A TAK1-Driven NLRP1 Inflammasome Pathway Revealed by Phosphatase-Targeting Environmental Toxins

51 Human NLRP1 is highly sensitive to phosphorylation-dependent activation, indicating a need for 52 tight phosphatase control to prevent excessive inflammasome activity. In this study, we 53 identified the phosphatases PP1 and PP2A as negative regulators of the NLRP1 inflammasome 54 in human keratinocytes. Accordingly, exposure to the environmental toxins Dinophysistoxin, 55 Okadaic acid, and Cantharidin, which inhibit PP1 and PP2A, triggered NLRP1 inflammasome 56 activation. Notably, this toxin-induced activation process relied on hyperactivation of the MAP3 57 kinase TAK1. Mechanistically, both TAK1 and its downstream effectors p38 kinases 58 phosphorylated and activated the NLRP1 inflammasome. Further studies underscored that TAK1 59 also contributed to the NLRP1 inflammasome activation during double-stranded RNA 60 stimulation and viral infection. Finally, human native skins exposed to environmental toxins 61 underlined the role of the PP1/PP2A-regulated TAK1/p38 axis in the development of skin 62 dermatitis. Thus, these findings reveal a novel pathway of phosphorylation-driven NLRP1 63 activation and expand our understanding of its regulation in epithelial immunity. 64 65 One Sentence Summary: PP1/PP2A phosphatases restrict TAK1-driven NLRP1 inflammasome 66 response. 67 68 69 70 71 72 73 74 75 76 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 4 77 Graphical abstract. Schematic representation of the phosphorylation- driven hNLRP178 inflammasome response upon exposure to environmental toxins and to dsRNA/viral infection. 79 80 81 82 83 84 85 86 87 88 89 P1 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 5

90 To cope with an environment filled with living microorganisms and (non)-biological threats such 91 as tissue injury, UV radiation, or chemical exposure, the innate immune system has evolved a 92 diverse repertoire of pattern-recognition receptors (PRRs) ( 1, 2). These receptors, expressed 93 across numerous cell types, play a central role in coordinating both inflammatory and non-94 inflammatory defenses ( 2, 3). Among PRRs, the inflammasome-forming sensors constitute a 95 specialized family capable of coupling the detection of danger signals to both cytokine release 96 and programmed inflammatory cell death (1, 2). 97 Inflammasomes are cytosolic signaling complexes that, upon activation by pathogens or cellular 98 stress, induce pyroptotic cell death and the secretion of the pro-inflammatory cytokines IL-1 β 99 and IL-18 ( 1). This process requires the protease caspase-1, which cleaves and activates IL-100 1β /IL-18 and processes gasdermin D to form membrane pores, ultimately culminating in 101 pyroptosis and Ninjurin-1–mediated cell lysis ( 1, 4–8). Although inflammasome activation is 102 essential for antimicrobial defense, dysregulation can drive chronic autoinflammation and severe 103 tissue injury (2). 104 Human NLRP1 is distinctive among inflammasome sensors because of its robust expression in 105 epithelial tissues, including keratinocytes, the airway epithelium, and the cornea ( 9). A hallmark 106 of NLRP1 activation is the release of its C-terminal CARD-containing fragment following 107 proteasome-dependent degradation of its N-terminal region, a mechanism termed “functional 108 degradation.” The liberated CARD fragment then oligomerizes to recruit and activate caspase-1 109 (10). 110 Recent studies have established NLRP1 as a versatile detector of redox and proteotoxic stress 111 (11–13), as well as viral proteases from picornaviruses and coronaviruses ( 14–16). Moreover, 112 NLRP1 is activated by ZAK α - and p38-driven phosphorylation in response to ribotoxic stress 113 response triggered by viral double stranded RNA, bacterial toxins, and environmental insults ( 9, 114 17–22). 