RAC2 gain-of-function variants causing inborn error of immunity drive NLRP3 inflammasome activation

Among the rat sarcoma (RAS)–like superfamily, 22 RAS homolog (RHO) GTPases have been identified in humans. Until now the best characterized are RHO, RAS-related C3 botulinum toxin substrate (RAC), and CDC42 (Ridley, 2006). These proteins were first described for their function as actin cytoskeleton dynamic regulators. Later they were found to be involved in critical cellular processes including cell cycle, cell growth, metabolism, as well as innate immune processes (Etienne-Manneville and Hall, 2002; Bokoch, 2005). RHO GTPases are one of the preferential targets of microbial virulence factors, probably because of their role in controlling innate immune responses (Boquet and Lemichez, 2003).

Among the RHO GTPases, three isoforms of RAC have been found in humans: RAC1, RAC2, and RAC3. While in humans RAC1 is expressed ubiquitously and RAC3 highly expressed in the brain, RAC2 expression is mostly restricted to the hematopoietic system (Burridge and Wennerberg, 2004). This restricted expression pattern suggests a specific immune function for the RAC2 GTPase isoform. Strikingly, human and mouse RAC2 proteins differ in only two amino acids (Kim and Dinauer, 2001). In vivo studies have demonstrated the RAC2-specific role in most of the cells in the hematopoietic system (Roberts et al., 1999; Yang et al., 2000; Yu et al., 2001; Troeger and Williams, 2013; Lougaris et al., 2020).

Like other RHO GTPases, RAC2 oscillates between an active GTP-bound and an inactive GDP-bound stage, which is regulated by the GTPase activating protein, guanine nucleotide exchange factor, and guanosine nucleotide dissociation inhibitor. Importantly, point mutations affecting the Switch domains of the Ras-like proteins affect the GTP cycling conferring to those mutants either a dominant loss or gain of function and were found associated with human diseases. RAC2 is following the same scheme, where single-point mutations in the G-boxes and/or the Switches domains have been described. Importantly, RAC2 mutations have been found in patients with immune disorders (Bourne et al., 1991; Ambruso et al., 2000; Accetta et al., 2011; Alkhairy et al., 2015; Caye et al., 2015; Hsu et al., 2019; Lagresle-Peyrou et al., 2021; Stern et al., 2021; Donkó et al., 2024). Clinically, RAC2 mutations frequently lead to an immune disorder associated with recurrent susceptibility to infections (Arrington et al., 2020; Donkó et al., 2024). The E62K mutation is a gain-of-function mutation localized in the G3/Switch II domain that was shown to be associated with a myeloid deficit, an altered neutrophil function, and lymphopenia (Hsu et al., 2019). RAC2 D63V is also a gain-of-function mutation, leading to a juvenile myelomonocytic leukemia, while RAC2 G12R was discovered in three patients suffering from bone marrow hypoplasia associated with a severe combined immunodeficiency syndrome of autosomal dominant inheritance. This G12R mutation with an autosomal dominant inherence has a major impact on susceptibility to infection since two G12R patients presented sepsis (Lagresle-Peyrou et al., 2021). In contrast, the D57N variant was associated with a RAC2 inhibitory effect (Ambruso et al., 2000). Located in the G3 box of RAC2, this variant caused a severe phagocyte immune deficiency. RAC2 D57N expression in macrophages inhibited membrane ruffling, inhibited the formation of macropinosomes, and induced an elongated, spread morphology (Hsu et al., 2019; Stern et al., 2021). Interestingly, among the patients with different forms of primary immunodeficiencies associated with RAC2 mutations, >75% (15/19) had a gain-of-function mutation (Stern et al., 2021).

We recently demonstrated that the RAC2 activation by the cytotoxic necrotizing factor 1 (CNF1) bacterial toxin triggered NLRP3 inflammasome activation (Dufies and Boyer, 2021; Dufies et al., 2021). This sensing mechanism was found to be critical for the innate immune response during bacteremia in mice and the signaling pathway was conserved in human macrophages (Diabate et al., 2015; Dufies et al., 2021). Mechanistically, the CNF1 induced a posttranslational modification consisting of the deamidation of glutamine 61 of RAC, converting this amino acid into a glutamate. This RAC2 Q61E protein modification within the G3/Switch II domain abolished the GTPase activity, thus confining the RAC GTPases into an active form (Boquet and Lemichez, 2003). Strikingly, this posttranslational modification is closely related to the reported RAC2 gain-of-function variants observed in patients with primary immunodeficiencies.

Here, we investigated whether RAC2 variants in patients with primary immunodeficiencies may have the potential to activate the NLRP3 inflammasome.

