Research ArticleImmunology

IRF2 transcriptionally induces GSDMD expression for pyroptosis

See allHide authors and affiliations

Science Signaling  21 May 2019:
Vol. 12, Issue 582, eaax4917
DOI: 10.1126/scisignal.aax4917

IRF2 induces gasdermin D

In response to activation of canonical and noncanonical inflammasomes, a subset of caspases processes the protein gasdermin D (GSDMD) to release N-terminal fragments that oligomerize and form pores in the plasma membrane. Assembly of the GSDMD pore leads to release of the inflammatory cytokine IL-1β and causes cell death by pyroptosis. Kayagaki et al. found that loss of the transcriptional regulator IRF2 reduced GSDMD mRNA and protein abundance in mice and in human cells, resulting in decreased IL-1β secretion and reduced pyroptosis in response to inflammasome activation. Given that loss of GSDMD in mice results in ameliorated disease in models of inflammasome-driven pathologies, these findings suggest that IRF2 might be a therapeutic target for the treatment of sepsis and other inflammasome-mediated diseases.


Gasdermin-D (GSDMD) is cleaved by caspase-1, caspase-4, and caspase-11 in response to canonical and noncanonical inflammasome activation. Upon cleavage, GSDMD oligomerizes and forms plasma membrane pores, resulting in interleukin-1β (IL-1β) secretion, pyroptotic cell death, and inflammatory pathologies, including periodic fever syndromes and septic shock—a plague on modern medicine. Here, we showed that IRF2, a member of the interferon regulatory factor (IRF) family of transcription factors, was essential for the transcriptional activation of GSDMD. A forward genetic screen with N-ethyl-N-nitrosourea (ENU)–mutagenized mice linked IRF2 to inflammasome signaling. GSDMD expression was substantially attenuated in IRF2-deficient macrophages, endothelial cells, and multiple tissues, which corresponded with reduced IL-1β secretion and inhibited pyroptosis. Mechanistically, IRF2 bound to a previously uncharacterized but unique site within the GSDMD promoter to directly drive GSDMD transcription for the execution of pyroptosis. Disruption of this single IRF2-binding site abolished signaling by both the canonical and noncanonical inflammasomes. Together, our data illuminate a key transcriptional mechanism for expression of the gene encoding GSDMD, a critical mediator of inflammatory pathologies.


The canonical and noncanonical inflammasomes activate caspase-1 and caspase-11 (caspase-4 in humans), respectively, in many infectious and sterile inflammatory states, ranging from periodic fever syndromes, gout, atherosclerosis, neurodegeneration, and, in particular, septic shock (14). Active caspase-1, caspase-4, and caspase-11 proteolytically cleave full-length gasdermin D (GSDMD) to generate an N-terminal pore-forming fragment, which leads to the passive release of inflammatory cytokines, including interleukin-1β (IL-1β) and IL-18, as well as cytolytic membrane rupture (pyroptosis) (510). In noncanonical inflammasome signaling, the N-terminal GSDMD pore triggers nucleotide-binding oligomerization domain–like receptor pyrin domain–containing-3 (NLRP3)–dependent caspase-1 activation. This, in turn, leads to the proteolytic activation of IL-1β and IL-18 (6). Genetic ablation of Gsdmd ameliorates disease development in several mouse models (6, 11, 12). Because of its pivotal role in multiple inflammatory pathologies, GSDMD is an attractive drug target. GSDMD is found in various cell types, in particular those of the innate immune system; however, the molecular mechanism for its expression at the transcriptional level has been lacking (5, 6, 13, 14).


Phenomics screen links IRF2 to inflammasome signaling

To identify essential mediators of inflammasome signaling, we performed an unbiased, forward genetic screen with N-ethyl-N-nitrosourea (ENU)–mutagenized mice (15). Pedigree IGL03671 was found to harbor a Mendelian recessive mutation that compromised IL-1β secretion from bone marrow–derived macrophages (BMDMs) that had been primed with the Toll-like receptor 2 (TLR2) agonist Pam3CSK4 and then stimulated with adenosine 5′-triphosphate (ATP), an activator of the canonical NLRP3 inflammasome (Fig. 1A and fig. S1A) (16). Exome sequencing of the founder first-generation (G1) male identified 38 single-nucleotide variants (SNVs). A combination of exome sequencing and SNV genotyping revealed that the trait was wholly correlated with the inheritance of a point mutation in the gene encoding interferon regulatory factor 2 (IRF2) (fig. S1, A to C, and table S1).

Fig. 1 A forward genetic screening in mice identifies an Irf2 mutation that abolishes inflammasome signaling.

(A) Screening of G3 offspring from ENU-treated C57BL/6 mice. The graph indicates the amount of IL-1β released from Pam3CSK4-primed BMDMs after culture for 2 hours with ATP or medium alone (cont). The green dot represents IGL03671 G3 12 (fig. S1, A and B). The gray dots represent other G3 mice in the same batch from multiple pedigrees. (B) WT and IGL03671 Irf2 genes. The exon 5 coding sequence is in uppercase, and the SNV mutation is in green and highlighted with an asterisk. Gray boxes represent exons. (C) RT-PCR analysis of full-length and truncated splicing variant Irf2 transcripts in BMDMs from the indicated mice. Rpl19, control housekeeping gene. (D) Western blotting analysis of IRF2 in extracts of BMDMs from the indicated mice. (E) Analysis of the amounts of IL-1β (top) and LDH (bottom) released from unprimed or Pam3CSK4-primed BMDMs that were untreated (cont), stimulated with ATP (for 4 hours), transfected with LPS (16 hours), or stimulated with FasL (16 hours). Data are means (bars) of at least three individual replicates (circles) from two mice per genotype. Data in (C) to (E) are representative of at least three independent experiments.

