Gene silencing dynamics are modulated by transiently active regulatory elements

Abstract

Transcriptional enhancers have been extensively characterized, but cis-regulatory elements involved in acute gene repression have received less attention. Transcription factor GATA1 promotes erythroid differentiation by activating and repressing distinct gene sets. Here, we study the mechanism by which GATA1 silences the proliferative gene Kit during murine erythroid cell maturation and define stages from initial loss of activation to heterochromatinization. We find that GATA1 inactivates a potent upstream enhancer but concomitantly creates a discrete intronic regulatory region marked by H3K27ac, short noncoding RNAs, and de novo chromatin looping. This enhancer-like element forms transiently and serves to delay Kit silencing. The element is ultimately erased via the FOG1/NuRD deacetylase complex, as revealed by the study of a disease-associated GATA1 variant. Hence, regulatory sites can be self-limiting by dynamic co-factor usage. Genome-wide analyses across cell types and species uncover transiently active elements at numerous genes during repression, suggesting that modulation of silencing kinetics is widespread.

Keywords: FOG1/NuRD complex; GATA1; erythroid differentiation; gene silencing; transcriptional regulation; transient enhancer.

Figure 1.. Intronic GATA1 binding sites transiently gain H3K27ac and eRNAs during Kit gene silencing.

A) Upper panel: cell lines and time points used in this study. Lower panel: Kit primary transcript (pT) levels normalized to Gapdh and relative to the mean of G1E-ER4 -E2 replicates. n=3 for G1E, n=4 for all other time points. Error bars represent standard error of the mean (SEM). One-way ANOVA: ****P < 0.0001. B) Anti-H3K27ac ChIP-seq tracks for G1E and G1E-ER4 -E2, +E2 13 hr and +E2 24 hr in green. n=2 for each condition, average of replicates depicted. Anti-GATA1 ChIP-seq track for G1E-ER4 +E2 13hr from Jain et al. in red. GATA1-ER occupied sites of interest are highlighted by boxes. C) PRO-seq tracks for the Kit locus. n=3 for G1E-ER4 -E2 and n=2 for G1E, G1E-ER4 +E2 13 and 24 hr, average of replicates depicted. The insets on the right show enlarged images of boxed GATA1 sites of interest. Red: plus strand; blue: minus strand. Related to Figure S1.

Figure 2.. Transient +49 kb element can delay Kit gene silencing and prevent heterochromatinization.

A) 4C-seq in G1E (purple) and G1E-ER4 +E2 13 hr (orange) using the Kit promoter as a viewpoint. Tracks are plotted on top of each other with the lowest one for each DpnII fragment depicted in grey. Significance of chromatin interactions was calculated using peakC (settings qWd:1.5, qWr:1.25, asterisk = significant interaction). B) Same as in (A) for G1E (purple) and G1E-ER4 +E2 13 hr (orange) using Kit +5 kb as a viewpoint. (C) Same as in (B) for G1E (purple) and G1E-ER4 +E2 24 hr (red). D) Kit primary transcript (pT) levels up to 24 hr in unedited control (n=3) and Δ+49 kb (n=3) subclones of clone A, normalized to Gapdh and relative to −E2 for each subclone. Error bars represent SEM. An unpaired, two-tailed t-test was done for each time point: P values indicated, * P < 0.05, ** P < 0.01. E) Same as in (D) for clone B. F) Same as in (D) for clone C. G) 4C-seq in clone A control subclone 1 at time points -E2 (yellow) and +E2 13 hr (orange) using the Kit +5 kb as a viewpoint. H) Same as in (G) for clone B control subclone 1. I) Same as in (G) for clone C control subclone 1. J) Left, schematic of the experimental set up. Right, Kit pT levels normalized to Gapdh and relative to the mean of G1E-ER4k control clones for G1E-ER4k clones (n=9), −114 kb enhancer deletion clones (n=11) and +49 kb with subsequent −114 kb enhancer deletion clones (n=6). Error bars represent SEM. K) Anti-H3K27ac (green) and anti-H3K27me3 (dark blue) ChIP-seq tracks. G1E-ER4k cells before and after 24 hr differentiation, two Δ−114 kb ~10% Kit clones and two Δ−114 kb 0% Kit clones are depicted. GATA1 sites of interest are highlighted by boxes. Exons are covered by a transparent white box to mask H3K27ac signal resulting from Kit cDNA overexpression. L) 4C-seq for G1E-ER4k (dark grey) and Δ−114 kb 10% clone 1 (yellow) using the Kit +5 kb site as a viewpoint. Related to Figure S2.

