Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 8 (5), 1595-606

The NF-κB Genomic Landscape in Lymphoblastoid B Cells


The NF-κB Genomic Landscape in Lymphoblastoid B Cells

Bo Zhao et al. Cell Rep.


The nuclear factor κB (NF-κΒ) subunits RelA, RelB, cRel, p50, and p52 are each critical for B cell development and function. To systematically characterize their responses to canonical and noncanonical NF-κB pathway activity, we performed chromatin immunoprecipitation followed by high-throughput DNA sequencing (ChIP-seq) analysis in lymphoblastoid B cell lines (LCLs). We found a complex NF-κB-binding landscape, which did not readily reflect the two NF-κB pathway paradigms. Instead, 10 subunit-binding patterns were observed at promoters and 11 at enhancers. Nearly one-third of NF-κB-binding sites lacked κB motifs and were instead enriched for alternative motifs. The oncogenic forkhead box protein FOXM1 co-occupied nearly half of NF-κB-binding sites and was identified in protein complexes with NF-κB on DNA. FOXM1 knockdown decreased NF-κB target gene expression and ultimately induced apoptosis, highlighting FOXM1 as a synthetic lethal target in B cell malignancy. These studies provide a resource for understanding mechanisms that underlie NF-κB nuclear activity and highlight opportunities for selective NF-κB blockade.

Conflict of interest statement


The authors declare they have no competing financial interests.


