Gasdermin D-mediated delivery of caspase inhibitors to suppress pyroptosis

Abstract

Caspase-1, -4, -5 and -11 activate gasdermin D (GSDMD) pores, causing pyroptotic cell death and the release of the interleukins IL-1β and IL-18 (ref. ¹). Blocking this pathway holds therapeutic promise for the treatment of inflammatory disorders, but cell-permeable caspase inhibitors have not proved successful in clinical trials². Here we describe covalent caspase inhibitors that selectively block pyroptosis and IL-1β secretion despite being excluded from healthy cells. These inhibitors did not prevent caspase-driven apoptosis, implying that GSDMD pores facilitated their uptake. Membrane-impermeable dyes entered the cells rescued from pyroptosis, consistent with transient membrane permeabilization by GSDMD pores. Caspase inhibition prevented rather than delayed cell death, consistent with membrane repair mechanisms neutralizing the initial GSDMD pores. Inhibiting caspase-1 and caspase-11 suppressed IL-1β and IL-18 production in a mouse model of endotoxic shock, underscoring the therapeutic potential of exploiting GSDMD pores for targeted caspase inhibition in inflammatory diseases.

Covalent, membrane-impermeable caspase inhibitors exploit transient GSDMD pore formation to enter pyroptotic cells, selectively blocking caspase-1 and caspase-11-driven pyroptosis and IL-1β and IL-18 release, enabling targeted suppression of inflammatory pathology in vivo.

Main

Cell death in multicellular organisms has pivotal roles in development, homeostasis and defence against pathogens¹. The programmed cell death pathways of apoptosis and pyroptosis are mediated by aspartate-specific cysteine proteases called caspases. The proinflammatory caspases (human caspase-1, -4 and -5, and mouse caspase-1 and -11) are activated within inflammasome complexes and cleave the pore-forming protein GSDMD to unleash pyroptosis, a lytic form of cell death. Distinct caspases, including caspase-8 and caspase-9, initiate apoptosis by cleaving and activating caspase-3 and caspase-7. In excess, both apoptosis and pyroptosis have been shown to cause inflammatory disorders in mice¹. The potential therapeutic benefit of caspase inhibition in human diseases that are characterized by uncontrolled cell death remains uncertain. Clinical trials of cell-permeable pan-caspase inhibitors have failed owing to toxicity or lack of efficacy². One potential liability of pan-caspase inhibition is that inactivation of caspase-8 sensitizes certain cell types to a lytic form of cell death termed necroptosis¹. The pursuit of caspase-selective inhibitors has been challenging because of pronounced structural homologies among the caspases³. Extracellular nanobodies targeting GSDMD can, under certain conditions, reduce pyroptotic cell death, possibly traversing early GSDMD pores and then limiting subsequent pore formation^4,5. These findings prompted us to test whether typically membrane-impermeable caspase inhibitors with selectivity for caspase-1, -4, -5 and -11 might also enter cells through GSDMD pores and prevent pyroptosis. If these caspase inhibitors only enter cells that have engaged the pyroptosis pathway, they might be tolerated better than cell-permeable pan-caspase inhibitors.

Substrate specificity of caspase-1, -4 and -5

As a first step towards generating libraries of potentially selective caspase inhibitors, we screened the substrate preferences of human caspase-1, -4 and -5 using hybrid combinatorial substrate libraries (HyCoSuLs) incorporating diverse tetrapeptides with Asp in the P1 position (Extended Data Fig. 1). HyCoSuL sublibraries P2, P3 and P4 were composed of tetrapeptides with an equimolar mixture of natural amino acids in two positions (P2 and P3) and a fixed natural or unnatural amino acid in the remaining space⁵ (P4). Human caspase-1, -4 and -5 resembled mouse caspase-1 and -11 (ref. ⁶) by favouring His in the P2 position (Extended Data Fig. 1). By contrast, human caspase-3 prefers small, aliphatic P2 residues such as Val and Ile⁷. All five caspases preferred Glu in the P3 position, but caspase-11 better tolerated bulky residues such as Bip (3-biphenyl-4-yl-l-alanine), Bpa (4-benzoyl-l-phenylalanine) and Tyr(2,6-Cl₂-benzyl)⁵ (Extended Data Fig. 1). Caspase-11 also accommodated small, aliphatic P4 residues (Tle, Val and Ile), whereas caspase-1, -4 and -5 favoured bulky P4 residues (Trp and Tyr). Thus, human caspase-4 and mouse caspase-11 have distinct amino acid preferences in their substrate-binding regions, despite being the core component of the non-canonical inflammasome that responds to cytoplasmic lipopolysaccharide (LPS) in both species^8,9.

Extended Data Fig. 1. Substrate specificity of human caspases 1, 4, and 5.

Hybrid combinatorial substrate libraries were screened using purified recombinant enzyme (P) or E. coli lysates (L) containing the caspase. Mix: equimolar mixture of 19 natural amino acids (AA): fixed natural or unnatural amino acid:reporter group (ACC) indicated by a yellow star. Results are the mean of 3 independent experiments.

Given that human caspase-1, -4 and -5 have similar cleavage preferences, we tested their ability to cleave a set of tetrapeptide caspase substrates described previously⁶. A tetrapeptide bearing the Trp-Val-His-Asp sequence found in human pro-IL-1β was among those most susceptible to cleavage by human caspase-1, whereas hydrolysis of the Phe-Leu-Thr-Asp sequence found in human GSDMD was approximately threefold less efficient (Extended Data Table 1). Interestingly, caspase-1 exhibited catalytic efficiencies (k_cat/K_M) that were an order of magnitude greater than those of caspase-4, -5 and -11 (ref. ⁶) (Extended Data Table 1). The incorporation of non-natural amino acids into the tetrapeptide sequences elevated the cleavage efficiencies, while concurrently enhancing the selectivity factor for proinflammatory over proapoptotic caspases.

Extended Data Table 1.

Caspase substrate specificity

Human caspase k_cat/K_M values (M⁻¹s⁻¹) for tetrapeptide substrates having the format Ac (acetyl)-P4-P3-P2-P1(Asp)-ACC (7-amino-4-carbamoylmethylcoumarin). n.d., not determined. See Methods for the details.

Screening a caspase inhibitor library

Taking into account the cleavage site preferences of caspase-1, -4, -5 and -11, we synthesized around 100 tetrapeptides bearing an acyloxymethyl ketone (AOMK) reactive group to covalently modify their targets in an irreversible manner^10,11. We determined their inhibition parameters on recombinant human caspase-1, -3, -4, -6, -7, -8, -9 and -10 (Supplementary Table 1). Tetrapeptides KGR-3, KGR-23 and KGR-72 (Extended Data Fig. 2a) stood out as potent and somewhat selective inhibitors of human caspase-1 (Supplementary Table 1) and were subsequently found to also inhibit mouse caspase-1 (Extended Data Table 2). The cell permeability of KGR-3, KGR-23 and KGR-72 was deemed very low when assessed in gMDCKI cells lacking the efflux pump MDR1 (ref. ¹²) (Extended Data Table 3).

Extended Data Fig. 2. Screening for inhibitors of nigericin-induced pyroptosis.

a, LC-MS analysis of representative tetrapeptide inhibitors. Upper panels show LC chromatograms recorded at 220 nm. Early peaks around 4 min correspond to DMSO/solvent, whereas major peaks between 10 to 14 min correspond to the indicated inhibitors. Lower panels show MS analysis of the corresponding inhibitor peak. All inhibitors were determined to be >98% pure. b-c, Percentage of LDH released from LPS-primed THP-1 cells after treatment with 6.25 (b) or 5 (c) µM nigericin for 2 h. Where indicated, cells were co-treated with 25 µM MCC950, 25 µM emricasan, or 25 µM tetrapeptide-AOMK (assigned the prefix KGR for tracking purposes). NT, no treatment. Bars indicate the mean of 3 replicates (circles). The horizontal, dashed yellow line in (c) indicates the degree of protection provided by MCC950.

