LRRC8A regulates hypotonicity-induced NLRP3 inflammasome activation

The NLRP3 inflammasome is a multi-molecular protein complex that converts inactive cytokine precursors into active forms of IL-1β and IL-18. The NLRP3 inflammasome is frequently associated with the damaging inflammation of non-communicable disease states and is considered a therapeutic target. However, there is much regarding the mechanism of NLRP3 activation that remains unknown. Chloride efflux is suggested as an important step in NLRP3 activation, but the identity of which chloride channels are involved is still unknown. We used chemical, biochemical, and genetic approaches to establish the importance of Cl- channels in the regulation of NLRP3 activation. Specifically we identify LRRC8A, an essential component of volume-regulated anion channels (VRAC), as a vital regulator of hypotonicity-induced, but not DAMP-induced, NLRP3 inflammasome activation. Although LRRC8A was dispensable for canonical DAMP-dependent NLRP3 activation, this was still sensitive to Cl- channel inhibitors, suggesting there are additional and specific Cl- sensing and regulating mechanisms controlling NLRP3.


Introduction
Inflammation is an important protective host-response to infection and injury, and yet is also detrimental during non-communicable diseases (1). Inflammasomes are at the heart of inflammatory responses. Inflammasomes are formed by a soluble pattern recognition receptor (PRR), in many cases the adaptor protein ASC (apoptosis-associated speck-like protein containing a CARD), and the protease caspase-1 (2). Inflammasomes form in macrophages in response to a specific stimulus to drive the activation of caspase-1, facilitating the processing of the cytokines pro-interleukin (IL)1β and pro-IL-18 to mature secreted forms, and the cleavage of gasdermin D to cause pyroptotic cell death (2). A number of different inflammasomes have been described, but potentially the inflammasome of greatest interest to non-communicable disease is formed by NLRP3 (NACHT, LRR and PYD domains-containing protein 3) (3). The mechanisms of NLRP3 activation remain poorly understood.
The NLRP3 inflammasome is activated through several routes which have been termed the canonical, non-canonical, and the alternative pathways (3). Activation of the canonical NLRP3 pathway, which has received greatest attention thus far, typically requires two stimuli; an initial priming step with a pathogen associated molecular pattern (PAMP), typically bacterial endotoxin (lipopolysaccharide, LPS), to induce expression of pro-IL-1β and NLRP3, and a second activation step usually involving a damage associated molecular pattern (DAMP), such as adenosine triphosphate (ATP) (4). In 1996, Perregaux and colleagues discovered that hypotonic shock was effective at inducing the release of mature IL-1β when applied to LPS treated human monocytes and suggested the importance of a volume regulated response (5).
It was later discovered that hypotonicity induced release of IL-1β via activation of the NLRP3 inflammasome (6), and that this was linked to the regulatory volume decrease (RVD), which is a regulated reduction in cell volume in response to hypo-osmotic-induced cell swelling, and was inhibited by the chloride (Cl -) channel blocker NPPB (5-nitro-(3-4 phenylpropylamino)benzoic acid) (6). The RVD is regulated by the Clchannel VRAC (volume regulated anion channel). The molecular composition of the VRAC channel was established to consist of an essential LRRC8A sub-unit in combination with other (B-E) LRRC8 sub-units (7,8). We recently reported that fenamate NSAIDs could inhibit the canonical NLRP3 inflammasome by blocking a Clchannel, which we suggested could be VRAC (9). We also further characterised the importance of Clflux in the regulation of NLRP3, showing that Clefflux facilitated NLRP3-dependent ASC oligomerisation (10). Given the poor specificity of many Clchannel inhibitors, we set out to systematically determine the importance of VRAC and the RVD to NLRP3 inflammasome activation. Using pharmacological and genetic approaches, we discovered that VRAC exclusively regulated RVD dependent NLRP3 activation in response to hypotonicity, and not NLRP3 activation in response to other canonical stimuli. Thus, we provide genetic evidence for the importance of Clin regulating NLRP3 via the VRAC dependence of the hypotonicity response, and suggest the presence of additional Clsensing mechanisms regulating NLRP3 in response to DAMPs.
The most favourably bound poses of these ligands were similarly found to block the pore in a cork-in-bottle manner (12) at the selectivity filter; the ligands' ionized acidic groups formed strong electrostatic interactions with Arg103 ( Figure 1D-H). Tamoxifen, a basic inhibitor of VRAC was also docked into the cryo-EM structure ( Figure 1I). Accordingly, tamoxifen docked with its cationic tertiary amino group remote to the Arg103 side-chains (Figures 1l and Supplementary Figure 1); these side-chains instead formed cation- interactions with the phenyl group of tamoxifen ( Figure 1l). The interaction of the tamoxifen pose was computed as having a calculated ligand-binding affinity of -6.5 kcal mol -1 via the molecular mechanics/generalized Born volume integration (MM/GBVI) method (15,16). The binding energies of the anionic ligands were also predicted as favourable, ranging from -5.1 (DCPIB) to -5.8 kcal mol -1 (MONNA). This range excludes the larger 4-sulfonic calix(6)arene (calixarene), which gave a binding energy of -9.9 kcal mol -1 ; we note that the GBVI implicit solvent model may be underestimating the high desolvation cost of this polyanionic ligand and therefore overestimating the magnitude of the corresponding binding energy of this compound.
