New Compound Induces Lytic Cell Death in Tumor Models

Compound DHN induces pyroptosis through caspase-8–mediated GSDME cleavage. Considering the inherent resistance of melanoma cells to apoptosis, lytic cell death may offer a more effective therapeutic approach by activating antitumoral immunity. To identify agents capable of inducing lytic cell death, we screened our in-house compound library using the lactate dehydrogenase (LDH) release assay, a well-established method for quantifying lytic cell death. This proprietary library was collaboratively developed by our research team and partners, and primarily consists of derivatives of Csn-B and THPN, compounds known to induce apoptosis and autophagic cell death, respectively (19, 20). Our results demonstrated that a compound named DHN was the most potent compound for inducing lytic cell death in A375 melanoma cells (Figure 1A). Further morphological assessment after DHN treatment revealed characteristic pyroptotic features, including cell swelling and the formation of large bubbles from the plasma membrane, as indicated by red arrowheads in Figure 1B. This DHN-induced characteristic pyroptotic morphology was closely associated with LDH release and cleavage of the pyroptotic executor GSDME in melanoma A375 cells and other cancer cell lines (Figure 1B and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI188872DS1). Knockdown of GSDME, but not other GSDM proteins, in A375 cells attenuated DHN-induced pyroptosis (Figure 1C and Supplemental Figure 1, B and C), demonstrating the involvement of GSDME-mediated pyroptosis. Notably, no DNA laddering or annexin V+/propidium iodide– cells, which are typical apoptotic markers, were observed upon DHN stimulation (Supplemental Figure 1D). Furthermore, pretreatment with Lip-1, Fer-1 (ferroptosis inhibitors), Necrosulfonamide, Nec-1 (necroptosis inhibitors), or Tetrathiomolybdate (a cuproptosis inhibitor) did not affect DHN-induced LDH release or pyroptotic morphology (Supplemental Figure 1E). Additionally, we investigated whether DHN could induce pyroptosis in nontumor cells and found that DHN exhibited a diminished capacity for pyroptotic induction in nontumor cells, including HK-2 human kidney proximal tubule epithelial cells, AC16 human cardiomyocytes, HEK293T cells, THP-1 human leukemia monocytic cells, HL-1 mouse cardiomyocytes, L929 mouse fibroblasts, and primary mouse BM-derived macrophages and BM-derived DCs, as compared with A375 melanoma cells (Supplemental Figure 1F). Collectively, these findings indicate that DHN predominantly induced GSDME-dependent pyroptosis rather than apoptosis, ferroptosis, necroptosis, or cuproptosis in tumor cells.

Cotreatment with Z-VAD, a pan-caspase inhibitor, abrogated DHN-induced pyroptosis (Figure 1D), suggesting the participation of a caspase protein in GSDME cleavage. In accordance with previous reports, incubation of immunoprecipitated GSDME protein with recombinant caspase-3 resulted in clear cleavage of GSDME in the in vitro assay (6, 7). However, we unexpectedly found that knockdown of caspase-3 in cells had no impact on DHN-induced pyroptosis (Supplemental Figure 1G), excluding the possibility of caspase-3–mediated GSDME cleavage upon stimulation with DHN. The treatment with DHN was found to markedly induce caspase-8 activation while only minimally activating caspase-3 (Supplemental Figure 1H). Cotreatment with Z-IETD, a specific inhibitor of caspase-8, or knockdown of caspase-8 effectively attenuated DHN-induced pyroptosis in A375 cells (Figure 1, E and F, and Supplemental Figure 1B), suggesting the involvement of caspase-8 in cleaving GSDME for pyroptotic induction. Although recombinant caspase-8 exhibited mild cleavage of GSDME under normal in vitro conditions, the appearance of the cleaved band of GSDME migrating at approximately 30–35 kDa was markedly enhanced as the pH value decreased to 6.5 in the in vitro cleavage buffer (Supplemental Figure 1I), implying that caspase-8 is capable of cleaving GSDME directly within an acidic environment.