IRF2, a member of the IRF family of transcription factors (1720), transcriptionally induces multiple direct target genes, including Tlr3 and Bcl11a (21, 22). The ENU-induced mutation (t→c, hereafter referred to as Irf2splice) was located at the noncoding 3′ splice donor site of Irf2 exon 5 (ex5) (Fig. 1B). Consequently, aberrant mis-spliced mRNAs (denoted as Δex5 and Δex5/6) and minor amounts of full-length transcript were detectable by reverse transcription polymerase chain reaction (RT-PCR) (Fig. 1, B and C). Western blotting analysis detected a 34-kDa truncated IRF2 protein but no detectable amounts of the full-length protein (Fig. 1D). This 34-kDa species is likely encoded by the IRF2 Δex5/6 transcript and lacks 55 amino acids after the DNA binding domain (DBD) due to an in-frame deletion (fig. S2, A and B). Irf2splice/splice BMDMs exhibited attenuated IL-1β secretion in response to cytoplasmic lipopolysaccharide (LPS), a potent stimulator of the noncanonical inflammasome (Fig. 1E) (23, 24). The severity of the defect was only partial compared to that of Gsdmd−/− BMDMs. GSDMD-dependent pyroptosis as measured by lactate dehydrogenase (LDH) release from ATP-stimulated or LPS-transfected Irf2splice/splice BMDMs was also attenuated (Fig. 1E). When cell death was examined by a more sensitive measurement (quantification of the uptake of the cell-impermeant nucleic acid dye, YOYO-1), Irf2splice/splice BMDMs were only partially protected from LPS-induced, GSDMD-dependent pyroptosis (fig. S2C), consistent with a partial reduction in IL-1β release (Fig. 1E). However, Irf2splice/splice BMDMs did not display an altered death response to the GSDMD-independent apoptotic stimuli Fas ligand (FasL) and cytochrome c (Cyto-c) (Fig. 1E and fig. S2C) (25, 26). Together, these data are consistent with Irf2splice being a hypomorphic, mis-splicing mutation.

IRF2 is essential for GSDMD expression

We confirmed a crucial role for IRF2 in both the canonical and noncanonical inflammasome pathways with CRISPR/guide RNA (gRNA) gene editing to generate two distinct Irf2 knockout (KO) immortalized macrophage (iMAC) lines. Both of the Irf2 KO iMAC lines were protected from LPS-induced pyroptosis to an extent equivalent to that of Gsdmd KO iMACs. They still, however, retained an intact death response to GSDMD-independent apoptotic stimuli (Fig. 2A and fig. S3, A to C).

Fig. 2 IRF2 is essential for Gsdmd transcript expression in iMACs.

(A) Measurement of LDH release from the indicated unprimed or Pam3CSK4-primed iMACs stimulated by transfection with LPS or treated with FasL for 16 hours. Luc, Luciferase. Data are means (bars) of three individual replicates (circles). (B) Volcano plot of RNA-seq data. Data are values (dots) of differentially expressed genes (log2 of the fold change was <−1 or >+1; −log10 P value: >1.3) between control Luc gRNA iMACs and IRF2 gRNA2 KO iMACs highlighted in red and blue. Data are from three individual replicates per line. (C) Measurement of the relative amounts of Gsdmd, Casp11, and Casp1 mRNAs (expressed as fold change) in the indicated iMACs stimulated with Pam3CSK4 for 6 hours. Data are means (bars) of four individual replicates (circles). (D) Western blotting analysis of IRF2, GSDMD, caspase-11, and caspase-1 in extracts from the indicated iMACs that were stimulated with Pam3CSK4 for 6 hours. (E) Analysis of cell death in the indicated Pam3CSK4-primed (LPS) or unprimed (FasL) iMACs as determined by measurement of the percentage of YOYO-1+ cells in a live-cell imaging analysis after Amaxa-based electroporation with LPS or stimulation with FasL for the indicated times. Data are means (circles) ± SD (shaded area) of three individual replicates. (F) Western blotting analysis of IRF2 and caspase-11 in the supernatants (sup) and extracts (ext) of the indicated iMACs 2 hours after they were electroporated with LPS. Data in (A) and (C) to (F) are representative of at least three independent experiments.

To determine the mechanism by which Irf2 deficiency conferred attenuated pyroptosis, we used RNA sequencing (RNA-seq) analysis to assess gene expression differences between control and Irf2 KO iMACs. IRF2 was originally discovered as a regulator of type I interferons (IFN-α and IFN-β) (27). IRF2 deficiency results in the enhanced production of type I IFNs by an ill-defined mechanism (28). Accordingly, many IFN target genes, including Oas3 and Rsad2 (29), were induced in Irf2 KO iMACs (Fig. 2B). Casp11 expression was also increased, albeit modestly, which is expected because it is a known IFN-inducible gene (30, 31). In contrast, Gsdmd was among the genes with the most attenuated expression. We confirmed these findings by quantitative PCR analysis and Western blotting of unprimed and primed iMACs (Fig. 2, C and D, and fig. S3A). Ectopic expression of Gsdmd in Irf2 KO iMAC cells almost fully restored GSDMD-dependent pyroptosis (Fig. 2E and fig. S3, D and E). Together, these results indicate that the reduced expression of Gsdmd in Irf2−/− cells was primarily responsible for attenuated pyroptosis. We next looked at caspase-11 activation, the most proximal event upstream of the proteolytic activation of GSDMD. Caspase-11 is auto-processed upon activation, which precedes GSDMD cleavage (32). Caspase-11 activation, as assessed by induction of the auto-processed caspase-11 p26 fragment, was comparable in LPS-stimulated wild-type (WT) and Irf2 KO cells (Fig. 2F). This result confirms that upstream caspase-11 activation was unaffected by Irf2 deficiency.

