Targeted 3′ Processing of Antisense Transcripts Triggers Arabidopsis FLC Chromatin Silencing

See allHide authors and affiliations

Science  01 Jan 2010:
Vol. 327, Issue 5961, pp. 94-97
DOI: 10.1126/science.1180278


Noncoding RNA is emerging as an important regulator of gene expression in many organisms. We are characterizing RNA-mediated chromatin silencing of the Arabidopsis major floral repressor gene, FLC. Through suppressor mutagenesis, we identify a requirement for CstF64 and CstF77, two conserved RNA 3′-end–processing factors, in FLC silencing. However, FLC sense transcript 3′ processing is not affected in the mutants. Instead, CstF64 and CstF77 are required for 3′ processing of FLC antisense transcripts. A specific RNA-binding protein directs their activity to a proximal antisense polyadenylation site. This targeted processing triggers localized histone demethylase activity and results in reduced FLC sense transcription. Targeted 3′ processing of antisense transcripts may be a common mechanism triggering transcriptional silencing of the corresponding sense gene.

Extensive noncoding RNA has been found in many organisms (1, 2). The degree and mechanism of processing of these transcripts is unclear, as studies of RNA processing have so far focused on coding (messenger) RNA transcripts in mammalian and yeast cells. The large protein complex mediating 3′-end processing and polyadenylation has recently been described (3, 4); however, its role in processing noncoding transcripts is unknown. Analysis of 3′ processing of unstable noncoding RNAs has identified independent termination pathways with different components (5).

We have been studying the function of RNA processing in chromatin regulation through analysis of the gene encoding the major Arabidopsis floral repressor FLOWERING LOCUS C (FLC) (6). The autonomous floral promotion pathway results in FLC transcriptional silencing through the activities of two RNA-binding proteins, FCA and FPA; a member of a cleavage and polyadenylation specificity factor (CPSF) RNA 3′-processing complex, FY, which interacts directly with FCA (7, 8); and FLD, a dimethylated histone H3 at lysine 4 (H3K4me2) demethylase and homolog of human LSD1 (9).

To better understand how RNA 3′-processing links to histone modification to result in FLC silencing, we sought to identify all the components necessary for FCA-mediated FLC repression through suppressor mutagenesis. We generated an Arabidopsis line sensitized to FCA action expressing 35S::FCAγ—a transgene overexpressing FCA, FRIGIDA—a strong activator of FLC expression and FLC::LUC, so FLC expression could be monitored with a bioluminescence detection assay (10, 11). Mutations that suppressed the ability of FCA to repress FLC, together with an epistasis analysis of their interaction with fca mutants, enabled us to identify components required for FCA action (fig. S1). These mutations, named suppressors of overexpressed FCA (sof) (9), identified additional alleles of known FCA pathway components FY and FLD (fig. S2) and new complementation groups sof2 and sof19 (Fig. 1A and fig. S2). Genetic mapping and sequencing revealed sof2 and sof19 carried mutations in CstF77 (At1g17760) and CstF64 (At1g71800), respectively (fig. S3, A to C).

Fig. 1

CstF64 and CstF77 regulate FLC levels. (A) Northern and flowering time analysis (total leaf number) of sof2 (cstf77-1 in FCA-sensitized background) and sof19 (cstf64-1 in FCA-sensitized background). Asterisk indicates FLC::LUC; APT is the loading control. (B) Northern analysis of FLC levels in cstf64 carrying no transgenes or cstf77 carrying the linked FRI transgene. β-TUB or APT is the loading control. (C) cstf64-2 is not additive with fca-9, fy-2, or fld-4 (all Col genotype). (A) and (C) graph values are means ± SD (n = 20).

CstF77 and CstF64 are the Arabidopsis homologs of two of the three components of the CstF RNA 3′-processing complex, conserved from yeast to plants and humans (12). CstF77 is a single-copy gene in the Arabidopsis genome. In sof2 (carrying the cstf77-1 allele), the splice acceptor site in intron 12 is mutated, which leads to an in-frame deletion of Tyr339 to Lys385 in exon 13 (fig. S3A). CstF77 is an essential gene in other organisms (13, 14), yet cstf77-1 plants are viable, but their flowering is delayed. cstf77-1 is likely to be a hypomorphic allele, because homozygous mutant progeny could not be identified from a mutation caused by a transferred DNA insertion into the 5′ end of the gene (cstf77-2) (fig. S3A). Further analysis suggested that cstf77-2 causes female gametophytic lethality (fig. S4, A and B), consistent with an essential function for CstF77. The sof19 mutation introduces a premature stop codon in the only full-length homolog of CstF64 in the Arabidopsis genome (fig. S3A). This allele (cstf64-1) also reduces fertility of the plants. A second, strong allele in Columbia (Col) background (cstf64-2) (fig. S3A) leads to reduced organ size, pale leaves, and sterility (fig. S4, C to E). Therefore, CstF64 function is generally required in plant growth and development. In mammals, CstF77 interacts physically with CstF64 through a proline-rich region (15). This interaction was found to be conserved in plants and to be unaffected by the in-frame deletion in cstf77-1 (fig. S4F).

