IκBα is required for full transcriptional induction of some NFκB-regulated genes in response to TNF in MCF-7 cells

Inflammatory stimuli triggers the degradation of three inhibitory κB (IκB) proteins, allowing for nuclear translocation of nuclear factor-κB (NFκB) for transcriptional induction of its target genes. Of these three, IκBα is a well-known negative feedback regulator that limits the duration of NFκB activity. We sought to determine whether IκBα’s role in enabling or limiting NFκB activation is important for tumor necrosis factor (TNF)-induced gene expression in human breast cancer cells (MCF-7). Contrary to our expectations, many more TNF-response genes showed reduced induction than enhanced induction in IκBα knockdown cells. Mathematical modeling was used to investigate the underlying mechanism. We found that the reduced activation of some NFκB target genes in IκBα-deficient cells could be explained by the incoherent feedforward loop (IFFL) model. In addition, for a subset of genes, prolonged NFκB activity due to loss of negative feedback control did not prolong their transient activation; this implied a multi-state transcription cycle control of gene induction. Genes encoding key inflammation-related transcription factors, such as JUNB and KLF10, were found to be best represented by a model that contained both the IFFL and the transcription cycle motif. Our analysis sheds light on the regulatory strategies that safeguard inflammatory gene expression from overproduction and repositions the function of IκBα not only as a negative feedback regulator of NFκB but also as an enabler of NFκB-regulated stimulus-responsive inflammatory gene expression. This study indicates the complex involvement of IκBα in the inflammatory response to TNF that is induced by radiation therapy in breast cancer.


