Cell fate in antiviral response arises in the crosstalk of IRF, NF-κB and JAK/STAT pathways

Abstract

The innate immune system processes pathogen-induced signals into cell fate decisions. How information is turned to decision remains unknown. By combining stochastic mathematical modelling and experimentation, we demonstrate that feedback interactions between the IRF3, NF-κB and STAT pathways lead to switch-like responses to a viral analogue, poly(I:C), in contrast to pulse-like responses to bacterial LPS. Poly(I:C) activates both IRF3 and NF-κB, a requirement for induction of IFNβ expression. Autocrine IFNβ initiates a JAK/STAT-mediated positive-feedback stabilising nuclear IRF3 and NF-κB in first responder cells. Paracrine IFNβ, in turn, sensitises second responder cells through a JAK/STAT-mediated positive feedforward pathway that upregulates the positive-feedback components: RIG-I, PKR and OAS1A. In these sensitised cells, the 'live-or-die' decision phase following poly(I:C) exposure is shorter-they rapidly produce antiviral responses and commit to apoptosis. The interlinked positive feedback and feedforward signalling is key for coordinating cell fate decisions in cellular populations restricting pathogen spread.

Fig. 1

LPS and poly(I:C) elicit divergent responses. a Schematic diagram of the regulatory system of three transcription factors, NF-κB, IRF3 and STAT1/2, activated upon stimulation with LPS or poly(I:C). The synthesis of cytokine IFNβ, mediating autocrine and paracrine signalling, requires activation of both NF-κB and IRF3. Arrow heads = activation, hammer heads = inhibition. b, c Protein levels of the system components in response to LPS or poly(I:C), characterised by western blotting and compared with numerical model simulations. WT MEFs were stimulated with 1 μg/ml LPS or 1 μg/ml poly(I:C). GAPDH and HDAC1 serve as loading controls. Trajectories show averages of 200 independent stochastic simulations; the colour key is located next to protein labels. b Whole-cell extracts were analysed using antibodies against phosphorylated (active) forms of IKKα/β and TBK1, as well as total TBK1, IκBα and A20. Representative experiments out of 2 for LPS and 4 for poly(I:C) are shown. (*) = IKK isoform-dependent phosphorylation sites: p-IKKα Ser176/180, p-IKKβ Ser177/181. c Cytoplasmic and nuclear fractions were analysed using antibodies against total RelA (NF-κB), IRF3 and c-Jun, as well as for phospho-forms (active forms) of IRF3 and c-Jun. Representative experiments out of 2 are shown. d mRNA levels of NF-κB inhibitors, IκBα and A20, in response to LPS, cycloheximide (CHX) with LPS, or poly(I:C). WT MEFs were stimulated with 1 μg/ml LPS in the absence or presence of 5 μg/ml CHX, or with 1 μg/ml poly(I:C). CHX was added 1 h prior to LPS stimulation starting at time = 0. Time profiles of relative mRNA levels were obtained with RT-PCR and then rescaled to absolute numbers using digital PCR measurements. Bars represent means ± s.e.m., n ≥ 2, see Supplementary Note for plots of all replicates compared with model simulations

Fig. 2

Nuclear dynamics of NF-κB and IRF3: pulse-like after LPS, switch-like binary after poly(I:C). Nuclear translocation of a transcription factor is defined here throughout as a normalised quantification of its nuclear fluorescence in confocal images. RelA and IRF3 use a normalisation relating nuclear to whole-cell fluorescence, which underlies the histograms, fractions of responding cells and scatter plots. Distinct normalisations were used for p-STAT1 (to account for its changing cellular level) and for trajectories in RelA-GFP MEFs (see Methods). a, b Time course of nuclear localisation of RelA (NF-κB), IRF3, IκBα and p-STAT1 in response to poly(I:C). Scale bars: 50 μm. WT MEFs were stimulated with 1 μg/ml poly(I:C), fixed and stained at given time-points with antibodies against RelA and IRF3, RelA and IκBα, or p-STAT1. a Histograms show the full time course. Representative excerpts from confocal images show cells after 0 (nt), 4 and 10 h of poly(I:C) stimulation. b Changes in fraction of cells responding to poly(I:C) in time. A cell was deemed responding if its nuclear translocation exceeded a threshold based on nuclear fluorescence of non-treated cells. c Nuclear RelA trajectories in RelA-GFP MEFs stimulated with 1 μg/ml or 3 μg/ml poly(I:C) for 24 h. Sample time-lapse confocal microscopy images showing nuclear translocation of GFP-tagged RelA in response to poly(I:C) are provided in Supplementary Movies 1 and 2. d Time course of nuclear localisation of RelA and IRF3 in response to LPS. WT MEFs were stimulated with 1 μg/ml LPS, fixed and stained at given time-points with antibodies against RelA and IRF3. Histograms show the full time course. Representative confocal images show cells after 0 (nt), 1 and 4 h of LPS stimulation. e Nuclear RelA trajectories in RelA-GFP MEFs stimulated with 1 μg/ml LPS for 10 h. Sample time-lapse confocal microscopy images showing nuclear translocation of GFP-tagged RelA in response to poly(I:C) are provided in Supplementary Movie 3. Histograms (n ≥ 500) show a representative experiment out of 3. See Supplementary Data 1 and 2 for corresponding uncropped immunostaining images

