ASC deglutathionylation is a checkpoint for NLRP3 inflammasome activation

Inflammasomes are key components of innate immune responses against danger signals such as pathogens and dead cells. Activation of inflammasome leads to cleavage-based activation of caspase-1 that induces pyroptosis and processes the proinflammatory cytokines IL-1β and IL-18 (McKee and Coll, 2020; Miao et al., 2011). NLRP3 inflammasome is the best-studied inflammasome, and its excessive activation relates to autoimmune diseases and inflammatory diseases. Given these consequences, activation of NLRP3 inflammasome is precisely controlled (Schroder and Tschopp, 2010). Activating NLRP3 inflammasome requires priming (signal 1) and activation (signal 2) signals. Activation signal, such as ATP, nigericin, and crystals, promotes assembly of NLRP3 inflammasome and the nucleation of the apoptosis-associated speck-like protein containing a CARD (ASC; Lu et al., 2014; Schroder and Tschopp, 2010). Nucleated ASC proteins recruit pro–caspase-1 and promote its cleavage. In this activation step, ASC oligomerization is an essential process (Dick et al., 2016). Both the N-terminal Pyrin domain and the C-terminal caspase recruitment domain (CARD) are required for ASC oligomerization, with ASC Pyrin domain assembling into filaments and CARD condensing ASC filaments into specks (Schmidt et al., 2016). ASC oligomerization is regulated by posttranslational modification, including ubiquitination, deubiquitination, and phosphorylation (Chung et al., 2016; Hoss et al., 2017). Whether other mechanisms govern ASC oligomerization and prevent unwanted NLRP3 inflammasome activation remains unclear.

GSTO1 belongs to the glutathione transferase (GST) enzyme family. Different from other GSTs, GSTOs could catalyze thioltransferase reactions as glutaredoxins (Board et al., 2000). Its capability of deglutathionylating protein implicates a role of GSTO1 in protein posttranslational modification (Menon and Board, 2013). Whether GSTO1-catalyzed deglutathionylation contributes to the precise control of ASC oligomerization and NLRP3 inflammasome activation remains unclear.

Here, we show an essential role of GSTO1 in promoting ASC oligomerization and NLRP3 inflammasome activation in macrophages via deglutathionylating ASC. In response to NLRP3 inflammasome activators, GSTO1 binds ASC and promotes ASC deglutathionylation at ER, under control of mitochondrial ROS and triacylglyceride synthesis, respectively (Hughes et al., 2019; Menon et al., 2017). In Asc^C171A macrophages expressing nonglutathionylated ASC protein, activation of NLRP3 inflammasome is GSTO1 independent, ROS independent, and signal 2 less dependent. Moreover, Asc^C171A mice display hyperinflammationupon activating NLRP3 inflammasome in vivo. Altogether, we propose that glutathionylation of ASC inhibits ASC oligomerization, and deglutathionylation of ASC at ER, under metabolic control, is a checkpoint for ASC oligomerization and NLRP3 inflammasome activation in macrophages.

We generated Gsto1^−/− mice and confirmed the deletion of GSTO1 protein in Gsto1^−/− bone marrow–derived macrophages (BMDMs; Fig. 1 A). These Gsto1^−/− mice were more resistant than Gsto1^+/+ mice in developing NLRP3 inflammasome-dependent arthritis, as indicated by less joint swelling gain (Fig. 1 B) after injecting NLRP3 inflammasome activator monosodium urate (MSU). Moreover, joints from the aforementioned Gsto1^−/− mice released much less IL-1β than those joints from control mice when cultured ex vivo (Fig. 1 C). Next, we activated NLRP3 inflammasome in BMDMs in vitro with LPS followed by ATP and found that IL-1β production was inhibited in Gsto1^−/− BMDMs (Fig. 1 D). However, IL-6 and TNF-α were not influenced by GSTO1 deficiency. Although Gsto1^−/− BMDMs produced less mature IL-1β after activation, they had similar level of pro–IL-1β as Gsto1^+/+ BMDMs (Fig. 1 E). These results indicated that deficiency of GSTO1 reduced processing of pro–IL-1β. Moreover, reduced active caspase-1 p20 but normal pro–caspase-1 was detected in Gsto1^−/− BMDMs (Fig. 1 E), indicating impaired activation of NLRP3 inflammasome. Additionally, we stimulated Gsto1^−/−and Gsto1^+/+ peritoneal macrophages with LPS followed by distinct NLRP3 inflammasomeactivators, including ATP, MSU, Nigericin, and imiquimod. ATP, MSU, and Nigericin activate NLRP3 inflammasome in a K⁺-dependent manner (Próchnicki et al., 2016), whereas imiquimod and CL097 activate in a K⁺-independent manner (Groß et al., 2016). Impaired cleavage of pro–IL-1β and pro–caspase-1 were detected in Gsto1^−/−peritoneal macrophages (Fig. 1 F) in response to both K⁺-dependent and K⁺-independent NLRP3 inflammasome activators. To further confirm the role of GSTO1 in activating NLRP3 inflammasome, we restored the GSTO1 expression in Gsto1^−/− BMDMs, and that successfully recovered IL-1β production (Fig. S1, A and B).

To test whether GSTO1 promoted NLRP3 inflammasome activation in an enzyme activity–dependent manner, we used inhibitor ML175 to inhibit the catalytic activity of GSTO1. ML175 reduced IL-1β production from LPS-ATP–stimulated BMDMs in a dose-dependent manner, whereas it showed no influence on IL-6 or TNF-α production (Fig. 1 G). Moreover, we showed that ML175 inhibited cleavage of pro–IL-1β and pro–caspase-1 in BMDMs stimulated with LPS followed by NLRP3 inflammasome activator ATP, MSU, Nigericin, imiquimod, and CL097, respectively (Fig. 1, H and I). Therefore, the catalytic activity of GSTO1 is required for both K⁺-dependent and K⁺-independent activation of NLRP3 inflammasome.