IRF2 directly drives GSDMD mRNA expression

We suspected that the enhanced production of type I IFNs in Irf2-deficient cells (Fig. 2B) (28) would inhibit Gsdmd expression. Instead, we found the opposite: Both GSDMD mRNA and protein were slightly increased in abundance in BMDMs in response to type I IFNs (IFN-α and IFN-β) (fig. S4, A and B). In addition, GSDMD abundance was unchanged in Ifnar−/−/Ifngr−/− BMDMs, which lack receptors for the type I IFNs and IFN-γ (fig. S4, C and D). We next hypothesized that, being a transcription factor, IRF2 could directly bind to a regulatory element within the GSDMD promoter to induce its expression. IRF2 chromatin immunoprecipitation sequencing (ChIP-seq) analysis identified an IRF2-binding site in the GSDMD locus. An especially prominent peak was evident upstream of the GSDMD transcription start site (TSS) (Fig. 3A). Genome-wide analysis of the top 100 IRF2-binding sequences revealed a highly conserved consensus binding motif present in the GSDMD TSS-proximal region (Fig. 3A and fig. S5, A and B). Specifically, the CACT core sequence within the IRF2 motif showed strong conservation with exceptionally high PhyloP scores among 100 vertebrates (fig. S5, B and C). Direct activation of the Gsdmd promoter by IRF2 was measured in a luciferase reporter assay. Overexpressed IRF2 activated the WT Gsdmd promoter, but promoter activation was almost completely abrogated upon deletion of the IRF2 motif or the CACT sequence (Fig. 3B). Single-nucleotide deletion or replacement within the CACT core (–ACT, C–CT, and C/AACT) strongly inhibited Gsdmd promoter activation, whereas other alterations (CA–T and CAC–) only partially reduced this activity. Hence, IRF2 can drive GSDMD promoter activation by directly engaging its binding motif.

Fig. 3 IRF2 directly drives GSDMD mRNA expression by binding to a consensus site proximal to the GSDMD TSS.

(A) Schematics of the GSDME (top) and GSDMD (bottom) loci with IRF2 ChIP-seq peaks. The IRF2-binding consensus sequence is in green, and the GSDMD TSS is underlined in the genomic sequence. (B) Analysis of GSDMD promoter activity in a reporter assay and schematics of the IRF2 motif mutations. HEK 293T cells were cotransfected with a Gsdmd promoter luciferase reporter gene with the indicated amounts of mock plasmid or of plasmid expressing IRF2. Reporter gene activation was measured by luciferase activity. Data are means (bars) of three individual replicates (circles). (C) CRISPR tiling screen to identify functional Gsdmd promoter regions. Each dot represents the fold enrichment in gRNA after electroporation with LPS. Data are from three individual replicates. (D) Analysis of the relative amounts of Gsdmd, Casp11, and Irf2 mRNAs in the indicated knock-in iMAC clones (fig. S6A) or parental cells stimulated with Pam3CSK4 for 6 hours. Data are means (bars) of four individual replicates (circles). (E) Cell death (percentage cytotoxicity as measured by CellTiter-Glo) of iMACs 16 hours after electroporation with LPS or Cyto-c. Data are means (bars) of three individual replicates (circles). Data in (B), (D), and (E) are representative of at least three independent experiments.

A comprehensive, unbiased CRISPR tiling screen was then undertaken to define the functional promoter region. The genomic sequence 10 kb upstream of the Gsdmd TSS was targeted for systematic mutagenesis using 1293 gRNAs (Fig. 3C). After cytoplasmic LPS pressure, surviving cells due to attenuated Gsdmd expression were recovered. The gRNAs around the IRF2 motif were uniquely enriched, underscoring a nonredundant and exclusive role for the IRF2 motif in Gsdmd expression. To confirm that this discrete IRF2 motif was responsible for Gsdmd expression in macrophages, motif knock-in mutant clones were generated by CRISPR (fig. S6A). Alteration of the IRF2 motif CACT to CGAC was confirmed to compromise Gsdmd promoter activity in the reporter assay (fig. S6B). In parallel, control lines harboring mutations external to the IRF2 motif were also established (fig. S6A). Only IRF2 motif mutant lines abolished the expression of GSDMD mRNA and protein (Fig. 3D and fig. S6C). These lines, however, retained expression of Casp11 and Irf2. Furthermore, the IRF2 motif mutant iMACs were resistant to LPS-induced cell death but retained susceptibility to GSDMD-independent apoptosis (induced by Cyto-c) (Fig. 3E). Together, these data demonstrate that the discrete conserved IRF2-binding core motif is essential for Gsdmd expression and pyroptosis induction in macrophages.

IRF2 is indispensable for Gsdmd expression and pyroptosis in BMDMs

The aberrant Irf2splice Δex5/6 derived truncated protein (Fig. 1D and fig. S2A) still partially activated the Gsdmd promoter (fig. S7A). Therefore, we generated gene-targeted Irf2−/− mice to confirm the role of IRF2 in Gsdmd expression in macrophages. Compared to WT BMDMs, Irf2−/− BMDMs had substantially reduced amounts of GSDMD protein and mRNA (Fig. 4, A and B, and fig. S7B). As expected, the type I IFN–inducible genes Oas3 and Casp11 were increased in expression in Irf2−/− BMDMs, whereas the amounts of Il1b and Nlrp3 mRNAs remained unchanged (Fig. 4B and fig. S7C). The deficiency in Irf2 expression also reduced Gsdmd mRNA abundance in all tissues examined, including the lung and liver (fig. S7D). Consistent with the reduced abundance of GSDMD in BMDMs, Irf2−/− BMDMs exhibited defects in GSDMD-dependent LDH release and IL-1β secretion in response to canonical (ATP) and noncanonical (LPS) inflammasome stimuli (Fig. 4C), confirming a major role for IRF2 in GSDMD expression in primary macrophages.

Fig. 4 IRF2 is indispensable for GSDMD expression and pyroptosis in BMDMs.

