A genome-wide screen identifies IRF2 as a key regulator of caspase-4 in human cells

Abstract

Caspase-4, the cytosolic LPS sensor, and gasdermin D, its downstream effector, constitute the non-canonical inflammasome, which drives inflammatory responses during Gram-negative bacterial infections. It remains unclear whether other proteins regulate cytosolic LPS sensing, particularly in human cells. Here, we conduct a genome-wide CRISPR/Cas9 screen in a human monocyte cell line to identify genes controlling cytosolic LPS-mediated pyroptosis. We find that the transcription factor, IRF2, is required for pyroptosis following cytosolic LPS delivery and functions by directly regulating caspase-4 levels in human monocytes and iPSC-derived monocytes. CASP4, GSDMD, and IRF2 are the only genes identified with high significance in this screen highlighting the simplicity of the non-canonical inflammasome. Upon IFN-γ priming, IRF1 induction compensates IRF2 deficiency, leading to robust caspase-4 expression. Deficiency in IRF2 results in dampened inflammasome responses upon infection with Gram-negative bacteria. This study emphasizes the central role of IRF family members as specific regulators of the non-canonical inflammasome.

Keywords: LPS; caspase-11; inflammasome; interferon regulatory factor; lipopolysaccharide.

Figure 1. A genome‐wide CRISPR/Cas9 screen identifies IRF2, gasdermin D, and caspase‐4 as the three main players in LPS sensing

1. A genome‐wide CRISPR/Cas9 screen based on LPS electroporation was performed in U937 cells. Four successive electroporation rounds were performed. DNA was extracted from surviving cells, and next‐generation sequencing (NGS) was performed to calculate the sgRNA enrichment factor.

2. The 10 most enriched sgRNA in the LPS‐electroporated samples are shown.

3. Graphical representation of the screen results with each gene identified in the output libraries (10,750) represented on the x‐axis and the corresponding adjusted P‐value on the y‐axis. Statistical analysis was performed using the Wald test. The colored line represents the 0.05 P‐value threshold. The size of the circle is inversely proportional to the P‐value.

4. U937 cell lines were knock‐out using CRISPR/Cas9, and cell death was quantified by LDH release assay 4 h after LPS electroporation. Each dot represents the average of three technical LDH replicates; means and SD of 3–9 independent experiments are shown. One‐way ANOVA with Dunnet's multiple comparisons test was performed. Adjusted P‐value is shown (from left to right ***P < 0.001, ***P < 0.001, ***P < 0.001, P = 0.84, P = 1, P = 0.092, *P = 0.012, **P = 0.0026, P = 1, **P = 0.0022, P = 0.98, P = 0.82, P = 0.97).

Data information: (A–C) Experiment was conducted once in quadruplicate. COP1 is also known as RFWD2.Source data are available online for this figure.

Figure EV1. 16 out of the 18 genes significantly enriched in the genome‐wide CRISPR/Cas9 screen are expressed above the RNA‐seq filtering threshold in U937

Each dot represents the transcript level of a biological U937 replicate as quantified by RNA‐seq; the bar represents the mean ± SD of a biological triplicate. The filtering threshold is shown as a horizontal dotted line and corresponds to absolute transcript counts below 30 in all the libraries.Source data are available online for this figure.

Figure EV2. IRF2 is required to promote F. novicida and E. coli LPS‐mediated cell death

1. IRF2 and β‐actin protein levels in WT and three independent IRF2 ^KO U937 cell lines (#1–3) were assessed by Western blotting analysis.

2. LDH release was quantified from three independent IRF2 ^KO cell lines 4 h after E. coli LPS electroporation. Each dot represents a technical LDH replicate; the bar represents the mean ± SD of 3 LDH replicates.

3. LDH release was quantified 4 h after electroporation with LPS for E. coli or F. novicida (5 μg for 4 × 10⁵ cells). Each dot represents a technical LDH replicate; the bar represents the mean ± SD of 3 LDH replicates.

Data information: (A–C) One experiment was performed.Source data are available online for this figure.

Figure 2. IRF2 is specifically required for caspase‐4‐mediated cell death

U937 cell lines were generated using CRISPR/Cas9.