Next, we studied the role of GSTO1 in regulating expression of NLRP3, NEK7, ASC, and pro–caspase-1, as well as their interactions. Deficiency of GSTO1 (Fig. 2 A) and inhibitor ML175 (Fig. 2 B) showed no influence on the levels of NLRP3, NEK7, ASC, or pro–caspase-1 in LPS-ATP–stimulated BMDMs. Although genetic deficiency and inhibitor exhibited no influence on NLRP3–NEK7 interaction or NLRP3–ASC interaction in LPS-ATP–stimulated BMDMs, they obviously decreased pro–caspase-1–ASC interaction (Fig. 2, A and B), in line with the reduced activation of caspase-1 (Fig. 1, E and H). Given the role of oligomerized ASC in recruiting and activating pro–caspase-1 (Schmidt et al., 2016), we next measured the oligomerization of ASC. We found that ASC oligomerization was disrupted in BMDMs either with deleted Gsto1 or treated with ML175 (Fig. 2, C and D). Notably, ML175 inhibited ASC oligomerization in LPS-ATP–stimulated Gsto1^+/+ BMDMs but not in LPS-ATP–stimulated Gsto1^−/− BMDMs (Fig. 2 E), confirming that ML175 inhibited ASC oligomerization via targeting on GSTO1. Additionally, when ATP, MSU, Nigericin, and imiquimod were used to activate NLRP3 inflammasome in LPS-stimulated peritoneal macrophages, ASC oligomerization was inhibited by deficiency of GSTO1 as well (Fig. 2 F). These results demonstrate that catalytic activity of GSTO1 promotes ASC oligomerization during the activation of NLRP3 inflammasome.

We found that ASC coimmunoprecipitated with GSTO1 at a low level in BMDMs, and LPS-ATP stimulation significantly increased their interaction (Fig. 3 A). Next, we investigated whether GSTO1 could deglutathionylate ASC. ASC was coimmunoprecipitated with antibody against GSH, indicating glutathionylation of ASC (Fig. 3 B). LPS-ATP stimulation reduced the level of glutathionylated ASC in BMDMs (Fig. 3 B), in company with an increase of GSTO1–ASC interaction (Fig. 3 A). GSTO1 deficiency increased the level of glutathionylated ASC in both stimulated and unstimulated BMDMs, suggesting that GSTO1 promoted ASC deglutathionylation in both resting and activated cells (Fig. 3 B). The result that GSTO1 could constitutively deglutathionylate ASC protein in unstimulated BMDMs is in line with the low level of GSTO1–ASC interaction in unstimulated BMDMs. The LPS-ATP stimulation–induced reduction of glutathionylated ASC in Gsto1^−/− BMDMs implied participation of other factors in mediating ASC deglutathionylation. In all conditions, we found no glutathionylated NLRP3 or glutathionylated pro–caspase-1. Although we detected glutathionylated NEK7, neither LPS-ATP stimulation nor GSTO1 deficiency influenced its glutathionylation. Therefore, the GSTO1 promotes deglutathionylation of ASC but not other NLPR3 inflammasome components.

ASC protein has only one cysteine residue at position 171, which is the potential glutathionylation site. Next, we generated Asc^C171A mice, in which the ASC^C171A protein could not be glutathionylated due to the substitution of alanine for cysteine. In coimmunoprecipitation experiments, we detected no ASC glutathionylation in Asc^C171A BMDMs (Fig. 3 C). To further confirm the glutathionylation of ASC at cysteine¹⁷¹ and reveal the glutathionylated ASC in situ, we performed proximity ligation assay (PLA), in which two PLA probe–labeled secondary antibodies bound the primary antibodies against ASC and GSH and generated fluorescence signal only if ASC was posttranslationally modified by glutathionylation. This imaging-based assay revealed glutathionylated ASC protein in Asc^wt BMDMs but not Asc^C171A BMDMs, further confirming the glutathionylation of ASC at cysteine¹⁷¹ in unstimulated cells (Fig. 3, D and E). Consistently, LPS-ATP stimulation reduced the amount of glutathionylated ASC in Asc^wt BMDMs, as indicated by reduced ASC-GSH PLA puncta numbers, and ML175 restored the amount of glutathionylated ASC (Fig. 3, D and E). These results further prove the role of GSTO1 in deglutathionylating ASC in activated BMDMs. Additionally, we compared the numbers of ASC-GSH PLA puncta in Gsto1^+/+ and Gsto1^−/− peritoneal macrophages. Again, we found that GSTO1 deficiency increased the amount of glutathionylated ASC in peritoneal macrophages stimulated with LPS and distinct NLRP3 inflammasome activators, including ATP, MSU, Nigericin, and imiquimod, as well as in unstimulated peritoneal macrophages (Fig. 3 F). These results further support that GSTO1 promotes ASC deglutathionylation in response to distinct NLRP3 inflammasome activators in macrophages.