We noticed that some of the activating RAC2 variants are interestingly located very close to the site modified by the CNF1 toxin that catalyzes the RAC2 posttranslational modification Q61E, thereby activating the NLRP3 inflammasome (Boyer et al., 2011; Dufies and Boyer, 2021; Dufies et al., 2021). The most striking example is the RAC2 Q61R variant that affects the same residue (Stern et al., 2021). We thus investigated whether this mutant behaved similarly to the mutant mimicking the CNF1 toxin-induced Q61E modification using the GST (glutathion S transferase)–PAK (p21 activated kinase)–RBD (RAC-binding domain) pull-down assay. PAK kinases are major downstream effectors that bind specifically the active form of Rac2. We measured similar binding strength of PAK for both Q61E and Q61R variants, but not for the RAC2 D57N, a variant associated with a RAC2 inhibitory effect found in patients suffering from immunodeficiency (Fig. 1 A). We thus extended our study to neighboring mutations affecting position 62 (E62K), 63 (D63V) (Fig. 1 B), as well as mutations in position 12 (G12R and G12V) (Fig. 1 C), a position known to be important for the GTPase activity of RAC2 (Illenberger et al., 2003). We measured for all these mutants a strong GST–PAK–RBD binding capacity, confirming that these mutations correspond to a gain of function and indicating their ability to bind constitutively PAK kinases (Fig. 1).

We next investigated whether the RAC2 mutants found in genetic disorders could activate the NLRP3 inflammasome. The deamidation of the glutamine 61 of RAC2 into a glutamate triggered by the CNF1 toxin confers to RAC2 the ability to bind and activate constitutively PAK1 leading to NLRP3 phosphorylation and inflammasome activation (Dufies et al., 2021). We thus compared the capacity of RAC2 Q61E mimicking the CNF1-triggered modification to the RAC2 mutations found in immune disorders using an NLRP3 inflammasome reconstitution system. In this system, the components of the NLRP3 inflammasome (NLRP3, ASC [apoptosis-associated speck-like protein containing a CARD domain], Caspase-1) and the pro-IL-1β cytokine are transfected in a non-professional immune cell (HEK293T) lacking the NLRP3 inflammasome. We determined that the RAC2 Q61R variant was able to trigger pro-IL-1β cleavage and secretion of the mature IL-1β, indicating its capacity to activate the NLRP3 inflammasome (Fig. 2 A). Similar results were found for RAC2 E62K and D63V, whereas the RAC2 D57N variant, associated with a RAC2 inhibitory effect, was similar to the control (empty vector) condition and lower than the wild-type (WT) form of RAC2 in terms of pro-IL-1β cleavage and secretion of IL-1β (Fig. 2 B). We detected an increased IL-1β secretion triggered by the RAC2 G12R compared with both the empty vector and the RAC2 WT controls (Fig. 2 C). Similar results were found with the RAC2 G12V mutant used here as a positive control (Fig. 2 C). Altogether, these results indicated that the NLRP3 inflammasome activation depends on RAC2 activation and PAK binding levels rather than the position of the mutated residue.

Our results indicated that the NLRP3 inflammasome activation was dependent on RAC2 activation and PAK binding. We next investigated using the same system whether we could block the signaling of the RAC–PAK–NLRP3–Caspase pathway. To address this point, we used inhibitors targeting different members of this signaling pathway. IPA3 blocks RAC/PAK binding, AZ13711265 is a PAK kinase inhibitor, MCC950 is an NLRP3 inhibitor, and Emricasan is a pan-Caspase inhibitor already used in Phase III clinical trials (Natori et al., 2003; Garcia-Tsao et al., 2020). RAC2 Q61E and RAC2 G12V mutants were included as positive controls. All the inhibitors tested blocked the IL-1β secretion triggered by the gain-of-function RAC2 variants Q61R, E62K, D63V, and G12R (Fig. 3).