Source data

Extended Data Table 2.

Kinetic parameters of the caspase inhibitors

Caspase k_obs/I values (M⁻¹s⁻¹) for caspase inhibitors having the format Ac-P4-P3-P2-P1(Asp)-AOMK, or emricasan (N-[2-(1,1-dimethylethyl)phenyl]-2-oxoglycyl-N-[(1S)-1-(carboxymethyl)-2-oxo-3-(2,3,5,6-tetrafluorophenoxy)propyl]-L-alaninamide). See Methods for how this parameter was calculated. n.d., not determined.

Extended Data Table 3.

Caspase inhibitor permeability

Apparent permeability (Papp) values for a subset of tetrapeptide caspase inhibitors in the apical to basolateral direction (A-to-B) using gMDCKI cells. Media in the apical and basolateral chambers were pH 7.4. Results are the mean of at least 4 technical replicates.

To determine whether any of our tetrapeptides might pass through GSDMD pores in the cell membrane to inhibit caspase-1, we tested their ability to block NLRP3- and caspase-1-dependent pyroptosis in LPS-primed human THP-1 monocytic cells treated with nigericin. Cell lysis was monitored by lactate dehydrogenase (LDH) release. As expected, nigericin-induced cell lysis was inhibited by the NLRP3 inhibitor MCC950 (refs. ^13–15), the pan-caspase inhibitor emricasan (IDN-6556)¹⁶, caspase-1 deficiency or GSDMD loss (Extended Data Fig. 2b). Several tetrapeptides, including KGR-3, KGR-23 and KGR-72, greatly reduced or inhibited nigericin-induced cell lysis (Extended Data Fig. 2c).

We also screened a subset of our tetrapeptides for their ability to block caspase-4- or caspase-11-driven pyroptosis, focusing on those with the highest potency against recombinant caspase-1 or that best blocked NLRP3-dependent pyroptosis in THP-1 cells. Rather than electroporate or transfect LPS into cells to activate caspase-4 or caspase-11 (refs. ^9,17,18), we engineered THP-1 cells to express either FK506 binding protein 12 (FKBP)–caspase-4 or FKBP–caspase-11 fusions in which FKBP replaced the caspase pro-domain. Cell-permeable B/B ligand was added to dimerize the FKBP fusions, resulting in GSDMD cleavage and pyroptosis that was inhibited by the pan-caspase inhibitor emricasan (Extended Data Fig. 3a–d). Cell death was determined by uptake of the membrane-impermeable dye propidium iodide (PI; Extended Data Fig. 3b), by CellTiter-Glo assay, which measures cellular ATP content (Extended Data Fig. 3c), or by LDH release (Extended Data Fig. 3d). We identified several tetrapeptides that were comparable to emricasan in suppressing cell death induced by caspase-4 and/or caspase-11 (Extended Data Fig. 3d). For example, KGR-3 blocked GSDMD processing and cell death induced by caspase-4 but was less effective against caspase-11 (Extended Data Fig. 3d). Others, including KGR-23, KGR-53 and KGR-72, suppressed cell death induced by caspase-4 or caspase-11. To exclude tetrapeptides that might access cells in a GSDMD-independent manner and be less selective for caspase-1, -4 or -11, we also screened for inhibition of staurosporine-induced apoptosis, which is thought to be dependent on BAX or BAK in the mitochondrial pathway that activates caspase-9 (ref. ¹⁹). However, unlike emricasan, none of the selected tetrapeptides suppressed staurosporine-induced cell death (Extended Data Fig. 3e).

Extended Data Fig. 3. Screening for inhibitors of caspase-4 or -11.

a, Western blots of THP-1 cells. Results representative of 2 independent experiments. b, Percentage of THP-1 cells exhibiting nuclear PI staining. Data are the mean ± SD of 3 replicates. c, Percentage of viable THP-1 cells as measured by CellTiter-Glo assay. d, Percentage of LDH released from THP-1 cells. Where indicated, cells were co-treated with 10 µM emricasan or tetrapeptide-AOMK. NT, no treatment. The horizontal, dashed yellow line indicates the degree of protection provided by emricasan. e, Percentage of viable THP-1 cells as measured by CellTiter-Glo assay. Where indicated, cells were co-treated with 10 µM tetrapeptide-AOMK. Bars in (c-e) indicate the mean of 3 replicates (circles). Results representative of 2 independent experiments.

Source data

Characterization of inhibitor KGR-3

We selected KGR-3, one of the more potent inhibitors of caspase-1 and caspase-4, for further characterization. To determine whether KGR-3 suppresses the release of IL-1β from pyroptotic cells, we primed THP-1 cells expressing FKBP–caspase-4 with the TLR1–TLR2 agonist Pam3CSK4 for 4 h to induce IL1B. The addition of B/B ligand for a further 24 h triggered pyroptosis and IL-1β secretion, whereas B/B ligand plus KGR-3 did not (Fig. 1a and Extended Data Fig. 4a). As expected, KGR-3 blocked GSDMD cleavage in the B/B-treated cells (Extended Data Fig. 4b). KGR-3 also suppressed IL-1β release from phorbol 12-myristate 13-acetate (PMA)-differentiated and Pam3CSK4-primed THP-1 cells treated with ATP to engage the NLRP3–ASC–caspase-1 inflammasome^20,21 (Fig. 1b and Extended Data Fig. 4c), and Ker-CT human keratinocytes treated with Val-boroPro (VBP) to activate the NLRP1–ASC–caspase-1 inflammasome²² (Fig. 1c and Extended Data Fig. 4d,e). Interestingly, VBP-induced IL-1β release from Ker-CT cells was mediated by GSDMD and GSDME (also called DFNA5; Extended Data Fig. 4f). These results indicate that KGR-3 can suppress gasdermin-dependent IL-1β secretion downstream of caspase-1 or caspase-4.

Fig. 1. KGR-3 inhibits caspase-1- or caspase-4-driven pyroptosis and IL-1β secretion.

a, Percentage of THP-1 cells stained with PI (left) and the amount of IL-1β released into the culture supernatant (right). Bars represent the mean of three replicates (circles). Results representative of two independent experiments. KO, knockout. b,c, IL-1β released from THP-1 (b) and Ker-CT (c) cells into the culture supernatant. Bars are the mean of three replicates (circles). Results representative of two independent experiments. WT, wild-type. d, Percentage of viable EA.hy926 cells by CellTiter-Glo assay. Bars are the mean of three replicates (circles). Results representative of three independent experiments. e, Percentage of EA.hy926 cells stained with PI. Data are mean ± s.d. of three replicates. f, EA.hy926 cells cultured in the presence of PI for 15 h. Yellow arrowheads highlight blebbing apoptotic cells. Black arrows indicate ballooning pyroptotic cells. PI⁺ cells are red. Results representative of three independent experiments. Scale bars, 40 μm. g, Viability of EA.hy926 cells determined by PI uptake (left) or CellTiter-Glo assay (right). Results representative of three independent experiments. Bars are the mean of three replicates (circles). NT, no treatment. h, IL-1β in the culture supernatant of Ker-CT cells. Bars are the mean of three replicates (circles) except for VBP (two replicates). Results representative of three independent experiments. i, Confluency of EA.hy926 cells. Cells were trypsinized on day 5 and replated in plain medium. Data are mean ± s.d. of four replicates. Results representative of two independent experiments.