Molecular dynamics (MD) simulations showed that residue cysteine¹⁷¹ localized at the ASC-CARD multimerization interface (Fig. S2 A). GSH covalently linking to cysteine¹⁷¹ at the interface clashed with the other ASC-CARD domain (Fig. S2 B). Additionally, binding of GSH at cysteine¹⁷¹ caused domain rotation and reduced area of CARD–CARD binding interface in dimeric and 16mer filament simulations (Fig. S2, C–H). These results indicate that GSH conjugation impairs ASC-CARD assembly. Assembly of ASC-CARD domains is required for ASC oligomerization and nucleation of pro–caspase-1 (Dick et al., 2016; Li et al., 2018). Therefore, deglutathionylating ASC at cysteine¹⁷¹ by GSTO1 would promote assembly of ASC-CARD domains and ASC oligomerization as a result. Notably, when extra GSH was added to ML175-treated BMDMs, it reduced the level of glutathionylated ASC and rescued the ASC oligomerization and ASC–pro–caspase-1 interaction (Fig. S3, A–C). To further investigate the role of GSTO1-mediated ASC deglutathionylation in activating NLRP3 inflammasome, we inhibited GSTO1 function with ML175 in Asc^C171A BMDMs and Asc^C171A peritoneal macrophages. We found that although ML175 inhibited oligomerization of ASC and processing of pro–caspase-1 and pro–IL-1β in LPS-ATP–stimulated Asc^wt BMDMs, LPS-Nigericin–stimulated Asc^wt BMDMs, and LPS-ATP–stimulated Asc^wt peritoneal macrophages, ML175 showed no effect in Asc^C171A BMDMs or Asc^C171A peritoneal macrophages activated by the above NLRP3 inflammasome activators (Fig. 3, G–I). Moreover, we showed that knockdown of Gsto1 reduced LPS-ATP stimulation–induced oligomerization of ASC and processing of pro–caspase-1 and pro–IL-1β in Asc^wt BMDMs but exhibited no influence in Asc^C171A BMDMs (Fig. 3 J). These results indicate that, given no ASC glutathionylation, GSTO1 catalytic activity is not required for ASC oligomerization and NLRP3 inflammasome activation in macrophages. In comparison with GSTO1 activity–impaired Asc^wt macrophages, GSTO1 activity–impaired Asc^C171A macrophages showed higher levels of ASC oligomerization and increased processing of pro–caspase-1 and pro–IL-1β after stimulation with LPS and NLRP3 inflammasome activators (Fig. 3, G–J), supporting the inhibitory effect of ASC glutathionylation on activation of NLRP3 inflammasome. Together, our results demonstrate that glutathionylation of ASC represses NLRP3 inflammasome activator–induced ASC oligomerization, and GSTO1 promotes NLRP3 inflammasome activation by deglutathionylating ASC.

Mitochondrial ROS is known to trigger NLRP3 inflammasome assembly (Sánchez-Rodríguez et al., 2021). We found that LPS-ATP stimulation induced ASC–GSTO1 interaction in BMDMs, as indicated by coimmunoprecipitation of ASC and GSTO1, and this interaction was inhibited by Mito-TEMPO, a mitochondria-targeted antioxidant, in a dose-dependent manner (Fig. 4 A). We further employed a PLA approach to view the GSTO1–ASC interaction in situ. Asc^−/− BMDMs were used as negative controls and showed no signal. We confirmed that LPS-ATP stimulation increased GSTO1–ASC interaction in BMDMs, whereas Mito-TEMPO inhibited the ASC–GSTO1 interaction (Fig. 4, B and C). Consequently, the level of ASC that coimmunoprecipitated with GSH increased with Mito-TEMPO treatment in a dose-dependent manner in LPS-ATP–stimulated BMDMs (Fig. 4 D). Additionally, in situ imaging of glutathionylated ASC by a PLA approach further confirmed that Mito-TEMPO treatment increased the level of glutathionylated ASC in LPS-ATP–stimulated BMDMs (Fig. 4, E and F). When another inhibitor, MitoQ, was used to inhibit mitochondrial ROS in LPS-ATP–stimulated BMDMs, similar reduction of ASC-GSTO1 interaction and elevation of glutathionylated ASC were observed (Fig. 4, G and H). On the other hand, EN460, which inhibits generation of ER ROS, showed no influence on ASC–GSTO1 interaction or ASC deglutathionylation (Fig. 4, G and H). Together, these results demonstrate that mitochondrial ROS promotes ASC–GSTO1 interaction and thus enhances GSTO1-mediated ASC deglutathionylation.

Moreover, mitochondrial ROS inhibitors Mito-TEMPO and MitoQ inhibited ASC oligomerization and processing of pro–caspase-1 and pro–IL-1β in LPS-ATP–stimulated Asc^wt BMDMs, but exhibited no effect in LPS-ATP–stimulated Asc^C171A BMDMs (Fig. 4, I and J). Notably, when NLRP3 inhibitor Oridonin was added to LPS-ATP–stimulated Asc^C171A BMDMs together with Mito-TEMPO, the ASC oligomerization and the processing of pro–caspase-1 and pro–IL-1β were completely inhibited (Fig. 4 K), confirming that, in Asc^C171A BMDMs, LPS-ATP stimulation–induced activation of inflammasome in the absence of mitochondrial ROS was NLRP3 dependent. Again, EN460, which inhibits generation of ER ROS, exhibited no influence on ASC oligomerization or processing of pro–caspase-1 and pro–IL-1β in both Asc^wt BMDMs and Asc^C171A BMDMs (Fig. 4 L). Our results, that activation of NLRP3 inflammasome is mitochondrial ROS independent in cells without ASC glutathionylation, demonstrate that mitochondrial ROS promotes activation of NLRP3 via induction of ASC deglutathionylation.

To reveal the site for ASC deglutathionylation, we first studied the location of GSTO1. GSTO1 was perfectly colocalized with ER marker Calnexin in BMDMs, indicating its location at ER (Fig. 5 A). Additionally, we found that glutathionylated ASC as shown by a PLA approach was largely colocalized with Calnexin in BMDMs as well (Fig. 5, B and C). In contrast, neither GSTO1 nor glutathionylated ASC was colocalized with mitochondrial marker Mito-Tracker (Fig. 5, A–C), although GSTO1 bound ASC in response to mitochondrial ROS (Fig. 4 A). LPS-ATP stimulation led to an increase of GSTO1–ASC interaction, as shown by the PLA approach, and the interaction site also colocalized with Calnexin rather than with Mito-Tracker in BMDMs (Fig. 5, D and E). These results suggest that GSTO1 deglutathionylates ASC, predominantly, at ER. To investigate the location of ASC oligomerization, we separated ER, crude mitochondrial, and cytosolic fractions from BMDMs with or without LPS-ATP stimulation. We found that LPS-ATP stimulation–induced oligomerization of ASC was predominantly detected in ER fraction, with no or a minor amount in the cytosolic or mitochondrial fraction (Fig. 5 F). In line with the role of GSTO1 in promoting ASC oligomerization during the activation of NLRP3 inflammasome, deficiency of GSTO1 reduced LPS-ATP stimulation–induced oligomerization of ASC in ER fraction (Fig. 5 G). Notably, FACL4, a marker for mitochondria-associated ER membranes (MAMs), was detected in the mitochondrial fraction but not in the ER fraction (Fig. 5 F), in line with reported location of MAMs in crude mitochondrial fraction (Wieckowski et al., 2009). Although we could not exclude the possibility that the minor amount of ASC oligomers detected in mitochondrial fraction are related to MAMs, our findings suggest that the majority of ASC oligomers are formed at ER and are not related to MAMs. When Nigericin was used to activate NLRP3 inflammasome in BMDMs, ASC oligomerization was detected predominantly in ER fraction and in a GSTO1-dependent manner as well (Fig. 5, H and I).