As previously mentioned, RAC2 expression is mostly restricted to the hematopoietic system (Burridge and Wennerberg, 2004). To further determine the consequences of RAC2 activating mutation on inflammasome activation, we directly monitored the Caspase-1 activity in blood circulating leukocytes of a patient harboring a RAC2 E62K mutation. To this aim, we used the FAM (carboxyfluorescein)–YVAD–FLICA (fluorescent-labeled inhibitors of Caspases) probe that binds to the active Caspase-1 and first analyzed the global Caspase-1 activity of circulating leukocytes by quantifying the FAM–YVAD–FLICA mean fluorescence intensity (MFI) of monocytes, granulocytes, and lymphocytes. We found in a RAC2 E62K patient an increased Caspase-1 activity specifically in monocytes and granulocytes that was not observed in lymphocytes compared with healthy controls (Fig. 4, A and B). To further demonstrate that this Caspase-1 activity was related to the inflammasome activation, we next set up an assay to analyze the speck formation, known to be specific to inflammasome oligomerization (Fig. S1). The speck quantification was primarily developed to monitor the number of cells with ASC specks by FACS using cells expressing ASC-EGFP or labeled with ASC antibodies (Sester et al., 2016). Here, we adapted this technic to further quantify by FACS directly in the blood the number of cells with FAM–FLICA specks. To this aim, we analyzed the number of cells harboring specks of activated Caspase-1 by measuring the area (A)/height (H) ratio (Fig. S1 A). We validated our approach by quantifying the increased number of FAM–YVAD–FLICA specks triggered by Nigericin, a well-known trigger of the NLRP3 inflammasome, directly in circulating monocytes of control healthy donors (Fig. S1, B and C). This Nigericin-triggered Caspase-1 speck formation was correlated with an increased FAM–YVAD–FLICA MFI in Nigericin-treated monocytes (Fig. S1, D and E). Next, we measured directly in the blood of a RAC2 E62K patient an increased number of specks in both monocytes and granulocytes compared with controls. Interestingly, the strongest signal was found in monocytes with up to 12% of monocytes presenting specks compared with a mean of 4% in the controls (Fig. 4, C–E). To further confirm the NLRP3 inflammasome activation triggered by RAC2 gain-of-function variants, we next expressed the RAC2 variants in monocyte-derived macrophages isolated from the blood of healthy donors. We expressed these mutants using a lentiviral expression system and measured the secreted level of IL-1β and IL-18, cytokines related to inflammasome activation, in primary human macrophages primed by LPS. Using these settings, we found that the overexpression of all tested gain-of-function variants induced higher levels of secreted IL-1β and IL-18 (Fig. 4, F and G). Conversely, we observed a decreasing trend in cytokines secretion associated with the expression of the D57N inactive variant of RAC2 as compared with the RAC2 WT (Fig. 4, F and G). Next, we overexpressed RAC2 variants in MOLM-13, a cell line derived from a patient with acute myeloid leukemia. To determine the transcriptional impact of this inflammasome activation, we introduced the activating mutant RAC2 E62K into MOLM-13 and compared it with MOLM-13 overexpressing RAC2 WT. For both conditions, RNA sequencing (RNAseq) analysis was performed and compared with MOLM-13 cells expressing the corresponding control lentivector. Consistent with our previous data, the RNAseq analysis of RAC2 E62K overexpressing cells showed an increased IL-1β cytokine-related gene signature (IL-1β and IL-1-RN encoding the IL-1 receptor antagonist protein), which was not observed in cells overexpressing RAC2 WT (Fig. 4 H).

During the course of our study, samples from a patient harboring a RAC2 A59S variant were addressed to us. This variant was recently clinically described and classified as a gain of function (Donkó et al., 2024). Due to the proximity to position 61 targeted by the CNF1 toxin and our findings concerning the RAC2 E62K patient, we tested whether this variant behaved like a gain-of-function mutant capable of activating the NLRP3 inflammasome. Using a GST–PAK–RBD pull-down assay, we first identified that the A59S mutant behaved like an active mutant, similar to the RAC2 E62K mutant used here as a positive control (Fig. 5 A). Next, we measured its capacity to activate the NLRP3 inflammasome in our inflammasome reconstitution system (Fig. 5 B). Consistent with our previous experiments, this newly revealed patient’s RAC2 variant showed gain-of-function effects, and the different tested inhibitors of the RAC–PAK–NLRP3–Caspase-1 pathway blocked IL-1β secretion (Fig. 5 C). We previously identified the threonine in position 659 of NLRP3 as a phosphorylation site for PAK1 that was critical for the CNF1-toxin-triggered IL-1β secretion (Dufies et al., 2021). We thus tested whether this phosphorylation site was important in the context of RAC2 A59S or E62K variants. Inflammasome reconstituted with the NLRP3 T659A mutant had an impaired IL-1β secretion compared with WT NLRP3 (Fig. 5, D and E). Next, since Caspase-1-triggered GasderminD (GSDMD) cleavage and the subsequent GSDMD pore formation have been shown to be important for IL-1β secretion (Shao, 2021), we investigated whether the RAC2 A59S or E62K variant–triggered IL-1β secretion was GSDMD dependent. To address this point, we took advantage of the U937 monocytic cell line. Consistent with what we observed in primary macrophages, the expression of RAC2 A59S or E62K variants in control U937 cells stimulated with LPS was sufficient to trigger IL-1β secretion (Fig. 5 F). Using U937 NLRP3 knock-out (KO) cells, we could confirm that the RAC2 A59S or E62K variant–triggered-IL-1β secretion was NLRP3 dependent. Additionally, we measured an impaired IL-1β secretion in GSDMD KO cells (Fig. 5 F). We next investigated the involvement of pyroptotic cell death in this phenomenon by investigating the impact of Ninjurin1 (NINJ1), a protein required for pyroptosis (Kayagaki et al., 2021; Fu et al., 2024). Using U937 NINJ1 KO cells, we observed that the IL-1β secretion triggered by RAC2 gain-of-function variants was not affected (Fig. S2 A). In addition, we did not measure a significant increase in cell death triggered by the expression of RAC2 A59S or E62K in U937 cells as well as in U937 cells KO for NLRP3, NINJ1 (Fig. 5 G), or GSDMD (Fig. S2 B). To investigate in depth the impact of the activating variant RAC2 A59S, we performed a single-cell RNAseq analysis of blood circulating cells (Fig. 5 H). This analysis showed increased numbers of both classical and non-classical monocytes as well as myeloid dendritic cells when compared with a healthy control (Fig. S3). In addition, and supporting our hypothesis, NLRP3 and IL-1β expression levels were increased in both monocytes and myeloid dendritic cells, while their expression in lymphocytes was not affected in our analysis (Fig. 5 I).

