Arabidopsis transcriptional repressor VAL1 triggers Polycomb silencing at FLC during vernalization

See allHide authors and affiliations

Science  29 Jul 2016:
Vol. 353, Issue 6298, pp. 485-488
DOI: 10.1126/science.aaf7354


The determinants that specify the genomic targets of Polycomb silencing complexes are still unclear. Polycomb silencing of Arabidopsis FLOWERING LOCUS C (FLC) accelerates flowering and involves a cold-dependent epigenetic switch. Here we identify a single point mutation at an intragenic nucleation site within FLC that prevents this epigenetic switch from taking place. The mutation blocks nucleation of plant homeodomain–Polycomb repressive complex 2 (PHD-PRC2) and indicates a role for the transcriptional repressor VAL1 in the silencing mechanism. VAL1 localizes to the nucleation region in vivo, promoting histone deacetylation and FLC transcriptional silencing, and interacts with components of the conserved apoptosis- and splicing-associated protein (ASAP) complex. Sequence-specific targeting of transcriptional repressors thus recruits the machinery for PHD-PRC2 nucleation and epigenetic silencing.

In Arabidopsis thaliana, prolonged cold exposure during winter promotes flowering through epigenetic silencing of FLOWERING LOCUS C (FLC) in a process called vernalization (1, 2). Cold exposure induces expression of antisense transcripts to FLC (collectively known as COOLAIR) (3) and a plant homeodomain (PHD) protein called VERNALIZATION INSENSITIVE 3 (VIN3) (4). COOLAIR facilitates FLC transcriptional silencing and coordinates the switching between chromatin states (5). VIN3 associates with a homologous PHD protein, VERNALIZATION 5 (VRN5), and a vernalization-specific Polycomb repressive complex 2 (PRC2) (6, 7), which accumulates at an intragenic nucleation region covering the first exon and part of the first intron of FLC. Quantitative accumulation of H3K27me3 at the nucleation region during cold exposure, and over the whole locus after cold exposure, reflects a cell-autonomous epigenetic switch affecting an increasing proportion of cells (8). The recruitment of PHD-PRC2 to the nucleation region is, therefore, a key step in the silencing process. In Drosophila, Polycomb response elements (PRE) have been identified as cis sites for PRC2 recruitment and provide sequence-specific “memory” modules for the activity of linked enhancers (9). In contrast, in mammals, CpG islands facilitate targeting of Polycomb machinery, with Polycomb complexes “sampling” chromatin to determine transcriptional states (10). In Arabidopsis, PRE-like elements have been identified (11), but whether they recruit Polycomb complexes has been unclear. We sought to determine what targets PHD-PRC2 to the nucleation region of FLC.

A forward genetic screen for impaired FLC-LUCIFERASE (FLC-LUC) silencing (12) identified vrn8 mutant (Fig. 1A and fig. S1, A to C). The progenitor plants showed characteristic cold-induced silencing of both the endogenous FLC and the FLC-LUC transgene (Fig. 1, B and C, and fig. S1A). In contrast, the FLC-LUC transgene expression remained high in vrn8 after cold exposure, whereas expression of the endogenous FLC was reduced as normal (Fig. 1, B and C, and fig. S1A). The different behavior of the two copies suggests that vrn8 does not encode a trans factor involved in vernalization.

Fig. 1 vrn8 disrupts an intronic RY element that is required for FLC silencing.

(A) The vrn8 mutant fails to silence FLC-LUC after cold exposure. Luciferase activity is depicted with false color from least (blue) to most intense (red). Ler, Landsberg erecta accession. (B and C) Expression of endogenous Ler FLC (B) and FLC-LUC transgene (C). The data are abundances relative to UBIQUITIN-CONJUGATING ENZYME 21 (UBC) and standardized to nonvernalized (NV) conditions. Numeral-W-numeral, number of weeks (W) of cold treatment followed by number of days of growth at 22°C. (D) Schematic representation of FLC genomic locus. Black boxes represent exons. The green dashed line represents the region analyzed in (F) and (G). Alignment of FLC intronic sequences from different species of Brassicaceae is shown below; RY motifs are in bold. The single nucleotide change in vrn8 is indicated (C585T, red). TSS, transcriptional start site. (E) Spliced FLC expression in FLC-WT and FLC-C585T transgenic lines. (F and G) ChIP analysis of H3K27me3 accumulation at the FLC locus in FLC-WT (F) and FLC-C585T (G) plants. Numbers on the x axes are distances to the TSS (TSS = 0). Throughout this figure, values are means ± SEM of three biological replicates. **P < 0.01; *P < 0.05; ns, not significant.