(A) Left: Western blotting analysis of GSDMD, IRF2, caspase-11, and caspase-1 in WT, Irf2−/−, and Gsdmd−/− BMDMs stimulated with Pam3CSK4 for 6 hours. Right: Quantified GSDMD band intensity values are means ± SEM from three independent experiments. ****P < 0.0001 by paired Student’s t test. (B) Analysis of the relative amounts of Gsdmd, Casp11, and Il1b mRNAs in BMDMs from the indicated mice treated with Pam3CSK4 for 6 hours. N.D., not detectable. Data are means (bars) of four individual replicates (circles) of two mice per genotype. (C) Measurement of IL-1β secretion and LDH release from the indicated BMDMs stimulated with ATP (4 hours) or after Amaxa-mediated electroporation with LPS (4 hours) or Cyto-c (16 hours). Data are means (bars) of at least three individual replicates (circles) from two mice per genotype. (D) Analysis of cell death in Pam3CSK4-primed BMDMs from the indicated mice as determined by measuring the percentage of YOYO-1+ cells in a live-cell imaging analysis after Amaxa-mediated electroporation with LPS or Cyto-c. Data are means (circles) ± SD (shaded area) of three individual replicates. Data in (A) to (D) are representative of at least three independent experiments.

IRF1 plays a compensatory role in the absence of IRF2

Unlike Irf2 KO iMACs, Irf2−/− BMDMs expressed low but detectable amounts of GSDMD protein (Figs. 2D and 4A). Accordingly, Irf2 deficiency failed to fully protect BMDMs from LPS-induced, GSDMD-dependent pyroptosis at later time points (Fig. 4D). This suggested that a compensatory mechanism drove Gsdmd expression in the absence of IRF2 in BMDMs. Several lines of evidence implicated IRF1 as the compensatory factor: Both IRF1 and IRF2 bind to similar or identical DNA elements (17, 18, 27, 33); IRF1 activated the Gsdmd promoter, albeit to a lesser extent than did IRF2 (Fig. 5A and fig. S8A); and Irf1 mRNA abundance was increased in Irf2−/− BMDMs (fig. S8B). Moreover, functional redundancy of IRF1 and IRF2 is not without precedent, as both play a role in the regulation of Tlr3 expression (21, 22). To address the possibility of an IRF1/2 redundant mechanism, we used CRISPR/gRNA to knock out these genes in human EA.hy926 endothelial cells, which express both IRF1 and IRF2 and undergo pyroptosis in response to LPS (6). Consistent with the BMDM studies, IRF2 deficiency alone markedly reduced GSDMD abundance (Fig. 5B). Note that a further reduction in GSDMD abundance was observed when both IRF2 and IRF1 were targeted, whereas IRF1 deficiency alone only slightly reduced the amount of GSDMD protein. Similar results were observed in GSDMD-dependent pyroptosis assays (Fig. 5C), confirming a compensatory role for IRF1 when IRF2 is absent. In contrast, IRF1/2 deficiency did not have any detectable effect on caspase-4 abundance (Fig. 5B).

Fig. 5 IRF1 plays a compensatory role in the absence of IRF2.

(A) Gsdmd promoter activity as determined by luciferase reporter assay. HEK 293T cells were transfected with the indicated amounts of IRF-expressing plasmid or mock plasmid. Reporter gene activation was measured by luciferase assay. Data are means (bars) of three individual replicates (circles). (B) Left: Western blotting analysis of GSDMD, IRF1, IRF2, and caspase-4 in the indicated human EA.hy926 cells. Right: Quantified GSDMD band intensity values are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01, ****P < 0.0001 by one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test. (C) Analysis of cell death as determined by measurement of the percentage of YOYO-1+ EA.hy926 cells in a live-cell imaging time-course analysis of the cells after transfection with LPS or treatment with etoposide. Data are means (circles) ± SD (shaded area) of three individual replicates. Data in (A) to (C) are representative of at least three independent experiments.


Caspase-1, caspase-4, and caspase-11, which are closely related members of the inflammatory caspase subfamily of intracellular cysteine proteases, cleave GSDMD in response to inflammasome activation. The resultant N-terminal GSDMD fragment assembles into a membrane pore, disrupting electrochemical gradients and terminating in pyroptosis (510). Despite advances in our understanding of the molecular mechanisms leading to GSDMD pore formation, little is known about the transcriptional mechanism responsible for GSDMD expression. Our data reveal that IRF2 is critical for GSDMD expression and pyroptosis induction in macrophages, endothelial cells, and multiple murine tissues. We identified a discrete, but unique, IRF2-binding motif proximal to the GSDMD TSS that was solely responsible for the expression of GSDMD mRNA. IRF2 deficiency resulted in attenuated pyroptosis induction and IL-1β secretion in response to both canonical and noncanonical inflammasome stimuli. Thus, IRF2 is a key master driver of inflammasome responses; in its absence, both the canonical and noncanonical pathways were attenuated. In contrast, whereas IRF family members also play roles in regulating inflammasome activation, they are restricted to specific pathways. For example, IRF8 drives the expression of genes encoding NLR family apoptosis inhibitory proteins (NAIPs) and NLR family CARD (caspase activation and recruitment domain) containing 4 (NLRC4), specialized sensors for the canonical inflammasome (20), and IRF3 transcriptionally induces expression of the gene encoding caspase-11 for the noncanonical pathway (31).

Interest in GSDMD as a therapeutic target has surged given the finding that deletion of Gsdmd strongly ameliorates disease development in multiple mouse models of inflammasome-driven pathologies (6, 11, 12). Similarly to Gsdmd−/− mice (6, 32), Irf2−/− mice are also protected from LPS-induced lethal septic shock (34). This finding in Irf2−/− mice was reported in 2003, but the mechanism for LPS unresponsiveness remains unsolved until our present study linked IRF2 to the direct transcriptional induction of GSDMD expression. Targeting IRF2 or as yet to be discovered upstream factors may provide new therapeutic modalities for the treatment of sepsis and other canonical and noncanonical inflammasome-mediated pathologies.