To further confirm our data, we compared monocytes isolated from peripheral blood mononuclear cells (PBMCs) of patients harboring the RAC2 A59S with monocytes isolated from RAC2 E62K mutation (Fig. 6 A). In both cases, the PBMCs were collected and frozen using the same protocol and PBMCs from patients and control healthy donors were processed in parallel. The RAC2 A59S and RAC2 E62K patient’s monocytes analysis showed, respectively, 11.4% and 14.3% of FAM–YVAD–FLICA positive specks while control monocytes isolated from healthy donors showed only 4%. We could repeat these experiments twice for blood from the same patients taken at different time points and observed similar results with an increased number of specks of activated Caspase-1 in RAC2 mutant patients compared with controls (Fig. 6 B). The Caspase-1 activation was also visualized by measuring the FAM–YVAD–FLICA MFI (Fig. 6 C). Importantly, we were able to differentiate monocytes isolated from the blood of both RAC2 A59S and RAC2 E62K variant patients into macrophages. For both RAC2 variants, macrophages treated with LPS showed cells positive for ASC specks colocalizing with NLRP3 while no ASC speck was found in macrophages treated with LPS isolated from healthy donors used as controls (Fig. 6 D). The analysis of the cytokines secreted by both RAC2 A59S or RAC2 E62K patient’s monocyte-derived macrophages treated with LPS showed an increased IL-1β secretion that was drastically reduced upon NLRP3 inhibitor (MCC950) treatment (Fig. 6, E and F) in accordance with our data generated in the NLRP3 inflammasome reconstitution experiments (Fig. 3 C and Fig. 5 C).

The precise mechanism of the immune phenotype associated with RAC2 variants in patients is still under debate especially because RAC2 is involved in several crucial cellular mechanisms (Lougaris et al., 2020). The most intriguing point is that both loss-of-function and gain-of-function variants show a large spectrum of immune phenotype (Hsu, 2023). Here, we show that several RAC2 gain-of-function mutations trigger NLRP3 inflammasome activation in a human cell reconstitution system as well as in monocytes and macrophages isolated from RAC2 A59S or RAC2 E62K patients. We show that this activation correlates with the capacity of the mutant proteins to bind the RBD of PAK. Furthermore, this RAC2-triggered activation of the NLRP3 inflammasome was blocked by the addition of PAK, NLRP3, or Caspase inhibitors.

Our major observation is that specific genetic RAC2 variants found in a monoallelic status in patients behave like the CNF1 toxin-triggered RAC2 posttranslational modification that occurs physiologically during bacteremia in mice. We discovered that the CNF1 toxin from uropathogenic Escherichia coli, by deamidating the Glutamine 61 of RAC2 into a glutamate, abolishes the GTPase activity converting the RHO GTPase into its constitutive active form and activates the NLRP3 inflammasome (Boyer et al., 2011; Dufies et al., 2021). Here, we show that RAC2 gain-of-function variants in patients activates the same signaling pathway, which subsequently leads to NLRP3 inflammasome activation. There is a major difference in the time frame of RAC2 activation in patients vs. RAC2 activation by CNF1. While in patients, RAC2 activation occurs as soon as the RAC2 variant is expressed in cells, CNF1-triggered RAC2 activation predominantly occurs in mature immune cells and only rarely in precursors. Thus, it will be interesting to determine the impact of the NLRP3 activation triggered by RAC2 variants during myelopoiesis, its potential consequences on the development of myeloid cells, and more broadly on hematopoiesis.