We further analyzed the involvement of CstF64 and CstF77 in FLC repression in genotypes not sensitized to FCA action. cstf64-1, cstf64-2, and cstf77-1 had elevated FLC mRNA levels and flowered later than their respective controls (Fig. 1, B and C, and fig. S5). We also undertook an epistasis analysis to determine whether the CstF components functioned in the same pathway with known FCA pathway components, namely FCA, FY, and FLD. cstf64-2 was not additive with fca-9, fy-2, or fld-4 mutations but was additive with fve-3 on the basis of analysis of both flowering time and FLC expression levels (Fig. 1C and fig. S6). FVE functions to repress FLC expression by histone deacetylation in an FCA-independent manner (16). In summary, our analysis demonstrates that CstF components do function in the same pathway as FCA in the repression of FLC.

Because the genetic analysis established that CstF components function with the histone demethylase FLD to down-regulate FLC, we tested whether the increased FLC in cstf mutants was an effect of defective transcriptional repression. cstf64 and cstf77 mutants showed elevated FLC nascent transcript levels (Fig. 2B and fig. S7); higher RNA polymerase II (Pol II) association at FLC (Fig. 2C); and higher levels of H3K4me3, a histone modification associated with active transcription (17) (Fig. 2D). The CstF complex is generally required for canonical mRNA 3′-end formation (12), but its loss did not affect FLC sense RNA 3′ processing; indeed, FLC sense levels accumulate and are functional as evidenced by the late flowering of cstf64 double mutants (Fig. 1C). Thus, CstF components mediate transcriptional repression of FLC levels but are unlikely to be required for FLC mRNA 3′-end formation.

Fig. 2

cstf-64 and cstf-77 increase FLC transcription. (A) Schematic of FLC, vertical bars denote exons. Horizontal bars (A, B, C, and G) denote regions analyzed by qPCR in chromatin immunoprecipitation (ChIP). Horizontal bars (a and b) denote the intron-containing FLC nascent transcripts analyzed by qRT-PCR. (B) Nascent FLC transcript analysis of fragment a (containing intron 1) and fragment b (containing introns 2 and 3). (C) Pol II ChIP assay in the different FLC regions in sof2 (gray) and sof19 (white) normalized to parental line C2 (black). (D) H3K4me3 ChIP assay in FLC regions in sof2 (gray) and sof19 (white) normalized to parental line C2 (black). Error bars represent standard errors derived from at least two biological and two technical repeats.

FCA directly associates with FLC chromatin (9), so one explanation for our observations is that the substrates of CstF64, CstF77, FCA, and FY are other nascent transcripts from the FLC locus. We have previously described alternatively polyadenylated FLC antisense transcripts (9, 18). One polyadenylation site of the FLC antisense RNA coincides with the location of FCA on FLC chromatin, whereas a distal polyadenylation site overlaps with the FLC sense promoter (9). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis revealed increased levels of FLC antisense transcription in fca, fy, and cstf mutants (fig. S8A). Further analysis has also revealed additional complexity in the antisense transcripts, so we developed an assay to detect RNA 3′ processing and polyadenylation at the specific polyadenylation sites shown in Fig. 3. qRT-PCR revealed that FLC antisense RNA 3′ processing was reduced at both proximal and distal polyadenylation sites in cstf64 and cstf77 (Fig. 3B), but FLC sense RNA polyadenylation was not reduced (fig. S8B). Thus, 3′ processing and polyadenylation of FLC antisense transcripts, but not FLC sense transcripts, appear sensitive to CstF complex activity. Rapid amplification of cDNA 3′ ends (RACE) analysis revealed that the same polyadenylation sites are used in the fca mutant and wild type; however, qRT-PCR, supported by strand-specific Northern analysis, showed that the relative usage of each site differed. In fca and fy, proximal site usage was reduced, and in fca, distal site usage increased (Fig. 3C and fig. S8C). This is not a feature of general FLC de-repression, because proximal site usage was not affected in fve (Fig. 3C) (16). To investigate the generality of this finding, we analyzed the effects of FPA, a second RNA recognition motif (RRM) protein that functions in the autonomous floral promotion pathway. FPA also requires FLD to repress FLC (19), but it functions independently of FCA (19). As in fca, proximal site usage in the FLC antisense transcript was reduced, and distal site usage increased in fpa-7 (Fig. 3C). Because the two RRM proteins trigger FLC silencing independently, we predicted that disruption of FPA function would cause increased FLC expression even in the sensitized FCA background. This was confirmed through the identification of sof34 as a novel fpa mutation (fig. S9). Taken together, these results reveal that both RNA-binding proteins independently function to promote 3′ processing at the proximal site in the FLC antisense transcript and that this activity triggers FLC sense strand transcriptional silencing.