115 Given the high sensitivity of hNLRP1 to phosphorylation-induced activation, we hypothesized 116 that cellular phosphatases might act as critical brakes to prevent uncontrolled NLRP1 signaling 117 (23–25). In this study, we identified the serine/threonine phosphatases PP1 and PP2A (23–25) as 118 key negative regulators of the NLRP1 inflammasome in human keratinocytes. Consistent with 119 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 6 this, exposure to environmental toxins such as dinophysistoxin, okadaic acid, and cantharidin, all 120 potent inhibitors of PP1 and PP2A, triggered robust NLRP1 activation. We found that this 121 response was driven by hyperactivation of the MAP3K TAK1. Mechanistically, both TAK1 and 122 its downstream effectors, the p38 MAP kinases, phosphorylated and activated hNLRP1. 123 Furthermore, we also unveiled that TAK1 contributed to NLRP1 activation in response to double 124 stranded RNA and viral infection, suggesting that this pathway extends beyond phosphatase 125 inhibition. Finally, experiments using human skin explant models exposed to environmental 126 toxins demonstrated that the PP1/PP2A-TAK1-p38 axis plays a central role in promoting toxin-127 induced dermatitis. Together, our findings uncover a previously unrecognized mechanism in 128 which inhibition of PP1/PP2A phosphatases unleashes a TAK1/p38-driven phosphorylation 129 cascade that activates the NLRP1 inflammasome. 130 131

132 133 A pharmacological screen identifies PP1/PP2A phosphatase inhibitors and toxins as 134 inducers of the human NLRP1 inflammasome 135 Given the high sensitivity of NLRP1 to phosphorylation-dependent activation, we hypothesized 136 that multiple phosphatases might tightly regulate the phosphorylation/dephosphorylation 137 dynamics governing hNLRP1 inflammasome activity. To explore this, we performed a 138 pharmacological screen coupled with fluorescence microscopy in HEK293 reporter cells 139 expressing hNLRP1 and ASC-GFP (HEK293 ASC-GFP/NLRP1). Cells were exposed to a small 140 library of phosphatase inhibitors, and inflammasome activation was quantified 8 hours later by 141 assessing ASC-GFP speck formation (Fig. 1A, B). Among all tested inhibitors, ten compounds 142 induced detectable hNLRP1 inflammasome assembly (Fig. 1A, B and Fig. S1A). 143 To validate these findings in a more physiological model, we treated WT and NLRP1-deficient 144 human TERT keratinocytes with the ten candidate compounds and measured IL-1 β release as a 145 readout of inflammasome activation (Fig. 1B). Eight of the ten inhibitors reproducibly induced 146 NLRP1-dependent IL-1β secretion. Analysis of their described molecular targets revealed that 147 seven compounds inhibit the PP1/PP2A family of serine/threonine phosphatases, whereas the 148 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 7 remaining three target the tyrosine phosphatase SHP2 (Fig. 1A, B). Although the potential 149 involvement of tyrosine phosphatases in NLRP1 regulation is intriguing, the strong enrichment 150 of PP1/PP2A-targeting compounds prompted us to focus on these phosphatases. 151 Among the identified PP1/PP2A-targeting compounds, were the environmental toxins 152 dinophysistoxin-1, okadaic acid, and calyculin A ( 26), and our bibliographic study added the 153 blister beetle toxin cantharidin ( 27), all able to induce skin dermatitis ( 28, 29). In this context, 154 using a panel of PP1/PP2A-targeting toxins, including the algal toxins dinophysistoxin-1, 155 okadaic acid, and calyculin A, and the blister beetle toxin cantharidin, we confirmed that all 156 these molecules robustly activated the NLRP1 inflammasome in human keratinocytes and in 157 reporter cells, as evidenced by ASC speck assembly, IL-1 β release, and gasdermin-D processing 158 (Fig. 1A–C, Fig. S1B-D). 159 Finally, in order to strengthen these observations, we genetically disrupted several catalytic 160 subunits of PP1 and PP2A (PP1 C α /Cγ and PP2A C α /Cβ ) in WT and NLRP1-KO TERT 161 keratinocytes using CRISPR–Cas9 (Fig. 1D) ( 23–25). Loss of PP1 C α or PP2A C α /Cβ were 162 sufficient to induce IL-1 β release in an NLRP1-dependent manner, indicating that multiple 163 PP1/PP2A phosphatase subunits could contribute to restraining NLRP1 inflammasome 164 activation. 