1. IRF2 and β‐actin protein levels were assessed in the lysate of the indicated U937 cells by Western blotting analysis. A non‐specific band (*) is observed in the IRF2 Western blot. One experiment representative of three experiments is shown.

2. Cell death induced by LPS electroporation or the indicated treatment was quantified in real time by measuring propidium iodide (PI) incorporation/fluorescence every 15 min. Cell death was normalized using untreated and TX‐100‐treated cells. The kinetics of one representative experiment and the areas under the curve (AUC) (normalized to the WT AUC) of three independent experiments are shown. Each point represents the mean of a biological triplicate of one experiment; the bar represents the mean ± SD of three independent experiments.

3. Cell death was measured by LDH release assay 4 h after LPS electroporation. Each dot corresponds to the LDH triplicate of one experiment; the bar represents the mean ± SD of four independent experiments.

4. Heat map representation of the enrichment factor for the 4 sgRNA targeting each IRF in the genome‐wide screen.

5. Cell death was measured by LDH release assay 4 h after LPS electroporation. Each dot corresponds to the LDH triplicate of one experiment; the bar represents the mean ± SD of three independent experiments.

6. Cell death was quantified by measuring PI incorporation/fluorescence every 15 min after treatment with nigericin, UCN‐01, or gliotoxin. The kinetics of one representative experiment and the areas under the curve (AUC) (normalized to the WT AUC) of three independent experiments are shown. Each point represents the mean of a biological triplicate of one experiment; the bar represents the mean ± SD of three independent experiments.

Data information: (B, E and F) One‐way ANOVA with Dunnet's multiple comparisons test was performed. (B‐WT vs. CASP4^KO: ***P < 0.0001, WT vs. IRF2^KO pEMPTY: ***P = 0.0004, WT vs. IRF2^KO pIRF2: P = 0.11; E‐left to right ***P < 0.0001; P = 0.36; ***P < 0.0001; P = 0.78; P = 0.43; P = 0.42; P = 0.98; P = 0.87; P = 0.68; P = 0.86; F‐ left to right, AUC Nigericin, P = 0.95, P = 0.98, P = 0.027; AUC‐UCN‐01, P = 0.012; P = 0.066; P = 0.99; AUC gliotoxin, P = 0.97, P = 1, P = 0.79). (C) One‐way ANOVA with Sidak's multiple comparisons test was performed. (WT vs. IRF2^KO pEMPTY: ***P = 0.0001, IRF2^KO pEMPTY vs. IRF2^KO pIRF2: **P = 0.0014).Source data are available online for this figure.

Figure 3. IRF2 regulates CASP4 transcript levels

1. A

   mRNA transcript levels in U937 cells WT vs. IRF2^KO cells were quantified by RNA‐seq.

2. B

   Venn diagram showing the number of genes with a transcript level fold change (FC) greater than 2 in U937 cells WT vs. IRF2^KO cells (yellow) and greater than 2 in IRF2^KO vs. IRF2^KO pIRF2 (Blue).

3. C

   CASP4, CASP1, CARD16, and GSDMD transcript levels were quantified by RNA‐seq in the indicated U937 cell lines.

4. D

   Chromatin immunoprecipitation sequencing (ChIP‐seq) in primary human monocytes from two healthy donors (HD) for IRF2 binding onto CASP1 and CASP4 promoters. Red arrows indicate promoter region with IRF2 binding.

5. E

   IRF2 and CASP4 protein levels were assessed by Western blotting in the indicated cell lines. The asterisk indicates a non‐specific band. For the long exposure of the CASP4 Western blot, the 2 left lanes were covered during revelation to prevent signal bleed through and were blacked out for representation purposes.

6. F, G

   Cell death was quantified by (F) LDH release assay 4 h after LPS electroporation or (G) by measuring propidium iodide (PI) incorporation/fluorescence every 5 min. Cell death was normalized using untreated and TX‐100‐treated cells. The kinetics of one representative experiment and the areas under the curve (AUC) (normalized to the WT AUC) of four independent experiments are shown. Each point represents the mean of a biological triplicate of one experiment; the bar represents the mean ± SD of four independent experiments.