ENU-mutagenized mouse strains

G3 offspring used for the phenotypic screen were from ENU-treated C57BL/6 mice, as described previously (15). All animals were housed under specific pathogen–free conditions at the Australian Phenomics Facility. Animals were used in this study in accordance with protocols approved by the Australian National University Animal Experimentation Ethics Committee.

Exome capture and sequencing

Exome-enriched, paired-end libraries were prepared from genomic DNA of G1 mice using the Agilent SureSelect XT2 Mouse All Exon Kit (Agilent) according to the manufacturer’s instructions. Each sample was prepared with an index and then pooled in a batch of eight in equimolar amounts before exome enrichment. Each 8-plex was run on four lanes of an Illumina HiSeq2500 with 100–base pair (bp), paired-end reads. ENU variants were identified as described previously (35). In brief, sequence reads were mapped to the GRCm38 assembly of the reference mouse genome using the default parameters of the Burrows-Wheeler Aligner ( Untrimmed reads were aligned, allowing a maximum of two sequence mismatches, and were discarded if aligned to the genome more than once. Sequence variants were identified with SAMtools ( and annotated using Annovar ( PolyPhen2 ( and SIFT ( were used for the calculation of variant effects. Variants were filtered to prioritize novel variants not in dbSNP (or in a list of common mouse variants identified by the pipeline) and predicted to be deleterious to the protein by PolyPhen, SIFT, or both. For variant validation, samples were genotyped for the ENU variants identified by exome sequencing using a competitive, allele-specific, dual fluorescence resonance energy transfer (FRET)–based assay, KASP (LGC). Primers were designed on the basis of the SNV locus flanking sequence. Assays were read on an ABI7500 real-time PCR instrument (Thermo Fisher Scientific). Genotyping primers used for ENU mutations were as follows: Irf2 (forward: 5′-GAAGGTGACCAAGTTCATGCTGAAGAGAGAGTTAAGCACATCAAGGT, FM: 5′-GAAGGTCGGAGTCAACGGATTAAGAGAGAGTTAAGCACATCAAGGC, and reverse: 5′-CAGGGGCTTTCTAGTCAACCAAAGAT), Armc6 (forward: 5′-GAAGGTGACCAAGTTCATGCTCCTGGATGCCCAGGGCTT, FM: 5′-GAAGGTCGGAGTCAACGGATTCTCCTGGATGCCCAGGGCTA, and reverse: 5′-GCTAGCGTGGCCACGAGAAGAT), and Mylk3 (forward: 5′-GAAGGTGACCAAGTTCATGCTCACAGACTTACTTACCTAGAACCAC, FM: 5′-GAAGGTCGGAGTCAACGGATTGCACAGACTTACTTACCTAGAACCAT, and reverse: 5′-CTTCTGCTGCAGAGGCTGCCAT).

Other mice

Gsdmd−/− (1632 bp deletion) mice with the C57BL/6N background were described previously (6). C57BL/6N mice were purchased from Charles River and used as controls. Ifnar1−/−/Ifngr1−/− C57BL/6J and control C57BL/6J were obtained from The Jackson Laboratory. For Irf2−/− mice, Irf2 exons 3 and 4 in C57BL/6N C2 embryonic stem cells were flanked by loxP sites and then deleted with Cre recombinase. Irf2−/− mice were genotyped with PCR primers (5′-TGAGCACTTAGCCAACTG, 5′-TTACCTTGATGTGCTTAACTCT, and 5′-TCTGAAGAGCGGAGCAT), yielding a 282-bp WT DNA fragment and a 342-bp mutant DNA fragment. The Genentech Institutional Animal Care and Use Committee approved all animal studies.

Reagents and antibodies

The following reagents were used to treat cells: Ultrapure LPS (Escherichia coli O111:B4, InvivoGen), Pam3CSK4 (InvivoGen), IFN-α (PBL Assay Science), IFN-β (PBL Assay Science), IFN-γ (eBioscience), ATP (Sigma), Cyto-c (bovine heart Cyto-c, Sigma), etoposide (Cell Signaling Technology), and FasL (MegaFasL, AdipoGen). Antibodies used in this study were specific for the following targets: caspase-1 (clone 4B4, Genentech), caspase-4 (clone 4B9, Enzo), caspase-11 (clone 17D9, Novus Biologicals), GSDMD [clone 17G2G9 (9), Genentech], IRF1 (clone D5E4, Cell Signaling Technology), and IRF2 (raised against full-length recombinant mouse IRF2, mouse IgG2b, clone 25, Genentech). In addition, horseradish peroxidase (HRP)–conjugated antibody against β-actin (AC-15, Novus Biologicals) and the FLAG epitope (M2 HRP, Sigma) were used.