An interesting hypothesis raised by our observation is that the inflammasome activation may be dependent on the expression of the RAC2 variant but also PAK, NLRP3, and Caspase-1. According to the expression level of these factors, the phenotype of myeloid cells could lead to secretion of IL-1β only or be combined with induction of pyroptotic cell death (Paerewijck and Lamkanfi, 2022). Altogether, our data suggest that the patient’s phenotype could range from immune deficiency associated with leukocyte pyroptotic cell death to inflammation associated with inflammasome dysregulation.

These different phenotypes might be determined by various parameters including the variants’ capacity to activate PAK as well as the level of expression of the members of the RAC–PAK–NLRP3 inflammasome pathway. A more precise immunophenotyping of the RAC2 patients together with markers for pyroptotic cell death will be necessary. The difficulty of these studies resides in the fact that RAC2 variants are rare, mostly found in young children with a low leukocyte count. Here, we show the involvement of GSDMD in the RAC2 A59S and RAC2 E62K triggered IL-1β secretion in U937 monocytic cells, which is not associated with cell death. It would be particularly interesting to consider the use of Gasdermin inhibitors to treat these patients (Shao, 2021). Additionally, this strategy might be considered for patients with a loss-of-function mutation in CDC42, which activates the PYRIN inflammasome and triggers a GSDMD–dependent IL-1β secretion (Nishitani-Isa et al., 2022; Spel et al., 2022). Given that 22 RHO GTPases have been identified in humans, it is likely that further studies will reveal additional inborn errors in other RHO GTPases activating other inflammasomes, further emphasizing the importance of RHO GTPases in inflammasome regulation and more widely in immunity.

Altogether, our data suggest that patients harboring RAC2 gain-of-function mutations might develop a new type of inflammasome-related disease. In this frame, it would be interesting to determine similarities with cryopyrin-associated periodic syndrome patients that have activating NLRP3 mutations. Furthermore, it would be important to consider in the future the therapeutical potential of NLRP3 inhibitors or other inhibitors targeting the RAC–PAK–NLRP3 inflammasome pathway for patients. Inhibiting the RAC2 mutant–triggered inflammasome activation might not be sufficient to cure these patients since RAC2 is important for cellular homeostasis by regulating the actin cytoskeleton as well as the NADPH oxidase. As an example, the lymphopenia found in patients might be related to a role of RAC2 in the lymphocyte differentiation and not related to inflammasome dysregulation but could also be linked to the perturbation of IL-1β secretion in the hematopoietic niche. Further studies will be necessary to determine the contribution of the inflammasome pathway in this RAC2 gain-of-function–related disease and whether NLRP3 might be considered a new target for treatment.

The study was conducted in accordance with French legislation and the principles of the Declaration of Helsinki. Informed consent was obtained from the patients’ parents or legal guardians, and the study protocol was approved by the regional independent ethics committee and the French Ministry of Research (2015-01-05 MS2; DC-2020-3994), as well as the French Advisory Committee on Data Processing in Medical Research (Comité Consultatif sur le Traitement de l’Information en matière de Recherche dans le domaine de la Santé, Paris, France; 15.297bis). Human blood from healthy donors was obtained from the Etablissement Français du Sang (13-PP-11/CCTIRS N°14.266).

HEK293T cells were obtained from American Type Culture Collection (ATCC) (CRL-3216) and maintained according to the ATCC instructions. The human acute monocytic leukemia cell line MOLM-13 was maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (Thermo Fisher Scientific). U937 monocytic cells lines were kindly provided by Thomas Henry (Centre International de Recherche en Infectiologie, Lyon, France). The different U937 cell lines were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum and differentiated with 10 ng/ml of Vitamin D3 for 72 h. U937 KO for NLRP3 or GSDMD were previously described (Lagrange et al., 2018). For generating the U937 NINJ1 KO, the single guide RNA (sgRNA) targeting NINJ1 were selected from the Brunello Library (Addgene) and cloned into the pKLV-U6gRNA(BbsI)-PGKpuro2ABFP vector (a gift from Kosuke Yusa; plasmid #50946; Addgene). The two sgRNAs used are listed in Table S2. sgRNA-expressing lentiviral particles were produced in 293T and used to transduce a Cas9-expressing U937 Clone by spinoculation as previously described (Lagrange et al., 2018). A polyclonal population was selected with 2 µg/ml puromycin (Sigma-Aldrich) at 72 h after transduction for 2 wk. Gene invalidation was verified by PCR sequencing followed by analysis using ICE CRISPR analysis tool (KO score 100; 100% indels). PBMCs, including monocytes and lymphocytes, were isolated using Ficoll (Cytiva), and red cells were lysed using lysing buffer (BD Biosciences). Then, monocytes were purified from PBMCs using anti-CD14 microbeads (Miltenyi) and autoMACS Pro Separator (Miltenyi) according to the manufacturer’s instructions. Both PBMCs and monocytes were cultivated in RPMI 1640 medium with glutamax-I (Life Technologies) supplemented with 10% (vol/vol) fetal bovine serum (Life Technologies). Macrophage differentiation was induced by addition of CSF-1 (100 ng/ml, Miltenyi) into the culture medium for 6 days. Inhibitors and reagents used in this study are 10 µM CP-456773 or MCC950 (Sigma-Aldrich), 5 µM IPA-3 (Tocris), 1 µM AZ13711265 (AGV Discovery), 5 µM Emricasan (Sigma-Aldrich), 10 ng/ml 1alpha,25-Dihydroxyvitamin D3 (Sigma-Aldrich), 100 ng/ml ultrapure LPS (Invivogen), and 5 μM Nigericin (Invivogen).

