Report

Widespread Role for the Flowering-Time Regulators FCA and FPA in RNA-Mediated Chromatin Silencing

See allHide authors and affiliations

Science  05 Oct 2007:
Vol. 318, Issue 5847, pp. 109-112
DOI: 10.1126/science.1146565

Abstract

The RRM-domain proteins FCA and FPA have previously been characterized as flowering-time regulators in Arabidopsis. We show that they are required for RNA-mediated chromatin silencing of a range of loci in the genome. At some target loci, FCA and FPA promote asymmetric DNA methylation, whereas at others they function in parallel to DNA methylation. Female gametophytic development and early embryonic development are particularly susceptible to malfunctions in FCA and FPA. We propose that FCA and FPA regulate chromatin silencing of single and low-copy genes and interact in a locus-dependent manner with the canonical small interfering RNA–directed DNA methylation pathway to regulate common targets.

Heterochromatin in many organisms is characterized by extensive DNA methylation and histone modifications (1). Plants display cytosine methylation in CG, CNG (N = any nucleotide), and CHH (H = A, C, or T) sequence contexts. In Arabidopsis, small interfering RNAs (siRNAs) are involved in localizing and maintaining these chromatin modifications in processes requiring RNA-DEPENDENT RNA POLYMERASE2 (RDR2), DICER-LIKE3 (DCL3), ARGONAUTE4 (AGO4), and the two RNA polymerase IV isoforms, Pol IVa and b (29).

To identify further components required for siRNA-mediated chromatin silencing, we used a reporter system in which the Arabidopsis phytoene desaturase (PDS) gene is silenced in response to a homologous inverted repeat (SUC-PDS) (10). Two mutants that partially suppressed the silencing of PDS (Fig. 1, A, B, C, and E) showed late flowering that was reversible by vernalization. The silencing and flowering phenotypes cosegregated, and the mutations mapped to chromosomes 2 and 4. The flowering phenotype suggested involvement of FPA and FCA, two members of the autonomous pathway (11), mapping to those genomic regions. Sequencing revealed a premature termination codon in FPA (Trp98*, G to A, fpa-8) and FCA (Gln537*, C to T, fca-11). The flowering defect was confirmed by complementation analysis with previously known flowering mutants (fca-9, fpa-7, and fve-3; Fig. 1F), which also showed PDS silencing (fig. S1). Thus, FCA and FPA are required for efficient PDS silencing in the presence of SUC-PDS.

Fig. 1.

FCA and FPA suppress SUC-PDS–induced silencing of PDS. (A to E) Leaf phenotypes in SUC-PDS background grown in long days. (A) Col, (B) fpa-8, (C) fca-11, (D) fca-11 fpa-8, (E) no transgene. Scale bar, 5 mm. (F) Complementation analysis: average flowering time (±SEM) of F1 progeny of crosses between the indicated mutations (white, selfed) and fpa-8 (black) or fca-11 (gray). (G) RNA gel blot analysis of PDS mRNA detecting endogenous PDS (PDS) and SUC-PDS mRNA (*). Numbers indicate relative expression of PDS averaged over two experiments. (H) RNA gel blot analysis of SUC-PDS siRNA. (I) Cytosine methylation at the endogenous PDS locus assayed by bisulfite sequencing.

FCA and FPA contain multiple RNA recognition motif (RRM) RNA binding domains (12, 13), which are known to bind single-stranded RNA, but share no other sequence homology. FCA negatively regulates its own expression through alternative polyadenylation site usage (14). Late flowering in fca and fpa is due to overexpression of the major repressor of flowering in Arabidopsis, FLOWERING LOCUS C (FLC) (11). FCA and FPA do not appear to regulate PDS in the absence of the silencing trigger, which suggests that the presence of the transgene makes the endogenous PDS a target of FCA and FPA.

Because FCA and FPA both contain RRM domains, we hypothesized that they act partially redundantly; consistent with this, an fca-11 fpa-8 double mutant showed no PDS silencing (Fig. 1D). Components of the siRNA chromatinsilencing pathway (the Pol IVa largest subunit NRPD1a and RDR2) also suppress PDS silencing completely (10). Both nrpd1a-5 and fca-11 fpa-8 double mutants showed reduced SUC-PDS and higher PDS mRNA levels (Fig. 1G and fig. S2) (10). PDS siRNA levels were reduced in rdr2-5, nrpd1a-5, and fca-11 fpa-8 mutants, but not in fca-11 or fpa-8 single mutants (Fig. 1H).