Fig. 3

Transcriptional activation and secretion of IFNβ. a Time profiles of Ifnb1 (INFβ) mRNA levels in response to different stimuli. WT MEFs were stimulated with 1000 U/ml IFNβ, 1 μg/ml LPS, or 1 μg/ml poly(I:C) in the absence (left) or presence (right) of IFNAR-blocking antibody (α-IFNAR). Stat1^–/– MEFs and RelA^–/– MEFs (right) were also stimulated with 1 μg/ml poly(I:C). α-IFNAR was added to cells at 10 μg/ml at 0, 3, 6 and 10 h after poly(I:C) transfection. Time profiles of relative mRNA levels were obtained with RT-PCR, and then rescaled to absolute numbers using digital PCR measurements (bars represent means ± s.e.m., n ≥ 2; see Supplementary Note for plots of all replicates compared with model simulations). b Secretion of IFNβ in response to TNFα, LPS or poly(I:C). WT MEFs were stimulated with 10 ng/ml TNFα (n = 2), 1 μg/ml LPS, 0.1, 0.3, 1 and 3 μg/ml poly(I:C), or 1 μg/ml poly(I:C) in the presence of α-IFNAR (n = 2). Both Stat1^–/– MEFs and RelA^–/– MEFs were stimulated with 1 μg/ml or 3 μg/ml poly(I:C). For each condition, IFNβ concentration after 2, 4, 6, 10 and 24 h was measured by ELISA in 200 μl of culture medium harvested from above 25,000 ± 5000 cells. Bars represent means ± s.e.m. (n = 3, except where stated otherwise, values of all replicates are provided in Supplementary Table 8)

Fig. 4

Paracrine regulation via the JAK/STAT pathway. a, c Protein levels of the pathway components in response to 1 μg/ml poly(I:C), characterised by western blotting and numerical model simulations. In separate experiments WT MEFs were stimulated with poly(I:C) in the presence of α-IFNAR, and Stat1^–/– MEFs were stimulated with poly(I:C), for the indicated times. For each experiment, its control, poly(I:C)-stimulated WT MEFs, is included. Whole-cell extracts were analysed. Representative replicates out of 3 are shown. Trajectories show averages of over 200 independent stochastic simulations; the colour key is located next to protein labels. a Mediators of positive feedback were analysed using antibodies against phosphorylated (active) forms of STAT1 (p-Tyr701), TBK1, IKKα/β and IRF3 (p-Ser396), as well as total STAT1, STAT2, RIG-I, TBK1, and IRF3. (*) = IKK isoform-dependent phosphorylation sites: p-IKKα Ser176/180, p-IKKβ Ser 177/181. c STAT1/2-regulated mediators of double negative feedback were analysed using antibodies against total OAS1A and PKR, as well as for a phosphorylated (active) form of PKR (p-Thr451). b, d mRNA levels of STAT1/2-regulated genes, Stat1, Stat2, Ddx58 (RIG-I), Socs1, Eif2ak2 (PKR) and Oas1a, in response to LPS, IFNβ or poly(I:C). WT MEFs were stimulated with 1000 U/ml IFNβ, 1 μg/ml LPS or 1 μg/ml poly(I:C) in the absence (left) or presence (right) of IFNAR-blocking antibody (α-IFNAR). Stat1^–/– MEFs and RelA^–/– MEFs (right) were also stimulated with 1 μg/ml poly(I:C). Time profiles of relative mRNA levels were obtained with RT-PCR, and then rescaled to absolute numbers using digital PCR measurements (bars represent means ± s.e.m., n ≥ 2, see Supplementary Note for plots of all replicates compared with model simulations)

Fig. 5

Inhibition of translation stabilises translocation of RelA and IRF3. a RelA (NF-κB) translocation and cytoplasmic IκBα levels in response to LPS or CHX + LPS. WT MEFs were stimulated with 1 μg/ml LPS in the absence or presence of 5 μg/ml CHX, fixed and stained at given time-points with antibodies for RelA and IκBα. Representative excerpt from confocal images show cells at 0 (nt), 30, 90 and 240 min after LPS stimulation. Histograms (n ≥ 700, from a representative experiment out of 2) show the full time course of RelA nuclear translocation, defined for Fig. 2. See Supplementary Data 3 for corresponding uncropped immunostaining images. Scale bar: 50 μm. b–d Protein levels in response to poly(I:C) upon PKR inhibition. WT MEFs were stimulated with 1 μg/ml poly(I:C) for 0 (nt), 2 and 4 h in the absence or presence of imidazolo-oxindole PKR inhibitor (1 μM/ml), C16, added at 1 h prior to poly(I:C) transfection. Culture medium for all conditions contained the C16 solvent DMSO (0.5% final concentration), and was FBS-free to prevent interference with C16. b Whole-cell extracts were analysed using antibodies against total PKR, A20 and IκBα, as well as against a phosphorylated (active) form of PKR (p-Thr451). c Nuclear and cytoplasmic fractions were analysed using antibodies against total RelA and IRF3, as well as against their phosphorylated (active) forms, p-RelA (Ser536) and p-IRF3 (Ser396). d Cells were fixed and immunostained for RelA and IRF3. Scatter plots show nuclear translocations of RelA vs. IRF3 (n = 500) based on confocal images analysis. Percentages indicate fractions of active cells; activity was defined by responding (see also Fig. 2b) with both RelA and IRF3 translocation, as illustrated in a mock plot at the top. See Supplementary Data 4 for corresponding uncropped immunostaining images

