Quantitative regulation of FLC via coordinated transcriptional initiation and elongation

Abstract

The basis of quantitative regulation of gene expression is still poorly understood. In Arabidopsis thaliana, quantitative variation in expression of FLOWERING LOCUS C (FLC) influences the timing of flowering. In ambient temperatures, FLC expression is quantitatively modulated by a chromatin silencing mechanism involving alternative polyadenylation of antisense transcripts. Investigation of this mechanism unexpectedly showed that RNA polymerase II (Pol II) occupancy changes at FLC did not reflect RNA fold changes. Mathematical modeling of these transcriptional dynamics predicted a tight coordination of transcriptional initiation and elongation. This prediction was validated by detailed measurements of total and chromatin-bound FLC intronic RNA, a methodology appropriate for analyzing elongation rate changes in a range of organisms. Transcription initiation was found to vary ∼ 25-fold with elongation rate varying ∼ 8- to 12-fold. Premature sense transcript termination contributed very little to expression differences. This quantitative variation in transcription was coincident with variation in H3K36me3 and H3K4me2 over the FLC gene body. We propose different chromatin states coordinately influence transcriptional initiation and elongation rates and that this coordination is likely to be a general feature of quantitative gene regulation in a chromatin context.

Keywords: COOLAIR; FCA; alternative polyadenylation; autonomous pathway; chromatin.

Fig. 1.

Large increases in RNA are associated with small changes in Pol II occupancy. (A) RNA fold up-regulation in fca-9 and fld-4 mutants compared with Col: spliced and unspliced FLC (∼25×), proximal (∼2×) and distal COOLAIR (∼13×). The model values are the fits to the experimental data. Experimental values are mean ± SEM from three to six independent samples. (B) Schematic illustration of a scenario where transcription initiation is the only difference between Col and fca-9, so that a 25× fold change in Pol II occupancy should be observed as illustrated on the Right. (C and D) ChIP experiments assaying Pol II occupancy across FLC using the antibodies anti CTD 8WG16 (C) and anti Ser² P CTD 3E10 (D). The bar charts at the Bottom indicate Pol II levels at various control genes. Three overlapping primer pairs are used to measure IGN5 expression (P1–P3). Values are mean ± SEM from two independent samples, with data presented as the ratio of Pol II at FLC/input at FLC to Pol II at ACT7 (−995)/input at ACT7 (−995).

Fig. S1.

Pol II levels at FLC are increased only approximately twofold in fca-9 and fld-4 compared with Col (related to Fig. 1). (A) ChIP experiments assaying Pol II (using anti-CTD 8WG16) across FLC and at internal controls. (B) ChIP experiments assaying Pol II (using anti-Ser²P CTD 3E10) across FLC and at internal controls. (C) Western blot detection of Pol II by using 8WG16 antibody in Arabidopsis. 8WG16 recognizes both hypophosphorylated (lla) and hyperphosphorylated (llo) CTD of NRPB1. Nuclear extract was separated on 6% SDS/PAGE gel. (A and B) These two assays were done by adopting less stringent washing steps after immunoprecipitation compared with the data shown in Fig. 1. The dashed line indicates an upper limit on the background level in this ChIP experiment, equalling the level at IGN5. Values are mean ± SEM from two independent samples, with data normalized to 1% of input.

Fig. S2.

Linearity of Pol II ChIP assay (related to Fig. 1). (A) Schematic illustration of the experimental procedure testing how the Pol II ChIP signal reflects the change of Pol II concentration locally at FLC. (B) 8WG16 Pol II ChIP signal from different FLC chromatin dilutions at 5′ region of FLC. (C) 3E10 Pol II ChIP signal from different FLC chromatin dilutions at 5′ region of FLC. (B and C) Theoretical values indicate the expected Pol II ChIP signal if the Pol II ChIP signal scales linearly with the local concentration of Pol II. The bar chart on the Right shows the ChIP signal at FLC and control genes from dilution 1. The signals are well above background (at IGN5). Values are mean ± SEM from two independent samples, with data normalized to Actin7 (Left) or %input (Right).

Fig. 2.

Small changes in Pol II occupancy can be explained by coordinated changes in transcription initiation and elongation. (A) Schematic of FLC locus and outline of the mathematical model for FLC transcription (details in Supporting Information). Black boxes indicate sense exons; gray boxes indicate proximal (Upper) and distal (Lower) antisense exons. (B) Total (sum of sense and antisense) model Pol II levels in Col and fca-9 across FLC. The fld-4 mutant model results are identical to fca-9. Shown on the Right is a schematic of the convolution process with experimental Pol II ChIP fragment size distribution (shown in Fig. S3). (C) Total Pol II levels in Col and fca-9 across FLC from the model convolved with experimental Pol II ChIP fragment size distribution. (D) Experimental and model Pol II fold up-regulation. Experimental values are mean ± SEM from two to five independent samples, including data shown in Fig. 1 C and D, and Fig. S1. Model fold changes are ratio of profiles shown in C.