Source data

Extended Data Fig. 4. Characterization of inflammasome signaling in THP-1, EA.hy926, and Ker-CT cells.

a, Percentage of THP-1 cells exhibiting nuclear PI staining. Circles, replicates (n = 3). b-d and f, Western blots of THP-1 (b, c) or Ker-CT (d, f). Cells in (b) were treated as indicated for 6 h. Results representative of 2 independent experiments. e-f, IL-1β released from Ker-CT cells into the culture supernatant. Bars are the mean of 3 replicates (circles). Results representative of 2 independent experiments. g and h, Percentage of viable Ker-CT (g) and EA.hy926 (h) cells as measured by CellTiter-Glo assay. Bars indicate the mean of 3 replicates (circles). i, Percentage of EA.hy926 cells exhibiting nuclear PI staining. Data are the mean ± SD of 3 replicates. j, Western blots of EA.hy926 and THP-1 cells. k, EA.hy926 cells treated with 20 µM VBP in the presence of 1 µg/ml PI. Scale bar, 40 µm. Black arrows, ballooning pyroptotic cells. Most GSDMD KO cells remained viable out to 22 h (top row), but some exhibited apoptotic blebbing with subsequent membrane rupture and PI uptake (bottom row). Results representative of 3 independent experiments. l, Western blots of EA.hy926 cells. m, Percentage of EA.hy926 cells with nuclear PI staining after treatment with 20 µM VBP in the presence of 1 µg/ml PI. Data are the mean ± SD of 3 replicates. n, Percentage of LDH released from EA.hy926 cells at 26 h after treatment with 20 µM VBP. Circles, results from 2 independent experiments.

Source data

As expected²², VBP was not a potent death stimulus for Ker-CT cells, killing less than 25% of cells after 24 h (Extended Data Fig. 4g). Human endothelial-like EA.hy926 cells were more susceptible to VBP, with cell death being driven by a CARD8–caspase-1 inflammasome (Fig. 1d,e and Extended Data Fig. 4h–j). Although EA.hy926 cells expressed NLRP1, they lacked detectable ASC (Extended Data Fig. 4j). We noted that neither KGR-3 nor GSDMD loss were as effective as caspase-1 deficiency at blocking VBP-induced EA.hy926 cell death (Fig. 1d,e), perhaps reflecting the ability of caspase-1 to also trigger apoptosis^23,24. We examined the morphology of VBP-treated EA.hy926 cells and observed caspase-1- and GSDMD-dependent swelling, cell lysis (pyroptosis) and nuclear PI staining (Fig. 1f and Extended Data Fig. 4k). However, dying GSDMD-deficient cells exhibited apoptotic blebbing (Fig. 1f) and caspase-3 cleavage (Extended Data Fig. 4l). Blebbing was followed by secondary membrane damage as measured by PI uptake (Extended Data Fig. 4k). Thus, caspase-1 can trigger GSDMD-dependent pyroptosis or GSDMD-independent apoptosis in EA.hy926 cells.

KGR-3 suppressed VBP-induced pyroptosis in EA.hy926 cells (Fig. 1d–f), exhibiting a half-maximal inhibitory concentration (IC₅₀) of approximately 2.5 μM when monitoring PI uptake (Extended Data Fig. 4m) or LDH release (Extended Data Fig. 4n). Notably, caspase-1-driven apoptotic blebbing and caspase-3 cleavage were resistant to KGR-3 (Fig. 1f and Extended Data Fig. 4l). The ability of KGR-3 to prevent caspase-1-dependent pyroptosis, but not caspase-1-dependent apoptosis, strongly suggests that KGR-3, similar to other tetrapeptide caspase inhibitors with poor membrane permeability²⁵, requires gasdermin pores to access caspase-1 in the cytoplasm.

Available evidence suggests that pyroptosis occurs when the assembly of GSDMD pores outpaces their elimination from membranes by the ESCRT (endosomal sorting complexes required for transport) machinery^4,26. Inhibition of caspase-1 by KGR-3 must favour membrane repair such that cell death is averted. By contrast, apoptotic cells that do not form GSDMD pores (for reasons that are unclear in wild-type cells) exclude KGR-3 and are not protected. Remarkably, addition of KGR-3 to EA.hy926 and Ker-CT cells several hours after VBP treatment arrested further cell lysis (Fig. 1g) and IL-1β release (Fig. 1h), respectively. Moreover, cell death was prevented rather than just delayed by KGR-3 because EA.hy926 cells continued to grow over the course of 12 days, whereas few cells emerged from the control cultures exposed to VBP alone (Fig. 1i and Extended Data Fig. 5a). Thus, the asynchronous nature of lethal GSDMD pore formation and pyroptosis provides a window for intervention after the initial death stimulus.

Extended Data Fig. 5. Disabling the ESCRT machinery neutralizes the protective effect of KGR-3 on caspase-4-driven pyroptosis.

a, WT EA.hy926 cells cultured in 1 µg/ml PI. Scale bar, 200 µm. b, Percentage of viable EA.hy926 cells by CellTiter-Glo assay. Bars are the mean of 3 replicates (circles). c, EA.hy926 cells expressing FKBP-caspase-4 and treated with 2 nM B/B ± 10 µM KGR-3 in the presence of 1 µg/ml PI (red). Scale bar, 40 µm. Results representative of 3 independent experiments. d-e, Graphs indicate the percentage of EA.hy926 cells exhibiting nuclear PI staining in the presence of 1 µg/ml PI. Cells were treated with 0.5 µg/ml doxycycline (Dox) or vehicle for 16 h followed by 2 nM B/B and/or 10 µM KGR-3 for 6 h. Bars are the mean of 3 replicates (circles). Western blots show EA.hy926 cells ± 0.5 µg/ml Dox for 16 h. FL, full length. Results representative of 2 independent experiments.

Source data

We also examined the effect of KGR-3 on the death of EA.hy926 cells expressing FKBP–caspase-4. KGR-3 suppressed GSDMD-dependent cell death, swelling and lysis in the B/B-treated cells (Extended Data Fig. 5b,c). Cell death induced by FKBP–caspase-4 was more rapid than that induced by VBP, with most of the cells dying within 6 h (Extended Data Fig. 5c). Consistent with the ESCRT machinery repairing GSDMD-mediated membrane damage in cells treated with KGR-3, dominant-negative versions of ESCRT proteins VPS4A (VPS4A(E228Q))²⁷ and CHMP3 (CHMP3-1–179)²⁸ overcame the protective effects of KGR-3 (Extended Data Fig. 5d,e).

Consistent with KGR-3 having low permeability in apoptotic cells, it did not affect TRAIL-induced caspase-8-dependent extrinsic apoptosis (Extended Data Fig. 6a,b) or intrinsic apoptosis induced by the MCL-1 inhibitor S63845 (Extended Data Fig. 6c–e). Necrosis induced by the pore-forming protein listeriolysin O was also unaffected by KGR-3 (Extended Data Fig. 6f).

Extended Data Fig. 6. KGR-3 does not suppress extrinsic apoptosis, intrinsic apoptosis, or necrosis.

a, Percentage of viable EA.hy926 cells by CellTiter-Glo assay. Bars are the mean of 3 replicates (circles). b, EA.hy926 cells treated with 100 ng/ml TRAIL ± 10 µM KGR-3 in the presence of 1 µg/ml PI. Scale bar, 40 µm. Results representative of 3 independent experiments. c, Percentage of viable EA.hy926 cells by CellTiter-Glo assay. Bars are the mean of 3 replicates (circles). d, EA.hy926 cells treated with 10 µM S63845 ± 10 µM KGR-3 in the presence of 1 µg/ml PI. Scale bar, 40 µm. Results representative of 2 independent experiments. e, Western blots of EA.hy926 cells. Results representative of 2 independent experiments. f, Percentage of EA.hy926 cells with nuclear PI staining after treatment with listerolysin O (LLO) in the presence of 1 µg/ml PI. Bars are the mean of 3 replicates (circles).