Bisulfite sequencing was used to investigate whether PDS silencing corresponded to DNA methylation at the endogenous PDS locus. We found non-CG methylation (CNG and CHH) at the endogenous PDS locus in a region complementary to the hairpin in SUC-PDS leaves, but not in wild-type leaves (Fig. 1I and table S1). CG sites were highly methylated in both the wild type and mutants. In nrpd1a-5 and fca-11 fpa-8 mutants, loss of siRNA coincided with loss of asymmetric DNA methylation (Fig. 1I and table S1). CHH methylation was also compromised in fca-11 and fpa-8 single mutants, although PDS siRNA levels were unaffected (Fig. 1, H and I). These results suggest a dual role for FCA and FPA:(i) They acttogether with NRPD1a and RDR2 and redundantly with each other to amplify siRNAs derived from the transgene locus; (ii) they act in the perception and interpretation of the silencing signal at the target locus. Mutants in two other members of the autonomous pathway—the MSI1 homolog FVE and the putative histone demethylase FLD (15, 16)—also suppressed PDS silencing (fig. S1), which indicates that multiple components of the autonomous pathway are involved in this process.

Transposons, retroelements, and intergenic transcripts are endogenous targets of chromatinsilencing pathways (58, 17). Expression of the AtSN1 retroelement and the AtMu1 DNA transposon were also controlled by FCA and FPA (Fig. 2A). AtSN1 was reactivated very strongly in fpa-8, fca-11 fpa-8, and nrpd1a-5 mutant seedlings, but not in fca-11. In contrast, AtMu1 was slightly derepressed in fca-11 and fpa-8 single mutants and more strongly in fca-11 fpa-8. AtMu1 reactivation in fca-11 fpa-8 was similar to that in nrpd1a-5. An intergenic transcript flanked by a solo long terminal repeat (LTR), IG/LINE, was also up-regulated in fca-11 fpa-8, albeittoa lesser extent than in nrpd1a-5 (Fig. 2A). Together, these findings indicate that FCA and FPA have a widespread role in the regulation of endogenous loci known to be silenced at the level of transcription and dependent on siRNA.

Fig. 2.

Reactivation of AtSN1, AtMu1, and IG/LINE in seedlings. (A) Quantitative reverse transcription polymerase chain reaction (RT-PCR) on Col, fpa-8, fca-11, fca-11 fpa-8, and nrpd1a-5. (B) RNA gel blot analysis of transacting siRNAs (siR255, siR1511), microRNAs (miR159, miR171), or siRNAs (all other) on Col, fca-9, fpa-7, fca-9 fpa-7, and nrpd1a-3. (C) Cytosine methylation for Col (black), nrpd1a-3 (light gray), and fca-9 fpa-7 (dark gray). (D) Quantitative RT-PCR (left: Col, nrpd1a-3, fpa-8, nrpd1a-3 fpa-8, fca-11, and nrpd1a-3 fca-11; right: Col, fca-9, rdr2-1, fca-9 rdr2-1, dcl3-1, and fca-9 dcl3-1). nrpd1a-3 is a weaker allele than nrpd1a-5 with respect to AtMu1 reactivation; error bars indicate SD.