165 Altogether, our results identify PP1 and PP2A phosphatases as key toxin-sensitive negative 166 regulators of the NLRP1 inflammasome in human keratinocytes. 167 168 PP1/PP2A-targeting toxins activate the NLRP1 inflammasome in a ZAK α -independent 169 manner 170 The robust NLRP1-dependent response induced by PP1/PP2A inhibition suggested that 171 PP1/PP2A phosphatases may normally restrain one or more kinases responsible for hNLRP1 172 phosphorylation and activation. hNLRP1 phosphorylation occurs within its disordered region 173 (DR, aa 86–275), which contains multiple serine and threonine residues highly sensitive to 174 phosphorylation, an essential step for phosphorylation-driven NLRP1 inflammasome activation 175 (Fig. 2A) (17, 18, 20–22). To investigate whether PP1/PP2A-targeting toxins promote 176 phosphorylation of this region, we used HEK293 cells expressing the hNLRP1 DR fused to GFP 177 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 8 (hNLRP1-DR–GFP) and assessed its phosphorylation status following toxin or compound 178 treatment (Fig. 2B). Anisomycin, a known inducer of hNLRP1 phosphorylation via the ribotoxic 179 stress response, served as a positive control. 180 Phos-tag analysis revealed that all PP1/PP2A-targeting toxins and compounds triggered marked 181 phosphorylation of hNLRP1-DR–GFP (Fig. 2B). These findings were corroborated in human 182 TERT keratinocytes, in which full-length hNLRP1 displayed time-dependent phosphorylation 183 following exposure to dinophysistoxin-1 (Fig. 2C). In contrast, Val-boroPro (VbP), a 184 phosphorylation-independent activator of NLRP1, did not induce detectable hNLRP1 185 phosphorylation (Fig. 2C). Together, these results indicate that PP1/PP2A inhibition unleashes 186 strong hNLRP1 phosphorylation, likely through activation of one or more yet-unidentified 187 protein kinases. 188 Because the stress-activated kinase ZAKα was recently shown to drive hNLRP1 phosphorylation 189 and activation in response to ribotoxic stress, we examined whether ZAKα contributes to NLRP1 190 activation induced by PP1/PP2A-targeting toxins. WT, ZAK α KO, and NLRP1 KO TERT 191 keratinocytes expressing a SNAP-tagged hNLRP1-DR (DR-SNAP) were treated with 192 dinophysistoxin-1, cantharidin, or VbP (Fig. 2D) ( 21). DR-SNAP phosphorylation was 193 comparable in WT and ZAK α KO cells after toxin treatment, whereas VbP did not induce 194 significant phosphorylation, demonstrating that toxin-induced NLRP1 phosphorylation occurs 195 independently of ZAKα (Fig. 2D). 196 These results were supported by measurements of plasma membrane permeabilization (Sytox 197 Green uptake) and IL-1 β secretion in WT, ZAK α KO, and NLRP1 KO keratinocytes following 198 PP1/PP2A inhibition (Fig. 2E). While anisomycin induced ZAK α -dependent IL-1β release, all 199 PP1/PP2A-targeting toxins triggered IL-1 β secretion and membrane permeabilization 200 independently of ZAK α (Fig. 2E). Similarly, ASC speck formation in WT and ZAK α KO 201 HEK293 ASC-GFP/NLRP1 cells showed that dinophysistoxin induces robust speck assembly 202 regardless of ZAKα status (Fig. 2D). 203 Altogether, these results demonstrate that PP1/PP2A-targeting toxins drive strong 204 phosphorylation of the NLRP1 disordered region, and that this process occurs independently of 205 the stress kinase ZAKα . 206 207 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 9 Multiple P38 MAPkinases contribute to NLRP1 inflammasome activation 208 Because PP1/PP2A-targeting toxins promote NLRP1 phosphorylation in a ZAK α -independent 209 manner, we next sought to identify the kinase(s) responsible for driving the hNLRP1 210 inflammasome response. To this end, we performed a PAMGENE kinase activity assay in 211 human keratinocytes exposed to PP1/PP2A-targeting toxins (Fig. 3A). This analysis revealed 212 robust activation of multiple MAP kinase family members (p38α , p38β , p38δ , JNK1/2, ERK1/2), 213 NF-κ B–related kinases (TBK1, IKK ε ), as well as CAM kinases and several cell-cycle kinases 214 (Fig. 3A). 