To further prove the ER as a site for ASC oligomerization, we viewed the ASC–ASC interaction in situ with a PLA approach, in which two distinct primary antibodies were used to target the different ASC proteins simultaneously, followed by incubation with two PLA probe–labeled secondary antibodies, and only interaction between two ASC proteins binding distinct primary antibodies could generate PLA signal. In Asc^−/− BMDMs, there was no PLA signal (Fig. 5 J). Although plenty of ASC existed in BMDMs, no PLA signal was detected without stimulation (Fig. 5 J), excluding the possibility that two primary antibodies targeted same ASC protein in unstimulated cells. Additionally, in LPS-ATP–stimulated BMDMs, if two primary antibodies against ASC were added to cells sequentially with an interval long enough to ensure blocking of ASC by the first antibody, we observed no PLA signal (Fig. 5 J). This result indicated that two antibodies competed for binding the same ASC protein, and thereby excluded the possibility of binding two distinct primary antibodies at the same ASC protein in ASC oligomers. In LPS-ATP–stimulated BMDMs, which were stained with two distinct ASC-specific primary antibodies simultaneously, we detected ASC-ASC PLA puncta, and that indicated oligomerization of ASC (Fig. 5 J). Moreover, these puncta were colocalized with Calnexin rather than with Mito-Tracker (Fig. 5, J and K), further confirming the ER location of ASC oligomerization. In addition to ASC oligomers, LPS-ATP stimulation increased levels of NLRP3 protein in ER fraction from BMDMs (Fig. 5 L). Although LPS-ATP stimulation increased levels of NLRP3 in mitochondrial fraction as well, the level was much lower than in ER fraction. This finding supports the activation of NLRP3 inflammasome at ER.

ER is the major intracellular organelle for lipid synthesis. After LPS-ATP stimulation, BMDMs increased the amount of triacylglyceride (Fig. 6 A) as well as the level of diacylglycerol acyltransferase 1 (DGAT1), a key enzyme for triacylglyceride synthesis (Fig. 6 B). Interestingly, DGAT1 inhibitor T863 reduced LPS-ATP stimulation–induced IL-1β production from BMDMs (Fig. 6 C). Moreover, we found that T863 did not influence the LPS-ATP stimulation–induced mitochondrial ROS (Fig. 6 D) or GSTO1-ASC interaction (Fig. 6, E and F), but inhibited ASC deglutathionylation, as indicated by increase of glutathionylated ASC in T863-treated BMDMs (Fig. 6, E and F). Moreover, T863 inhibited the oligomerization of ASC and the processing of pro–caspase-1 and pro–IL-1β in Asc^wt BMDMs but not in Asc^C171A BMDMs (Fig. 6 G). These results indicate that T863 interferes with the activation of NLRP3 inflammasome by inhibiting ASC deglutathionylation. To further confirm the functional impact of triacylglyceride synthesis, the key enzyme DGAT1 was knocked down in BMDMs by shRNA. We showed that knockdown of Dgat1 inhibited oligomerization of ASC, processing of pro–caspase-1 and pro–IL-1β, and ASC deglutathionylation in Asc^wt BMDMs, after LPS-ATP stimulation (Fig. 6, H–J). Again, knockdown of Dgat1 did not influence the LPS-ATP stimulation–induced GSTO1–ASC interaction (Fig. 6, I and J). Due to the absence of ASC glutathionylation in Asc^C171A BMDMs, knockdown of Dgat1 in these cells showed no influence on LPS-ATP stimulation–induced oligomerization of ASC or processing of pro–caspase-1 and pro–IL-1β (Fig. 6 K). Notably, although LPS-ATP stimulation increased the amount of cholesterol, another lipid synthesized in ER as well, inhibiting cholesterol synthesis by Simvastatin showed no influence on IL-1β production (Fig. S4, A–C). Together, these results demonstrate that triacylglyceride synthesis in ER promotes deglutathionylation of ASC during the activation of NLRP3 inflammasome.