We next investigated whether this transcriptional reactivation correlated with loss of corresponding siRNA. AtSN1 and AtMu1 siRNAs were detected at wild-type levels in fca-9, fpa-7, and fca-9 fpa-7, but were absent from nrpd1a-3 mutant seedlings (Fig. 2B). Corresponding results were obtained for other siRNAs. Thus, despite their role in the amplification of PDS siRNA, FCA and FPA do not generally act in NRPD1a-dependent siRNA production. There was no change in DNA methylation at the AtSN1 locus in fca fpa (Fig. 2C, fig. S3, A and B, and table S2). However, bisulfite sequencing indicated a reduction of ∼50% in asymmetric (CHH) DNA methylation at AtMu1 in fca fpa, whereas CG and CNG methylation were not affected (Fig. 2C, fig. S3B, and table S2). Like-wise, asymmetric DNA methylation at the solo LTR was reduced (fig. S3C). Maintenance of asymmetric DNA methylation requires the continued presence of the trigger, whereas symmetric DNA methylation can be maintained through cell divisions in the absence of the trigger. Silencing at these loci is also associated with changed histone tail modifications such as increased H3 K9 dimethylation and reduced H3 K4 dimethylation (5, 8, 17). Using chromatin immunoprecipitation, we did not find any pronounced alteration in these marks in fca-9 fpa-7.

Heterochromatic loci are targeted by multiple silencing pathways, and their contribution at individual loci differs considerably (1820). This is corroborated by our finding that silencing of AtSN1, AtMu1, IG/LINE, and PDS in the presence of SUC-PDS differentially requires FCA and FPA. PDS silencing is associated with target DNA methylation and siRNA production through mechanisms that are dependent on both the siRNA chromatin-silencing pathway and fca fpa. Derepression of AtMu1 and IG/LINE in fca fpa mutants coincides with loss of DNA methylation but not siRNAs, whereas both are lost in mutants of the siRNA chromatin-silencing pathway. Despite much stronger reactivation of AtSN1 in fca fpa, neither DNA methylation nor siRNA accumulation was affected. Our findings are consistent with the idea that transcription can be reactivated in the presence of DNA methylation, as was established for the morpheus' molecule 1 (mom1) mutation (19, 21). Despite this similarity, it seems unlikely that FCA and FPA generally act together with MOM1, because AtSN1 and AtMu1 are not misregulated in mom1 (22).

To investigate how FCA and FPA relate to the chromatin siRNA amplification pathway including Pol IVa, RDR2, and DCL3, we analyzed the release of silencing in double mutants (Fig. 2D). All double mutants showed much higher reactivation of AtSN1 and AtMu1 than any of the single mutants, which suggests that FCA and FPA do not act downstream of the siRNA amplification pathway, but rather in parallel. Similarly, transposon reactivation was greatly enhanced in fve nrpd1a double mutants relative to either of the single mutants (fig. S3D). Strikingly, although FCA is dispensable for AtSN1 silencing in the wild type, the loss of FCA in nrpd1a, rdr2, or dcl3 mutant backgrounds greatly enhanced the release of AtSN1 silencing.

Our findings predict that perturbation of DNA methylation in fca fpa mutants will affect reactivation of target loci differently. At AtSN1, where the effect of FCA and FPA is uncoupled from DNA methylation, enhanced loss of silencing in the presence of the DNA methylation inhibitor 5-aza-deoxycytidine (aza-dC) would be expected. Conversely, at AtMu1, where fca fpa mutants show reduced DNA methylation, the additional effect of the inhibitor would be small. Our results (Fig. 3A and table S3) are consistent with this prediction, because fca-9 fpa-7 mutants were more sensitive than the wild type to aza-dC with respect to AtSN1 reactivation, but less sensitive than the wild type with respect to AtMu1 reactivation. Also, development of fca-9 fpa-7 seedlings was strongly perturbed when exposed to aza-dC at concentrations where development of wild-type or fca-9 seedlings was not abnormal and development of fpa-7 seedlings was only very slightly abnormal (Fig. 3B and table S4) (23).

Fig. 3.

(A and B) Aza-dC treatment. (A) Quantitative RT-PCR on Col and fca-9 fpa-7 seedlings grown on aza-dC (white, mock; gray, 2 μM; black, 4 μM) normalized to the expression level after mock treatment (±SD). (B) Seedlings (Col, fpa-7, fca-9, and fca-9 fpa-7) grown for 14 days on aza-dC. (C) An FPA-YFP fusion protein localizes to the nucleus of transgenic Arabidopsis seedling roots. Scale bar, 50 μm. (D) Chromatin immunoprecipitation from two independent FPA-YFP lines. Lane 1, Col; lane 2, FPA-YFP line 2; lane 3, FPA-YFP line 5. (E and F) RT-PCR assaying spliced and unspliced transcripts of FLC and AtMu1.