Figure 1
Figure 1. NF-κB subunit genome-wide distribution and consensus motif
(A) NF-κB subunit ChIP-seq signals at the BCL2 locus. The y-axis is scaled between zero and twenty times the median signal value of the surrounding 100 kB. (B) Genome-wide NF-κB subunit distribution across chromatin states, as defined by GM12878 histone modifications, CTCF and Pol2 occupancy. Each horizontal bar shows the fraction of NF-κB subunit peaks that were assigned to each chromatin state. The total number of NF-κB subunit peaks is shown to the right of each bar. (C) Consensus de novo motif for each NF-κB subunit.
Figure 2
Figure 2. NF-κB subunit binding profiles
(A) 10 promoter and (B) 11 enhancer peak clusters with distinct NF-κB subunit binding profiles were identified by k-means clustering of ChIP-seq signals at regions bound by at least one subunit. Red values indicate higher ChIP-seq signal intensity. The total number of promoter vs. enhancer NF-κB binding sites is shown at the lower right. The heat map to the right of the peak clusters displays the extent of consensus de novo NF-κB subunit motif enrichment in each cluster. See also Figure S3A. (C) NF-κB subunit ChIP-seq signals at the TRAF1 locus illustrate the co-occurrence of multiple SBPs at an NF-κB target gene. Red boxes enclose promoter-associated peaks and orange boxes enclose enhancer-associated peaks. (D) Gene set enrichment analysis of NF-κB clusters using GO Biological Process (BP) terms, as determined by GREAT analysis. Each row corresponds to a unique GO BP term with a false discovery rate (FDR) < 0.01. A subset of highly enriched terms is highlighted. See also Table S2.
Figure 3
Figure 3. Effect of an 11 bp motif with a 3′ cytosine on p50 recruitment to NF-κB sites
(A) ROC curves show that the p50 motif 11th base pair cytosine predicts p50 ChIP occupancy with high specificity. In the comparisons shown, peaks from clusters E5 and E2 (top; differential p50 binding) and E5 and E9 (bottom; both bound by p50) are compared. The maximum p50, but not RelA motif match score for each sequence, predicts whether peaks belong to E5 (defined as a positive) or to E2 (negative). Meanwhile, neither motif has predictive power when comparing the p50-bound clusters E5 and E9. See also Figures S3B–D. (B) p50 signals are significantly higher at NFKB2 locus binding sites that contain the 11 bp motif. Other NF-κB subunits bind more uniformly. Motif match scores are scaled from a minimum match threshold of 0.0 to a maximum of 1.0. (C) The functional importance of the p50-preferred 11 bp κB site 3′ cytosine is supported by its evolutionary conservation across 33 mammalian species. Each bar shows the mean GERP++ score (higher values indicate more evolutionary constraint) at various positions of κB motif instances. Error bars indicate 1 s.e. of the mean. See also Figure S3E.
Figure 4
Figure 4. Alternative motifs enriched in clusters that lack κB sites
De novo motif discovery revealed enrichment for E-box, CTCF, composite PU.1/IRF4, and ZNF143 motifs at several clusters with low κB site enrichment. The discovered motif, AUC and p-value motif enrichment values, and ChIP-seq signals at representative loci are shown.
Figure 5
Figure 5. Co-occurrence of NF-κB subunits and GM12878 TF ChIP-seq peaks across enhancer and promoter clusters
Red intensity indicates the extent of NF-κB co-occupancy with indicated TFs in GM12878 at promoter and enhancer clusters, normalized for the number of cluster elements and the number of total peaks for each TF. Basal TFs names are indicated in green, DNA looping factors are in blue.
Figure 6
Figure 6. NF-κB and FOXM1 are present in DNA-bound protein complexes at κB sites
(A) Venn diagram showing the extent to which NF-κB and FOXM1 ChIP-seq peaks overlap in GM12878. The number of sites occupied by NF-κB, FOXM1, or co-occupied by both, and the consensus de novo motif for sites co-occupied by NF-κB and FOXM1, is shown. (B) FOXM1 ChIP-seq signals at the TNFAIP3 locus show a high degree of co-occupancy with NF-κB. See also Figure S4. (C) ChIP-re-ChIP identified FOXM1 and RelA as present together in DNA-bound protein complexes at the PLK1 and BCL2 loci. FOXM1 ChIP was followed by RelA ChIP. Mean fold enrichment −/+ SD for replica experiments is shown. *p<0.05, **p<0.01. (D) EMSA identified that both FOXM1 and NF-κB from GM12878 LCL nuclear extract (N.E.) bind a PLK1 promoter DNA probe that contains a κB site, but no forkhead recognition site. Incubation with a cold probe (C.P.) is indicated. *FOXM1 super-shift. **p50 super-shifted band. A representative EMSA of three independent experiments is shown. (E) A PLK1 promoter probe with a central κB motif (WT κB), but not a probe with a mutant κB motif (MT κB), competed for binding in EMSA (see Supplemental Experimental Procedures for full details). (F) Inducible expression of an IκBα super-repressor (IκBα S.R.) in IB4 LCLs diminished RelA and FOXM1 ChIP signals at the PLK1 and AURKA promoters. Mean + SD for replica experiments is shown. *p<0.05, **p<0.01, ***p<0.001
Figure 7
Figure 7. FOXM1 cooperates with NF-κB to regulate GM12878 target gene expression
(A) Three independent shRNAs against FOXM1 reduced expression of key NF-κB target gene mRNAs by 48 hours after delivery. Mean −/+ SEM effects from three independent experiments are shown. p < 0.05 for TNFAIP3, p<0.01 for all other comparisons between control and FOXM1 shRNAs. (B) FOXM1 shRNAs did not reduce expression control genes (which were not identified as NF-κB or FOXM1 targets in GM12878) at 48 hours post-shRNA delivery. (C) FOXM1 depletion inhibited LCL proliferation and caused accumulation of cells with <2n DNA content by Propidium Iodide (PI) analysis. See also Figure S5. (D) At 120 hours after shRNA delivery, FOXM1 depletion triggered cleavage of caspase 3, 7, and their substrate PARP. Tubulin load control is shown. Representative western blot of three independent experiments is shown. (E) FOXM1 expression in DLBCL tumor samples correlated with worse clinical outcome. Retrospective analysis of FOXM1 expression in microarray datasets obtained from 414 DLBCL tumor samples, and its relationship with patient survival. Wald test p = 0.0037.

Similar articles

See all similar articles

Cited by 67 articles

See all "Cited by" articles

Publication types

Associated data