In addition to NLRP3 inflammasome, cleavage of pro–caspase-1 and pro–IL-1β by Pyrin inflammasome, AIM2 inflammasome, and NLRC4 inflammasome are ASC dependent (Hornung et al., 2009; Xu et al., 2014; Zhao et al., 2011). Next, we activated Pyrin inflammasome, AIM2 inflammasome, and NLRC4 inflammasome in BMDMs with LPS-TcdB (Clostridium difficile toxin B), LPS-poly(deoxyadenylic-deoxythymidylic [dA:dT]), and LPS-Flagellin, respectively. We found that deficiency of GSTO1 showed no influence on LPS-TcdB stimulation–induced oligomerization of ASC or processing of pro–caspase-1 and pro–IL-1, whereas it diminished LPS-poly(dA:dT) stimulation–induced and LPS-Flagellin stimulation–induced oligomerization of ASC and processing of pro–caspase-1 and pro–IL-1β (Fig. S5 A). These results demonstrate that GSTO1 is not required for activation of Pyrin inflammasome, but it promotes activation of AIM2 inflammasome and NLRC4 inflammasome. Notably, the inhibitory effects of GSTO1 deficiency on activation of AIM2 inflammasome and NLRC4 inflammasome were less than those on activation of NLRP3 inflammasome. Moreover, Gsto1 knockdown reduced LPS-poly(dA:dT) stimulation–induced and LPS-Flagellin stimulation–induced oligomerization of ASC and processing of pro–caspase-1 and pro–IL-1β in both Asc^wt BMDMs and Asc^C171A BMDMs similarly (Fig. S5, B and C). Therefore, GSTO1 promotes activation of AIM2 inflammasome and NLRC4 inflammasome in an ASC deglutathionylation–independent manner. Notably, Gsto1 knockdown Asc^wt BMDMs and Gsto1 knockdown Asc^C171A BMDMs showed similar levels of ASC oligomerization and processing of pro–caspase-1 and pro–IL-1β in response to LPS-poly(dA:dT) stimulation or LPS-Flagellin stimulation (Fig. S5, B and C), implying no influence of ASC glutathionylation on activation of AIM2 inflammasome or NLRC4 inflammasome. Moreover, we showed that LPS-poly(dA:dT)–induced ASC oligomerization and LPS-Flagellin–induced ASC oligomerization were predominantly in mitochondrial fraction (Fig. S5, D and E), where glutathionylated ASC was absent (Fig. 5 B). GSTO1 deficiency reduced the LPS-poly(dA:dT)–induced and LPS-Flagellin–induced ASC oligomerization in mitochondrial fraction (Fig. S5, D and E), confirming the requirement of GSTO1 for activating AIM2 inflammasome and NLRC4 inflammasome. Together, these findings demonstrate that GSTO1 promotes activation of AIM2 inflammasome and NLRC4 inflammasome via other mechanisms rather than ASC deglutathionylation, and ASC deglutathionylation is not required for activating these inflammasomes.

Our findings indicate that deglutathionylation of ASC in response to mitochondrial ROS serves as a derepression signal triggering the activation of NLRP3 inflammasome. It is well known that NLRP3 inflammasome activator as signal 2 induces mitochondrial ROS (Mills et al., 2016) and promotes NLRP3 inflammasome assembly (Amores-Iniesta et al., 2017). Due to the absence of glutathionylated ASC in Asc^C171A BMDMs, mitochondrial ROS was not required for NLRP3 inflammasome activation (Fig. 4, I and J). Consistently, we found that oligomerization of ASC and processing of pro–caspase-1 and pro–IL-1β were induced by LPS alone without the presence of ATP in Asc^C171A BMDMs (Fig. 7 A). Via inhibiting NLRP3 with Oridonin or knocking down Nlrp3, we further proved that IL-1β production induced by LPS alone in these Asc^C171A BMDMs was NLRP3 inflammasome dependent (Fig. 7, B and C). In Asc^C171A BMDMs, adding ATP further increased IL-1β production (Fig. 7, B and C), indicating that ATP as signal 2 may regulate other pathways in addition to ASC deglutathionylation. Together, our findings suggest that without repressive effect of ASC glutathionylation, NLRP3 inflammasome could be activated by signal 1 alone, implying hyperactivation of NLRP3 inflammasome. Moreover, when LPS was injected into Asc^C171A and Asc^wt mice to induce septic shock, the Asc^C171A mice produced higher levels of proinflammatory cytokines including IL-1β, IL-6, and TNF-α, and the production of these cytokines was abrogated by NLRP3 inhibitor Oridonin (Fig. 7 D). These findings exhibited a NLRP3 inflammasome-dependent hyperinflammatory phenotype of Asc^C171A mice. In line with the hyperinflammation caused by LPS injection, Asc^C171A mice exhibited higher mortality than Asc^wt mice (Fig. 7 E). Additionally, when NLRP3 specific activator MSU was injected into joints of Asc^wt mice and Asc^C171A mice to induce arthritis, we found that Asc^C171A mice showed more severe joint swelling than Asc^wt mice, and NLRP3 inhibitor MCC950 significantly reduced joint swelling in both groups (Fig. 7, F and G). Consistently, joints from the MSU-injected Asc^C171A mice released higher amounts of IL-1β than joints from MSU-injected Asc^wt mice when cultured ex vivo, and IL-1β production was inhibited by MCC950 (Fig. 7 H). We also measured IL-1β production in sera from these MSU-injected Asc^C171A and Asc^wt mice and confirmed the higher level of IL-1β in Asc^C171A mice (Fig. 7 H). Again, MCC950 inhibited MSU-induced IL-1β production in sera of these mice (Fig. 7 H). These results demonstrate the hyperinflammation in Asc^C171A mice after activating NLRP3 inflammasome. It is possible that glutathionylation of ASC serves as a brake to prevent NLRP3 inflammasome-related hyperinflammation in vivo.

Despite significant progress in understanding NLRP3 inflammasome activation, the mechanisms governing NLRP3 inflammasome activation and prevention of excessive inflammation are largely unknown. Our results indicate that glutathionylation of ASC at its cysteine¹⁷¹ residue inhibits ASC oligomerization in the resting state and thereby blocks unwanted activation of NLRP3 inflammasome. This repressive effect is relieved by GSTO1-mediated deglutathionylation of ASC in response to mitochondrial ROS. Thus, deglutathionylation of ASC functions as a major checkpoint for triggering the NLRP3 inflammasome activation. Our finding that GSTO1 promotes NLRP3 inflammasome activation by deglutathionylating ASC rather than other proteins is contrary to those of another study that indicated a role of GSTO1 in deglutathionylating NEK7 (Hughes et al., 2019). It is possible that the NEK7 overexpression strategy used in that study is responsible for the different result. Under stress conditions, glutathionylation of other inflammasome proteins has been reported to regulate inflammasome activation (Meissner et al., 2008). Deficiency of superoxide dismutase 1 inhibits NLRP3 inflammasome activation potentially by increasing the extent of glutathionylation on pro–caspase-1 protein. In the same study, both inhibitory glutathionylation sites and activating glutathionylation sites on overexpressed pro–caspase-1 have been indicated. Stressed conditions might increase the complexity of glutathionylation-based regulation on NLRP3 inflammasome activation. Notably, GSTO1 is not the only factor promoting ASC deglutathionylation in activated BMDMs, as indicated by LPS-ATP stimulation–induced ASC deglutathionylation in GSTO1-deficient BMDMs. Whether glutaredoxins contribute to ASC deglutathionylation in these GSTO1-deficient cells remains to be explored. In the absence of ASC glutathionylation, ATP as signal 2 was not essential for NLRP3 inflammasome activation. This result is particularly interesting in light of a previous study that found NLRP3 inflammasome activation in dendritic cells (DCs) to be signal 2 independent (Kang et al., 2013). In addition to the higher levels of pro–IL-1β and NLRP3 in DCs than in macrophages, distinct ASC glutathionylation/deglutathionylation in DCs might also make a contribution. As shown by the structural simulations, conjugation of GSH to ASC cysteine¹⁷¹ interfered with ASC–CARD interactions and inhibited ASC oligomerization. Notably, the ASC-CARD domain also binds CARD domain of pro–caspase-1. We did not exclude the possibility that ASC glutathionylation might influence the ASC–pro–caspase-1 interaction as well.

