Short tandem repeats are important contributors to silencer elements in T cells

Abstract

The action of cis-regulatory elements with either activation or repression functions underpins the precise regulation of gene expression during normal development and cell differentiation. Gene activation by the combined activities of promoters and distal enhancers has been extensively studied in normal and pathological contexts. In sharp contrast, gene repression by cis-acting silencers, defined as genetic elements that negatively regulate gene transcription in a position-independent fashion, is less well understood. Here, we repurpose the STARR-seq approach as a novel high-throughput reporter strategy to quantitatively assess silencer activity in mammals. We assessed silencer activity from DNase hypersensitive I sites in a mouse T cell line. Identified silencers were associated with either repressive or active chromatin marks and enriched for binding motifs of known transcriptional repressors. CRISPR-mediated genomic deletions validated the repressive function of distinct silencers involved in the repression of non-T cell genes and genes regulated during T cell differentiation. Finally, we unravel an association of silencer activity with short tandem repeats, highlighting the role of repetitive elements in silencer activity. Our results provide a general strategy for genome-wide identification and characterization of silencer elements.

Figure 1.

CapSTARR-seq for silencer assessment. (A) Schematic of the CapSTARR-seq strategy to assess the silencer activity in the P5424 mouse T cell line. (B) Distribution of CapSTARR-seq signal (log₂) in the different conditions for DHS and random captured regions. The threshold for putative silencers (log₂(FC) ≤ −1) and enhancers (log₂(FC) ≥ 1) is indicated. Statistical analysis was performed using Wilcoxon test, P-values are displayed (ns: not significant). (C) The UCSC genomic track of Mouse NCBI37/mm9 around DHS19266 showing the log₂ STARR-seq signal of individual clones, the captured DHS region, the core silencer and the silencer score of the region. (D) Bar plot showing the number of unique and shared silencer candidates identified with the three different CapSTARR-seq libraries. (E) Venn diagram displaying the overlap between the silencer candidates identified with the three different CapSTARR-seq libraries. (F) Example of a putative silencer identified with the three promoter-based CapSTARR-seq strategies. The signal for each CapSTARR-seq experiment and the corresponding Input are displayed. (G) Bar plot displaying the proportion of silencer candidates that are proximal (<1 kb) or distal (>1 kb) to the closer TSS. (H) Genomic distribution of silencer candidates compared to the whole set of DHSs. Bar plots represent the –log₁₀(P-value) of the negative binomial test computed by OLOGRAM. Depletion is represented by negative values.

Figure 2.

Validation of the CapSTARR-seq approach. (A) Luciferase reporter assays in P5424 cells of DHSs defined as putative silencers by CapSTARR-seq (green) or with log₂ STARR-seq signal close to zero (black). The promoter-based CapSTARR-seq where the silencer was identified is indicated in the bottom panel. Data represent the normalized fold change over the pSV40 vector control. Error bars show s.d. from three independent transfections (***P-values < 0.001, **P-values < 0.01, *P-values < 0.05; two-sided Student's t-test). (B) Comparison of luciferase activity between silencer candidates and control regions. The two-sided Student's t-test is shown. (C) Assessment of orientation-dependent silencer activity for a subset of identified silencers (F: forward; R: reverse). (D) FACS analysis for GFP expression assessing DHS12366 silencer activity.

Figure 3.

Chromatin features associated with silencers. (A) Average profiles of H3K27ac and H3K4me3 (ChIP-seq), H3K9me3 (CUT&RUN), H3K27me3 (CUT&Tag) signal from P5424 cells centered on putative silencers (red), control regions (black) and putative enhancers (green) with a window of ±5kb. (B) Average profiles and heatmaps of pPGK silencers clustered in the function of the signal of different histone marks and CTCF in four distinct groups. (C) Comparison of pPGK silencer activity within each cluster. Statistical analysis was performed using the Wilcoxon test, significant P-values are displayed.

Figure 4.

Gene expression and GO term analysis. (A) Violin plot comparing the expression in P5424 cells of genes associated with putative silencers (red), control regions (black) and putative enhancers (green). The mean values are indicated by a dot. Statistical analysis was performed using the Wilcoxon test; P-values are displayed. (B, C) Heatmap of top five GO terms analysis for mouse phenotype (B) and MSigDB pathway (C) enriched in genes associated with silencers or SCP1 enhancers.

Figure 5.