Fig. 3.

Combination of increased initiation and elongation, with cotranscriptional splicing and lariat degradation, leads to distinct RNA profiles along FLC intron1. (A) Schematic indicating intronic nascent RNA, RNA_nasc (blue lines), arising from Pol II (blue circles) elongating through the intron and from unspliced RNA_s with full-length intron. Once Pol II has passed the intron acceptor site (IA), splicing can occur. Initiation, elongation, and splicing rates are F, v, and k_s, respectively. Analytic expression for RNA_nasc is shown below. (B) Schematic (Left) indicating model profiles of nascent RNA along FLC intron1 in fca-9 and Col. Between fca-9 and Col, F and v are coordinately increased, but with the same k_s. This generates a characteristic pattern of intronic nascent RNA fold changes between fca-9 and Col (Right) with analytic expression shown. (C) Modeled and experimentally measured chromatin-bound RNA fold changes along FLC intron1. The lower increase toward the 3′ end in fld-4 is due to increased splicing rate as shown experimentally in D. Crosses indicate positions where data are from three different, overlapping primer sets that each show similar results (Fig. S4). (D) Estimate of FLC intron1 splicing efficiency (intron cleavage rate) in fld-4 and fca-9, normalized to the level in Col. Values are mean ± SEM from three independent samples. Asterisks indicate statistical significance: for all of the figures in this study, *P < 0.05, **P < 0.01, ***P < 0.001, two-sided unpaired t test, unless specified otherwise. (E) Schematic showing effect of 5′ to 3′ intronic RNA degradation on lariat RNA levels (RNA_lariat). Full-length lariat RNA results from splicing and is degraded with rate k_i; ID: intron donor site. These degradation intermediates, together with the nascent RNA described in A, make up total intronic RNA. Fold up-regulation then generates the characteristic profiles shown. Analytic expressions for RNA_lariat and total intronic RNA fold changes are shown. (F) Modeled and experimentally measured total RNA fold changes along FLC intron1. (C and F) Experimental values are mean ± SEM from at least three independent samples. Absolute levels are shown in Fig. S4.

Fig. S4.

Intronic RNA levels increase gradually along FLC intron1 (related to Fig. 3). (A) Chromatin-bound RNA levels in fca-9, fld-4, and Col, along FLC intron1. (B) Total RNA levels in fca-9, fld-4, and Col along FLC intron1. Primers along Actin7 were checked as a control. (A and B) Experimental values are mean ± SEM from three independent samples. (C) Proteins in different fractions obtained during chromatin-bound RNA preparation. Same volume from each fraction was loaded and separated in 4–20% gradient NuPAGE gel and stained with Coomassie blue. (D) The size distribution of chromatin-bound RNA. No obvious band of ribosomal RNA was detected. (E) Enrichment of unspliced FLC (intron2 and intron3 unspliced) over spliced FLC (intron4, intron5, and intron6 spliced) in total, nuclear, and chromatin-bound RNA fractions. Values are mean ± SD from three independent samples. (F) Majority of Pol II (8WG16) and H3 was preserved in the chromatin fraction. Different fractions obtained during chromatin-bound RNA preparation were gel separated (as in C), followed by Western blot detection of Pol II and H3. (G) Model and experimentally measured chromatin-bound RNA fold up-regulation in fca-9 and fld-4 compared with Col at FLC exon1 and exon1–intron1 junctions. The related mathematical analysis can be found in Computational Modeling.

Fig. S5.