Human RAC2 mutations (G12R, G12V, D57N, A59S, Q61E, Q61R, E62K, and D63V) were obtained using the QuickChange Site-Directed Mutagenesis Kit (Agilent Technologies) using the pCDNA3.1-HA3-RAC2 as a template and using the primers listed in Table S1. Human RAC2 was subcloned from pCDNA3.1 RAC2 using InFusion HD Cloning Kit (Takara Bio) into the HIVSFFV-IRESGFP lentiviral vector construct previously described (Girard-Gagnepain et al., 2014). Briefly, human RAC2 WT and mutants were PCR amplified using primers oligo 1F and oligo 2R. PCR products were further cloned into HIVSFFV-IRESGFP previously digested using BamHI/XhoI restriction enzymes. All plasmid constructs were verified by sequencing (Eurofins Genomics). For HA-RAC2 lentiviral particles production, HEK293T cells were cotransfected using calcium phosphate precipitation with pHIVSFFV-GFP-HA-RAC2 WT or mutants, PAX2 (gift from Malin Parmar, #35002; Addgene) and pCMV-VSV-G (gift from Bob Weinberg, #8454; Addgene). Supernatants containing lentiviral particles were collected 48 h after transfection and filtered through a 0.45-μm pore-size filter before overnight centrifugation at 3,000 g, 4°C. The supernatant was carefully removed by aspiration to obtain 100–150× concentration. To determine the titer of the lentiviral vectors, HEK293T cells were infected with increasing dilutions of lentiviral particles for 72 h and GFP expression was detected by FACS. To transduce macrophages, human monocytes were seeded in 24-well plates in complete RPMI medium and differentiated into macrophages by adding CSF-1 for 6 days. Cells were then coinfected for 72 h with a multiplicity of infection of 10 particles/cell for HIVSFFV-HARAC2-IRES-GFP lentiviral vector and an equivalent quantity of Vpx-VLP particles as previously described (Berger et al., 2011). U937 cell lines were transduced using the same protocol. To induce IL-1β secretion, fresh medium containing LPS was added for 8 h. Supernatants were collected for cytokines assays.

Total protein extracts were prepared by lysing cells with LDS sample buffer (Pierce) supplemented with 50 mM DTT (Euromedex). Lysates were then fractionated on poly-acrylamide gels using SDS-PAGE, transferred to nitrocellulose membranes (BioRad), and incubated with specific antibodies. Immobilon Western Chemiluminescent HRP Substrate (Millipore) and a PXi4 GeneSys imaging system (Syngene) were used for detection. The primary antibodies used for immunoblotting were mouse anti-Caspase-1 (AG-20B-0042, clone Casper-1; Adipogen), mouse anti-NLRP3 (clone Cryo-2; Adipogen), mouse anti-β-actin (AG-20B-0014, AC-74; Sigma-Aldrich), mouse anti-Flag (F1804, clone M2; Sigma-Aldrich), mouse anti-GFP (11814460, clone 7.1, 13.1; Roche), mouse anti-HA (901515, clone 16B12; BioLegend), and mouse anti-GST (2624, clone 26H1; CST). The secondary antibody was a polyclonal goat anti-mouse immunoglobulins conjugated with HRP (P044701-2; Agilent Technologies).

HEK293T cells were transfected using Lipofectamine 2000 (Life Technologies) with pCDNA3.1-HA3 plasmids encoding RAC2 (WT) and the constitutively active mutants as indicated in the figure legend. 16 h after transfection, cells were lysed at 4°C using a lysis buffer (Tris 25 mM pH 7.5, NaCl 150 mM, MgCl₂ 5 mM, TritonX100 0.5%, glycerol 4%) and pull-down assays were performed on 1 mg of proteins lysate using 30 μg of GST-PAK 70–106 (GST–PAK–RBD). Total and activated HA-RAC2 were revealed by immunoblotting anti-HA. Equal amounts of proteins engaged in the pull-down assays were confirmed by immunoblotting anti-β-actin. Equal amounts of GST–PAK–RBD proteins engaged in the pull-down assays were confirmed by immunoblotting anti-GST.

