Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 26 (11), 1478-1489

Binding of Nuclear Factor κB to Noncanonical Consensus Sites Reveals Its Multimodal Role During the Early Inflammatory Response

Affiliations

Binding of Nuclear Factor κB to Noncanonical Consensus Sites Reveals Its Multimodal Role During the Early Inflammatory Response

Petros Kolovos et al. Genome Res.

Abstract

Mammalian cells have developed intricate mechanisms to interpret, integrate, and respond to extracellular stimuli. For example, tumor necrosis factor (TNF) rapidly activates proinflammatory genes, but our understanding of how this occurs against the ongoing transcriptional program of the cell is far from complete. Here, we monitor the early phase of this cascade at high spatiotemporal resolution in TNF-stimulated human endothelial cells. NF-κB, the transcription factor complex driving the response, interferes with the regulatory machinery by binding active enhancers already in interaction with gene promoters. Notably, >50% of these enhancers do not encode canonical NF-κB binding motifs. Using a combination of genomics tools, we find that binding site selection plays a key role in NF-κΒ-mediated transcriptional activation and repression. We demonstrate the latter by describing the synergy between NF-κΒ and the corepressor JDP2. Finally, detailed analysis of a 2.8-Mbp locus using sub-kbp-resolution targeted chromatin conformation capture and genome editing uncovers how NF-κΒ that has just entered the nucleus exploits pre-existing chromatin looping to exert its multimodal role. This work highlights the involvement of topology in cis-regulatory element function during acute transcriptional responses, where primary DNA sequence and its higher-order structure constitute a regulatory context leading to either gene activation or repression.