BMDM and iMAC culture and stimulation

Bone marrow cells were differentiated into macrophages in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) low-endotoxin fetal bovine serum (FBS; Omega Scientific) and 20% (v/v) L929-conditioned medium at 37°C with 5% CO2. Unless otherwise specified, BMDMs were harvested on day 5. iMACs were maintained in RPMI 1640 medium supplemented with 10% (v/v) low-endotoxin FBS, murine granulocyte-macrophage colony-stimulating factor (GM-CSF) (20 ng/ml) (eBioscience), and 1 μM β-estradiol (Sigma) and then were differentiated with L929-conditioned medium for 5 days at 37°C with 5% CO2. For stimulation, cells were plated overnight at ~1.0 × 106 cells/ml in 100 μl on 96-well plates or in 500 μl for 24-well plates. For 96-well–based inflammasome stimulations, cells were primed with Pam3CSK4 (1 μg/ml) for 5 hours, which was followed by stimulation with 5 mM ATP, or else the cells were transfected with LPS (5 μg/ml) and 0.25% (v/v) FuGENE HD (Promega) in Opti-MEM I media (Gibco) (24). For FasL stimulations, cells were left unprimed and were stimulated with FasL (1 μg/ml) in Opti-MEM I. For electroporation, cells were collected and resuspended at 0.3 × 106 to 0.5 × 106 cells/10 μl of R buffer from the Neon Transfection System (Thermo Fisher Scientific). Stimuli were mixed with cells at the following ratios: LPS 2 μg/1 × 106 cells; Cyto-c 50 μg/1 × 106 cells. Cells were electroporated using the 10-μl Neon tip with 1720 Voltage, 10 Width, 2 Pulse settings according to the manufacturer’s instructions, added to medium to yield ~0.5 × 106 to 1.0 × 106 cells/ml, and then plated in 100 μl for analysis on 96-well plates. Where indicated in the figure legends, cells were electroporated with the Amaxa 4D-Nucleofector Y Unit (Lonza) in Opti-MEM I medium with LPS (5 μg/ml) or Cyto-c (100 μg/ml). For IFN treatments, cells were cultured with IFN-α (50 ng/ml), IFN-β (50 ng/ml), or IFN-γ (100 ng/ml).

Other cell cultures

Human EA.hy926 Cas9+ cells (6) were maintained in DMEM supplemented with 10% (v/v) low-endotoxin FBS. Cells were plated overnight at ~1.5 × 105/ml in 100 μl in 96-well plates and stimulated the following day with 500 μM etoposide or transfected with LPS (5 μg/ml) and 0.2% (v/v) Lipofectamine LTX (Thermo Fisher Scientific) in Opti-MEM I medium.

Cell death measurements

For all stimulations that were subjected to imaging analysis, medium containing YOYO-1 (491/509) dye (Thermo Fisher Scientific) at a final concentration of 200 nM was added at the time of stimulation. Images were scanned in the green channel every 30 min to about 1 hour for at least 16 hours with IncuCyte ZOOM (Essen BioScience) at ×10 magnification. Nuclear-ID Red DNA stain (Enzo Life Sciences) was added at a 500-fold dilution at the last time point and scanned in the red channel. IncuCyte software was used to determine the total number of dead YOYO+ cells and Nuclear-ID+ cells (live and dead). Percentage death (or % YOYO+) was calculated as the number of YOYO+ cells divided by the total number of Nuclear-ID+ cells. Culture medium was analyzed for the percentage of LDH release with the CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) and for IL-1β secretion with the Mouse IL-1β Tissue Culture Kit (Meso Scale Discovery). For iMAC cytotoxicity assays, the CellTiter-Glo reagent (Promega) was used.

Sample preparation for Western blotting

For standard Western blotting analysis (cell extracts only), 0.5 × 106 cells were lysed in 75 μl of NP-40 buffer [10 mM tris (pH 7.5), 150 mM NaCl, 1% NP-40, 2.5 mM MgCl2, and 0.5 mM CaCl2]. For Western blotting analysis of the caspase-11 p26 fragment, cell culture medium and extracts from LPS-stimulated iMACs were prepared as reported previously (32). Briefly, ~8.0 × 105 cells were electroporated as described earlier with the Neon Transfection System and then were incubated in 75 μl of Opti-MEM I medium for 2 hours. Cell culture medium was collected, and the cells were lysed in NP-40 buffer with 1× cOmplete Protease Inhibitor (Roche Applied Science) and then combined with the cell culture medium with additional protease inhibitor.


GSDMD band intensities were quantified using ImageJ software where indicated in the figure legends. Statistical analysis was performed by paired Student’s t test or by one-way ANOVA followed by Tukey’s multiple comparison test using GraphPad Prism software.

Plasmids and transient transfection

The complementary DNAs (cDNAs) encoding C-terminally FLAG-tagged full-length mouse IRF2, IRF2 Δexon5, IRF2 Δexon5/6 (fig. S2A), and IRF2 ΔDBD (ΔP2-K120) were synthesized and subcloned into pcDNA3.1 Zeo(+) (Thermo Fisher Scientific) for transient expression in human embryonic kidney (HEK) 293T cells (American Type Culture Collection). HEK 293T cells were cultured overnight in 96-well plates at 1.2 × 105 cells/ml and then transfected with a total 60 ng of plasmid with 0.16 μl of Lipofectamine 2000 (Thermo Fisher Scientific) per well. Cells were lysed with radioimmunoprecipitation assay buffer and protease inhibitor 24 hours after transfection.