fca fpa double mutant plants are late flowering but otherwise largely normal. However, closer examination of fca-11 fpa-8 siliques revealed that ∼20% of developing seeds aborted and ∼70% of ovules did not initiate development (fig. S4A and Table 1). When pollinating double mutants with wild-type pollen, no seeds aborted, but the high proportion of undeveloped seeds persisted; this finding suggested that the embryonic lethality was zygotic, whereas the undeveloped seed phenotype was caused by the genotype of the mother plant. When fca/fca FPA/fpa ovules were pollinated with wild-type pollen, 34% of seeds appeared undeveloped (Table 1). Microscopic examination of mature ovules did not reveal any abnormalities (fig. S4, B and C), which suggests that the genotype of the female gametophyte determined the undeveloped seed phenotype. Thus, (female) gametophytic and early embryonic development is extremely sensitive to loss of FCA and FPA. Once these stages are passed successfully, development can proceed largely independently of FCA and FPA. Whether misregulation of a few key genes or more global genome misorganization causes these defects remains to be investigated.

Table 1.

Percentage of aborted and undeveloped seed in fca-11 fpa-8 mutant lines.

View this table:

We propose that the increased transcript levels measured for the targets in fca fpa reflect transcriptional reactivation rather than increased cytoplasmic RNA stability. This is supported by the subcellular localization of FPA and FCA: A fully complementing FPA–yellow fluorescent protein (YFP) fusion protein localized to the nucleus (Fig. 3C and fig. S5); FCA is a nuclear protein that interacts with the SWI/SNF chromatin remodeler SWI3B (14, 24). Both proteins associate with the chromatin of their target genes: The FPA-YFP fusion protein localized to the chromatin of AtMu1 and FLC (Fig. 3D); FCA localized to FLC chromatin (25). Lastly, using an established assay for transcriptional activity (26), FLC and AtMu1 unspliced (nascent) transcripts were up-regulated in all backgrounds that caused up-regulation of the spliced transcript, and both unspliced and spliced transcripts were increased similarly (Fig. 3, E and F). Together, these data all indicate that silencing does not occur posttranscriptionally but rather cotranscriptionally before any processing occurs.

Taken together, our results show that the nuclear proteins FCA and FPA have a much more widespread role in development and gene silencing than previously anticipated. We propose a model in which FCA and FPA cotranscriptionally recognize aberrant RNA and mark it for silencing (fig. S6). A nascent RNA may be made aberrant by the presence of low levels of complementary siRNAs or misconducted processing events. FCA and FPA would then facilitate silencing by recruiting or stabilizing effector complexes. Although the common result of FCA and FPA action is silencing of a target locus, the identity of these effector complexes presumably varies with the contribution of different pathways at individual loci, thus leading to somewhat different silencing signatures. Whereas the majority of functionally characterized RRM-domain proteins act in posttranscriptional RNA processing (27), FCA and FPA appear to integrate the state of the nascent RNA with transcription. That this might be a novel function of some RRM-domain proteins is supported by two other reports. The yeast Set1 histone methyltransferase has an RRM domain thought to bind nascent RNA and thereby regulate the methyltransferase activity (28). Furthermore, three RRM-domain proteins are required for transcriptional silencing in Caenorhabditis elegans cosuppression (26).

Although the canonical siRNA-directed chromatin-silencing pathway has been described for repetitive loci, FCA and FPA silence mainly single-copy loci and do not affect silencing of the highly repetitive 5S loci (fig. S7). At a subset of targets, however, these pathways clearly interact. The canonical chromatin-silencing/siRNA amplification pathway involves amplification of siRNAs and shuttling of silencing information between the locus and a nucleolar RNA processing center (29, 30), thereby silencing any sufficiently homologous locus in the genome. In contrast, FCA and FPA may bypass the siRNA amplification step, thereby restricting it to acting in cis. Unraveling the interactions between the different pathways will ultimately enable us to understand what properties in a target commit it to being silenced in a particular way.

Supporting Online Material

www.sciencemag.org/cgi/content/full/318/5847/109/DC1

Materials and Methods

Figs. S1 to S7

Tables S1 to S5

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article