HEK293T cells were transfected using Lipofectamine 2000 with plasmids encoding the NLRP3 inflammasome components as previously described (Shi et al., 2016). Briefly, HEK293T cells were transfected for 16 h with plasmids encoding myc-NLRP3 (WT or T659A), ASC-GFP, pro-Caspase-1, and pro-IL-1β-Flag, and various HA-RAC2 mutants. The monitoring of IL-1β cleavage was performed using supernatant immunoblotting. Equal expression of transfected proteins in the cell lysates was confirmed by immunoblotting.

Cells were fixed in 4% paraformaldehyde (PFA) for 15 min, PFA was neutralized with 50 mM NH4Cl for 15 min, and cells were permeabilized with 0.5% Triton X-100 for 5 min and blocked with 2% TBS-BSA. Cells were incubated with mouse anti-NLRP3 (AG-20B-0014, clone Cryo-2; Adipogen) and rabbit anti-ASC (AG-25B-0006; Adipogen) antibodies for 1 h followed by incubation with secondary antibodies TexasRed anti-mouse IgG (TI-2000; Vector Laboratories), Alexa Fluor 488 anti-rabbit IgG (A21206; Life Technologies), Phalloidin iFluor 647 conjugate (ab176759; Abcam), and Hoechst 33342 (H1399; Thermo Fisher Scientific) for 30 min. Cells were imaged using a Nikon A1R confocal microscope.

IL-1β secretion in differentiated macrophages or in U937 cell lines was induced with 100 ng/ml LPS for 8 h. Human IL-1β and human total IL-18 cytokine concentrations in cell supernatants were determined by DuoSet ELISA (DY201 and DY318, respectively) according to the manufacturer’s instructions (R&D Systems). Briefly, after overnight coating of capture antibody on 96-well microplates, supernatants and standards were added for 2 h, followed by a 2-h incubation with detection antibody. Streptavidin-HRP was then added for 20 min before incubation in substrate solution for another 20 min. Optical density was determined immediately after addition of stop solution using a microplate reader set to 450 and 540 nm to correct optical imperfections in the plate. The concentrations of cytokine were then determined using the generated standard curve.

The culture medium of LPS-stimulated U937 cell lines was collected and centrifuged at 300 g for 5 min to remove cellular debris. Lactate dehydrogenase (LDH) measurement was performed using the LDH Cytotoxicity Assay Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Data were plotted as the percentage of cytotoxicity considering a TritonX100-treated well as 100%.

Whole blood or PBMCs isolated from healthy donors or from patients with RAC2 mutation were used. Caspase-1 activation was detected using the FAM–YVAD–FLICA probe according to the manufacturer’s instruction (ICT098; Bio-Rad). Cells were collected and analyzed by flow cytometry using a Cytek Aurora cytometer (Cytek). Cytometry data were analyzed using FlowJo v.10.8.1. Doublets were excluded using a side scatter SSC-A (area) and SSC-H (height) plot. For whole blood, cells were incubated for 10 min with antibodies (APC-Vio 770 anti-CD10 (130-114-505; Miltenyi Biotec), Vioblue anti-CD14 (130-110-524; Miltenyi Biotec), PE anti-CD16 (130-113-393; Miltenyi Biotec), APC anti-CD45 (130-110-633; Miltenyi Biotec), and PE-Vio 770 anti-CD66b (130-119-768; Miltenyi Biotec) before red blood cells lysis. For PBMCs, Vioblue anti-CD14 (130-110-524; Miltenyi Biotec) antibody was incubated for 10 min with cells. Cells were then analyzed for Caspase-1 activation and formation of specks by checking the FAM–FLICA probe signal area (Caspase-1-A) and height (Caspase-1-H). Cells positive for Capase-1 specks were defined with a high Caspase-1-A/Caspase-1-H ratio as compared to control cells.

Overnight cultured MOLM-13 cells were transduced during 7 h in the presence of lentiboost (Sirion) and with the appropriated lentiviral supernatant at a multiplicity of infection of 30. 2 days after the transduction step, cells were collected and pellet processed for transcriptomic analysis. The lentiviral constructs contain a GFP reporter gene allowing to check whether >92% of the cells were transduced. For single-cell RNAseq, blood samples were processed for Chromium Single Cell Gene Expression Flex analysis according to the manufacturer’s protocol. Whole blood cells from control (n = 8,078 cells) and RAC2 A59S patient (n = 14,524 cells) were analyzed. Raw sequencing data were processed using the 10× Chromium CellRanger “multi” analysis pipeline (version 7.0.0). Reads were aligned to the human reference genome (GRCh38-3.0.0) (10x Genomics). Rstudio (version 4.3.1) and Seurat (version 5.0) were used to merge, scale, and normalize gene expression data, as well as for clustering, differential gene expression analysis, and visualization. Raw data are accessible through the following link: https://doi.org/10.5061/dryad.p8cz8w9zx (Batistic and Boyer 2024). We used scType Cell Marker Database for cell-type annotation.