Source data

Small-molecule uptake via GSDMD pores

To support the notion that inhibitors such as KGR-3 enter living cells through GSDMD pores, we determined whether small amounts of membrane-impermeable, fluorescent small molecules such as PI and YOYO-1 enter the cytoplasm of live cells in a GSDMD-dependent manner. When cultured in 10 μg ml⁻¹ PI (which is 10 times more PI than routinely used to monitor plasma membrane integrity), EA.hy926 cells treated with VBP and KGR-3 had detectable PI in the cytoplasm (Fig. 2a,b). Importantly, VBP-induced PI uptake in the presence of KGR-3 required both caspase-1 and GSDMD. Similar results were obtained with YOYO-1 (Fig. 2c,d). These data suggest that the initial GSDMD pores compromise the integrity of the cell membrane towards small molecules such as PI and YOYO-1, but inhibition of caspase-1 prevents the irrevocable commitment to pyroptosis. Attempts to image KGR-3 uptake directly were unsuccessful. Tagging of KGR-3 with different fluorophores reduced its potency, enhanced its cell permeability, or caused cellular toxicity. Visualization of caspase inhibitors passing through endogenous gasdermin pores is a future goal.

Fig. 2. Enhanced membrane permeability in live cells forming GSDMD pores.

a, EA.hy926 cells cultured in 10 μg ml⁻¹ PI with 20 μM VBP and/or 10 μM KGR-3 for 6 h. Results representative of three independent experiments. Scale bar, 40 μm. b, Graph indicates the PI fluorescence intensity in a after normalizing to the fluorescence at 0 h. Data are the mean ± s.d. of three replicates. c, EA.hy926 cells cultured in 20 μM VBP and/or 10 μM KGR-3 for 20 h. YOYO-1 dye (10 μg ml⁻¹) was added just before imaging. Results representative of three independent experiments. Scale bar, 40 μm. d, YOYO-1 fluorescence in the cells in c relative to vehicle-treated EA.hy926 cells. Circles, individual cells (KGR-3 alone, WT n = 32, CASP1 KO n = 26, GSDMD KO n = 23; VBP + KGR-3, WT n = 41, CASP1 KO n = 31, GSDMD KO n = 31). Lines indicate the mean.

Source data

Characterization of inhibitor KGR-53P

Caspase-11-dependent pyroptosis is an important mediator of LPS-induced endotoxic shock in mice^8,29; we therefore characterized KGR-53, a more potent inhibitor of caspase-11 than KGR-3 (Extended Data Table 2). KGR-53 suppressed but did not prevent caspase-11-dependent pyroptosis in LPS-transfected mouse bone marrow-derived macrophages (BMDMs) (Extended Data Fig. 7a). Given that GSDMD pores may transport negatively charged compounds less efficiently³⁰, we derived KGR-53P (Extended Data Fig. 7b,c), which lacks the negative charge of the P3 Glu. KGR-53P was better than KGR-53 at inhibiting LPS-induced pyroptosis in BMDMs (IC₅₀ approximately 4 μM; Fig. 3a,b), despite having very low cell permeability when assessed in healthy gMDCKI cells (Extended Data Table 3). KGR-53P blocked LPS-induced GSDMD cleavage, as well as caspase-11-dependent secretion of IL-1β and IL-18 (Fig. 3c–e). KGR-53P reduced cell death and IL-18 secretion even when added several hours after LPS transfection (Fig. 3e,f). By contrast, KGR-53P did not protect BMDMs from venetoclax-induced intrinsic apoptosis and thymocytes from Fas ligand (FasL)-induced extrinsic apoptosis (Extended Data Fig. 7d–h).

Extended Data Fig. 7. Characterization of KGR-53 and KGR-53P.

a, Percentage of PI⁺ BMDMs after transfection with 5 µg/mL LPS ± 30 µM KGR-53 (mean ± SD of cells from 3 different mice). b, Synthetic scheme for KGR-53P. c, LC-MS analysis of KGR-53P. Upper panel, LC chromatogram recorded at 220 nm. Early peaks around 4 min correspond to DMSO/solvent, whereas the peak at ~15 min corresponds to KGR-53P. Lower panel, MS analysis of the KGR-53P peak. d, Percentage of PI⁺ BMDMs after treatment with 25 µM Venetoclax ± 10 µM KGR-53P (mean ± SD of cells from 3 different mice). e, Percentage of viable BMDMs by CellTiter-Glo assay after treatment with 25 µM Venetoclax ± 10 µM KGR-53P. Circles, cells from different mice (n = 3). Bars indicate the mean. f, Western blots of BMDMs. Results representative of 2 independent experiments. g, Percentage of PI-negative thymocytes after treatment with 100 ng/mL FasL ± 10 µM KGR-53P for 20 h. Circles, cells from different mice (n = 3). Bars indicate the mean. h, Western blots of thymocytes treated with 100 ng/mL FasL ± 10 µM KGR-53P or emricasan for 2 h. i, Concentrations of KGR-53P detected in mouse whole blood, lung, or liver after a single intraperitoneal dose of 200 mg KGR-53P per kg body weight. Circles, individual mice (n = 3). The black line connects the mean blood concentrations. j, Liver enzyme levels (AST, Aspartate Aminotransferase; ALT, Alanine Aminotransferase) in the serum of mice in Fig. 4. Circles, individual mice (n = 6). Lines indicate the mean. P-values determined by one way ANOVA with Dunnett’s T3 multiple comparisons test.

Source data

Fig. 3. KGR-53P inhibits caspase-11-dependent pyroptosis.

a, Percentage of PI⁺ BMDMs (mean ± s.d. of cells from three different mice) after transfection with 5 μg ml⁻¹ LPS and then culture in KGR-53P. Results representative of two independent experiments. b, Percentage of viable BMDMs by CellTiter-Glo assay. Bars indicate the mean of cells from different mice (circles, n = 3). c, Western blots of BMDMs transfected with 5 μg ml⁻¹ LPS and then cultured in 10 μM KGR-53P for 16 h. Results representative of two independent experiments. d,e, IL-1β (d) and IL-18 (e) secreted from BMDMs after transfection with 5 μg ml⁻¹ LPS and culture in 10 μM KGR-53P for 18 h. Bars indicate the mean of replicate wells (circles, n = 3). Results representative of two independent experiments. f, Percentage of viable BMDMs in e by CellTiter-Glo assay.

Source data

Next, we determined whether KGR-53P suppressed caspase-1, caspase-11 and GSDMD-dependent IL-18 and IL-1β production in C57BL/6N mice dosed with LPS from Escherichia coli O111:B4 (refs. ^8,17,29). Pharmacokinetic analyses of mice dosed intraperitoneally with KGR-53P suggested that near-micromolar concentrations could be achieved in blood, but repeat dosing was warranted owing to rapid compound clearance (Extended Data Fig. 7i). By contrast, the poor pharmacokinetic properties of KGR-3 precluded its use in vivo. As expected, mice receiving 20 mg kg⁻¹ LPS had significantly increased IL-18 and IL-1β in their serum after 12 h compared with mice given vehicle alone (Fig. 4). However, markedly less IL-18 and IL-1β was observed when mice were treated with KGR-53P at 30 min before and 6 h after LPS dosing. CCL5, which exits cells via the classical secretory pathway, was suppressed to a lesser extent in mice receiving KGR-53P (Fig. 4). We noted that KGR-53P increased liver enzyme levels in the serum even in the absence of LPS (Extended Data Fig. 7j), but whether this is an off-target effect of the tool compound is unclear. Further optimization of KGR-3 and KGR-53P is needed to facilitate a broader assessment of caspase-1 and/or caspase-11 inhibition in pre-clinical disease models. Regardless, the data indicate that an inhibitor of caspase-1/11 that is impermeable in healthy cells can nonetheless limit the GSDMD-dependent secretion of proinflammatory IL-1β and IL-18 in vivo, presumably accessing its intracellular target caspases through GSDMD pores. This finding indicates that caspase inhibitors with poor cell permeability profiles may yet have therapeutic potential in the treatment of diseases characterized by excessive inflammasome activation, including sepsis, acute respiratory distress syndrome and COVID-19 (refs. ^31–33).