The vrn8 mutation was a cytosine-to-thymine change in intron 1 of the FLC-LUC transgene, at position +585 downstream of the transcriptional start site (hereafter, we term this mutation C585T; Fig. 1D and fig. S1D). C585T maps to the first of a pair of RY cis elements (TGCATG, RY-1 and RY-2; R, purine; Y, pyrimidine), which are recognized by B3 DNA binding domains (13). Alignment of FLC intronic sequences from different species of Arabidopsis and Brassica shows 100% sequence conservation of both RY motifs (Fig. 1D). To confirm the effect, we regenerated the C585T mutation and compared plants carrying wild-type (FLC-WT) and mutated (FLC-C585T) transgenes (fig. S2A). Cold-induced repression of FLC was impaired in FLC-C585T transgenic lines (Fig. 1E and fig. S2B), and the plants flowered later (fig. S2C). The proximity of the C585T change to the PHD-PRC2 nucleation region prompted an analysis of cold-induced chromatin changes in FLC-C585T. The quantitative increase in H3K27me3 and equivalent decrease in H3K36me3 at FLC-WT (Fig. 1F and fig. S2D) (14) were not found at FLC-C585T (Fig. 1G and fig. S2E), suggesting that the C585T mutation prevents PHD-PRC2 nucleation.

Identification of the C585T mutation raised the question of what bound to the RY elements. Potential candidates included the B3 transcriptional regulators belonging to the LAV family (13): LEAFY COTYLEDON 2 (LEC2), ABSCISIC ACID INSENSITIVE 3 (ABI3), FUSCA3 (FUS3), and the VIVIPAROUS1/ABI3-LIKE factors (VAL1 VAL2 and VAL3). The low levels of expression of ABI3, LEC2, FUS3, and VAL3 in 10-day-old seedlings (fig. S3, A and B) argued against a function of the corresponding transcriptional regulators in FLC silencing during vernalization. In contrast, VAL1 and VAL2 were expressed at higher levels than other LAV family genes in seedlings (fig. S3A) and continued to be expressed during vernalization (fig. S3, C and D). VAL proteins repress late seed maturation genes and promote the switch from embryonic to vegetative development. val1 val2 double-mutant seedlings express many embryonic-specific transcripts and also show synergistically increased expression of FLC compared with each single mutant alone (15). We crossed val1 and val2 single mutants with the Columbia FRIGIDA (Col FRI) line to assess whether VAL genes are required for FLC regulation during vernalization. val1 FRI mutants flowered later than Col FRI and val2 FRI plants (Fig. 2A and fig. S4A), and this was reflected in higher FLC expression levels before and during cold exposure (Fig. 2B). val1 FRI mutants also showed reduced sensitivity to vernalization. The cold-induced reduction in nonspliced FLC transcript (probably reflecting transcription; figs. S4B and S5A) and FLC mRNA (Fig. 2C) was slower in val1 FRI than in wild-type plants, but COOLAIR induction was unaffected (fig. S5B). The mutant phenotype was complemented by expression of a hemagglutinin (HA)–tagged VAL1 (fig. S6). VAL1 not only modulated FLC transcriptional shutdown but also appeared to influence the FLC homologs MADS AFFECTING FLOWERING 1 and 2 (MAF1 and MAF2; fig. S4, C and D). Loss of VAL1 also attenuated the cold-induced H3K27me3 accumulation at FLC: Starting levels before cold exposure were lower and failed to reach wild-type levels after 4 weeks of cold (Fig. 2, D and E). The nonreactivation of FLC after cold exposure in val1 FRI (Fig. 2C and fig. S4B) implies that nucleation is defective (Fig. 2E) but that the components of the longer-term Polycomb memory are not perturbed. The phenotype of val1 FRI (Fig. 2E) was not as strong as that caused by C585T (Fig. 1G), which is consistent with VAL1 and VAL2 functioning redundantly in FLC regulation (15).