Fig. 3

The 3′ end processing of FLC antisense RNA. (A) In this FLC schematic, vertical bars denote exons; transcription start site of FLC sense RNA is indicated by an arrow. FLC sense (top) and antisense (bottom) transcripts with dashed lines indicating introns, arrows with symbol (A) are polyadenylation sites, and shading shows the region of multiple antisense RNA start sites. (B) The 3′ processing of FLC antisense transcripts is reduced at both distal and proximal sites in sof2 and sof19. FLC antisense transcript levels are normalized to total FLC antisense transcript and given as fold change compared with the parental line C2. (C) The 3′ processing of FLC antisense transcripts at the proximal (top) and the distal sites (bottom) in different mutants (all in the Col genotype without transgene). FLC antisense transcript levels are normalized to total FLC antisense transcript and given as fold change compared with the Col wild type. The raw qPCR data are given in table S1. In (B) and (C), averages and standard errors of three biological repeats with two technical repeats are shown.

The position of the proximal 3′-processing site on the antisense transcript coincides with the site where FCA associates with FLC chromatin (9). This suggests a model whereby FCA, interacting with a component of the CPSF complex, targets CstF-dependent 3′ processing to the proximal site on FLC antisense transcripts (fig. S10). An indirect effect, via CstF regulation of an intermediary regulator, is possible, but unlikely, because of the direct association of FCA with FLC. Proximal FLC antisense transcript 3′ processing is promoted by FCA and FY (Fig. 3), which is reminiscent of the FCA-FY interaction–dependent 3′-processing site choice in FCA negative-feedback regulation (7). FPA also enhances usage of this proximal 3′-processing site. These activities then appear to converge to trigger FLD-dependent demethylation of H3K4me2 in the body of the gene, downstream of the proximal polyadenylation site. We speculate that this may involve termination-associated processes, perhaps cotranscriptional decay of the antisense RNA downstream of the cleavage site (20). Whatever the mechanism, down-regulation of both sense and antisense transcription is the net result. We suspect the activity of DICER-LIKE3, previously shown to function in this process (9), is required to couple the 3′ processing to FLD histone demethylase function (21); however, RNA polymerases involved in silencing mediated by small interfering RNA (siRNA) (18) only weakly suppress FCA activity (fig. S11). A role for small RNAs in the targeting of the histone demethylase activity to the central part of the locus may lead to trans effects on homologous sequences in the genome, perhaps explaining the small effect of an fca mutation on an FLC transgene lacking the 3′ region (22).

The proteins involved in this silencing mechanism—two RRM proteins, a factor associated with 3′ processing components, and a conserved histone demethylase—also silence transposons and transgenes in Arabidopsis (20, 23). This mechanism may therefore play a more general role, rather than only regulating the floral repressor gene, FLC. This is consistent with the role for CPSF complex components in suppressing viral amplicon-induced gene silencing in Arabidopsis (24). This mechanism may also be very important in nonplant genomes. Genome-wide antisense transcripts, identified in many organisms, function in chromatin silencing (2, 25, 26) and RNA 3′-processing components mediate gene silencing in Caenorhabditis elegans (27). Therefore, the 3′ processing of antisense transcripts may be a general mechanism triggering chromatin silencing in eukaryotes.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11

Table S1


  • * These authors contributed equally to this work.

  • Present address: Institute of Biochemistry and Biophysics, Polish Academy of Sciences, 02-106 Warsaw, Poland.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank our colleagues for comments and advice and G. Szittya for comments on the manuscript. Supported by a UK Biotechnology and Biological Sciences Research Council (BBSRC) Core Strategic grant to John Innes Centre; BBSRC grant BB/D010799/1 (C.D.); European Union (EU) Marie Curie studentship MEST-CT-2005-019727 to S.M.; EU Framework VI program Integrated Project LSHG-CT-2006-037900.
View Abstract

Navigate This Article