Fig 2: Small molecule inhibitors of H3 acetylation and RNAPII pause release alter TNF-mediated changes in transcriptional bursting ASchematic of A-485 inhibition of the histone acetyl transferase p300/CBP, which is recruited by NF-?B, and of JQ1 inhibition of BET bromodomains, which recruit the positive transcription elongation factor b (P-TEFb).BChange in enrichment of histone H3 and AcH3 after treatment with 300 nM A-485 for 4 h measured by ChIP-qPCR and shown as % input (non-IP control) normalized to the uninhibited control for each gene. Data are presented as mean ± SD for two or four biological replicates.CBar graphs of mean mRNA level for basal condition, 1-h TNF, 1-h TNF + 4 h pretreatment with 300 nM A-485, and 1-h TNF + 62.5 nM JQ1 cotreatment measured by smFISH for Tnfaip3 (left) and Tnf (right). Cells were combined from two replicates (Tnfaip3 A-485, JQ1; Tnf A-485) or one replicate (Tnf JQ1). Error bars indicated bootstrapped 95% CIs. Samples with non-overlapping CIs are significant.DChange in enrichment of total and Serine-5-phosphorylated RNAPII after 20 ng/ml TNF treatment for 1 h with 62.5 nM JQ1 measured by ChIP-qPCR and shown as % input (non-IP control) normalized to the uninhibited control for each gene. Data are presented as mean for two biological replicates.E, FBurst size (E) and burst frequency (F) parameter fits from the two-state model for the same conditions in (C) measured by smFISH for Tnfaip3 (left) and Tnf (right) for the data described in (C). Error bars indicated bootstrapped 95% CIs. Samples with non-overlapping CIs are significant.G, HProbability density of mRNA distributions measured by smFISH for 20 ng/ml TNF for 1-h (blue) vs 1-h TNF + 4 h pretreatment with 300 nM A-485 (green) (G) and 1-h TNF (blue) vs 1-h TNF + 62.5 nM JQ1 cotreatment (yellow) (H) for Tnfaip3 (left) and Tnf (right).

Fig 3: TNF treatment induces differential changes in noise across target genes that is not explained by cellular sources of variation (related to Fig 2) Bar graphs of coefficient of variation (CV) of basal and TNF-stimulated smFISH distributions presented in Fig 2B and C for the indicated genes at the indicated time points. Error bars indicate bootstrapped 95% confidence intervals (CIs). Significant differences indicated by non-overlapping CIs. Sample sizes are displayed in Fig 1E and 2C.Fano factor of mRNA counts in the nuclear vs cytoplasmic fractions of cells before and after 20 ng/ml TNF stimulation for the indicated genes. Data points are labeled with gene and time point. The diagonal is shown as a solid line to guide the eye.CV of basal and TNF-stimulated smFISH distributions for Tnfaip3, Tnf, and Il6 presented in Fig 2B and C unnormalized (light gray) or normalized to cellular area (dark gray) or nuclear area (blue). Lower bars indicating the CV of cellular area (dark gray) and nuclear area (blue) in basal and TNF-stimulated smFISH images are included with each gene for reference. Data are presented as mean ± SD of CVs from two replicate experiments for each gene and condition.Scatter plots comparing mRNA counts for Tnf and Nfkbia in the same cells from multiplex smFISH for basal, 1 h and 2 h of 20 ng/ml TNF stimulation. Pearson correlation r, P-value, and no. of cells n are indicated.

Fig 4: TNF induces gene-specific changes in transcript distributions at NF-?B targets AMaximum intensity projections of smFISH fluorescence microscopy z-stacks of Jurkat T cells stained for the indicated genes after 1-h (Nfkbia, Tnfaip3, Tnf, Il8) or 4-h (Il6, Csf2) treatment with 20 ng/ml TNF. Nfkbia, Tnfaip3, Tnf, and Il6 were labeled with Quasar 670, and Il8 and Csf2 were labeled with fluorescein. All images were filtered as described in Materials and Methods. Brightness and contrast were enhanced for visualization. Scale bars: 10 µm.BBar graphs of mean of smFISH distributions before and after TNF treatment for the indicated genes. Cells were combined from three replicates (Nfkbia 1 h; Tnfaip3 1 h; Tnf 1, 2 h; and Il6 2, 4 h) two replicates (Nfkbia 2 h) or one replicate (Tnfaip3 2 h; Il8 1, 2 h; Csf2 2, 4 h). Basal data are same as in Fig 1E. Error bars indicate bootstrapped 95% CIs for the samples in (C). Significant differences indicated by non-overlapping CIs.CProbability density plots of single-cell mRNA distributions from smFISH as described in (B) before and after treatment with 20 ng/ml TNF for the indicated time points.D, EGraph of log10(mean) vs log10(CV2) for endogenous gene targets that maintain CV2 (D) or decrease CV2 (E) after treatment with 20 ng/ml TNF in Jurkat T cells (black), HeLa cells (red), or murine bone marrow–derived macrophages (green). Gray shading indicates 95% CI of basal trend line. Poisson trend line indicated by dashed line. HeLa data from Lee et al (2014).

Fig 5: Mathematical modeling predicts differences in cell-to-cell heterogeneity resulting from TNF positive feedback on burst size- vs burst frequency-mediated transcript activation ASchematic of the two-state model of transcription coupled to translation of a protein (TNF) that positively feeds back on its own transcription rate.B, CModel simulation of cell-population averages of Tnf mRNA (B) and TNF protein (C) vs time with and without positive feedback. Data are presented as mean (dark line) and SD (shaded region) of 1,000 simulated cells.D, EViolin plots of simulated single-cell mRNA (D) and protein (E) numbers with and without positive feedback for either an increase in burst size (blue) or burst frequency (purple) in response to exogenous TNF treatment. Data are presented as number of molecules in 1,000 simulated cells. Red line indicates threshold of 20 protein molecules designating activated cells.F, GCell-population averages from stochastic simulations with positive feedback of Tnf mRNA at 1 h (F) and TNF protein at 4 h (G) after TNF treatment across a parameter space with increasing burst frequency (k a) and decreasing burst size (k i) chosen to produce levels of Tnf mRNA similar to experimental observations.