Fig. 6

Mathematical model: analysis of the effect of IFNβ pre-stimulation on responses to poly(I:C). a Network diagram. The components of the model which do not influence other components are listed next to the diagram. Blue arrow-headed lines = activation, red hammer-headed lines = inhibition, green arrow-headed lines = positive regulation of transcription. b, c The three-day long stochastic simulations are preceded by 3 × 10⁶ s (~35 days) of stochastic equilibration phase (not shown) rendering different values of model variables at time = 0 for the 5 trajectories shown in each panel. b After 2 days of quiescence, WT MEFs were stimulated with 1 μg/ml poly(I:C) for 1 day. c After 1 day of quiescence, WT MEFs were prestimulated with 1000 U/ml IFNβ for 1 day, followed by 1 day stimulation with 1 μg/ml poly(I:C)

Fig. 7

Model validation: IFNβ pre-stimulation increases the fraction of responding cells and response strength. Following 24 h of quiescence (a, b) or pre-stimulation with 1000 U/ml IFNβ (a, d), WT MEFs were stimulated with 1 μg/ml poly(I:C) for 0–24 h, or they were only stimulated with 1000 U/ml IFNβ for 0–24 h (c). a Protein levels of model components, characterised by western blotting and numerical model simulations. Whole-cell extracts were analysed with antibodies against NF-κB–IRF3–STAT1/2 pathways components. (*) = IKK isoform-dependent phosphorylation sites: p-IKKα Ser176/180, p-IKKβ Ser 177/181. Representative experiments out of 2 are shown. Trajectories show averages of over 200 independent stochastic simulations; the colour key is located next to protein labels. b–d Cells were fixed and immunostained for (b, d) RelA (NF-κB) and IRF3 or RelA and IκBα, or (c) p-STAT (Tyr701). Representative excerpts from confocal images show cells at the indicated times after stimulation. c, d Histograms (n ≥ 600, from one out of 2 experiments) show the full time course of translocation, as defined for Fig. 2. Scale bars: 50 μm. e, f Following 24 h of quiescence or pre-stimulation with 1000 U/ml IFNβ, WT and Stat1^–/– MEFs were stimulated with 1 μg/ml poly(I:C) for 0–24 h, fixed and stained with antibodies against RelA and IRF3. e Scatter plots show nuclear translocations of RelA vs. IRF3 (as defined for Fig. 2b, n = 300); ρ is the Pearson correlation coefficient. f Fractions of active cells (see Fig. 5d) calculated from the scatter plots in e. See Supplementary Data 5, 6 and 7, 8 for uncropped immunostaining images corresponding to c, d and scatter plots for Stat1^–/– MEFs in e

Fig. 8

Apoptotic cell fraction in response to LPS, TNFα, poly(I:C) or IFNβ. Bars show mean fractions of early apoptotic (pink) and late apoptotic/dead cells (purple), estimated by Annexin V/PI staining (bars represent means ± s.d., n ≥ 3, data for all replicates are given in the Supplementary Data 14). See Supplementary Fig. 8 for density plots of Annexin V/PI staining. a Fraction of apoptotic WT MEFs in response to 1 μg/ml LPS, 10 ng/ml TNFα or 1 μg/ml poly(I:C) after 24 h of treatment. b Fraction of apoptotic RelA^–/– MEFs in response to 24-hr stimulation with 10 ng/ml TNFα or 1 μg/ml poly(I:C). c Fraction of apoptotic WT MEFs in response to 24-hr treatment with 1 μg/ml poly(I:C) following 24 h of quiescence or 1000 U/ml IFNβ pre-stimulation. d Fraction of apoptotic Stat1^–/– MEFs in response to 24 h of 1000 U/ml IFNβ stimulation, 1 μg/ml poly(I:C) 24 h treatment, or response to 24 h treatment with 1 μg/ml poly(I:C) after 24 h of 1000 U/ml IFNβ pre-stimulation. e Apoptotic fraction time course for WT MEFs in response to 1 μg/ml poly(I:C). f Apoptotic fraction time course for WT MEFs in response to 1 μg/ml poly(I:C) following 24 h of 1000 U/ml IFNβ pre-stimulation