215 We next assessed the contribution of these kinase families to hNLRP1 regulation using a 216 pharmacological approach (Fig. 3B). Inhibitor screening showed that the p38 α /β inhibitor 217 SB203580 and the pan-p38 inhibitor doramapimod ( 30) both reduced IL-1 β release in TERT 218 keratinocytes treated with PP1/PP2A-targeting toxins, suggesting that p38 kinases are involved 219 in NLRP1 inflammasome activation (Fig. 3B). In contrast, inhibition of the other activated 220 kinases did not significantly affect toxin-induced IL-1 β release. The greater efficacy of 221 doramapimod compared with SB203580 further suggested that additional p38 isoforms beyond 222 p38α /β contribute to this process. 223 Consistent with this idea, our PAMGENE data also identified p38 δ as strongly activated in 224 keratinocytes exposed to dinophysistoxin (Fig. 3A). Phos-tag analyses confirmed that all three 225 p38 isoforms, p38α , p38β , and p38δ , undergo robust phosphorylation in response to PP1/PP2A-226 targeting toxins (Fig. 3B, C). These results prompted us to assess the individual and collective 227 roles of these isoforms in hNLRP1 activation. Using CRISPR-Cas9, we generated single-isoform 228 knockout TERT keratinocytes and measured IL-1 β release following PP1/PP2A inhibition (Fig. 229 3D). Loss of p38 δ or p38 α /β alone caused minimal or only modest reductions in IL-1 β release, 230 whereas combined deletion of all three isoforms (p38 TKO) led to a str ong, though incomplete, 231 inhibition (Fig. 3D). These findings indicate that p38 α , β , and δ cooperatively regulate hNLRP1 232 inflammasome activation in response to PP1/PP2A-targeting toxins. 233 To further validate the requirement for p38 kinases in this pathway, we used reporter cell lines 234 expressing either WT hNLRP1 or p38-insensitive hNLRP1 variants in which important p38-235 responsive residues in the disordered region (Ser107 and Thr-Ser-Thr 112–114/178–180) were 236 replaced with alanines (A) (Fig. 3E) (20 , 22). ASC speck formation assays showed that 237 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 10 PP1/PP2A inhibition failed to induce speck assembly in cells expressing p38-insensitive 238 hNLRP1, whereas cells expressing WT hNLRP1 displayed robust ASC oligomerization (Fig. 239 3E). Importantly, Val-boroPro (VbP), a p38-independent activator of NLRP1, induced similar 240 ASC speck formation in all cell lines, confirming that the introduced mutations do not impair 241 other activation pathways (Fig. 3E). 242 Together, these results demonstrate that the p38 α , β , and δ isoforms act collectively as key 243 regulators of hNLRP1 inflammasome activation following PP1/PP2A inhibition. 244 245 TAK1 apical MAP kinase both activates P38 kinases and directly contributes to triggering 246 the hNLRP1 inflammasome in various contexts 247 Our observation that multiple p38 isoforms (p38 α , β , and δ ) contribute to hNLRP1 activation in 248 response to PP1/PP2A-targeting toxins led us to hypothesize that PP1/PP2A might regulate 249 upstream MAP kinases involved in p38 activation. To test this, we examined whether any apical 250 MAP3Ks participate in the hNLRP1 inflammasome response induced by PP1/PP2A inhibition. 251 We performed a pharmacological screen targeting the major druggable MAP3Ks in the HEK293 252 ASC-GFP/NLRP1 reporter system (Fig. 4A). Strikingly, only HS-276, a preclinical inhibitor of 253 the MAP3K7 kinase TAK1 ( 31), robustly suppressed ASC speck formation following toxin 254 exposure (Fig. 4A), suggesting that TAK1 is a key upstream regulator of hNLRP1 activation 255 under PP1/PP2A-inhibiting conditions. 