Data information: (A–C) One experiment performed with three biological replicates. (C) Each point represents the transcript count per million in one biological replicate; the bar represents the mean ± SD of the three replicates. Unpaired t‐tests were performed. WT vs. IRF2^KO P = 0.0009, 0.0003, 0.0003, 0.41; IRF2^KO vs. IRF2^KO pIRF2 P = 0.0001, 0.0019, 0.0063, 0.18, for CASP4, CASP1, CARD16, and GSDMD, respectively. (E) Data are representative of three independent experiments. (F) Each dot corresponds to the LDH triplicate of one experiment; the bar represents the mean ± SD of three independent experiments. One‐way ANOVA with Sidak's multiple comparisons test was performed. (WT vs. CASP4^KO, WT vs. IRF2^KO pEMPTY, IRF2^KO pEMPTY vs. IRF2^KO pCASP4 all ***P < 0.0001, WT vs. IRF2^KO pCASP4: P = 0.90 NS: not significant) (G) One‐way ANOVA with Dunnet's multiple comparisons test was performed. (WT vs. CASP4^KO: ***P = 0.0003, WT vs. IRF2^KO: ***P = 0.0008, WT vs. IRF2^KO pCASP4: P = 0.078).Source data are available online for this figure.

Figure EV3. IRF2 regulates GSDMD cleavage and Casp1 level in U937 cells

1. Caspase‐4, gasdermin D, IRF2, and β‐actin protein levels in the indicated U937 cell lines were assessed by Western blotting analysis at 1 h post‐mock (−) or LPS (+) electroporation. Processing of gasdermin D (31 kDa band) is clearly observed in WT U937 cells but is barely detectable in IRF2^KO cells. One experiment representative of three independent experiments is shown.

2. IRF2, caspase‐1, and β‐actin protein levels in the indicated U937 cell lines were assessed by Western blotting analysis at steady state. The asterisk indicates a non‐specific band. One experiment representative of two independent experiments is shown.

3. IL‐1β level in the supernatant of the indicated PMA‐differentiated U937 cell lines was quantified by ELISA at 2 h post‐nigericin treatment. Each dot represents the value of a biological replicate; the bar represents the mean ± SD of 3 biological replicates from one experiment representative of two independent experiments.

Source data are available online for this figure.

Figure 4. IFN‐γ treatment induces IRF1 to redundantly act with IRF2 to regulate caspase‐4 levels and LPS‐mediated pyroptosis

1. A–C

   Cell death was assessed (A, B) by measuring LDH release 4 h after LPS electroporation or (C) by measuring propidium iodide (PI) incorporation/fluorescence every 10 min in the indicated U937 cell lines primed (A‐as indicated, B, C‐all samples) or not with IFN‐γ. (C) Cell death was normalized using untreated and TX‐100‐treated cells. One real‐time cell death experiment (mean and SD of triplicate) representative of three independent experiments is shown. The areas under the curve (AUC) (normalized to the WT AUC) of three independent experiments are shown. Each point represents the mean of a triplicate of one experiment; the bar represents the mean (and SD) of the three independent experiments.

2. D

   Caspase‐4, IRF2, and IRF1 protein levels were assessed in the lysate of the indicated U937 cell lines primed or not with IFN‐γ. Data are representative of three independent experiments.

Data information: (A,B) Each dot corresponds to the mean of LDH triplicate of one experiment; the bar represents the mean ± SD of three (B) to four (A) independent experiments. (A): One‐way ANOVA with Sidak's multiple comparisons test was performed (untreated WT vs. IRF2^KO: ***P < 0.0001, WT vs. GSDMD^KO: ***P < 0.0001; IFN‐γ‐treated WT vs. IRF2^KO: P = 0.3, WT vs. GSDMD^KO: ***P < 0.0001). (B) One‐way ANOVA with Dunnet's multiple comparisons test was performed (All P‐values = 1 except WT vs. IRF1/2^DKO: **P = 0.0048 and WT vs. CASP4^KO: ***P < 0.0001). (C) One‐way ANOVA with Dunnet's multiple comparisons test was performed (WT vs. IRF2^KO: P = 0.84, WT vs. IRF1/2^DKO: *P = 0.03, WT vs. CASP4^KO: ***P = 0.0004).Source data are available online for this figure.