Figures

Figure 1.
Figure 1.
Analysis of gene expression at the level of intronic RNA. HUVECs were treated with TNF for 0–90 min, total RNA sequenced, and changes in intronic RNA levels assessed using the iRNA-seq pipeline. (A) Log2-normalized mean counts of reads mapping to introns at the different times. Only genes up- (orange) and down-regulated (blue) by TNF are shown. (B) Browser views of typical RNA-seq coverage (reads per million) along a TNF-induced (HIVEP2) and a TNF-repressed gene (ARHGAP18). (C) Log2-fold changes of intronic RNA levels for up- (≥0.6) and down-regulated genes (≤−0.6) at the different times compared to 0 min. (*) Significantly different mean; P < 0.01, two-tailed Student's unpaired t-test. (D) The five most significantly enriched GO terms relevant to up- and down-regulated genes. (E) Venn diagrams showing shared and unique up- (left) and down-regulated genes (right) at the different times post-stimulation. (F) Log2-fold changes of intronic RNA levels (30- versus 0-min levels) for TNF-induced/-repressed genes lying alone (“1”), in pairs (“2”), or in groups of three or more (“≥3”) in a given TAD (the total number of genes in each group are shown on top of or below each box plot). (*) Significantly different mean; P < 0.01, two-tailed Student's unpaired t-test.
Figure 2.
Figure 2.
RELA binds already-active enhancers “with” or “without” a canonical binding motif. HUVECs were treated with TNF for 0 or 30 min and NF-κB binding assessed (via RELA ChIP-seq). (A) ChIP- and RNA-seq coverage (reads per million) of the responsive NFKBIA locus; “motifs” refers to canonical NF-κB binding motifs. Closed and open arrows (bottom): a RELA peak overlapping a canonical and a noncanonical recognition motif, respectively. (B) (Top) The overlap between 0-min H3K4me1 (gray), H3K27ac (teal), and 30-min RELA (orange) ChIP-seq peaks; 5366 peaks are bound by RELA at 30 min. (Bottom) Percent RELA-bound peaks that carry (or not) the canonical motif, are intragenic, or overlap a TSS or a CTCF peak. (C) Relative enrichment of transcription factor (TF) binding motifs in DNase-hypersensitive footprints around “with” (blue) or “without” (red) canonical RELA binding motifs. TFs induced or repressed by TNF are colored orange and blue, respectively. (D) Occupancy (read tags per bp) for MAX, FOS, JUN, and GATA2 ChIP-seq in the 2 kbp around RELA peaks “with” (orange) or “without” (black) canonical motifs. (E) Conservation (among 17 vertebrates) in the 5 kbp around “with”/”without” RELA or CTCF peaks. (F) The most significantly enriched GO terms (biological process) associated with genes assigned to 30-min RELA peaks “with”/”without” motifs in the same TAD; GO terms for up-regulated genes are shown for comparison. Note that any gene linked to both “with” and “without” peaks is classified as “with.” (G) Heat maps illustrating RNAPII signal (grouped hierarchically into five clusters) in the 4 kbp around RELA peaks “with”/”without” motifs. (H) Changes in eRNAs levels (data from Caudron-Herger et al. 2015) around RELA peaks from “with” (top) and “without” (bottom) cluster 4. (AS) Antisense-strand signal, (S) sense-strand signal. (I) The fraction of RELA peaks per cluster (as in G) overlapping H3K27ac peaks at the different times.
Figure 3.
Figure 3.
Global analysis of contacts made by RELA-bound enhancers “with” or “without” motif. HUVECs were stimulated with TNF for 0 or 30 min, and ChIA-PET was performed after pulling-down chromatin complexes associated with active RNA polymerase II isoforms. (A) Browser views of typical ChIA-PET interactions at gene loci induced (NFKBIA) or repressed (CTGF) by TNF. 0-min ChIA-PET contacts are shown alongside ChIP-seq data for RELA (30 min) and CTCF (0 min), and positions of canonical NF-κB motifs. RELA-bound contacts are highlighted (magenta). (B) More than half of the 5366 RELA-bound enhancers (defined as in Fig. 2B) are involved in 0-min ChIA-PET interactions, and are connected to other already-active enhancers (orange), gene TSSs (circles), or CTCF cites (hexagons) to different extents. Fifty-seven percent of RELA-bound enhancers carry noncanonical motifs (gray highlight). (C) Boxplots showing H3K27ac levels (using ChIP-seq ±TNF) at each of the 2791 RELA-bound enhancers present in the 0-min ChIA-PET data. Enhancers are grouped according to their underlying motif as “with” (orange) or “without” (gray). (*) Significantly different mean; two-tailed, unpaired, Student's t-test. (D) Log2-fold changes of intronic RNA levels (30- or 60- versus 0-min; only changes of ±0.6 or more are shown) for genes connected with RELA-bound enhancers in 0-min ChIA-PET data. Genes associated to “with” and “without” RELA peaks are denoted by orange and gray points, respectively, and their numbers per quartile are shown (right). (E) Pie chart showing that 73% of the 305 TSSs of up-regulated genes in 0-min ChIA-PET are prelooped to already-active enhancers. Boxplots depict the log2 fold-change in intronic RNA levels for genes contacting “with” (orange), “without” (gray) or non-RELA-bound peaks (light green). (F) Pie charts and boxplots as in E, but for the 39% of the 235 TSSs of down-regulated genes in 0-min ChIA-PET that are prelooped to already-active enhancers. (*) Significantly different mean; two-tailed, unpaired, Student's t-test.
Figure 4.
Figure 4.
Spatial interactions in a TNF-responsive locus. HUVECs were stimulated with TNF for 0, 30, or 60 min, and 3C-seq performed using the TSSs of the BMP4, CDKN3, and SAMD4A genes as viewpoints. (Top) Browser view, including ChIP-seq data, of 2 Mbp on Chromosome 14 (ideogram on top). Rectangles 1–4: TADs. Colored triangles: 3C-seq viewpoints. Arrows (bottom): RELA-bound enhancers contacted by BMP4 (white), CDKN3 (blue), and SAMD4A (maroon). (Bottom) Zoom-in on the interactions formed by TNF-suppressed BMP4 and TNF-induced SAMD4A TSSs. 3C-seq profiles are shown alongside JDP2 and RELA ChIP-seq data at 0 and 30 min post-stimulation.
Figure 5.
Figure 5.
C646 treatment represses the proinflammatory expression program. HUVECs were treated with C646 for 1 h, followed by TNF stimulation for 0 or 30 min and analyzed. (A) Top: cumulative separation of RELA-bound (30 min) and CTCF peaks (0 min) in the 2.8-Mbp SAMD4A locus. (Bottom) Browser view of RNA- and ChIP-seq data (in reads per million) along SAMD4A and CGRRF1. The RELA-bound super-enhancer (magenta rectangle) and peaks “with” (black arrows) or “without” NF-κΒ motif (white arrows) are demarcated. (B) Browser view of 3C-seq data (in reads per million) generated using the SAMD4A TSS as a viewpoint (yellow triangle) at 0 and 30 min in the presence or absence of inhibitors C646 and DRB. (C) Log2-normalized mean counts of reads mapping to introns at 30 min post-stimulation in the presence or absence of C646. Only genes up- (orange) and down-regulated (blue) are shown. Pie chart: 15% and 9% of the 967 down-regulated genes associate with “with” and “without” RELA peaks, respectively. (D) The most significantly enriched GO terms (biological process) associated with C646-down-regulated genes assigned to 30-min RELA peaks “with”/”without” motifs in the same TAD.
Figure 6.
Figure 6.
Inserting the RN7SK promoter in SAMD4A alters NF-κB–driven spatial interactions. Wild-type (wt) or 7SKi HUVECs (a single cell-derived clone with the RN7SK promoter inserted 13.5 kbp downstream from the SAMD4A TSS) were treated with TNF for 0 or 30 min. (Purple triangles) Insertions, (magenta boxes) intronic SAMD4A enhancer cluster. (A) T2C contact maps. (White circles) Two exemplary interactions of the SAMD4A TSS that differ in the two cell populations. One (left) involves contacts with the intronic enhancer cluster, the second (right) with the CTCF site immediately downstream from the enhancer cluster. RELA and CTCF ChIP-seq tracks are also shown below the SAMD4A gene model. (B) 3C- and ChIP-seq data (in reads per million) from wt (gray background) and 7SKi HUVECs. (Yellow triangle) 3C-seq viewpoint at the SAMD4A TSS, (purple triangle) 3C-seq viewpoint at the RN7SK insertion. (C) Total RNA- and ChIP-seq data (in reads per million) from wt (gray background) and 7SKi HUVECs along the SAMD4A locus. (D) Interactions captured by T2C between the 13 RELA-bound enhancers at 0 and 30 min in the two cell populations; RELA ChIP-seq levels for each peak are also shown (right). (E) (Top) Interactions captured by T2C between the intronic SAMD4A enhancer cluster (“SE”) and gene TSSs within 3 Mbp. All interactions with ≥0.013 rpm are indicated by the spider plot. (Bottom) Log2-fold changes in intronic RNA levels after 30 min of TNF stimulation of genes in the extended locus (iRNA-seq data); noncontacted TNFAIP3 and CXCL3 serve as controls.

Similar articles

See all similar articles

Cited by 8 articles

See all "Cited by" articles

Publication types

LinkOut - more resources

Feedback