Generation of stable cell lines

ER-Hoxb8–immortalized WT and Gsdmd−/− C57BL/6N mouse–derived iMACs were previously reported (6, 36, 37). For CRISPR gRNA KO stable lines, parental Cas9+ iMAC lines were generated by retroviral transduction with human codon–optimized Streptococcus pyogenes Cas9 [cloned in pMX–green fluorescent protein (GFP) (Cell BioLabs)], which was followed by sorting and subcloning of GFP-positive cells. Cas9+ cells were transduced with gRNAs by lentiviral delivery with the pLKO.1 vector (Sigma), which was followed by the selection of gRNA-expressing cells with puromycin (10 μg/ml) (Gibco) and testing after 14 days. The gRNA sequences are as follows: Luciferase: 5′-GCATGCGAGAATCTCACGC, Gsdmd: 5′-AGAGGCGATCTCATTCCGG, Irf2 gRNA1: 5′-TAAATTCCAATACGATACC, and Irf2 gRNA2: 5′-GGATGCATGCGGCTCGGCA. For reconstitution of Gsdmd, the cDNA encoding mouse GSDMD was subcloned into the piggyBac vector (BH1.11, Genentech). IRF2 gRNA2 KO iMAC cells were co-electroporated with GSDMD/BH1.11 and the transposase vector (pBO, Transposagen Biopharmaceuticals) using the Amaxa shuttle with Solution V (Lonza). Cells were selected with blasticidin (6.25 μg/ml) (Gibco). For CRISPR-mediated, homology-directed repair to generate the IRF2-binding motif knock-in mutants, a protocol from IDT (Integrated DNA Technologies) was followed, in which a Cas9+ iMAC parental line was electroporated with synthetic [annealed crispr RNA (crRNA) + trans-activating crRNA (tracrRNA) from IDT] gRNA (5′-ACGTCGGGCTGAAGCTTTA) and single-stranded DNA (ssDNA) donor template (Ultramer Oligo DNA from IDT) with ~40 bp of homology arms on both sides containing the site mutation introducing a Sal I restriction site in place of the IRF2 consensus sequence and mutation to disrupt the protospacer adjacent motif (PAM) site preventing further CRISPR cuts (5′-CTCCTTGAAAGCACCGCGCCCGCTACGTCGGGCTGAAGCTTTACaGTgTCgacTTTGTGTCCTGCCGCCTGAGTTCCGCTCTTGGTCGTGGCTCCC; bold italics indicate the gRNA recognition sequence; lowercase letters represent mutations; the Sal I site is underlined). After 5 days, cells were plated to identify single-cell clone lines with integrated mutations. In brief, cells were collected in 96-well plates and genomic DNA was harvested with the Quick-DNA 96 Kit (Zymo Research). Genomic DNA was subjected to PCR with the Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific) with either WT or mutant primers (WT forward: 5′-GCTGAAGCTTTACGGTTTCACTTTTG: mutant forward: 5′-GGCTGAAGCTTTACAGTGTCGAC, which were paired with the reverse primer: 5′-CGGGGAGGGGTCTATTGTCT) to identify a ~100-bp product corresponding to the WT or mutant sequences. For potential clones, a larger, 250-bp region around the target IRF2 motif was amplified by PCR with the following primers: forward: 5′-GAAAGCGAAGCTCCCGGAT; reverse: 5′-CGGGGAGGGGTCTATTGTCT. The amplified fragment was subjected to Sal I (New England Biolabs) digestion for prescreening and then subcloned into Zero Blunt TOPO (Thermo Fisher Scientific) for Sanger sequencing. From the same gRNA/ssDNA donor–transfected parental pool, three control cell lines and three IRF2 motif mutant lines were established (fig. S6A). For EA.hy926 CRISPR gRNA KO lines, parental Cas9+ EA.hy926 cells were electroporated with synthetic gRNA (annealed crRNA:tracrRNA from IDT) using the Neon Transfection System. The crRNA sequences are as follows: Luciferase: 5′-GCATGCGAGAATCTCACGC, GSDMD: 5′-GGTAGTCCGGAGAGTGGTCC, IRF1_a: 5′-TAATCTGCATCTCTAGCCA, IRF1_b: 5′-CATGCTGCCAAGCATGGCT, IRF2_a: 5′-GTTATCTGCTCCTCCAGCCA, and IRF2_b: 5′-GTCTAGCCGCATGCATCCAG.

Gsdmd promoter CRISPR tiling array screen

For systematic mapping of the functional Gsdmd promoter with CRISPR/Cas9, 1293 gRNAs tiling across a total of 10 kb of Gsdmd genomic sequence (from −10 kb to −1 bp) were designed using a custom gRNA design algorithm. The designed guides were cloned into TRC2-pLKO-puro (Sigma) by Cellecta Inc. Cas9+ iMACs were infected with a pooled lentiviral library at an MOI (multiplicity of infection) of 0.3, with sufficient cell numbers plated to obtain a 1500-fold coverage of gRNA in triplicate. After 13 days of selection, the cells were subjected to Neon electroporation with or without LPS and incubated for 5 days to enable the growth of survivors. Surviving cells were subjected to genomic DNA extraction with Puregene reagents (Qiagen). PCR was performed with Phusion High-Fidelity DNA Polymerase using the following primers: forward: 5′-TCTTGTGGAAAGGACGAGGTACCG; reverse: 5′-TCTACTATTCTTTCCCCTGCACTGT. The PCR product was then purified using AMPure XP (Beckman Coulter). Next-generation sequencing was performed on a MiSeq system (Illumina) with 50 million sequence reads per sample. The mean accumulation of gRNAs in the LPS-treated samples compared to that in the controls was plotted over the corresponding cutting site with a uniform average smoothing window of 25 points.

Quantitative RT-PCR

Total RNA from BMDMs and iMACs was prepared with an RNeasy kit (Qiagen). The entire Irf2 coding region was amplified with a SuperScript One-Step RT-PCR for long templates kit (Thermo Fisher Scientific) and subcloned into pCR2.1 TOPO (Thermo Fisher Scientific) for sequencing. The primers used were 5′-GTGGCTGGAGGAGCAGATAA and 5′-CAACAACCACCAGGGAGAGT. TaqMan quantitative PCR (Applied Biosystems) was performed on the QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific) with the following primer and probe sets (Applied Biosystems): Gsdmd (Mm00509958_m1), Irf2 (Mm00515204_m1), Casp1 (Mm01243908_m1), Casp11 (Mm00432304_m1), Nlrp3 (Mm00840904_m1), Il1b (Mm00434228_m1), Oas3 (Mm00460944_m1), Irf1 (Mm00515191_m1), Gapdh (Mm99999915_g1), and Hprt (Mm03024075_m1). Samples were normalized by quantification of Gapdh and Hprt mRNAs.