Figure 3.. Genome-wide analysis reveals many more transient regulatory elements at downregulated genes.

A) Genes with significant ANOVA results across timepoints grouped using unsupervised hierarchical clustering. Colors correspond to the average normalized expression of replicates for each time point (number of standard deviations from G1E) with blue: lower and red: higher. B) Same as in (A) for bidirectionally transcribed dREG peaks (non-TSS sites only). C) PRO-seq tracks for the Myc and Mitoferrin enhancers as well as a newly identified transient intronic site for Prmt3. n=3 for G1E-ER4 -E2 and n=2 for G1E, G1E-ER4 +E2 13 and 24 hr, average of replicates depicted. D) Same as in (C) for the Atp1a1 locus. E) 4C-seq for G1E (purple) and G1E-ER4 +E2 13 hr (orange) using the Atp1a1 promoter as a viewpoint. Tracks are plotted on top of each other with the lowest one for each DpnII fragment depicted in grey. Significance of chromatin interactions was calculated using peakC (settings qWd:1.5, qWr:1.25, asterisk = significant interaction). F) Correlation between dREG cluster 7 and all gene clusters. Colors represent the odds ratio of observed over expected values before and after fitting a linear model with blue: lower than 1 and red: higher than 1. Number of asterisks indicates statistical significance of the enrichment based on a Fisher’s exact test: * P < 0.05, ** P < 0.01 and *** P < 0.001. Arrows indicate significant correlations between dREG cluster 7 peaks and downregulated gene clusters 7 and 11. Related to Figure S3 and Table S1.

Figure 4.. Transient elements become persistent in the absence of FOG1/NuRD engagement.

A) Fraction of dREG peaks bound by GATA1-ER for each cluster. B) Anti-H3K27ac ChIP-seq tracks for G1E-ER4 (-E2, +E2 13 hr and +E2 24 hr) in green and G1E-GATA1(V205M)-ER (-E2, +E2 13 hr and +E2 24 hr) in black at the Kit locus. n=2 for each condition, average of replicates depicted. GATA1 bound regions of interest are highlighted by boxes. C) Kit primary transcript (pT) levels normalized to Gapdh and relative to the mean of -E2 samples for G1E-ER4 (n=3) and G1E-GATA1(V205M)-ER (n=3) cells. Error bars represent SEM. One-way ANOVA: ****P < 0.0001. D) Quantification of H3K27ac ChIP-seq enrichment at +49 kb transient element after GATA1-ER and GATA1(V205M)-ER activation for two independent replicates. Difference (calculated by subtraction) between rlog normalized read counts of indicated timepoints and -E2 are depicted. Error bars represent standard deviation (SD). E) H3K27ac dynamics at all GATA1-ER and GATA1(V205M)-ER bound cluster 7 dREG peaks in G1E-ER4 and G1E-GATA1(V205M)-ER cells. Relative enrichment represents number of standard deviations from G1E condition. Two-tailed, paired t-test: **** P < 0.0001. Related to Figure S4.

Figure 5.. Co-factors define the output of GATA1-occupied sites.

A) Anti-CHD4 ChIP-seq tracks for -E2, +E2 13 hr and +E2 24 hr after GATA1-ER (G1E-ER4, red) as well as GATA1(V205M)-ER (G1E-GATA1(V205M)-ER, dark green) activation. n=2 for each condition, average of replicates depicted. B) Quantification of CHD4 ChIP-seq enrichment at +49 kb transient element depicted in (A) for G1E-ER4 and G1E-GATA1(V205M)-ER. Difference (calculated by subtraction) between rlog normalized read counts of indicated timepoints and -E2 are depicted. Error bars represent SD. C) Same as in (A) for LDB1. D) Same as in (B) for LDB1. Related to Figure S5.

Figure 6.. Transient regulatory elements in primary murine cells during erythroid differentiation.

A) Strategy to isolate and analyze primary mouse fetal liver cells at different stages of differentiation based on cell surface markers KIT (CD117) and CD71. B) Anti-H3K27ac ChIP-seq tracks at the Atp1a1 locus for populations from mouse fetal liver cells described in (A). C) Same as in (B) for the Kit locus. D) RT-qPCR for Kit primary transcripts (pT) in populations described in (A). E) Same as in (B) for the Eng locus. F) Same as in (B) for the Pex14 locus. G) Same as in (B) for the Mrpl23 locus. Related to Figure S6 and S7.