Fig. 4. KGR-53P reduces serum IL-1β and IL-18 in mice treated with LPS.

Serum cytokines in female C57BL/6N mice at 12 h after LPS dosing. Lines indicate the mean. Circles represent individual mice (n = 6). P values determined by one way ANOVA with Dunnett’s T3 multiple comparisons test.

Source data

Methods

Reagents

All chemicals were from commercial sources and used without further purification. Fluorenylmethyloxycarbonyl (Fmoc)- and tert-butyloxycarbonyl (Boc)-protected amino acids were from Iris Biotech GmbH, Sigma Aldrich, Bachem, Creosalus and PE Biosciences Limited. Rink amide AM resin (200-300 mesh, loading 0.74 mmol g⁻¹), 2-chlorotrityl chloride resin (100–200 mesh, loading 1.59 mmol g⁻¹), biotin, HBTU (hexafluorophosphate benzotriazole tetramethyl uronium), HATU (hexafluorophosphate azabenzotriazole tetramethyl uronium), piperidine, diisopropylcarbodiimide, 2,2,2-trifluoroethanol and trifluoroacetic acid (TFA) were from Iris Biotech GmbH. Anhydrous HOBt (1-hydroxybenzotriazole) was from Creosalus. 2,4,6-Collidine (2,4,6-trimethylpyridine), acetonitrile (ACN, high-performance liquid chromatography (HPLC) gradient grade), triisopropylsilane (TIPS), hydrobromic acid solution (Hbr, 30% wt in acetic acid), N-methylmorpholine, tetrahydrofuran (anhydrous), isobutylchloroformate and 2,6-dimethylbenzoic acid (2,6-DMBA) were from Sigma Aldrich. N,N′-dimethylformamide (DMF, pure for analysis), methanol, dichloromethane (DCM), acetic acid, diethyl ether and phosphorus pentoxide (P₂O₅) were from POCh. Diazomethane used for the synthesis of AOMK inhibitors was generated according to the Aldrich Technical Bulletin (AL-180) protocol. All compounds (peptides, 7-amino-4-carbamoylmethylcoumarin (ACC) fluorescent substrates, aggregation-induced emission fluorescent substrates and inhibitors) were purified by reverse-phase HPLC on a Waters system (Waters M600 solvent delivery module and Waters M2489 detector system) using a semi-preparative Discovery C8 column (particle size 10 μm). Compound purity was confirmed by HPLC using an analytical Discovery C8 column (particle size 10 μm). Phase A was water/0.1% TFA. Phase B was ACN/0.1% TFA. For purification and compound analysis, the assay was run for 30 min in a linear gradient (from 5% phase B to 100% phase B). The molecular weight of each synthesized compound was confirmed on a WATERS LCT Premier XE High Resolution Mass Spectrometer with electrospray ionization (ESI) and a time of flight (TOF) module.

Synthesis of caspase substrates

Substrates were synthesized as described⁵. In brief, Fmoc-P2-OH (2.5 eq.) was preactivated with HATU (2.5 eq.) and 2,4,6-collidine (2.5 eq.) in DMF, then added to a cartridge containing H₂N-Asp(O-tBu)-ACC-resin. The mixture was gently agitated for 3 h, filtered and washed with DMF. The Fmoc-protecting group was subsequently removed using 20% piperidine in DMF, with a ninhydrin test performed after each coupling and deprotection step. A solution of Fmoc-P3-OH (2.5 eq.), HATU (2.5 eq.) and 2,4,6-collidine (2.5 eq.) in DMF was added to the resin, followed by agitation for 3 h. Next, the Fmoc group was deprotected, and Fmoc-P4-OH amino acid was conjugated in a similar manner. After Fmoc deprotection from the P4 position, the N terminus was protected with an acetyl group using acetic acid (5 eq.), HBTU (5 eq.) and DIPEA (5 eq.) in DMF. Following solvent removal, the resin was washed with DMF, DCM and methanol, then dried over P₂O₅ and cleaved from the resin with a mixture of TFA/TIPS/H₂O. The crude product was purified by HPLC and lyophilized. Purity and molecular weight were confirmed with liquid chromatography–mass spectrometry (LC–MS)⁶. Substrates were dissolved in peptide-grade DMSO to a concentration of 20 mM and stored at −80 °C.

Synthesis and characterization of caspase inhibitors

Ac-P4-P3-P2-P1-AOMK inhibitors were synthesized as described^34,35. Boc-l-Asp(tBu)-OH was converted to Boc-Asp(tBu)-CH₂N₂ using a solution of diazomethane in diethyl ether. The resulting diazoketone was subsequently transformed into Boc-Asp(tBu)-CH₂Br by treatment with HBr solution (30% wt. in acetic acid) and water. The crude bromomethyl ketone product, obtained as a pale-yellow oil, was then reacted with 2,6-dimethylbenzoic acid (2,6-DMBA) in the presence of KF in DMF to afford Boc-Asp(tBu)-AOMK. After removal of the Boc protecting group using 50% TFA in DCM, the resulting Asp-AOMK intermediate was used in subsequent syntheses without further purification. In parallel, selected Ac-P4-P3-P2-OH peptide fragments bearing appropriate side-chain protecting groups were synthesized on 2-chlorotrityl chloride resin and used without further purification; for example, Ac-Tle-hGlu(tBu)-hCha-OH. The peptide fragments were then coupled with Asp-AOMK in DMF using HATU and DIPEA as coupling reagents. The crude products were purified by HPLC using a Waters system equipped with an M600 solvent delivery module and an M2489 UV/Vis detector, fitted with a semi-preparative Discovery C8 column, 10 µm particle size. Purification was performed using a water-acetonitrile gradient (0 to 100% in 25 min) at a flow rate of 15 ml min⁻¹, followed by lyophilization. KGR-53P, Ac-Tle-hGlu(tBu)-hCha-Asp-AOMK, was used without further deprotection. The remaining inhibitors were deprotected using 50% TFA in DCM, after which TFA and DCM were removed under reduced pressure. The resulting products were subjected to a second round of HPLC purification, lyophilized and dissolved in DMSO to obtain stock solutions at a final concentration of 10 mM. Each inhibitor was analysed by LC–MS using a Waters system equipped with an ACQUITY QDa detector, Arc HPLC module and 2489 UV/Vis detector. Analyses were performed on an analytical C8 column, 10 µm particle size, using a water-acetonitrile gradient (0 to 100% in 20 min) at a flow rate of 1 ml min⁻¹. The purity of each compound was above 95%, as determined by UV absorbance at 220 nm. The molecular mass of each inhibitor was confirmed by mass spectrometry using either positive or negative ionization mode on the QDa detector (Supplementary Table 2 and Supplementary Fig. 1).