RNA sequencing

Total RNA was extracted from control Luc gRNA iMACs and Irf2 gRNA2 KO iMAC lines (n = 3) using an RNeasy kit (Qiagen) with on-column deoxyribonuclease digestion. Quality control of total RNA was performed to determine sample quantity and quality. The concentration of RNA was determined using a NanoDrop 8000 (Thermo Fisher Scientific), and the integrity of the RNA was determined by Fragment Analyzer (Advanced Analytical Technologies). Total RNA (100 ng) was used as an input material for library preparation using the TruSeq Stranded Total RNA Library Prep Kit (Illumina). The sizes of the libraries were confirmed using High Sensitivity D1K screen tape (Agilent Technologies), and their concentrations were determined with a quantitative PCR–based method using a Library Quantification kit (KAPA). The libraries were multiplexed and sequenced on an Illumina HiSeq4000 (Illumina) to generate 30 million single-end, 50-bp reads. For RNA-seq analysis, the raw FASTQ reads were aligned to the mouse reference genome (GRCm38-mm10) using GSNAP (with parameters -M 2 -n 10 -B 2 -i 1 -N 1 -w 200000 -E 1 --pairmax-rna = 200000 --clip-overlap). Reads were filtered to include only the uniquely mapped reads. Differential expression analysis was performed using the voom/limma R package (38). Genes were considered to be differentially expressed if the log2 of the fold change was >1 or <−1 and the adjusted P value was <0.05. RNA-seq data generated are available under the following Gene Expression Omnibus (GEO) submission ID: GSE130195.

ChIP-seq and IRF2 motif analysis

ChIP-seq data for IRF2 in human K562 cells were obtained from the ENCODE project (ENCSR376WCJ) (39). The fold change over control bigWig file from both the first and third isogenic replicates was parsed using pyBigWig ( and was smoothed with a one-dimensional Savitzky-Golay filter with a window size of 15. IRF2 binding (filtered fold change over control) was plotted for the given chromosome region using Matplotlib. For IRF2 motif analysis, the BED hg19 peaks file from the ChIP-seq experiment described earlier (GEO: GSE91709) was used to identify the top 100 peaks. The DNA sequences corresponding to the top 100 peak locations served as the input for the MEME (Multiple EM for Motif Elicitation) (40) server to identify the IRF2-binding motifs. Using the MEME server’s FIMO motif–scanning feature, the top motif was used to detect putative IRF2-binding sites within the human genome. The obtained IRF2 motif sequence (fig. S5A) is similar to the previously reported mouse IRF2 motif (22). For IRF2 motif conservation at the GSDMD TSS-proximal region across species, the basewise conservation PhyloP score (41) from 100 vertebrates for the region corresponding to the IRF2 motif was plotted using Matplotlib. Both the PhyloP scores and the corresponding 100-vertebrate MULTIZ alignment (42), with species not harboring the GSDMD gene filtered out, were obtained from the University of California, Santa Cruz (UCSC) table browser. GSDMD TSS determination was identified by a cap analysis of gene expression (CAGE) from the FANTOM5 database (43).

Reporter assay and plasmids

The WT and mutant genomic sequences of Gsdmd containing the TSS (−950 bp to +50 bp, Fig. 3B) were synthesized and subcloned into pGL3-Basic (Promega). In a 96-well plate, HEK 293T cells were cotransfected with Gsdmd (−950 bp to +50 bp) pGL3-Basic vector (25 ng) and pRL-TK Renilla Luciferase control (5 ng) together with (0.01 to 1 ng) of C-ter FLAG-Myc–tagged mouse IRF2 in pCMV6-Entry vector (TrueORF, OriGene) or mock plasmid. Twenty-four hours later, luciferase activity was measured by Dual-Luciferase Reporter Assay (Promega). For other IRFs, C-ter FLAG-Myc–tagged mouse IRF cDNAs in pCMV6-Entry vector (TrueORF, OriGene) were used.


Fig. S1. Homozygosity for the Irf2 point mutation correlates with low responsiveness to ATP.

Fig. S2. Irf2splice is a hypomorphic mis-splice mutation.

Fig. S3. Characterization of Irf2 gRNA KO iMACs.

Fig. S4. IFNs are dispensable for the attenuation of GSDMD expression.

Fig. S5. The GSDMD TSS-proximal IRF2 consensus motif is conserved across species.

Fig. S6. IRF2 motif mutant alleles.

Fig. S7. Characterization of Irf2−/− BMDMs and mice.

Fig. S8. Expression of IRF1.

Table S1. Bioinformatic analysis of ENU-induced SNVs present in IGL03671 pedigree.


Acknowledgments: We thank the staff of the Australian Phenomics Facility, the Genentech Transgenic Technology, FACS Cores, Antibody Discovery, Sequencing Laboratory, and Virus Laboratory Core facility. We thank the ENCODE Consortium and the ENCODE production laboratory for generating datasets. Special thanks to C. Espinosa, L. Tam, and S. Warming. Author contributions: N.K., B.L.L., I.B.S., O.S.K., K.O., K.M.M., B.H., C.W., Z.M., Y.Z., V.C., L.X.M., and E.M.B. designed and performed experiments. T.D.A., S.K., R.R., and O.S.K. provided computational analysis. M.R.-G. generated the Irf2–/– mice. N.K., B.L.L., I.B.S., O.S.K., E.M.B., and V.M.D. prepared the manuscript. N.K., C.C.G., E.M.B., and V.M.D. contributed to the study design and data analyses. Competing interests: N.K., B.L.L., I.B.S., O.S.K., K.O., K.M.M., B.H., C.W., Z.M., S.K., R.R., M.R.-G., and V.M.D. are employees of Genentech Inc. Data and materials availability: All data are available in the main text or the Supplementary Materials. RNA-seq data have been deposited in the GEO database ( with the identifier GSE130195. Cell lines generated through the Hoxb-8 system are part of a material transfer agreement with Murdoch Children’s Research Institute.
View Abstract

Navigate This Article