Detection of prematurely terminated FLC transcripts. (A) Schematic illustration of premature termination and mathematical analysis of its influence on the Pol II occupancy across FLC. (B) Schematic illustration of prematurely terminated transcripts detected from 3′-RACE. The upper panel indicates the major FLC transcript with regions omitted after exon2. The lower panel indicates the splicing pattern of prematurely terminated FLC transcripts. Exons and introns are indicated by black boxes and dashed lines, respectively. The gray box indicates a region containing polyadenylation sites detected in 3′-RACE. Numbers above the schematic indicate positions of intron or exon boundaries (in base pairs post-TSS). (C) Number of clones mapped at each pA site detected by 3′-RACE in Col, fca-9, and 35S::FCA. All clones have a small intron spliced as shown in B. Position of pA site (in base pairs post-TSS) is marked below. (D) Percentage of polyadenylated, nonpolyadenylated, and splicing intermediate RNAs mapped in Col, fca-9, and 35S::FCA. Apart from clones with poly(A) tail, we also mapped many transcripts without a poly(A) tail. Some of these are likely splicing intermediates, as their 3′ ends correspond to the 3′ of an exon with the previous introns removed and exons ligated. Other unpolyadenylated transcripts are likely to be Pol II-bound as their 3′ ends mapped to the middle of an exon or intron. At least 100 clones were sequenced for each genotype. (E) Abundance of transcripts with the alternatively spliced intron (B) in different genotypes. Levels were measured using a primer set with one primer spanning exon–exon junction. Values are mean ± SEM from three independent samples; data are shown as abundance relative to intronic FLC level (612–698 bp post-TSS) in each genotype.

Fig. S6.

Relative RNA expression level of FLC spliced and unspliced in Col, xrn3-3, and xrn2-1. Level in Col was set as 1. Values are mean ± SD from three independent samples.

Fig. 4.

FLD enrichment at the FLC locus is associated with changed histone modifications. (A) FLD-TAP ChIP enrichment across FLC in Col and FLD-TAP/fld-4. Values are mean ± SEM from two independent samples, with data presented as enrichment at FLC relative to enrichment at STM. (B–F) ChIP across FLC in Col, fca-9 and fld-4 measuring H3K4me2 (B), H3K4me3 (C), H3Ac (D), H3K36me3 (E), and H3K27me3 (F). Values are mean ± SEM from two independent samples, with data normalized to H3. Values at the control genes STM, ACT7, and TUB8 are shown on the Right. H3/input values can be found in Fig. S7.

Fig. 5.

Coordination of initiation and elongation at FLC in the H3K36 methyltransferase-deficient sdg8 mutant. (A) Total RNA levels along FLC intron1. Model is as described in Fig. 2. All values are relative to fca-9. Experimental values are mean ± SEM from three independent samples and are averaged from overlapping primer sets (Fig. S8). (B) Working model of how FLC expression is quantitatively regulated through coordination of transcription initiation and elongation. In the absence of FCA/FLD, H3K36me3 is increased at FLC through SDG8 function, and this promotes fast transcription initiation and elongation. In the presence of FCA/FLD, antisense processing triggers a reduction of H3K4me2, loss of H3K36me3, and an increase in H3K27me3, which reduces transcription initiation and slows elongation.

Fig. S7.

The FLD-TAP transgene is functional, and H3 levels at FLC are not affected in fld-4 and fca-9 mutants (related to Fig. 4). (A) Expression level of the FLD-TAP transgene compared with endogenous FLD. (Top) Northern blot probed with an FLD cDNA. (Bottom) Ethidium bromide staining of the gel showing equal loading. Genotypes are labeled at Top of images. (B) FLD-TAP transgene fully complemented the fld-4 mutation. Total leaf number of plants grown under long day conditions is shown. Genotypes are labeled on the x axis. Data shown are means ± SDs of 20 plants of each genotype. (C) Western blot detection of FLD-TAP before and after immunoprecipitation. FLD-TAP can be detected as a single band at around 100 kDa in the nucleus, and it was enriched after immunoprecipitation following the same procedure as in the ChIP. (D) Histone H3 levels do not differ significantly among Col, fld-4, and fca-9 at FLC. Values at control genes, STM, ACT7, and TUB8 are shown on Right. Values are mean ± SEM from two independent samples, with data normalized to %input.

Fig. S8.

(A) Relative RNA expression level of FLC spliced and unspliced in Col, sdg8, sdg8 fca-9, and sdg8 fld-4. Level in Col was set as 1. Values are mean ± SD from three independent samples. (B) Total RNA levels in fca-9, sdg8 fca-9, and Col along FLC intron1. Level in Col was set as 1. Values are mean ± SEM from three independent samples.

Fig. S3.

Pol II levels obtained in ChIP are affected by the fragment size (related to Fig. 2). (A) Fragment size intensity in Pol II ChIP assay. The DNA was recovered from sonicated chromatin and was resolved on a gel with DNA size markers running in parallel. (B) Fragment size distribution of lengths ranging between 100 bp and 1 kb, as extracted from A. See Determining the Pol II ChIP Fragment Size Distribution for details. (C) Normalized Pol II occupancy resulting from the experimental fragment distribution shown in B, in presence of a single Pol II located at the origin.