Fig. 2 VAL1 is a component of the vernalization mechanism.

(A) Flowering time after 6 weeks of cold. Each gray triangle represents a single plant (n = 36); means (black horizontal lines) ± SD (error bars) are shown. ***P < 0.001. (B) Spliced FLC expression in val1 FRI, Col FRI, and val2 FRI plants before (NV) and during (6W0) cold exposure. (C) Dynamics of spliced FLC down-regulation during vernalization. Data in (B) and (C) are abundances, expressed as in Fig. 1. (D and E) H3K27me3 accumulation (from ChIP analysis) along the FLC locus in Col FRI (D) and val1 FRI (E) plants. Numbers on the x axes are distances to the TSS (TSS = 0). The schematic of the FLC locus is shown below each panel. Values in (B) to (E) are means ± SEM of three biological replicates.

ABI3, FUS3, and LEC2 bind in a sequence-specific manner to RY elements in vitro (1618). This is also the case for the VAL1 B3 domain, which binds in vitro to the FLC RY-1 element (Fig. 3A), with the C585T mutation sufficient to disrupt binding. Competition experiments showed specificity of VAL1 B3 binding to the FLC RY motif (fig. S7, A and B) and not to a different B3 binding cis element (RAV) (19). VAL1 B3 can also bind to the second RY site (RY-2) that occurs just downstream of RY-1; competition experiments revealed that mutation of both RY elements is required to block VAL1 B3 binding to the nucleation region in vitro (fig. S7C). Consistent with this, in vivo mutation of RY-2 also attenuated FLC silencing (fig. S7D), suggesting that both RY elements contribute functionally. Chromatin immunoprecipitation (ChIP) confirmed in vivo binding of VAL1-HA to the FLC nucleation region during cold exposure (Fig. 3B). These data raise interesting parallels with the cooperative binding of auxin response factors (ARFs) to tandem binding sites in vivo (20).

Fig. 3 VAL1 binds to the FLC nucleation region.

(A) Electrophoretic mobility shift assay testing GST-VAL1B3 binding to an RY cis element (GST, glutathione S-transferase). Wild-type (RY-1) or mutated [RY-1(C585T)] RY probes were combined with increasing amounts of GST-VAL1B3 (1 = 25 ng/μl). (B) ChIP–quantitative polymerase chain reaction of VAL1-HA binding along FLC. ACTIN2 (ACT), UBC, and SHOOT MERISTEMLESS (STM) were used as negative controls. The schematic of the FLC locus is shown below. Values are means ± SEM of one biological replicate for samples from six independent transgenic lines.

The association of VAL1 with the nucleation region as a prerequisite for PHD-PRC2 activity at the locus raised two questions. First, how does VAL1 binding within intron 1 repress FLC transcription? Second, what is the link between VAL1 and PHD-PRC2? Affinity purification of HA- and green fluorescent protein (GFP)–tagged VAL1 from Arabidopsis seedlings revealed VAL1 in vivo interactors: the PRC1 RING finger homolog AtBMI1A, the co-repressor SIN3-associated protein SAP18 (AtSAP18), and its two partners, the RNA-binding protein SR45 and the SAP-domain protein ACINUS (Fig. 4A, fig. S6D, and tables S1 and S2) (21). AtBMI1A has previously been found to interact with VAL1 to repress seed maturation genes through H2A lysine 121 ubiquitination (H2Aub) (22). We therefore tested accumulation of H2Aub at FLC during vernalization. Although H2Aub was detected at FLC after 4 weeks of cold, there was no difference between Col FRI and val1 FRI (fig. S8). However, H2Aub is not always required for efficient PRC1 repression (23), which can directly repress genes by triggering chromatin compaction (24).

Fig. 4 VAL1 nucleates silencing at the FLC locus.