Given that caspase-8 cleaves its substrate after an Asp residue (21), different Asp residues around the hinge region of GSDME were mutated into Ala. The in vitro results showed that GSDMED270A completely blocked caspase-8–mediated cleavage (Supplemental Figure 1J), suggesting that caspase-8 may target GSDME Asp270 for cleavage, similar to caspase-3 (6, 7). When different GSDME mutants were separately expressed in the GSDME-knockdown A375 cells, only GSDMED270A completely blocked DHN-induced GSDME cleavage (Figure 1G), thereby attenuating pyroptotic induction (Figure 1H). Taken together, these results demonstrated that DHN serves as a compound capable of inducing pyroptosis through caspase-8–mediated cleavage of GSDME.

DHN targets mitochondrial protein CypD to promote the opening of mPTP. We aimed to elucidate the target of DHN by synthesizing a photoactive DHN probe (referred to as DHN-P) that exhibited similar properties to DHN in inducing pyroptosis (Supplemental Figure 2A). We detected the subcellular localization of DHN-P using click chemistry (Figure 2A) and found that DHN-P predominantly colocalized with Tom20, a mitochondrial marker protein, whereas minimal colocalization was observed with CALR (an ER marker), GM130 (a Golgi apparatus marker), or LAMP2 (a lysosomal marker) (Figure 2B), indicating that mitochondria may be the organelle for DHN function. Our previous studies have demonstrated that the induction of pyroptosis is associated with an upregulation of various ROS (2, 7, 8, 22). We thus investigated whether mitochondrial ROS (mito-ROS) are involved in DHN-induced pyroptosis. Treatment with DHN indeed resulted in a significant increase in mito-ROS levels; however, the scavenging of mito-ROS by mito-TEMPO or mitoQ failed to rescue DHN-induced pyroptosis (Supplemental Figure 2, B and C), which excluded the association between mito-ROS and DHN-induced pyroptosis. To elucidate the crucial mitochondrial functions underlying DHN-induced pyroptosis, we employed various inhibitors, and the results indicated that inhibition of the electron transport chain by antimycin A, rotenone, or oligomycin; suppression of the TCA cycle by CPI-613 or dimethyl malonate; and attenuation of fatty acid oxidation by ranolazine had no impact on DHN-induced pyroptosis. Furthermore, modulation of mitochondrial fission through Mdivi-1 treatment or regulation of mitochondrial calcium homeostasis via MCU-i4 did not affect DHN-induced pyroptosis either (Supplemental Figure 2C). However, cyclosporin A (CsA), an inhibitor targeting cyclophilin D (CypD) within the mPTP complex, markedly blocked DHN-induced caspase-8 activation and GSDME-mediated pyroptotic cell death (Figure 2C). Similar results were also obtained in CypD knockdown cells (Figure 2D). Given that DHN-P could effectively pull down CypD but not other components in the mPTP complex, such as adenine nucleotide translocator1 (ANT1) and voltage-dependent anion channel 1 (VDAC1), or proteins in the ER (CALR) or Golgi apparatus (TGN46) (Figure 2E), it is likely that DHN targets CypD for pyroptotic induction.

To further verify that CypD is the direct target of DHN, we conducted surface plasmon resonance experiments and confirmed the direct interaction between DHN and CypD, with a KD of 1.19 ± 0.068 μM (Figure 2F). We additionally performed fluorescence labeling–based differential scanning fluorimetry (FL-DSF), a well-established method for evaluating protein-ligand interactions (23). The FL-DSF assay yielded a KD value of 0.69 ± 0.33 μM (Supplemental Figure 2D). We also performed cellular thermal shift assays to detect drug-target interactions through analyzing melting temperature shifts. Addition of DHN significantly enhanced the thermal stability of CypD (Figure 2G), indicating direct binding of DHN to CypD. Molecular docking indicated the theoretical binding mode of DHN to CypD (PDB: 5CBV), in which the naphthalene ring of DHN formed a distinct cationic π-interaction with R97 of CypD, and Q105 formed a hydrogen bond with the oxygen atom while the hydrophobic carbon chain of DHN lay flat in the pocket (Supplemental Figure 2E). When these 2 critical residues were mutated (CypDR97A/Q105A), DHN could no longer bind to CypD (Figure 2G). As a result, DHN failed to induce GSDME cleavage and pyroptosis in CypDR97A/Q105A-expressing cells (Figure 2H). Therefore, DHN induces pyroptosis through binding to CypD.