Although the ASC protein has been reported as a cytosolic protein, our results indicated that glutathionylated ASC was enriched at intracellular compartments, particularly at ER, rather than at the cytosol or mitochondria (Fig. 5 B). It is possible that the nonglutathionylated ASC in cytosol and mitochondria makes no/low contribution to NLRP3 inflammasome activation. Indeed, we showed that oligomerized ASC was detected mainly at ER rather than at cytosol or mitochondria while activating NLRP3 inflammasome (Fig. 5, F and H). These results suggest that other ER-related factors in addition to nonglutathionylated ASC control NLRP3 inflammasome activator–induced ASC oligomerization. A previous study proposed that NLRP3 and ASC may translocate to MAMs after LPS-ATP stimulation (Zhou et al., 2011). In our study, ASC oligomers induced by NLRP3 inflammasome activators were predominantly enriched in the ER fraction that contained no MAMs, as indicated by the absence of FACL4 (Fig. 5, F and H). Interestingly, AIM2 and NLRC4 inflammasome activators induce ASC oligomerization predominantly at mitochondria where ASC is not glutathionylated. These findings explain the dispensable role of ASC deglutathionylation in activating AIM2 and NLRC4 inflammasomes. Therefore, GSTO1-promoted ASC deglutathionylation is not generally required for activation of all ASC-dependent inflammasomes.

ASC deglutathionylation induced by NLRP3 inflammasome activators is promoted by mitochondrial ROS and triacylglyceride synthesis at different steps, in line with reported essential roles of ROS and fatty acid synthesis in activating NLRP3 inflammasome (Moon et al., 2015). Given the previously reported metabolic control on priming process, activation of NLRP3 inflammasome is regulated by cellular metabolism at multiple steps. Still, further studies are required to understand the mechanisms of how triacylglyceride controls GSTO1-mediated ASC deglutathionylation.

It is well known that NLRP3 inflammasome closely relates to autoimmune diseases and inflammatory diseases, including experimental autoimmune encephalitis (Inoue et al., 2012), colitis (Bauer et al., 2010), and insulin resistance (Vandanmagsar et al., 2011). Unlike GSTO1 deficiency, ASC^C171A mutant results in hyperinflammation in mouse models. Therefore, we propose a link between GSTO1, deglutathionylation of ASC, and inflammasome-related diseases. Whether the differential expression and/or genetic polymorphisms in Gsto1 or Asc may relate to these inflammatory diseases remains unclear. Further studies are required to explore the roles of this regulatory pathway in human diseases and evaluate its potential as a drug target.

Gsto1^−/− mice were generated by deleting exons 1–4. Asc^C171A mice expressing ASC^C171A mutant were generated using CRISPR/Cas9-mediated genome editing to replace the cysteine at position 171 with alanine in the third exon of ASC. All mice used were on the C57BL/6J background and were 8–12 wk old. Cohoused littermate controls were used in our experiments. Mice were housed under specific pathogen–free conditions. Food and water were available ad libitum. All animal procedures were approved by the Animal Care and Use Committee of University of Science and Technology of China, and all experiments were performed in accordance with approved guidelines.

To induce arthritis, MSU (Sigma-Aldrich, U0881) was injected into the joints of Gsto1^+/+, Gsto1^−/−, Asc^wt, and Asc^C171A mice at a dose of 0.5 mg per mouse, and size of joint was measured at different time points. MCC950 (Selleckchem, S8930) was injected into the joints at a dose of 50 mg kg⁻¹. To measure the production of IL-1β from the joints, patellae from these mice were isolated after 24 h and cultured in 200 µl Opti-MEM (Gibco) for 1 h. To induce lethal septic shock, mice were challenged i.p. with LPS (Sigma-Aldrich, L4391) at a dose of 15 mg kg⁻¹. Oridonin (Selleck) was injected into mice at a dose of 20 mg kg⁻¹. To measure the production of IL-1β, IL-6, and TNF-α, sera were collected from these mice after 24 h.