Statistical analyses were performed using GraphPad Prism v.9.5.0.

Fig. S1 shows gating strategies and validation of the speck gating protocol. Fig. S2 shows NINJ1-independent and GSDMD-dependent IL-1 β secretion triggered by RAC2 A59S and E62K. Fig. S3 shows an increased number of circulating monocytes and dendritic cells in the blood of the RAC2 A59S variant patient. Table S1 lists the primers used to generate RAC2 mutants. Table S2 shows sgRNA sequences used to generate the U937 491 NINJ1 KO. Data S1 is the dataset used to generate Fig. 4 F.

Data are available in the article itself and its supplementary materials. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. The sequencing data supporting the conclusions of this article have been deposited in Dryad and are accessible through the following link: https://doi.org/10.5061/dryad.p8cz8w9zx.

We thank staff from the C3M and the UCAGenomiX core facilities. We thank Thomas Henry, Pierre-Simon Rohrlich, Bernard Mary, Valentine Freschi, Guillaume Beaumont, Isabelle André, and Marie Irondelle for providing tools, technical assistance, and fruitful discussion. We acknowledge the Office of International Scientific Visibility of Université Côte d’Azur for professional language editing. We acknowledge the facilities of SFR Biosciences (UAR3444/Centre National de la Recherche Scientifique, US8/Institut National de la Santé et de la Recherche Médicale, École normale supérieure de Lyon, Université Claude Bernard Lyon) especially Gisèle Froment, Didier Nègre, and Caroline Costa, and the AniRA lentivectors production facility from the CELPHEDIA Infrastructure.

This work was supported by grants from Fondation pour la Recherche Médicale (EQU202403018053), the Agence Nationale de la Recherche (ANR) (ANR-17-CE15-0001, ANR-22-CE15-0032, ANR-22-CE14-0003), Investments for the Future programs LABEX SIGNALIFE ANR-11-LABX-0028-01, IDEX UCAJEDI ANR-15-IDEX-01, Fondation ARC (RAC15014AAA), Université Côte d’Azur. The authors acknowledge the financial support from the France Génomique National Infrastructure, funded as part of “France 2030” program managed by the ANR (ANR-10-INBS-09). The platform is supported by 3IA@cote d’azur (ANR-19-P3IA-0002), Conseil départemental 06, and Canceropole PACA. Part of this work was supported by grants from the French government managed by the ANR under the France 2030 program (ANR-23-IAHU-0007), “Investissments d’avenir” program (ANR-10-IAHU-01), and grant ANR-19-CE17-0012 (ANR-AID) as well as state funding from the ANR under ANR-PRTS 2013 “Immune-Rep,” and E-Rare Joint Transnational Call for Proposals 2015 “EuroCID.” P. Bronnec is supported by a grant from ANR ANR-21-CE17-0046 (SolvingMEFVariants).

Author contributions: A. Doye: Formal analysis, Investigation, Methodology, Validation, Visualization, Writing—review & editing, P. Chaintreuil: Formal analysis, Investigation, Methodology, Supervision, Validation, Visualization, Writing—review & editing, C. Lagresle-Peyrou: Formal analysis, Resources, Writing—review & editing, L. Batistic: Data curation, Formal analysis, Investigation, Visualization, Writing—original draft, Writing—review & editing, V. Marion: Methodology, P. Munro: Investigation, Methodology, C. Loubatier: Investigation, Methodology, Resources, R. Chirara: Formal analysis, Methodology, N. Sorel: Investigation, B. Bessot: Investigation, P. Bronnec: Resources, J. Contenti: Writing—review & editing, J. Courjon: Writing—review & editing, V. Giordanengo: Conceptualization, Resources, Supervision, Validation, A. Jacquel: Resources, P. Barbry: Methodology, Software, Validation, Writing—review & editing, M. Couralet: Investigation, N. Aladjidi: Resources, A. Fischer: Data curation, Validation, M. Cavazzana: Conceptualization, Resources, Supervision, Validation, Writing—review & editing, C. Mallebranche: Resources, O. Visvikis: Investigation, Writing—review & editing, S. Kracker: Formal analysis, Funding acquisition, Resources, Writing—review & editing, D. Moshous: Investigation, Resources, E. Verhoeyen: Conceptualization, Formal analysis, Methodology, Supervision, Writing—review & editing, L. Boyer: Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing.