress transcriptional read-through at endogenous FCA but is not an absolute requirement for transcription termination over the whole Arabidopsis genome.

Rnt 1, a dsRNA-specific ribonuclease III, mediates cotranscriptional cleavage (CoTC) of nascent RNA downstream of the polyadenylation site in Saccharomyces cerevisiae (5). The domain structure of Rnt 1 is very similar to the C terminus of DCL4—an RNase III domain followed by a dsRNA binding domain. To investigate if DCL4 could be involved in repressing read-through of endogenous FCA through a similar mechanism, we mapped the 3′ ends of the FCA read-through RNA. To do this, we used hybrid selection circular rapid amplification of cDNA ends, a technique that enables read-through transcripts to be specifically enriched (fig. S8) (18). In Ler WT plants, the majority of the read-through products ended in the region 122 to 561 nt downstream of the FCA polyadenylation site, with most (66 to 84%) between 230 to 390 nt (Fig. 3 and fig. S9). In dcl4-15, no read-through transcripts terminated in the 230- to 390-nt window, and more transcripts ended further downstream compared with the wild type (Fig. 3 and fig. S9). This DCL4-dependent cleavage site was at one end of the domain of DCL4 association defined by ChIP (Fig. 2E). DCL4 might recognize a secondary RNA structure (19), which can be formed by the nascent transcript read-through product (fig. S10). In human cells, a very short polyA tail is added to the 3′ end of the CoTC product followed by exosome-mediated trimming (18). However, the mapped 3′ ends of read-through products of endogenous FCA did not contain short A tails, so may represent trimmed products following a DCL4-dependent CoTC activity. These data suggest that DCL4 enhances transcription termination of endogenous FCA by facilitating CoTC of the nascent RNA downstream of the polyadenylation site.

Fig. 3

3′ ends of endogenous FCA read-through transcripts in the wild type and dcl4. The sequence shown here is the endogenous FCA gene starting downstream of the polyadenylation site. Numbers on the right of each line indicate the nucleotide position relative to pA. The biotin-labeled antisense probe used to affinity purify the FCA transcripts was upstream of the sequence shown (-375 to -234 relative to the pA site). The sequence used for oligonucleotides for RNase H digestion, RT and PCR reactions is indicated. The mapped 3′ ends of the FCA transcripts are shown below the sequence. Dots indicate 3′ ends from the Ler wild type, whereas crosses indicate 3′ ends from dcl4-15. The numbers under the arrows indicate number of clones; symbols without a number indicate a single clone. The sequence to which most of the 3′ ends mapped in the Ler wild type is shown in bold. The data are combined from three biological repeats. Analysis of same data set with individual biological repeats is shown in fig. S9.

The transgene does not contain the sequences to which DCL4 associated, and, consistent with this, no DCL4 enrichment was detected in the 3′ region of the transgene (fig. S11). Therefore, we hypothesized that it was the read-through transcription of the endogenous FCA gene that triggered the transgene silencing. Supporting this model, a transferred DNA (T-DNA) insertion blocking the endogenous FCA transcriptional read-through (fig. S12) attenuates the transgene silencing in dcl4 (Fig. 4A). The mechanism of transgene silencing appears to involve both Dicer-like 2 and 3 (DCL2 and DCL3), as revealed through accumulation of 22- and 24-nt small RNAs homologous to the transgene in dcl4-15 (Fig. 4B). A dcl2, dcl4 double mutant fully restored expression of the 35S::FCAγ transgene (Fig. 4C), removed the 22-nt small RNAs (Fig. 4D), and prevented RDR6-dependent copying of sense RNA (fig. S13). These data suggest that a DCL2-dependent posttranscriptional gene silencing is the predominant silencing mechanism (20, 21). A minor role for DCL3 in silencing the 35S::FCAγ transgene was indicated by the partial restoration of expression in a dcl3, dcl4 double mutant (Fig. 4, C and D, and fig. S13). The 24-nt small RNAs also induced DNA and H3K9 methylation on the 35S::FCAγ transgene (figs. S14 and S15).

Fig. 4

Read-through transcripts from endogenous FCA trigger an RNAi-mediated silencing of the 35S::FCAγ transgene in a dcl4 background. (A) A T-DNA insertion (Flag_105D1) that prevents read-through from endogenous FCA attenuates the 35S::FCAγ transgene silencing. (B and D) Small RNA blots hybridized with dsDNA probes over the 35S::FCAγ coding region. The blots were reprobed with an antisense oligonucleotide against siRNA255 (a tasiRNA). (C) Total RNA from seedlings of different genotypes was hybridized with a single-stranded antisense probe of 35S::FCAγ. The blots were reprobed with β-TUBULIN (β-TUB) as a loading control. Col (Columbia) is the cognate wild type of dcl4-2t.

We conclude that DCL4 can promote expression of endogenous genes by contributing to the “tidying up” of 3′ end formation at genes where, for some reason, effective termination and polyadenylation has not occurred (fig. S16). The relatively weak canonical polyadenylation at the endogenous FCA 3′ region would lead to read-through RNA that potentially folds into structures recognized by DCL4. DCL4 would function as a CoTC enzyme enhancing transcription termination. Loss of this activity would increase levels of aberrant RNA from the endogenous gene. In the presence of high levels of FCA RNA from the 35S::FCAγ transgene, the aberrant RNA from the endogenous gene would efficiently trigger robust RNAi-mediated gene silencing (2022). Our work has thus revealed a previously unrecognized cotranscriptional function for Arabidopsis DCL4 in transcription termination of an endogenous gene.

Supporting Online Material

Materials and Methods

Figs. S1 to S16

References (2325)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: This work was supported by a UK Biotechnology and Biological Sciences Research Council (BBSRC) Core Strategic grant to John Innes Centre, UK BBSRC grant BB/D010799/1 (C.D.), and EU Framework VI program Integrated Project LSHG-CT-2006-037900. We thank C. Melnyk and I. Searle in D. C. Baulcombe’s group at Cambridge University for dcl3-7 seeds; O. Voinnet (ETH Zurich) for seeds of dcl4-3, dcl4-4, and dcl4-6; T. Fukuhara (Tokyo University of Agriculture and Technology) for the DCL4 antibody; and G. Szittya for excellent comments on the manuscript.
View Abstract

Navigate This Article