We then tested the ability of six of the compounds described in Figure 1 Figure 3A). At the dose used in this assay, DCPIB did not consistently inhibit ATP-induced IL-1β release ( Figure 3A), but at higher concentrations did inhibit NLRP3 activation (Supplementary Figure 2). Likewise, pyroptosis, as measured by LDH release, was significantly reduced by MONNA and FFA, and was unaffected by tamoxifen or calixarene ( Figure 3B). The inhibitors that blocked ATP-induced IL-1β release also inhibited ASC oligomerisation, caspase-1 activation, and gasdermin D cleavage ( Figure 3C). These data show that some very effective VRAC inhibitors failed to inhibit activation of the NLRP3 inflammasome and release of IL-1β, suggesting that VRAC may not be the molecular target of these molecules inhibiting the inflammasome.
However, ASC oligomerisation, caspase-1 activation and gasdermin D cleavage induced by imiquimod were not affected by FFA and NS3728 pre-treatment ( Figure 3F). Increased extracellular KCl (25 mM) was sufficient to block nigericin-induced activation, but not imiquimod, demonstrating the K + dependency of nigericin ( Figure 3F). These data suggest that these Clchannel inhibiting compounds exclusively target the K + -dependent canonical pathway of NLRP3 activation.
Many Clchannel inhibiting drugs are known to inhibit multiple Clchannels, and we established that very effective VRAC inhibitors (tamoxifen and DCPIB) had negligible effect 8 on NLRP3 activation at VRAC inhibiting concentrations. Thus, to conclusively determine the role of VRAC in NLRP3 inflammasome activation we generated a macrophage specific LRRC8A knockout (KO) mouse using CRISPR/Cas9 ( Figure 4A). The generation of a macrophage specific LRRC8A KO was required as whole animal LRRC8A KO mice do not survive beyond four weeks and have retarded growth (19). Lrrc8a fl/fl mice were bred with mice constitutively expressing Cre under the Cx3cr1 promoter, as previously shown to be expressed in monocyte and macrophage populations (20). This generated mice with the genotype Lrrc8a fl/fl :Cx3cr1 cre (KO) with littermates Lrrc8a fl/fl :Cx3cr1 WT (WT). Cell lysates were prepared from BMDMs and peritoneal macrophages isolated from WT and KO mice and were western blotted for LRRC8A confirming that Lrrc8a KO cells were knocked out for LRRC8A ( Figure 4B). Functional loss of LRRC8A was confirmed using the calcein RVD assay described above. BMDMs were subjected to a hypotonic shock and changes in calcein fluorescence measured over time. In WT cells there was a characteristic RVD ( Figure 4C). However, in Lrrc8a KO cells there was complete loss of the RVD response ( Figure 4C, D). The absence of RVD was also strikingly evident by observation of the cells by phase contrast microscopy ( Figure 4E, Supplementary videos 1, 2). These data confirm functional KO of the VRAC channel in macrophages.
We next used the Lrrc8a KO macrophages to test the hypothesis that VRAC and the RVD were important for NLRP3 inflammasome activation and IL-1β release in response to DAMP stimulation. WT BMDMs and Lrrc8a KO BMDMs were primed with LPS (1 µg mL -1 , 4 h) and then treated with the NLRP3 inflammasome activators ATP (5 mM, 2 h), nigericin (10 µM, 2 h), silica (300 µg mL -1 , 2 h), or imiquimod (75 µM, 2 h). Knocking out LRRC8A had no effect on the release of IL-1β ( Figure 5A) or cell death ( Figure 5B). We then used western blotting to determine ASC oligomerisation and caspase-1 activation. In response to the NLRP3 inflammasome activators nigericin, ATP, and imiquimod, there was no effect of LRRC8A KO on ASC oligomerisation or caspase-1 activation ( Figure 5C). Furthermore, IL-1β release in response to ATP or nigericin was still inhibited by the VRAC inhibitors flufenamic acid (FFA, 100 µM), and NS3728 (10 µM) in the Lrrc8a KO BMDMs, confirming that these inhibitors are inhibiting NLRP3 inflammasome activation by a VRAC-independent mechanism ( Figure 5D).
Flufenamic acid and NS3728 also inhibited ASC oligomerisation and caspase-1 activation as determined by western blot in the Lrrc8a KO BMDMs to the same extent as in the WT ( Figure   5E). We then used a murine peritonitis model described previously (9) Figure 5I). Moreover, similar to our in vitro findings, NS3728 was still effective at inhibiting this response in the absence of LRRC8A ( Figure 5H, I). These data suggested that VRAC was dispensable for NLRP3 activation by DAMP stimulation, and that the VRAC inhibitors are effective at inhibiting NLRP3 in the absence of VRAC, suggesting the presence of another target.