Recombinant caspases and kinetic assays

Caspases were expressed, purified and active site titrated using z-VAD-FMK (Promega) as described³⁶. Kinetic assays were conducted at 37 °C using an fMax fluorimeter (Molecular Devices) operating in kinetic mode. Excitation for ACC-based substrates was at 355 nm, while emission was detected at 460 nm with a cutoff set at 455 nm. The assay buffer was 1 M sodium citrate, 10% w/v sucrose, 20 mM Pipes, 10 mM NaCl, 1 mM EDTA, and 10 mM DTT, pH 7.2–7.4, except for caspase-3 and caspase-7, when 1 M sodium citrate was omitted. In Extended Data Fig. 1, caspases were incubated at 37 °C for 15 min prior to mixing with 100 μM substrate in 100 μl. Reactions contained 3 nM caspase-1, 100 nM caspase-4 or 120 nM caspase-5, and were monitored for 30 min. Alternatively, E. coli expressing one of the caspases were resuspended in 50 mM Tris pH 7.4, 100 mM NaCl and sonicated for 2 min at full power with 50% duty cycle (on for 0.5 s, then off for 0.5 s, Branson Sonifier 250). Lysates were centrifuged for 30 min at 18,000g. Supernatants were filtered (0.45 μm) and diluted twofold with 2× buffer (described above) prior to measurements. The total protein concentration in the lysates varied between 150–300 mg ml⁻¹, and the final concentration of the lysate was adjusted during caspase activity measurements with an appropriate substrate (for example, Ac-DEVD-ACC for caspase-3, Ac-WEHD-ACC for caspase-1, -4 and -5). HyCoSuLs were screened with 100 μM substrate. Velocity calculations were performed using the linear phase of the progress curve, typically observed between 15–30 min. Experiments were performed in triplicate, and reported results are mean values, with variation between individual measurements consistently <10%. Data analysis involved normalizing total relative fluorescence units (RFUs) for each sub-library, with the highest value set to 100% and adjustments made accordingly.

In Extended Data Table 1, reactions contained 2–15 nM caspase-1, 3–35 nM caspase-3, 35–200 nM caspase-4, 20–250 nM caspase-5, 40–400 nM caspase-7, 6–42 nM caspase-8, or 20–150 nM caspase-9, and were performed using 8 different substrate concentrations between 0.05 and 500 μM. DMSO in the reactions was kept below 2% (v/v). Each reaction was done in triplicate. k_cat/K_m calculations were performed using GraphPad Prism v8.0.2.

In Supplementary Table 1, the k_obs/I (observed pseudo-first-order rate constant/inhibitor concentration) parameter was determined using the equation: Y = (V₀ − V_S) × (1 − exp(−k_obs[app]))/k_obs[app] + V_S × X, where Y denotes enzyme activity, V₀ stands for initial enzyme activity, [app] is the apparent inhibitor concentration and V_S represents enzyme activity at equilibrium. Initially, we screened all inhibitors in triplicate at concentrations of 10 nM, 100 nM, 1 μM and 10 μM against each caspase. Subsequently, we calculated the k_obs[app]/I values from the plots that exhibited the best fitting to the equation curve. The k_obs[app]/I values for the most effective inhibitors were derived from the analyses conducted at 10 nM and 100 nM inhibitor, while those for less effective inhibitors were determined using 1 μM and 10 μM inhibitor. We calculated k_obs/I using the formula k_obs/I = k_obs[app]/I × (1 + K_m/[S]), where K_m represents the Michaelis–Menten constant for the substrate used, and [S] denotes the substrate concentration. Each caspase was incubated with an appropriate ACC-labelled substrate to monitor the reaction: caspase-1, Ac-WEHD-ACC, [S] = 50 μM, K_m = 11 μM; caspase-3, Ac-DEVD-ACC, [S] = 50 μM, K_m = 20 μM; caspase-4, Ac-WEHD-ACC, [S] = 100 μM, K_m = 43 μM; caspase-6, Ac-VEID-ACC, [S] = 100 μM, K_m = 42 μM; caspase-7, Ac-DEVD-ACC, [S] = 100 μM, K_m = 48 μM; caspase-8, Ac-LEHD-ACC, [S] = 50 μM, K_m = 8 μM; caspase-9, Ac-LEHD-ACC, [S] = 200 μM, K_m = 102 μM; caspase-10, Ac-LEHD-ACC, [S] = 50 μM, K_m = 27 μM.

In Extended Data Table 2, inhibitors were examined with greater rigour against both human and mouse caspases. Mouse caspase-1, Ac-WEHD-ACC, [S] = 50 μM, K_m = 12.5 μM; mouse caspase-1, Ac-DEVD-ACC, [S] = 50 μM, K_m = 25 μM; mouse caspase-8, Ac-LEHD-ACC, [S] = 50 μM, K_m = 6.5 μM; mouse caspase-11, Ac-WEHD-ACC, [S] = 200 μM, K_m = 132 μM. Reactions contained 1.9 nM human caspase-1, 2.3 nM human caspase-3, 43 nM human caspase-4, 4.1 nM human caspase-8, 14 nM human caspase-9, 2.2 nM mouse caspase-1, 1.8 nM mouse caspase-3, 3.2 nM mouse caspase-8, or 38 nM mouse caspase-11, ensuring that the lowest inhibitor concentration was at least fourfold higher than the caspase concentration. Inhibitors were diluted in 384-well plates, mixed with substrate and preincubated at 37 °C. Caspases were preincubated separately at 37 °C for 15 min, and then added to the mix of inhibitor and substrate. Fluorescence was monitored for 20 min. k_obs values were calculated as before, and k_obs/I was determined by plotting k_obs values against [I] and applying linear regression. Parameters were derived from at least three independent experiments and presented as average values.

Cells

THP-1 cells (ATCC; mycoplasma tested but not authenticated) were maintained in RPMI-1640 medium supplemented with 2 mM l-glutamine, 10% fetal bovine serum (FBS; Omega Scientific), 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin (Gibco). Where indicated, THP-1 cells were differentiated in 100 ng ml⁻¹ PMA (Sigma Aldrich) overnight. EA.hy926 cells (ATCC; mycoplasma tested, but not authenticated) were grown in the high-glucose version of Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10 mM HEPES pH 7.4, 1× GlutaMAX (Gibco), 100 U ml⁻¹ penicillin, and 100 μg ml⁻¹ streptomycin, 1× non-essential amino acids (Gibco), 1 mM sodium pyruvate (Gibco) and 10% FBS. Ker-CT cells (ATCC CRL-4048; mycoplasma tested, but not authenticated) were cultured in equal parts Ham’s F12 (ThermoFisher) and DMEM (high-glucose, no glutamine, no calcium; ThermoFisher) supplemented with 1× non-essential amino acids, 10 mM HEPES, 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin, 0.5% EpiLife defined growth supplement (ThermoFisher), 25 µg ml⁻¹ bovine pituitary extract (ThermoFisher), 20 ng ml⁻¹ epidermal growth factor (Peprotech) and 0.1 mM CaCl₂ (PromoCell).

Cell line generation

To generate GSDMD, CASP1 and ASC KO THP-1 cells, THP-1 cells stably expressing Cas9 were electroporated with single guide RNAs (sgRNAs) (IDT) using the Neon electroporation system (1,400 V, 10 ms, 3 pulses; ThermoFisher). The following sites were targeted: GSDMD (CACAAGCGTTCCACGAGCGA and ACGCGCACCCACAAGCGGGA), CASP1 (CGGCTTGACTTGTCCATTAT and GACCTCTGACAGCACGTTCC), ASC (GCTGGAGAACCTGACCGCCG and ATGGACAGCAGCGCGCCCCG). CARD8-knockout EA.hy926 cells were generated by electroporating Cas9-expressing cells with sgRNAs targeting the sequences AAAATGTGAGATGTCACAA and GGTATCTGTGGTCAGCCAC. CASP1 or GSDMD-knockout EA.hy926 cells were generated from Cas9⁺ cells using the lentiviral vector pLKO.1 (Sigma) for guide RNA delivery (the target sequence in human CASP1 was CCTCAAACTCTTCTGTAGT). Guide RNA-expressing cells were selected with 2 μg ml⁻¹ puromycin (Gibco). NLRP1, ASC, CASP1, GSDMD and GSDME-knockout Ker-CT cells were generated by electroporation of sgRNAs (IDT) and recombinant Cas9 (IDT) using program DS-138 of the P3 Primary Cell 4D-Nucleofector kit (Lonza). The following sites were targeted: NLRP1 (ACTCCAAGTAACAGGCCAGG and ATTCCTGACGTTTCATCCAG), ASC (GCTGGAGAACCTGACCGCCG and ATGGACAGCAGCGCGCCCCG), CASP1 (TAATGAGAGCAAGACGTGTG and CAATGCTGCTACAAAATCTG), GSDMD (CACAAGCGTTCCACGAGCGA and ACGCGCACCCACAAGCGGGA), GSDME (GTCGGACTTTGTGAAATACG and GTATAACTCAATGACACCGT).

To express FKBP–caspase-4 and FKBP–caspase-11 fusions, the dimerizer domain (DmrB) of a modified FK506 binding protein (FKBP) was added as an N-terminal fusion to the catalytic domains of caspase-4 (amino acids 88–377) and caspase-11 (amino acids 88–373). cDNAs encoding the fusion proteins were generated in a lentiviral vector for EA.hy926 cells or piggyBac vector BH1.11 (Genentech) for THP-1 cells. After lentiviral transduction, EA.hy926 cells were selected with 2 μg ml⁻¹ puromycin (Gibco). THP-1 cells expressing Cas9 were co-electroporated with FKBP–caspase-4/BH1.11 or FKBP–caspase-11/BH1.11 and the transposase vector pBO (Transposagen Biopharmaceuticals) using the Neon system. THP-1 cells were selected with 10 μg ml⁻¹ blasticidin S HCl (Gibco). B/B homodimerizer (Takara Bio) was used at 50 nM for THP-1 cells and 2 nM for EA.hy926 FKBP–caspase-4 cells.

VPS4A(E228Q) and CHMP3-(31–179) were cloned into a doxycycline inducible lentiviral vector conferring blasticidin resistance. Transduced FKBP–caspase-4 or FKBP–caspase-4; GSDMD-knockout EA.hy926 cells were selected with 25 µg ml⁻¹ blasticidin.

Bone marrow cells from C57BL/6N mice were differentiated into BMDMs in DMEM supplemented with 10% FBS, 1× GlutaMAX, 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin and 30% L929-conditioned medium for 5–6 days. Thymocytes from C57BL/6N mice were cultured in the high-glucose version of DMEM supplemented with 10 mM HEPES pH 7.4, 1× GlutaMAX, 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin, 1× non-essential amino acids, 50 μM 2-mercaptoethanol and 10% FBS.

In vitro permeability assays

Permeability assays in gMDCKI cells (mycoplasma tested but not authenticated) were performed as described¹². In brief, cells were seeded at 7 × 10⁵ cells per ml in 96-well transwells in 100 μl DMEM containing phenol red, 10% BSA, 1% pen/strep, 1× GlutaMAX and 5 μg ml⁻¹ plasmocin, grown until confluent (around 4 days), and then placed in Hank’s Balanced Salt Solution containing 25 mM sterile filtered HEPES buffer, pH 7.4 (Invitrogen). 25 μM caspase inhibitor was added to the donor compartment together with the monolayer integrity marker Lucifer Yellow (Sigma Aldrich). Inhibitor concentrations in the donor and receiver compartments after 3 h at 37 °C were determined by LC–MS/MS. The apparent permeability (P_app) of test compounds was determined using the equation P_app = (dQ/dt) × (1/AC₀), where dQ/dt is the rate of compound appearance in the receiver compartment, Q is the quantity of compound, C₀ is the concentration in the donor compartment and A is the surface area of the insert. Efflux ratio was calculated as (P_app, B-to-A)/(P_app, A-to-B), where A is apical and B is basolateral.

Cell death and cytokine assays

For screening of tetrapeptide-AOMK compounds, THP-1 cells were seeded in Opti-MEM (Gibco) supplemented with 0.1% FBS. For Nigericin stimulation, cells were primed for 4 h with 1 μg ml⁻¹ ultrapure LPS (E. coli O111:B4; InvivoGen) and then stimulated for 2 h with 5 μM nigericin (InvivoGen) with or without 25 μM emricasan (Abcam), 25 μM tetrapeptide-AOMK, or 25 μM MCC950 (InvivoGen). THP-1 cells were treated for 24 h with 1 μM staurosporine with or without 10 μM emricasan or tetrapeptide-AOMK. Unprimed FKBP–caspase-4 or FKBP–caspase-11 THP-1 cells were treated for indicated times with 50 nM B/B with or without 10 μM emricasan or tetrapeptide-AOMK. LDH release into the culture medium was measured with a CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega). Results were calculated as a percentage of maximum LDH release by 0.1% Triton X-100. Cell viability was determined using a CellTiter-Glo Luminescent Cell Viability Assay (Promega).

For characterization of KGR-3 in THP-1 cells, experiments were performed in complete media with a seeding density of 50–100,000 cells per well in a 96-well plate. FKBP–caspase-4 THP-1 cells were primed for 5 h in 1 µg ml⁻¹ Pam3CSK4 before stimulation with 50 nM B/B ± the indicated concentration of KGR-3. For ATP treatment, THP-1 cells were differentiated in 100 ng ml⁻¹ PMA overnight (day 1), replated in complete media (day 2), and then stimulated the following day (day 3). Differentiated cells were primed for 5 h in 1 µg ml⁻¹ Pam3CSK4 before stimulation with 5 mM ATP with or without 20 µM KGR-3. IL-1β ELISA measurements were performed on cell culture supernatants using a Human IL-1 beta/IL-1F2 DuoSet ELISA kit (R&D Systems). For percent PI measurements, cells were stained with 1 µg ml⁻¹ PI (ThermoFisher) and imaged with a 10× objective in an IncuCyte S3 (Essen BioScience). Incubation with 1% Triton X-100 was used to record the total number of cells at the final time point. Data are presented as the percentage of total cells that have PI⁺ nuclei.

Ker-CT cells (25,000 per well) were seeded in 96-well plates overnight and then stimulated in complete media containing 2 µM VBP (MilliporeSigma) with or without 20 µM KGR-3 for 24 h, unless indicated otherwise. IL-1β ELISA measurements were performed using a Human IL-1 beta/IL-1F2 DuoSet ELISA kit.

EA.hy926 cells (8,000-10,000 per well) were seeded in 96-well plates overnight and then stimulated in Opti-MEM containing 0.1% FBS with either 20 μM VBP, 10 µM S63845 (Chemietek), 20 ng ml⁻¹ Lysteriolysin O (U.S. Biological) or 100 ng ml⁻¹ TRAIL (Enzo Life Sciences) with or without 10 μM tetrapeptide-AOMK. Viability was measured by CellTiter-Glo assay or by PI uptake (1 µg ml⁻¹; ThermoFisher). For experiments utilizing VPS4A(E228Q) and CHMP3-(31–179), cells were seeded overnight and then treated with 0.5 µg ml⁻¹ doxycycline (Takara). At 16 h after doxycycline addition, cells were treated with 2 nM B/B in the presence or absence of 10 µM KGR-3 in Opti-MEM + 0.1% FBS. Cell viability was determined by measuring percent PI⁺ cells at 6 h after B/B and KGR-3 addition.

BMDMs (100,000 per well) were plated in 96-well plates overnight. For inflammasome stimulations, cells were primed with 1 μg ml⁻¹ Pam3CSK4 (InvivoGen) for 5 h to induce expression of Casp11. Primed cells were transfected with 5 μg ml⁻¹ LPS using Fugene-HD transfection reagent (Promega). Unprimed BMDMs were stimulated with 25 μM venetoclax (R&D Systems). Where indicated, BMDMs were treated with 10 μM tetrapeptide-AOMK. IL-1β in culture supernatants was measured using a 384-well multi-array mouse IL-1β assay (Meso Scale Diagnostics). IL-18 secretion was determined using a mouse IL-18 ELISA kit (MBL International Corporation).

Mouse thymocytes were stimulated with 1 μg ml⁻¹ FasL (Adipogen) with or without 10 μM emricasan or KGR-53P. Cells were stained with 2.5 μg ml⁻¹ PI (BD Biosciences) and analysed in a FACSCanto II cytometer (BD Biosciences) with BD FACSDiva8.0. Viable cells were PI-negative.

Western blots

In Extended Data Fig. 3a, cells were lysed in 10 mM Tris-HCl pH 7.5, 150 mM NaCl, 2.5 mM MgCl₂, 0.5 mM CaCl₂, 1% NP40, phosphatase and protease inhibitors (Roche), and DNase (approximately 80 U ml⁻¹, Qiagen). For all other western blots, cells were lysed in 20 mM Tris-HCl pH 7.5, 135 mM NaCl, 1.5 mM MgCl₂, 10% glycerol, 1 mM EGTA and 1% Triton X-100 supplemented with protease and phosphatase inhibitors (ThermoFisher). The soluble lysate was recovered after centrifugation for 15 min at 20,000 g and 4 °C. Westerns blots used antibodies recognizing human and mouse GSDMD (ab219800, Abcam), ASC (AG-25B-0006-C100, Adipogen), CARD8 (ab24186, Abcam), GSDME/DFNA5 (ab215191, Abcam), caspase-1 (3866, Cell Signaling Technology), caspase-3 (9662, Cell Signaling Technology), cleaved caspase-3 (9664, Cell Signaling Technology), caspase-7 (9492, Cell Signaling Technology), cleaved caspase-7 (9491, Cell Signaling Technology), caspase-8 (4927, Cell Signaling Technology), cleaved caspase-8 (8592, Cell Signaling Technology), caspase-9 (9508, Cell Signaling Technology), caspase-11 (NB-120-10454, Novus), FKBP12 (ab2918, Abcam), HRP-conjugated GAPDH (3683, Cell Signaling Technology), HRP-conjugated β-actin (5125S, Cell Signaling Technology), NLRP1 (679802, Biolegend), α-tubulin (66031-1-Ig, Proteintech), VPS4A (14272-1-AP, Proteintech) and CHMP3 (ab175930, Abcam). HRP-conjugated secondary antibodies were from Jackson Immunoresearch. For gel source data, see Supplementary Fig. 2.

Live-cell imaging

EA.hy926 cells (3×10⁴ per well) were plated in 8-well chamber slides (Ibidi #80806). Brightfield and fluorescence images were acquired every 15 min using a Nikon TI-E spinning disk confocal equipped with a 20× Apo 0.75 NA DIC, WD1 objective (Nikon) and a Photometrics Prime95B sCMOS camera (Teledyne Photometrics). Samples were maintained at 37 °C and 5% CO₂ throughout imaging using an environmental control chamber and gas mixer (Oko Labs). Cells were treated in Opti-MEM containing 0.1% FBS and 10 µg ml⁻¹ PI. Images were processed in the Nikon Elements software or Fiji. For quantification of cytoplasmic PI entry, cells were segmented using the Trackmate plugin on Fiji³⁷. Fluorescence intensity was measured after thresholding and background fluorescence subtraction. Relative intensity change was calculated using the fluorescence intensity ratio F/F₀, where F₀ is the cellular fluorescence at time zero.

For single time-point experiments utilizing YOYO-1, EA.hy926 cells (10,000 per well) were plated on a glass-bottom 96-well plate (Greiner Bio-One) and allowed to adhere overnight. Cells were incubated in 20 µM ValBoro-Pro with or without 10 µM KGR-3 diluted in Opti-MEM + 0.1% FBS for 20 h. Immediately before imaging, 10 µg ml⁻¹ YOYO-1 Iodide (ThermoFisher) was added to each well. Images were acquired using a Leica SP8 laser scanning confocal microscope equipped with a 40× 1.3 HC PL APO CS2 oil immersion objective (Leica). Samples were maintained at 37 °C and 5% CO₂ throughout imaging using an environmental control chamber and gas mixer (Oko Labs). Images were processed in Fiji software. For fluorescence quantification of YOYO-1, cells from brightfield images were first segmented using CellPose³⁸ and then quantified for fluorescence intensity using CellProfiler³⁹. Data are presented as YOYO-1 fluorescence intensity normalized to cells treated with vehicle alone.

Plasma membrane integrity was assessed by culturing cells in 1 μg ml⁻¹ PI in an IncuCyte S3 (Essen BioScience). Incubation with 1% Triton X-100 was used to record the total number of cells at the final time point. Data are presented as the percentage of total cells that have PI⁺ nuclei. For representative images of PI⁺ cells, cells were plated on a glass-bottom 96-well plate (Greiner Bio-One) and imaged with a 20× objective on the IncuCyte S3.

Mice

All mouse studies complied with relevant ethics regulations and were approved by the Genentech Institutional Animal Care and Use Committee (IACUC) in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited facility. No sample size calculations were performed. Mice were randomly assigned to experimental groups. Studies were performed unblinded. In Fig. 4 and Extended Data Fig. 7j, female C57BL/6N mice aged 10 weeks (Charles River Laboratories) were dosed intraperitoneally with 125 mg kg⁻¹ KGR-53P in vehicle (20 mM phosphate pH 7 with 3.75% Solutol HS 15) or vehicle alone at 0 h, 20 mg kg⁻¹ LPS (E. coli O111:B4, Sigma Aldrich) in PBS or PBS alone at 30 min, and 63 mg kg⁻¹ KGR-53P or vehicle alone at 6 h. Serum IL-18 after 12 h was determined by mouse IL-18 ELISA (7625, MBL International Corporation). Serum IL-1β was determined by the V-PLEX Plus Mouse IL-1β kit (K152QPG, Meso Scale Diagnostics). Serum CCL5 was determined by Luminex assay (Bio-Rad). Serum AST and ALT levels were measured in a serum chemistry analyser (Beckman Coulter AU480). In Extended Data Fig. 7i, female C56BL/6N mice aged 10 weeks (Charles River Laboratories) were dosed intraperitoneally with 200 mg kg⁻¹ KGR-53P in 20 mM phosphate pH 7 with 3.75% Solutol HS 15.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-026-10957-y.

Author contributions

This work was primarily carried out by K.M.G., a participant in the Genentech Postdoctoral Program, and M.E.T. The caspase inhibitors were designed by K.M.G. under the supervision of M.P. K.M.G., J.N. and M.K. synthesized inhibitors as part of the HyCoSuL screening library, a method developed by M.D. K.M.G., J.N. and M.K. performed kinetic analyses under the supervision of M.P. K.M.G., M.E.T., H.B., C.M.R., R.S.J., N.K., K.N., M.P. and V.M.D. contributed to experiment design. K.M.G., M.E.T. and K.E.W. performed cell culture experiments. M.E.T., I.S., K.E.W., B.L. and S.B. generated cell lines. R.S.J. and E.P. acquired and analysed drug metabolism and pharmacokinetic data. E.S.L. and P.K. optimized inhibitor formulations. W.P.L. and J.Z. performed mouse studies. K.N. wrote the paper with input from all co-authors.

Peer review

Peer review information

Nature thanks the anonymous reviewers for their contribution to the peer review of this work.

Funding

K.M.G., M.E.T., K.E.W., I.S., B.L., R.S.J., E.S.L., P.K., W.P.L., J.Z., N.K., K.N. and V.M.D. were funded by Genentech. M.P. was supported by the National Science Centre in Poland (OPUS grant, UMO-2018/29/B/NZ1/02249).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information files. Source data are provided with this paper.

Competing interests

K.M.G., M.E.T., I.S., K.E.W., B.L., S.B., R.S.J., E.P., E.S.L., P.K., W.P.L., J.Z., H.B., C.M.R., N.K., K.N. and V.M.D. were employees of Genentech. The other authors declare no competing interests.

Extended data

is available for this paper at https://doi.org/10.1038/s41586-026-10957-y.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41586-026-10957-y.