To induce BMDMs, bone marrow cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Biological Industries), 50 µM β-mercaptoethanol (Sigma-Aldrich), and 50 µg ml⁻¹ M-CSF (Novus Biologicals) for 5–6 d. For NLRP3 inflammasome activation, BMDMs or peritoneal macrophages were primed with 50 ng ml⁻¹ LPS (Sigma-Aldrich, L4391) for 4 h and then stimulated with 3 mM ATP (Sigma-Aldrich, A2383) or 5 µM Nigericin (Sigma-Aldrich, N7143) for 30 min, with 150 µg ml⁻¹ MSU (Sigma-Aldrich, U0881) for 3 h, or with 45 µg ml⁻¹ CL097 (InvivoGen, tlrl-c97) or 45 µg ml⁻¹ imiquimod (InvivoGen, tlrl-imq) for 1 h. To inhibit the enzyme activity of GSTO1, BMDMs were treated with ML175 (AOBIOUS AOB1209) at 4 µM. To inhibit the production of mitochondrial ROS, BMDMs were treated with Mito-TEMPO (Enzo ALX-430-150) at 0.5 or 1 mM, or with MitoQ (MedChemExpress, HY-100116A) at 1 µM. To inhibit the production of ER ROS, BMDMs were treated with EN460 (MedChemExpress, HY-12837) at 10 µM. To inhibit activation of NLRP3 inflammasome, BMDMs were treated with Oridonin (Selleckchem, S2335) at 2 mM. To inhibit the synthesis of triacylglyceride, BMDMs were treated with T863 (MedChemExpress, HY-32219) at 10 µM. To inhibit the synthesis of cholesterol, BMDMs were treated with Simvastatin (MedChemExpress, HY-17502) at 10 or 20 µM. These inhibitors were added to cells together with LPS. IL-1β, IL-6, and TNF-α in supernatants and sera were measured with cytometric beads array kit.

Anti-ASC (Cell Signaling Technology, 67824, 1:1,000), anti-NLRP3 (Cell Signaling Technology, 15101, 1:1,000), anti-cleaved IL-1β (Cell Signaling Technology, 52718, 1:1,000), anti-pro–IL-1β (Cell Signaling Technology, 31202, 1:1,000), anti-DGAT1 (Gene-Protein Link, 1:1,000), anti–β-actin (Cell Signaling Technology, 3700, 1:2,000), anti-pro–caspase-1/cleaved caspase-1 p20 (AdipoGen Life Sciences, AG-20B-0042, 1:1,000), anti-NEK7 (Santa Cruz, sc-50756, 1:1,000), anti-GSTO1 (Proteintech, 15124-1-AP, 1:1,000), anti-GSH (ViroGen, 101-A, 1:1,000), anti-calnexin (LifeSpan Biosciences, LS-B9772-200, 1:1,000), anti-FACL4 (ABclonal, A20414, 1:1,000), peroxidase-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch, 111-035-144, 1:5,000), anti-mouse IgG (Jackson ImmunoResearch, 111-035-146, 1:5,000), peroxidase-conjugated donkey anti-goat IgG (Abcam, ab97110, 1:5,000), and Alexa Fluor 488–conjugated chicken anti-goat IgG (Invitrogen, 21467, 1:1,000) were used in our experiments.

In coimmunoprecipitation experiments, cells were harvested and lysed with cell lysis buffer (50 mM Tris, pH 8.0, 280 mM NaCl, 0.5% NP-40, 0.2 mM EDTA, 2 mM EGTA, 10% glycerol, 1 mM dithiothreitol, and protease inhibitor cocktail), and then centrifuged at 13,000 g at 4°C for 10 min. The supernatants were incubated with Dynabeads Protein G (Invitrogen) and corresponding antibodies overnight at 4°C. The proteins in the precipitated complexes were analyzed by Western blot. To measure the glutathionylation of proteins in NLRP3 inflammasome, anti-ASC, anti-NLRP3, anti-NEK7, and anti-pro–caspase1 antibodies were used to detect corresponding proteins coimmunoprecipitated with anti-GSH antibody, respectively. To detect the oligomerization of ASC, cells were lysed with a lysis buffer (20 mM Hepes-KOH, pH 7.5, 150 mM KCl, 1% NP-40, 0.1 mM PMSF, and protease inhibitor cocktail) and centrifuged at 6,000 rpm at 4°C for 10 min. The pellets were resuspended in CHAPS buffer (20 mM Hepes-KOH, pH 7.5, 5 mM MgCl₂, 0.5 mM EGTA, and 0.1% CHAPS). The resuspended pellets were cross-linked with BS³ cross-linker (Thermo Fisher Scientific) at 4 mM concentration for 30 min at room temperature, and then were dissolved in SDS sample buffer. To detect ASC oligomerization in distinct cell fractions, mitochondria, ER, and cytosol were isolated from BMDMs using Endoplasmic Reticulum Isolation kit (Sigma-Aldrich). Cell fractions were then cross-linked with BS³ cross-linker (Thermo Fisher Scientific) as described.

BMDMs on glass-bottom dishes were washed and fixed with 4% paraformaldehyde in PBS buffer for 15 min and then permeabilized using 0.1% Triton X-100 for 15 min. After blocking with 2% BSA for 1 h, cells were incubated with primary antibodies overnight at 4°C, followed by staining with secondary antibodies. Nuclei were stained with DAPI. To stain the mitochondria, BMDMs were added to prewarmed (37°C) staining solution containing Mito-Tracker probe (Invitrogen) using a working concentration of 400 nM for 30 min under growth conditions. Cells were imaged by an LSM 880 laser scanning microscope (Zeiss).

To detect glutathionylated ASC and ASC–GSTO1 interaction in situ, PLA was performed using Duolink In situ PLA kit (Sigma-Aldrich). Briefly, BMDMs on glass-bottom dishes were fixed and permeabilized as described. Cells were then incubated with primary antibodies against the two targets at 4°C overnight. After washing, a pair of PLA probe–labeled secondary antibodies targeting the primary antibodies were added to cells to allow signal amplification only if the two targets interacted with each other. To view the ASC–ASC interaction in situ with the PLA approach, two distinct ASC-specific primary antibodies were used to target the different ASC proteins simultaneously, followed by incubation with two PLA probe–labeled secondary antibodies, and only interaction between two ASC proteins binding distinct primary antibodies could generate PLA signal. To validate the staining method, two distinct primary antibodies against ASC were added to stimulated cells sequentially, with an interval long enough to ensure blocking ASC by the first antibody, followed by incubation with two PLA probe–labeled secondary antibodies. The absence of PLA signal indicated that two primary antibodies competed for binding same ASC protein. Therefore, we could exclude the possibility of binding two distinct primary antibodies at the same ASC protein in ASC oligomers. Nuclei were stained with DAPI. Confocal image acquisition was performed using an LSM 880 laser scanning microscope (Zeiss). Quantification of PLA signal was performed with ImageJ software.