Figure 5. IRF1 cooperates with IRF2 to regulate non‐canonical inflammasome activation in PMA‐differentiated macrophages

U937 cell lines were differentiated with PMA and primed with IFN‐γ.

1. A

   Cell death was quantified by LDH assay 2 h after LPS electroporation.

2. B

   Cells were primed with Pam3CSK4. IL‐1β levels were assessed by ELISA 4 h after LPS electroporation.

3. C

   Cell death was quantified by measuring propidium iodide (PI) incorporation/fluorescence every 5 min after infection with F. novicida or (F) E. coli. Cell death was normalized using untreated and TX‐100‐treated cells. One real‐time cell death experiment (mean and SD of triplicate) representative of three independent experiments is shown. The areas under the curve (AUC) (normalized to the WT AUC) of three independent experiments are shown. Each point represents the mean of a triplicate of one experiment; the bar represents the mean (and SD) of the three independent experiments.

4. (D–H)

   (D, G) IL‐1β and (E, H) TNF levels were assessed by ELISA 6 h after WT or ΔFPI mutant F. novicida (D, E) or E. coli (G, H) infection.

Data information: (A, B) Each dot corresponds to the mean of LDH triplicate of one experiment; the bar represents the mean ± SD of three independent experiments. (D–E, G–H) Each dot corresponds to the mean of biological triplicates of one experiment, the bar represents the mean ± SD of three (D, G, H) to four (E) independent experiments. (A–H) One‐way ANOVA with Dunnet's multiple comparisons test was performed. (A) WT vs. IRF1/2^KO **P = 0.0084, WT vs. CASP4^KO **P = 0.0019 (B) WT vs. IRF1^KO *P = 0.012, WT vs. IRF1/2^KO ***P = 0.0003, WT vs. CASP4^KO ***P < 0.0001 (C) (WT vs. IRF2^KO: P = 0.26, WT vs. IRF1^KO: **P = 0.0015, WT vs. IRF1/2^DKO: ***P = 0.0002, WT vs. CASP4^KO: **P = 0.0067). (D) (WT vs. IRF2^KO: ***P = 0.001, WT vs. IRF1^KO: ***P < 0.0001, WT vs. IRF1/2^DKO: ***P < 0.0001, WT vs. CASP4^KO: ***P = 0.0005). (E) (WT vs. IRF2^KO: P = 0.045, WT vs. IRF1^KO: *P = 0.19, WT vs. IRF1/2^DKO: P = 1, WT vs. CASP4^KO: P = 0.57). (G) (WT vs. IRF2^KO: *P = 0.016, WT vs. IRF1^KO: *P = 0.024, WT vs. IRF1/2^DKO: **P = 0.0032, WT vs. CASP4^KO: P = 0.57). (F) WT vs. IRF1^KO *P = 0.045, WT vs. IRF1/2^KO **P = 0.0017, WT vs. CASP4^KO *P = 0.0107 (H) (WT vs. IRF2^KO: P = 0.86, WT vs. IRF1^KO: P = 1, WT vs. IRF1/2^DKO: P = 0.79, WT vs. CASP4^KO: P = 1).Source data are available online for this figure.

Figure 6. IRF2 controls caspase‐4 levels in human induced pluripotent stem cell (iPSC)‐derived macrophages

1. A, B

   Macrophages were derived from human iPSC clones knock‐out for the indicated genes. Two clones were selected for IRF2 ^KO cells. (A) Caspase‐4, caspase‐1, IRF2, and β‐actin protein levels were assessed by Western blotting analysis. One experiment representative of two independent experiments is shown. (B) The indicated mRNA levels were quantified by qRT–PCR normalized to β‐actin mRNA levels and expressed as fold change compared to WT levels. Each dot represents the average of a technical RT–PCR triplicate from one experiment; the bar represents the mean of three independent experiments. One‐way ANOVA with Dunnet's multiple comparisons test was performed (WT vs. IRF2 ^KO #I17 **P = 0.0020; WT vs. IRF2 ^KO #32 **P = 0.0013 for CASP4 transcript. WT vs. CASP1 ^KO ***P < 0.0001 for CASP1 transcript levels).

Source data are available online for this figure.