First responders shape a prompt and sharp NF-κB–mediated transcriptional response to TNF-α

Abstract

<jats:title>Summary</jats:title><jats:p>NF-κB acts as the master regulator of the transcriptional response to inflammatory signals by translocating into the nucleus upon stimuli, but we lack a single-cell characterization of the resulting transcription dynamics. Here we show that transcription of NF-κB target genes is strongly heterogeneous in individual cells but dynamically coordinated at the population level, since the average nascent transcription is prompt (i.e. occurs almost immediately) and sharp (i.e. increases and decreases rapidly) compared to NF-κB nuclear localization. Using an NF-κB-controlled MS2 reporter we confirm that the population-level transcriptional activity emerges from a strongly heterogeneous response in single cells as compared to NF-κB translocation dynamics, including the presence of a fraction of “first responders”. Mathematical models show that a combination of NF-κB mediated gene activation and a gene activity module including a gene refractory state is enough to produce sharp and prompt transcriptional responses. Our data and models show how the expression of the target genes of a paradigmatic inducible transcription activator upon stimuli can be time-resolved at population level and yet heterogeneous across single cells.</jats:p>

TNF stimulation primarily modulates transcriptional burst size of NF‐κB‐regulated genes

“…This simple promoter model sufficiently captured differences in transcriptional bursting "modes" following TNF treatment in Jurkat T cells that was a main focus of our study. However, two recent studies analyzing transcriptional bursting in response to stimulation of NF-κB by TNF or LPS reported somewhat different results (Bagnall et al, 2020;Zambrano et al, 2020). Bagnall et al studied activation of Tnf and Il1b following LPS stimulation in macrophages and found gene-specific mean-noise trends for Tnf vs Il1b; while a two-state promoter model was sufficient to reproduce Tnf distributions, a three-state model with an additional unproductive (or "refractory") state was required to fit Il1b distributions.…”

“…Bagnall et al studied activation of Tnf and Il1b following LPS stimulation in macrophages and found gene-specific mean-noise trends for Tnf vs Il1b; while a two-state promoter model was sufficient to reproduce Tnf distributions, a three-state model with an additional unproductive (or "refractory") state was required to fit Il1b distributions. Zambrano et al similarly demonstrated that a promoter model with a third refractory state, combined with variability in upstream NF-κB signaling, was necessary to explain their observation of a subset of "first responder" cells that produced higher levels of Nfkbia, Tnf, and HIV in HeLa cells (Zambrano et al, 2020). We have previously demonstrated that variability in upstream NF-κB signaling is correlated with transcript levels in individual cells for the targets in our study even though they exhibit variations in noise (Wong et al, 2019), and thus, we do not think our results are inconsistent.…”

Small transcriptional differences among cell clones lead to distinct NF-κB dynamics

“…Beyond differences within the first 30 minutes, our clonal populations also show differences at later time points, with clone R having a sharper NF- κ B response as compared to clone B, i.e. its NF- κ B nuclear localization decays faster (Zambrano et al, 2020). We investigated if transcriptomic data could also shed light on the origin of these differences.…”

“…We then chose 3 clones with archetypical dynamics reminiscent of those observed in the literature and in cells in our original MEF population ( Figure 1B ): a clone with more persistent nuclear localization of NF- κ B (clone B, blue), one with a first well-defined sharp peak (clone R, red) (in other words, a nuclear localization that decreases fast (Zambrano et al, 2020)) and a clone characterized by a low activation of NF- κ B (clone G, green) ( Figure 1E ).…”

TNF stimulation primarily modulates transcriptional burst size of NF-κB-regulated genes

“…To model transcriptional bursting, we used a simple two-state model in which a promoter can occupy either an active ‘on’ state or inactive ‘off’ state, which fit our data from Jurkat T cells well. Including a third refractory promoter state that an active promoter transitions to before transitioning back to the inactive state has produced better fits in some systems, including bursting of the HIV LTR promoter in HeLa cells (Li et al , 2018; Suter et al , 2011; Zambrano et al , 2020). However, the extremely low transition rates we observed from the inactive to the active state (burst frequencies on the order of 1 per hour) might be slower than (and thus mask) transitions from a refractory to the inactive state.…”