293T cells were transfected with the vector containing shRNA Dgat1 (5′-CCG​GGC​TGA​GTC​TGT​CAC​C-3′; 6 µg), shRNA Gsto1 (5′-CCG​GCC​TTC​TAT​GGT​TGA​TTA​TCT​TCT​CGA​GAA​GAT​AAT​CAA​CCA​TAG​AA-3′), and shRNA Nlrp3 (5′-CCG​GCC​GGC​CTT​ACT​TCA​ATC​TGT​TCT​CGA​GAA​CAG​ATT​GAA​GTA​AGG​CC-3′; 6 µg) along with PMD2.G (2 µg) and PsPAX2 (4 µg), respectively. The lentivirus-containing media were collected and passed through 0.45-µm filters after 48 h. BMDMs were seeded on 6-well plates with 80 × 10⁴ cells per well, and the lentivirus-containing medium was then added to cells for 48 h to knock down Dgat1, Gsto1, and Nlrp3. Scrambled shRNA (5′-TTC​TCC​GAA​CGT​GTC​ACG​T-3′) was used as a negative control.

The cryo-EM structure of human ASC-CARD filament (PDB accession number 6N1H; Li et al., 2018) was used to generate monomeric, dimeric, and 16mer filament structures of mouse ASC-CARD with the MUTATE plugin in VMD (Humphrey et al., 1996). Only the dimeric configuration with residue cysteine¹⁷¹ located at the binding interface is considered. The initial positions of GSH in monomeric and dimeric structure were obtained by docking to the vicinity of residue cysteine¹⁷¹ with Autodock Vina (Trott and Olson, 2010), and the docking result that favored the formation of cysteine-GSH disulfide bond was selected. The initial positions of GSH in 16mer filament structure were modeled from that in the docked dimeric structure. These initial models with or without GSH conjugation were then processed with VMD PSFGEN plugin to add hydrogen and other missing atoms. The resulting systems were solvated in rectangular water boxes with a TIP3P water model. K⁺ and Cl⁻ ions were then added to these solvated systems to neutralize the systems and maintain salt concentration at 150 mM.

All simulations were performed with NAMD2 (Phillips et al., 2005) software using CHARMM36m additive protein forcefield with CMAP correction (MacKerell et al., 1998) under the periodic boundary condition. The force field of GSH was generated from CHARMM general forcefield (CGenFF; Vanommeslaeghe et al., 2010). All systems underwent energy minimization with four steps: (1) 10,000 steps with the heavy atoms of proteins fixed; (2) 10,000 steps with the backbone atoms of proteins fixed; (3) 10,000 steps with the backbone atoms of proteins fixed; and (4) 10,000 steps with all atoms free. Then each system was equilibrated for 10 ns with 1-fs time steps. Finally, production simulations were performed with 2-fs time steps under rigid bond algorithms, and the snapshots were saved every 100 ps for further analysis. For each system, the temperature was controlled at 310 K by Langevin dynamics, and the pressure was maintained at 1 atm by the Nosé-Hoover Langevin piston method during the simulation. Particle mesh Ewald summation was used for computing the electrostatic interactions, and a 12-Å cutoff with 10–12-Å smooth switching was used for short-range nonbounded interactions. The system preparations, trajectory analyses, and graphic illustrations were performed with VMD.

For AIM2 inflammasome activation, BMDMs were primed with 50 ng ml⁻¹ LPS (Sigma-Aldrich, L4391) for 4 h and then stimulated by 1 µg ml⁻¹ poly(dA:dT) (InvivoGen, tlrl-patn-1) transfection for 4 h. For NLRC4 inflammasome activation, BMDMs were primed with 50 ng ml⁻¹ LPS (Sigma-Aldrich, L4391) for 4 h and then stimulated with 100 nM Flagellin for 6 h. For Pyrin inflammasome activation, BMDMs were primed with 50 ng ml⁻¹ LPS (Sigma-Aldrich, L4391) for 4 h and then stimulated with 500 ng ml⁻¹ TcdB (Abcam, ab124001) for 1 h.

Statistical analyses were performed with Mann–Whitney U test, unpaired Student’s t test, and log-rank test, using GraphPad Prism 8.0.1 software. Unpaired Student’s t test was used for normally distributed parameters in two unpaired groups, Mann–Whitney U test was used for non–normally distributed parameters in two unpaired groups, and log-rank test was used for survival analysis. *, P < 0.05, **, P < 0.01, ***, P < 0.001, and ****, P < 0.0001 were considered statistically significant.

Fig. S1 shows recovery of IL-1β production in GSTO1-deficient BMDMs by restoring GSTO1 expression. Fig. S2 shows impairment of ASC-CARD assembly by GSH conjugation. Fig. S3 shows GSH-induced reduction of ASC glutathionylation and elevation of ASC oligomerization in ML175-treated BMDMs. Fig. S4 shows the dispensable role of cholesterol synthesis in NLRP3 inflammasome activation. Fig. S5 shows the dispensable role of ASC deglutathionylation in activating AIM2 and NLRC4 inflammasomes.

This work was supported by National Natural Science Foundation of China (91942310, 81771671, and 91954122), National Key R&D Program of China (2017YFA0505300), and Fundamental Research Funds for the Central Universities (WK9110000149 and WK9100000001).

Author contributions: S. Li, L. Wang, Z. Xu, Y. Huang, R. Xue, T. Yue, L. Xu, H. Zhang, F. Gong, S. Bai, Q. Wu, J. Liu, and B. Lin performed experiments. Y. Xue, P. Xu, J. Hou, X. Yang, T. Jin, R. Zhou, and T. Xu provided materials or discussed experiments and manuscript. J. Lou performed MD simulations. S. Li and L. Bai conceived the idea, designed the experiments, and wrote the manuscript.