Supplementary Figure 1. Difference in the basal level of nuclear NFkB abundance between presence and absence of IkBa
Time course nuclear NFkB abundance of two biological replicates were max-normalized together from 0 to 100.  Chromatin accessibility was normalized to z-score, where red shows high chromatin accessibility and blue shows low chromatin accessibility in the heatmap. In addition to the enrichment of kB sites, these clusters also showed enrichment of AP-1 and IRF4 binding sites. Chromatin accessibility was normalized to z-score, where red shows high chromatin accessibility and blue shows low chromatin accessibility in the heatmap. Statistical tests were performed for chromatin accessibility between Ctrl and siIkBa at these regions in Ctrl and siIkBa (*: p-value < 0.01, **: p-value < 0.0001 and n.s.: p-value ≥ 0.01 by one-tailed Wilcoxon rank sum test). The center line indicates the median, the upper and lower hinges indicate the first and third quartiles, the upper whisker extends from the hinge to the largest value no further than 1.5 × IQR (interquartile range) from the hinge, the lower whisker extends from the hinge to the smallest value at most 1.5 × IQR of the hinge, and the points indicate the outliers. center line indicates the median, the upper and lower hinges indicate the first and third quartiles, the upper whisker extends from the hinge to the largest value no further than 1.5 × IQR (interquartile range) from the hinge, the lower whisker extends from the hinge to the smallest value at most 1.5 × IQR of the hinge, and the points indicate the outliers.  Supplementary Figure 8. The scaled fold change of nuclear NFkB abundance as the input data for mathematical modeling For replicate 1, fold change for all time points in each condition were calculated. Fractional nuclear RelA abundance (nuc cell^-1) in Ctrl and siIkBa were normalized together to span a range of 2 -100 (50-fold) to avoid assay specific reductions of the dynamic range. For replicate 2, after calculating the fold change for all time points in each condition, fractional nuclear RelA abundance (nuc cell^-1) in Ctrl was normalized to span a range of 2 -100 to standardize the maximum activity with the scaled Ctrl data in replicate 1. Then, the measured data in siIkBa was normalized using the same scale used in Ctrl.   Schematic diagram of the simple mathematical model. Heatmaps of the time course gene expression from experimental results and data-fit from the simple model. Genes highlighted with yellow and grey indicate whether the model is acceptable or not (definition of acceptable genes: nRMSD in Ctrl > 0.5, nRMSD in siIkBa > 0.39 and AUC of fold change in expression should be Ctrl > siIkBa or max-fold induction should be Ctrl > siIkBa). Line graphs of representative genes from ERGs in subcluster 2, IRGs in subcluster 2 and DRGs in subcluster 2 showing good and bad fits between the simulation results and the experimental data. Yellow color bars show that simulation were acceptable for both replicates, and grey color bars show that simulation was not acceptable for at least either one of the replicates based on the definition.   TAP1  TBC1D9  TNFAIP2  TRAF3  TRIP10  TUBD1  Fold change of expression level in Ctrl TET2  TICAM1  TNFRSF11B  TXNRD1  VMP1   ABAT  ABTB2  AOX1  APOL2  ARID5B  B4GALT5  BAZ1A  CHST15  CLK4  CREBZF  CRIM1  DAXX  DDX58  DRAM1  EFEMP1  FAM117A  FGD6  FOXO1  HIVEP2  HMGCR  IER5L  IFNGR2  IL6ST  IRF2  ITGB8  KLHL5  KYNU  LAMC2  LINC00052  LINC02015  LTB  MGAT4A  NFATC2  NFKB1  NFKB2  OPTN  PHF23  PHLDB2  PLEKHG3  PPP1R18  PRKAR2B  RAB3IP  RBM25  RELB  RFX5  RSRC2  RUNX2  SEMA3C  SEMA4B  SERPINB8  SGPP2  SLC6A14  STAT5A  TAP1  TBC1D9  TNFAIP2  TRAF3  TRIP10  TUBD1     Fold change of expression level in Ctrl TET2  TICAM1  TNFRSF11B  TXNRD1  VMP1   ABAT  ABTB2  AOX1  APOL2  ARID5B  B4GALT5  BAZ1A  CHST15  CLK4  CREBZF  CRIM1  DAXX  DDX58  DRAM1  EFEMP1  FAM117A  FGD6  FOXO1  HIVEP2  HMGCR  IER5L  IFNGR2  IL6ST  IRF2  ITGB8  KLHL5  KYNU  LAMC2  LINC00052  LINC02015  LTB  MGAT4A  NFATC2  NFKB1  NFKB2  OPTN  PHF23  PHLDB2  PLEKHG3  PPP1R18  PRKAR2B  RAB3IP  RBM25  RELB  RFX5  RSRC2  RUNX2  SEMA3C  SEMA4B  SERPINB8  SGPP2  SLC6A14  STAT5A  TAP1  TBC1D9  TNFAIP2  TRAF3  TRIP10  TUBD1 ARRDC3  DLC1  EFNA1  ELF3  FAM46A  GADD45A  IER3  IER5  IRF1  IRS2  JUNB  KLF10  KLF3  LIF  MAP3K1  MAP3K8  MXD1  NRIP1  NUAK2  PHLDA1  PIM1  PLEKHF2  PLK2  PPP1R15A  PPP1R3C  PTGER4  RND1  SALL4  SOX9  SPRY4  SPSB1  TNF  TNFAIP3  ZC3H12A   ADGRF4  ADGRG6  AKR1C2  AMIGO2  ANKRD18B  ANKRD33B  ANXA8  AREG  ATP1B1  BAG1  BCL2L11  BCL3  BID  BIRC3  CDKN2B  CLIC4  DMXL2  EIF5  ETS2  FAM107B  FNBP4  ICAM1  IL17C  ITGAV  ITPKC  LYPD6B  MAFF  NCOA7  NEDD9  NFKBIE  NKX3-1  NMD3  PCNA  PLAU  PLPP6  RAB9A  RBM3  REL  RHOV  S100A9  SDC4  SDCBP  SPTSSB  TET2  TICAM1  TNFRSF11B  TXNRD1  VMP1   ABAT  ABTB2  AOX1  APOL2  ARID5B  B4GALT5  BAZ1A  CHST15  CLK4  CREBZF  CRIM1  DAXX  DDX58  DRAM1  EFEMP1  FAM117A  FGD6  FOXO1  HIVEP2  HMGCR  IER5L  IFNGR2  IL6ST  IRF2  ITGB8  KLHL5  KYNU  LAMC2  LINC00052  LINC02015  LTB  MGAT4A  NFATC2  NFKB1  NFKB2  OPTN  PHF23  PHLDB2  PLEKHG3  PPP1R18  PRKAR2B  RAB3IP  RBM25  RELB  RFX5  RSRC2  RUNX2  SEMA3C  SEMA4B  SERPINB8  SGPP2  SLC6A14  STAT5A  TAP1  TBC1D9  TNFAIP2  TRAF3  TRIP10  TUBD1   Motif analysis results at ATAC-seq peaks in promoter regions in Ctrl (focused on kB motifs) Motif analysis results at ATAC-seq peaks in promoter regions in siIkBa (focused on kB motifs)   Figure 1a and 1b represent the means of two biological replicates.

Supplementary Table 2. Final parameters from IFFL model which showed the best-fit with this model
Concordant parameter values from IFFL model are shown for genes which showed the best-fit with this model (red). The color corresponds to the background color in Figure 3a, bar color in Figure 3b and color of the heatmap in Figure 3c.

Supplementary Table 3. Final parameters from model v4 which showed the best-fit with this model
Concordant parameter values from model v4 are shown for genes which showed the best-fit with this model (purple). The color corresponds to the background color in Figure 3a, bar color in Figure 3b and color of the heatmap in Figure 3c.  Figure 3a, bar color in Figure 3b and color of the heatmap in Figure 3c.

Supplementary Table 5. Final parameters from simple model which showed the best-fit with this model
Concordant parameter values from simple model are shown for genes which showed the best-fit with this model (blue). Genes that are in grey indicate that none of the 4 models were able to recapitulate their fold change in expression, and thus these parameters were used to create Figure 3c. The color blue corresponds to the background color in Figure 3a, bar color in Figure 3b and color of the heatmap in Figure 3c.
Final parameters from simple model