Functional validation of silencers by CRISPR–Cas9 genomic editing. (A) Genomic tracks displaying the indicated RNA-seq and chromatin signals as well as DHS and TADs surrounding the silencer regions. The relative expression values based on P5424 RNA-seq (TPM) are indicated in bracket. (B) Gene expression analysis of genes around the silencer region in wild-type (WT) and deleted P5424 clones. The expression of the genes was normalized to Rlp32 and with respect to the WT value. Error bars, s.d.: ***P-values < 0.001, **P-values < 0.01, *P-values < 0.05, two-sided Student's t-test. (C) Heatmap displaying the relative gene expression of genes around DHS12366 locus in T cell populations and other cell types.

Figure 6.

TFBS associated with silencers and site-directed mutagenesis. (A) Heatmap displaying the enrichment score of the top 10 clustered TF motifs enriched in each of the silencer sets, as well as, in the SCP1 enhancers. (B) Dot plots displaying the mean activity of silencers carrying a given TFBS against the enrichment score for the same TFBS. Only significantly enriched TFBS are displayed. TFs of interest are highlighted. (C–E) Validation of REST (C), SMAD3 (D) and MAFK (E) binding sites impact on silencer activity. The left panels display the mutated nucleotides. The right panels display the luciferase reporter assay in wild-type and mutated silencers (***P-values < 0.001, **P-values < 0.01, *P-values < 0.05; two-sided Student's t-test). (F) Bar plots showing the proportions of silencers harboring REST binding sites among the SCP1, pPGK or shared between the two libraries. (G) Expression level of genes associated with silencers containing or not REST binding sites. Significance was assessed by the Wilcoxon test. (H, I) The proportion of genes associated with all DHS, all silencers or silencers containing or not REST binding sites that are tissue-specifics excluding T-cells (C) or T-cell regulated genes (I).

Figure 7.

Functional validation of DHS23650 silencer by CRISPR/Cas9 system. (A) Genes ranked in the function of their expression variance across T-cell differentiation. The top 5% of variable genes are highlighted in blue. The Nfkbid and Hcst genes associated with DHS23650 silencer are also shown. (B) All putative silencers were associated with genes located in a window of 100 kb upstream and downstream. Of these, 516 genes (corresponding to 615 silencers) were part of the top 5% variable genes during T-cell differentiation. (C) Relative expression of the Nfkbid and Hcst genes during T-cell differentiation. (D) Top panel: Hi-C data and TADs in DP thymocytes surrounding the DHS12366 silencer. Bottom panel: genomic tracks displaying the indicated RNA-seq and ChIP-seq signals in P5424 cells stimulated or not with PMA and Ionomycin (36). (E) Gene expression analysis of genes around the DHS23650 locus in wild-type, ΔDHS23650 R1 and ΔDHS23650 R2 P5424 clones. (F) Gene expression analysis of Nfkbid in wild-type, ΔDHS23650 R1 and ΔDHS23650 R2 P5424 clones stimulated or not with PMA and Ionomycin. The expression of the genes was normalized to Rlp32 and with respect to the WT value as 1. Error bars, s.d.: ***P-values< 0.001, **P-values < 0.01, *P-values < 0.05 two-sided Student's t-test.

Figure 8.

Analysis and validation of simple tandem repeats. (A) UCSC genomic track of mouse NCBI37/mm9 around the DHS23650 silencer displaying the individual clonal activity, the DHS region (silencer), the core silencer, the RepeatMasker track and the repeat units. (B) Luciferase reporter assay of wild-type and mutated DHS23650 silencer. The impact of STR repeats on silencer activity was assessed by either deleting the STR region or by mutating the indicated number of STR binding sites. Data represent the normalized fold change over the pSV40 vector. Error bars show s.d. from three independent transfections (***P-values < 0.001, **P-values < 0.01, *P-values < 0.05; two-sided Student's t-test). (C) Enrichment of repetitive elements at putative silencers. Bar plots represent the –log₁₀(P-value) of the negative binomial test computed by OLOGRAM. (D, E) Activity of silencers found with the pPGK library associated with the presence of SINE (D) or STR (E) repetitive elements. Significance was assessed by a Wilcoxon test. (F) Average profiles of silencers with or without STR (left panels) or SINE (right panels) with a window of ±5 kb showing the enrichment of different histone marks in P5424 (H3K4me3, H3K27ac, H3K9me3, H3K27me3) or hematopoietic lineages (H4k20me3). (G) STARR-seq signal of silencers in the function of the number of repeat units within the STRs. Significance was assessed by a Wilcoxon test. (H) Comparison of the proportion of regions from pPGK with different lengths of STRs. Control enhancers are from the SCP1 library. (I) Activity of pPGK silencers ranked by the pattern of the STR. (J) Motif enrichment and associated TFs found in the subset of pPGK silencers containing STR with a log₂(signal) ≤ −1.5 (square in the panel I).