(A) List of proteins identified by VAL1-HA (IP1 and IP2) and GFP-VAL1 (IP3 and IP4) affinity purification (IP, immunoprecipitation). (B and C) ChIP analysis of histone H3 acetylation (acetylH3) along the FLC locus in Col FRI (B) and val1 FRI (C). Numbers on the x axes are distances to the TSS (TSS = 0) and correspond to the schematic below each panel. (D) Days to flower after 12 weeks of vernalization for F2 plants from val1 FRI crosses with vin3 FRI and vrn2 FRI (table S6). WT indicates at least one functional allele for VAL1, VRN2, and VIN3. Lowercase indicates that both alleles are nonfunctional (val1, vin3, and vrn2). All F2 individuals are in FRI background. DNF, did not flower. (E) Schematic depicting sequence-specific binding of VAL proteins to the FLC nucleation region, targeting ASAP, HDA19, and potentially PRC1 activities to shut down transcription and thereby enable PHD-PRC2 nucleation. After this process, H3K27me3 covers the FLC locus to help maintain epigenetic silencing. Small double arrows indicate protein interactions identified in this work.

SAP18 is a component of the SIN3–histone deacetylase complex (HDAC) that in humans is required to enhance SIN3-mediated repression of transcription (25). VAL1 contains a plant-specific ethylene-responsive element binding factor–associated amphiphilic repression (EAR) motif that can physically interact with AtSAP18 to mediate histone deacetylase 19 (HDA19) recruitment (26). We hypothesized that VAL1 could bring HDA19 activity to the FLC locus. H3 acetylation is reduced at the FLC nucleation region during cold exposure (Fig. 4B) (27); this reduction was blocked in val1 FRI (Fig. 4C), as has been observed in vin3 (27). SAP18 has also been linked to RNA processing and degradation (28) and is a subunit of the conserved apoptosis- and splicing-associated protein (ASAP) complex, together with SR45 and ACINUS (21). Affinity purification of GFP-AtSAP18 from Arabidopsis confirmed association with AtSR45 and ACINUS (table S3). This suggests that the conserved function of ASAP and HDA19 is required for VAL1-mediated FLC silencing. Accordingly, hda19 and sr45 mutants, and to a lesser extent sap18 mutants, showed FLC up-regulation (fig. S9). Importantly, AtSAP18, AtSR45, and HDA19 were immunopurified with both VRN5-GFP (table S4) and VIN3-GFP (table S5), linking VAL1 association with a specific DNA sequence to Polycomb silencing. Combination of val1 FRI with mutants with loss of function in VRN2 (vrn2 FRI) and VIN3 (vin3 FRI) confirmed the close functionality of VAL1 and PHD-PRC2. val1 vin3 and val1 vrn2 plants were very delayed in flowering (Fig. 4D; table S6; and fig. S10, A to D) after 12 and 18 weeks of cold, and FLC expression remained high (fig. S10, E and F). This synergistic interaction is best explained by VAL1-VAL2 redundancy, with both proteins functioning through PHD-PRC2.

We propose that VAL1 binds sequence-specifically, probably as a homodimer, but potentially as a heterodimer with VAL2, to the RY motifs in the FLC nucleation region (Fig. 4E). This recruits the ASAP complex and potentially PRC1, resulting in the shutdown of transcription and reduced histone acetylation. In turn, these activities allow PHD-PRC2 nucleation and long-term epigenetic silencing of the locus. We cannot exclude the possibility of effects of VAL1 and VAL2 on FLC expression that are independent of the binding to FLC intron 1. The association of HDA19 with the PHD proteins suggests that it has multiple roles in the process, potentially interacting with VAL1 through SAP18 (26) for transcriptional repression and with VIN3 (and VRN5) to facilitate +1 nucleosome stabilization (27). It will now be important to investigate which components confer the switchlike on-off property that has been proposed for the nucleation event (29).

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 to S7

References (3039)

References and Notes

  1. Acknowledgments: We thank all members of the Dean and Howard research groups for discussions. We are grateful to G. Saalbach for assistance with mass spectrometry, M. Suzuki for providing val1 and val2 lines, J. Irwin for the Brassica FLC sequences, J. Long for the hda19-1 line, and A. Pendle for the 35S::GFP-AtSAP18 plasmid. We thank V. Coustham, B. Rutjens, and J. Mylne for contributions at the early phase of this project.The Dean laboratory is supported by the UK Biotechnology and Biological Sciences Research Council and a European Research Council Advanced Investigator grant. The supplementary materials contain additional data.
View Abstract

Stay Connected to Science

Navigate This Article