Considering the crucial role of CypD in mPTP (24), we propose that DHN may regulate mPTP opening by targeting CypD. Indeed, DHN augmented mPTP opening, which was effectively inhibited by either CypD knockdown or CsA treatment (Figure 2I). The essential role of mPTP opening in DHN-induced pyroptosis is further supported by our observation that knockdown of ANT1, another constituent of the mPTP, also impaired DHN-induced mPTP opening, GSDME cleavage, and subsequent pyroptosis (Supplemental Figure 2, F and G). Together, these findings suggest that DHN binding to CypD facilitates mPTP opening, ultimately leading to pyroptosis.

DHN-induced mtDNA release activates the cytosolic cGAS. It has been reported that several NLRP3 inflammasome activators can induce the oxidation of mtDNA, resulting in the release of 500–650 bp fragments into the cytosol through mPTP- and VDAC-dependent channels (25). Upon DHN stimulation, we also observed a significant increase in the amount of mtDNA in the cytoplasmic fraction that was free from mitochondrial contamination (Figure 3A and Supplemental Figure 3A). This release of mtDNA was effectively suppressed by treatment with CsA or knockdown of ANT1 or CypD (Figure 3A and Supplemental Figure 3A), emphasizing the crucial role played by mPTP in DHN-induced mtDNA release. However, no obvious oxidation of mtDNA was detected upon DHN stimulation (Supplemental Figure 3B), and DNA fragments in the cytosolic samples within the range of 500–700 bp were barely observed via agarose gel electrophoresis (Supplemental Figure 3C). Moreover, while an approximately 600 bp mtDNA fragment in the cytosolic fraction was clearly detected by PCR after DHN treatment, PCR also successfully amplified an approximately 5,000 bp mtDNA fragment (Supplemental Figure 3D), suggesting that DHN may induce the release of mtDNA fragments exceeding 5,000 bp in length. Given that only mtDNA fragments smaller than 700 bp can be released upon mPTP opening (26), it is unlikely that DHN induces direct mtDNA release from mitochondria via the mPTP channel. It has been documented that prolonged mPTP opening leads to mitochondrial rupture (27, 28). Indeed, DHN-induced mitochondrial rupture was clearly observed by transmission electron microscopy. Furthermore, the release of HSP60, a mitochondrial matrix protein, into the cytosol upon DHN stimulation confirmed mitochondrial rupture (Supplemental Figure 3E). The DHN-induced release of HSP60 and mtDNA could be effectively suppressed by CsA treatment or CypD knockdown (Supplemental Figure 3, F and G). Therefore, it can be concluded that DHN may promote prolonged mPTP opening, leading to mitochondrial rupture and subsequent mtDNA release.

Consistent with a previous report that cytoplasmic DNA triggers phase transition of cGAS to activate its activity (29), we did observe the formation of cGAS puncta in response to DHN stimulation (Figure 3B), and this DHN-induced cGAS puncta formation could be abolished by treatment with 1,6-hexanediol (1,6-HD) (Supplemental Figure 3H), a small molecule known for melting phase–separated condensates, indicating the phase transition of cGAS upon DHN stimulation. Furthermore, CsA treatment or knockdown of CypD or ANT1 abrogated the formation of these DHN-induced cGAS puncta (Figure 3, B and C). The ability of DHN to induce cGAS puncta formation was lost when the interaction between CypD and DHN was disrupted by the R97A/Q105A mutation in CypD (Supplemental Figure 3I). These findings demonstrate a direct link between mPTP-mediated mtDNA release and activation of cGAS upon binding of DHN to CypD.