Discussion
Pharmacological and biochemical evidence supporting an important role of Clions in the activation of the NLRP3 inflammasome has been provided by various studies over the years (5,6,9,10,21), although conclusive genetic evidence has been lacking. The promiscuous nature of many Clchannel inhibiting drugs, and an unresolved molecular identity of major Clchannels, have prevented the emergence of conclusive genetic proof. However, the discovery that the Clchannel regulating the RVD (VRAC) was composed of LRRC8 sub-units, and that LRRC8A was essential for channel activity, offered us the opportunity to investigate the direct importance of VRAC in the regulation of NLRP3. Hypotonicity induces cell swelling which is corrected by the VRAC-dependent RVD (7,8). The RVD was previously linked to NLRP3 activation (6). Thus by knocking out LRRC8A, and thus VRAC, we would discover that VRAC was essential for RVD-induced NLRP3 inflammasome activation, providing strong evidence for the direct requirement of a Clchannel in NLRP3 inflammasome activation.
However VRAC was only essential for RVD induced NLRP3 activation and was not involved in the NLRP3 response to DAMP stimulation. The fact that our VRAC channel inhibiting drugs block DAMP-induced NLRP3 activation suggests that additional Clchannels are involved in coordinating NLRP3 responses to other stimuli. Chloride intracellular channel proteins (CLICs 1-6) form anion channels and regulate a variety of cellular processes (22,23). Localisation of CLIC1 and 4 to membrane fractions in macrophages is increased by LPS stimulation, and RNAi knockdown of both CLIC1 and 4 impaired LPS and ATP-induced IL-1β release from macrophages (24). In addition to CLICs 1 and 4, CLIC5 is also implicated in NLRP3-dependent IL-1β release (25). Knockdown of CLICs 1, 4, and 5 inhibits NLRP3 inflammasome activation in response to the soluble agonists ATP and nigericin, and also the particulate DAMP monosodium urate crystals (25). Thus, it appears that multiple Clchannels encode diverse signals arising from DAMP stimulation, or from altered cellular homeostasis, to trigger NLRP3 inflammasome activation.
Inhibiting the NLRP3 inflammasome has become an area of intense research interest due to the multiple indications of its role in disease (3). The inhibitor MCC950 is now thought to bind directly to NLRP3 to cause inhibition (26,27), although it has also been reported to inhibit Clflux from macrophages treated with nigericin (28), and was found to bind directly to CLIC1 (29), so it is possible that some of its inhibitory activity may be attributable to an effect on Cl -.
We found that Clchannel inhibition blocked IL-1β release in a NLRP3-dependent model of peritonitis, and previously reported protective effects of the fenamate NSAIDs in rodent models of Alzheimer's disease that we attributed to an effect on Clchannel inhibition (9). Thus, it is possible that targeting Clchannels offers an additional route to inhibit NLRP3-dependent inflammation in disease.
In summary, we have reported that hypotonicity induced NLRP3 inflammasome activation depends exclusively on the Clchannel VRAC, and that different Clsensing and regulating

Inflammasome activation assays
Primary mOsm kg -1 ) buffer containing either drug or vehicle was then added and cells were incubated for a further 5 min. NaI (200 mM, 25 µL) was then added directly to the well, and fluorescence readings were take every 2 seconds using the FlexStation3 plate reader.

Generation of Lrrc8a fl/fl mice
We used CRISPR-Cas9 to generate the floxed LRRC8A allele on C57BL/6J background.
LRRC8A is a 4 exon gene spanning 26kb on mouse chromosome 2. Only 2 of these exons contain coding sequence, with exon 3 harbouring > 85% of the coding sequence and possessing large introns, and thus an ideal candidate for floxing. We initially attempted the 2-sgRNA, 2-oligo approach described previously (30), but failed to obtain mice with both loxP integrated on the same allele (31). Instead, a colony from a single founder with the 5' LoxP integrated was established, bred to homozygosity, and used as a background to integrate the second 3' loxP. For both steps, we used the Sanger WTSI website (http://www.sanger.ac.uk/htgt/wge/, (32)) to design sgRNA that adhered to our criteria for off target predictions (guides with mismatch (MM) of 0, 1 or 2 for elsewhere in the genome were discounted, and MM3 were tolerated if predicted off targets were not exonic). sgRNA sequences were purchased as crRNA oligos, which were annealed with tracrRNA (both oligos supplied IDT; Coralville, USA) in sterile, RNase free injection buffer (TrisHCl 1mM, pH 7.5, EDTA 0.1mM) by combining 2.5 µg crRNA with 5 µg tracrRNA and heating to 95 o C, which was allowed to slowly cool to room temperature.

Quantification and statistical analysis
Data are presented as mean values plus the SEM. Accepted levels of significance were *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Statistical analyses were carried out using GraphPad Prism (version 8). Data where comparisons were made against a vehicle control, a one way ANOVA was performed with a Dunnett's post hoc comparison was used.
Experiments with two independent variables were analysed using a two-way ANOVA followed by a Tukey's post hoc corrected analysis. Equal variance and normality were assessed with Levene's test and the Shapiro-Wilk test, respectively, and appropriate transformations were applied when necessary. n represents experiments performed on individual animals or different passages for experiments involving HeLa cells.