The activation of cGAS is crucial for DHN-induced pyroptosis, as evidenced by the effective attenuation of DHN-induced GSDME cleavage and pyroptosis through silencing cGAS expression or inhibiting cGAS activity using inhibitor G140 (30) (Figure 3, D and E). Combined with the finding that neither knockdown of cGAS nor G140 treatment affected the DHN-induced mtDNA release (Supplemental Figure 3J), these experiments indicate that it is the release of mtDNA that activates cGAS and induces pyroptosis. It is well accepted that STING serves as a downstream effector of cGAS (10). Indeed, knockdown of STING markedly impaired DHN-induced GSDME cleavage and pyroptosis (Figure 3F). However, DHN treatment failed to induce the phosphorylation of TBK1 and IRF3 (Supplemental Figure 3K), the downstream kinases in the classical cGAS/STING pathway (10), or to regulate the transcription levels of classical downstream target genes associated with cGAS/STING signaling (including CXCL10, IFNB, RSAD2, ISG15, and RIG1) (Supplemental Figure 3L). Moreover, inhibition of TBK1 by GSK8612 (31) did not affect DHN-induced pyroptosis (Supplemental Figure 3M). These results appear to show that DHN may activate an alternative pathway within the cGAS/STING axis to induce pyroptosis.

The aggregation of STING in the ER provides the platform for GSDME cleavage. We also discovered that DHN could induce the formation of punctate structures of STING in a manner dependent on cGAS activity (Figure 4A) and mPTP opening (Supplemental Figure 4, A and B). However, these STING puncta were not colocalized with cGAS (Supplemental Figure 4C), suggesting that the STING puncta are distinct from cGAS puncta. Knockdown of STING did not affect DHN-induced formation of cGAS puncta (Supplemental Figure 4D). Notably, these STING puncta were exclusively localized to the ER rather than the Golgi apparatus or mitochondria (Figure 4B and Supplemental Figure 4E), resulting in obvious puncta appearance within the ER in a cGAS- and STING-dependent manner (Supplemental Figure 4F). Transmission electron microscopy revealed condensed membranous structures resembling aggregates within the ER upon DHN stimulation (Figure 4C), which could be abolished by cGAS inhibitor G140 (Supplemental Figure 4G) or knockdown of STING (Supplemental Figure 4H). Additionally, in cells expressing STING-APEX fusion protein, the APEX signal was prominently observed within these tangled ER structures upon DHN stimulation (Figure 4D). Therefore, it is likely that DHN induces the aggregation of STING in the ER through activating cGAS.

It has been reported that phase separation of STING within the ER can prevent excessive activation of classical cGAS/STING signaling (32). However, treatment with 1,6-HD failed to disrupt DHN-induced punctate aggregates of STING (Supplemental Figure 4I), and expression of STINGEE/GG mutants, known to abolish the STING phase separator (32), had no effect on these aggregates either (Supplemental Figure 4J). These results suggest that the DHN-induced STING aggregate in the ER (termed ER-STING aggregate) is distinct from the STING phase separator reported in another study (32) and may play a role in pyroptosis induction.

To investigate the function of these ER-STING aggregates in pyroptotic induction, we employed a detergent-free immunoprecipitation technique using an anti-HA antibody to selectively isolate ER-STING aggregates in HA-STING–expressing A375 cells. The immunoprecipitants were found to contain the ER protein CALR, while Golgi protein GM130 and mitochondrial protein Tom20 were not detected (Figure 4E), confirming the absence of contamination from the Golgi apparatus or mitochondria. Western blotting showed the activated caspase-8, full-length GSDME, and GSDME N-terminal within these ER-STING aggregates upon DHN stimulation (Figure 4E). Confocal microscopy consistently indicated the colocalization of caspase-8 and GSDME with STING puncta in the presence of DHN (Figure 4F). These results suggest the recruitment of caspase-8 and GSDME into the ER-STING aggregates. Since protein aggregates are typically resistant to mild detergents like Triton X-100, we fractionated these ER-STING aggregates into a Triton X-100–insoluble (TI) fraction. DHN stimulation obviously increased STING levels in the TI fraction, in which active caspase-8, full-length GSDME, and cleaved-GSDME were also detected (Figure 4G). The full-length GSDME appeared to be more enriched in the TI fraction compared with the cleaved-GSDME, which is consistent with previous findings that upon cleavage, cleaved-GSDME tends to localize to the plasma membrane for pyroptosis execution (6). Knockdown of CypD, cGAS, or STING or inhibition of cGAS activity using G140 resulted in loss of active caspase-8 and GSDME within the STING aggregates even in the presence of DHN (Figure 4G and Supplemental Figure 4K). Proximity labeling assays also demonstrated that DHN enhanced proximity between STING and caspase-8/GSDME (Figure 4H). Considering the role of STING as a scaffold protein (33), it is proposed that DHN-induced ER-STING aggregate represents a large protein complex, potentially serving as a platform to recruit and process the cleavage of GSDME by caspase-8 for pyroptosis induction.

We further explored the underlying mechanism of caspase-8 activation within the ER-STING aggregates. Fas-associated death domain (FADD), an adaptor protein essential for death receptor–mediated caspase-8 activation (34), was investigated for its role in this process. Our results demonstrated that DHN stimulation substantially enhanced the interaction between STING and FADD (Supplemental Figure 4L). This STING-FADD interaction is essential for the recruitment of caspase-8 by STING in response to DHN stimulation, as knockdown of FADD completely abolished the interaction between STING and caspase-8, even in the presence of DHN (Supplemental Figure 4L). Consequently, no caspase-8 was detected in STING-dependent protein aggregates upon FADD knockdown. Furthermore, FADD knockdown also eliminated DHN-induced caspase-8 activation, GSDME cleavage, and pyroptosis (Supplemental Figure 4M). The death effector domain plays a pivotal role in the mutual interaction between FADD and caspase-8. In FADD-knockdown cells, reexpression of WT FADD but not FADDΔDED restored DHN-induced caspase-8 activation, GSDME cleavage, and pyroptosis (Supplemental Figure 4N). Similarly, in caspase-8 knockdown cells, reexpression of WT caspase-8, instead of caspase-8ΔDED, restored DHN-induced GSDME cleavage and pyroptosis (Supplemental Figure 4O). Collectively, these findings indicate that DHN-induced interaction of STING with FADD promotes the recruitment of caspase-8 into ER-STING aggregates, where caspase-8 is activated to cleave GSDME for pyroptosis execution.

An acid environment promotes the polymerization of STING for the formation of ER-STING aggregates. Since the formation of protein aggregates typically results from protein polymerization (35), we investigated whether DHN-induced ER-STING aggregate is associated with the polymerization of STING. As anticipated, DHN markedly enhanced the formation of STING dimers and polymers, which could be abolished by knockdown of CypD and ANT1 (Figure 5A and Supplemental Figure 5A) or inhibition of cGAS by G140 (Figure 5A), indicating that the opening of mPTP and cGAS activity are essential for DHN-induced STING polymerization. Cys148 and Cys206 in STING play critical roles in the occurrence of STING polymers (36, 37). Mutation at Cys206 but not at Cys148 abrogated DHN-induced STING polymerization (Figure 5B and Supplemental Figure 5B). Although C206S mutation did not influence STING dimerization, it impaired the formation of ER-STING aggregates (Figure 5C) and subsequent pyroptotic cell death (Supplemental Figure 5C). It is likely that DHN-induced opening of mPTP and subsequent activation of cGAS facilitate the polymerization of STING, and that this STING polymerization, rather than dimerization, leads to the formation of ER-STING aggregates.

cGAMP is the product of activated cGAS (38). Interestingly, incubation of cGAMP with A375 cells induced the dimerization of STING in accordance with previous reports (36); however, it barely induced the polymerization of STING, the formation of ER-STING aggregates, and pyroptosis (Supplemental Figure 5, D–F), suggesting that cGAS activation alone is insufficient for STING polymerization and that other unknown factors may be involved in DHN-induced pyroptosis. Considering our finding that caspase-8 exhibited unique enhanced activity in an acidic environment to cleave GSDME (Supplemental Figure 1I) and that DHN, a naphthol derivative, has the property of being a weak acid, we hypothesized that DHN might acidify the intracellular environment that benefits pyroptotic induction. Indeed, treatment with DHN, but not cGAMP, effectively induced dose-dependent acidification of the intracellular milieu (Figure 5D and Supplemental Figure 5G). NH4Cl significantly rescued DHN-induced decline of intracellular pH (Figure 5D), which was closely associated with a series of events, including inhibited STING polymerization (Figure 5E), the suppression of ER-STING aggregates (Figure 5, F and G), the decrease of caspase-8 activation and GSDME cleavage in TI fraction (Figure 5H), and the inhibition of DHN-induced pyroptosis (Figure 5I). Clearly, an acidic intracellular environment induced by DHN is a prerequisite for the formation of ER-STING aggregates and pyroptosis.

Considering the ER localization of the ER-STING aggregates, we speculated that the pH level of the ER might be altered. To this end, we first developed a reliable method for assessing ER pH, wherein the double ratio variation in different pH values served as an indicator of the detection system (Supplemental Figure 5H). Utilizing this approach, we demonstrated that DHN did acidify the ER environment in a dose-dependent manner and this acidification could be rescued by NH4Cl treatment as expected (Supplemental Figure 5I), in accordance with the results observed in the cytoplasm. These findings align with previous knowledge that the ER lacks an intrinsic pH regulatory system and readily equilibrates its pH to cytoplasmic levels (39).

Since NH4Cl failed to regulate DHN-induced cGAS puncta formation (Supplemental Figure 5J), and the decline in intracellular or ER pH caused by DHN remained unaffected by inhibition of CypD or STING (Supplemental Figure 5, K and L), we postulated that cGAS activation and ER acidification are 2 independent pathways that synergistically contribute to the formation of ER-STING aggregates and subsequent pyroptosis. To further validate this hypothesis, we simulated cGAS activation with cGAMP or diABZI treatment while inducing a decrease in intracellular and ER pH through lactic acid treatment (Supplemental Figure 5M), which is abundant in the tumor microenvironment and known to lower intracellular pH levels (40, 41). We demonstrated that either cGAMP/diABZI treatment alone or lowering pH by lactic acid alone was insufficient to induce pyroptosis; however, simultaneous treatment with cGAMP/diABZI and lactic acid clearly induced GSDME cleavage and subsequent pyroptosis (Figure 6A). Similar phenomena were also observed upon cotreatment of HCl, but not sodium lactate, with cGAMP (Supplemental Figure 5, N and O). Knockdown of LDHA/LDHB (which are essential for lactate metabolism) or AARS1/AARS2 (involved in protein lactylation) (34, 42) had no effect on pyroptosis induced by cGAMP plus lactic acid (Supplemental Figure 5P), thereby excluding the involvement of lactate metabolism or protein lactylation. Moreover, infection with HSV1, a DNA virus known to activate cGAS (43), could also trigger GSDME-mediated pyroptosis in the presence of lactic acid or HCl (Figure 6B). In conclusion, when both cGAS activation and intracellular acidification are simultaneously achieved, STING-dependent GSDME-mediated pyroptosis occurs.

PERK-induced STING phosphorylation facilitates the polymerization of STING. The mechanism by which an acidic environment facilitates STING polymerization and the formation of ER-STING aggregates remains unclear. We unexpectedly found that DHN induced a time-dependent phosphorylation of STING, as evidenced by the appearance of an upshifted band in a Phos-tag gel, which was abolished when cell lysates were incubated with calf intestinal alkaline phosphatase (CIAP) (Figure 7A). Notably, treatment with NH4Cl substantially impaired DHN-induced STING phosphorylation (Figure 7A), demonstrating the association between intracellular acidity and STING phosphorylation. Considering the subcellular localization of STING in the ER, we employed various inhibitors, including GSK2656157 targeting PERK, GSK650394 targeting SGK1, GSK2850163 targeting IRE1α, LRRK2-IN-1 targeting LRRK2, sorafenib targeting FLT3, lenvatinib targeting KDR, and HG-9-91-01 targeting SIK to identify the specific ER-resident protein kinase involved in STING phosphorylation. Among these inhibitors, only GSK2656157 effectively impaired DHN-induced STING phosphorylation (Figure 7B and Supplemental Figure 6A), indicating that PERK was a crucial mediator of STING phosphorylation upon DHN stimulation. Knockdown of PERK inhibited DHN-induced STING phosphorylation (Figure 7B and Supplemental Figure 6B). It has been reported that PERK can be activated through autophosphorylation at the Thr982 residue (44). DHN treatment effectively enhanced PERK phosphorylation (Figure 7C), and mutation of Thr982 or inhibition of PERK activity by GSK2656157 eliminated DHN-induced PERK phosphorylation and the association of PERK with STING (Figure 7, C and D). Therefore, these findings suggest that DHN promotes autophosphorylation and activation of PERK, leading to the association of PERK with STING and subsequent STING phosphorylation.

Recent studies have suggested that cGAS/STING activation directly triggers PERK activation (18). In contrast, inhibition of mtDNA release (CsA treatment or CypD knockdown) or suppression of cGAS/STING activity (G140 treatment or STING knockdown) did not affect PERK phosphorylation upon DHN stimulation (Supplemental Figure 6D), excluding the involvement of cGAS/STING in PERK activation during DHN treatment. Instead, NH4Cl treatment profoundly suppressed DHN-induced PERK activation (Figure 7E), and a decrease in intracellular pH caused by lactic acid, HCl, or citric acid was sufficient to induce PERK phosphorylation and activation (Figure 7E and Supplemental Figure 6E). Combined with the fact that inhibition of PERK by GSK2656157 or knockdown of PERK did not affect DHN-induced acidification (Supplemental Figure 6F), it could be concluded that PERK is activated when there is a decline in intracellular pH induced by DHN treatment.

Although inhibition of PERK did not affect cGAS puncta formation (Supplemental Figure 6G), it markedly impaired DHN-induced STING polymerization (but not dimerization), formation of ER-STING aggregates, cleavage of GSDME by caspase-8 within ER-STING aggregates, and subsequent pyroptosis (Figure 7, F–H, and Supplemental Figure 6, H–K), demonstrating the crucial role of PERK in STING polymerization but not in cGAS activation. According to the PhosphoSitePlus database, there are 4 potential phosphorylated Ser/Thr residues in STING, including Thr84, Ser345, Ser358, and Ser366 (45, 46). Substituting alanine for either Ser345 or Ser358 weakly impaired DHN-induced STING phosphorylation (Supplemental Figure 6L), while combined mutation of these 2 residues (STINGS345A/358A) resulted in the abrogation of DHN-induced STING phosphorylation in A375 cells and PERK-induced STING phosphorylation in vitro (Figure 8A), leading to elimination of the polymerization of STING, formation of ER-STING aggregates, GSDME cleavage by caspase-8 in the TI fraction, and subsequent pyroptosis (Figure 8, B–D, and Supplemental Figure 6M). Considering that the Cys206 mutation, which abolishes STING polymerization (Figure 5B), had no impact on DHN-induced STING phosphorylation (Supplemental Figure 6N), it is plausible that PERK-mediated phosphorylation of STING serves as an upstream event, which facilitates the assembly of STING polymers upon DHN stimulation, thereby contributing to pyroptosis.