256 To validate this genetically, we disrupted TAK1 in TERT keratinocytes using CRISPR–Cas9 and 257 assessed IL-1β release after stimulation with dinophysistoxin, okadaic acid, or cantharidin (Fig. 258 4B). While anisomycin and VbP elicited strong IL-1 β secretion in both wild-type and TAK1-259 deficient cells, thus confirming their TAK1-independent mode of action, TAK1 KO cells failed 260 to release IL-1β in response to any PP1/PP2A-targeting toxin (Fig. 4B), establishing TAK1 as a 261 central mediator of toxin-induced NLRP1 inflammasome activation. 262 Because TAK1 is also known to activate p38 kinases ( 32, 33), we next asked (1) whether 263 PP1/PP2A inhibition leads to TAK1 activation and (2) whether TAK1 is required for toxin-264 induced p38 activation (Fig. 4C–E). We found that all three toxins induced robust TAK1 265 phosphorylation (Fig. 4C–E). Moreover, toxin-induced p38 phosphorylation was completely 266 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 11 abolished in TAK1 KO cells (Fig. 4E), indicating that TAK1 not only becomes activated 267 following PP1/PP2A inhibition but also is essential for downstream p38 activation. 268 To dissect the respective roles of TAK1 and p38 in NLRP1 inflammasome activation, we 269 evaluated IL-1β release, NLRP1-DR phosphorylation, and gasdermin-D cleavage (Fig. 4C–E). 270 Both p38 TKO and TAK1 KO keratinocytes showed markedly reduced IL-1 β production and 271 gasdermin-D processing in response to the toxins (Fig. 4C, D). Notably, the reduced IL-1 β 272 production was stronger in TAK1 KO than in p38 TKO cells, indicating that TAK1 exerts both 273 p38-dependent and p38-independent effects on hNLRP1 activation (Fig. 4C, D). Consistently, 274 combined deletion of TAK1 in the p38 TKO background (p38 TKO/TAK1 KO) resulted in a 275 complete loss of toxin-induced IL-1 β release and gasdermin-D cleavage (Fig. 4C, D). In 276 addition, NLRP1-DR phosphorylation was slightly impaired in either TAK1 KO or p38 TKO 277 cells, and further reduced in p38 TKO/TAK1-KO cells (Fig. 4E), supporting a cooperative yet 278 partially independent contribution of p38 and TAK1 kinases. 279 Given previous work showing that the apical kinase ZAK α can directly phosphorylate NLRP1, 280 we next asked whether TAK1 might also directly phosphorylate hNLRP1 in addition to 281 activating p38. Using recombinant active TAK1 and p38 α , we found that both kinases 282 phosphorylated full-length hNLRP1 and its DR region in vitro (Fig. 4F), suggesting that TAK1 283 can directly phosphorylate and activate the hNLRP1 inflammasome (Fig. 4G). 284 The strong involvement of TAK1 in toxin-induced hNLRP1 activation led us to examine its 285 potential role in other NLRP1-activating contexts. Alphavirus infections and dsRNA have been 286 shown to induce a phosphorylation-dependent, albeit only partially ZAK α -dependent, NLRP1 287 response, suggesting the involvement of additional apical kinases ( 19, 20). We therefore 288 hypothesized that TAK1 could contribute to dsRNA- and virus-induced NLRP1 activation in 289 keratinocytes. Wild-type, TAK1 KO, ZAK α KO, p38 TKO, p38 TKO/TAK1 KO, and NLRP1 290 KO TERT keratinocytes were either transfected with high-molecular-weight dsRNA (poly:IC) or 291 infected with VSV, oncolytic VSV-M51, or Sindbis virus (Fig. S4A, B) (20). At the exception of 292 VSV, all stimuli triggered NLRP1-dependent IL-1β release (Fig. S4A, B). Whereas single loss of 293 ZAKα , TAK1, or of the p38 isoforms partially reduced IL-1 β secretion, combined TAK1 294 KO/p38 TKO or ZAK α KO/TAK1 KO cells completely impaired IL-1 β release upon dsRNA 295 transfection or viral infection (Fig. S4A, B). Thus, these results suggest that TAK1 and TAK1-296 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 12 associated pathways are key additional regulators of the hNLRP1 inflammasome in 297 keratinocytes. 298 Because PP1/PP2A inhibition activates TAK1, we finally tested whether PP2A activation might 299 conversely dampen TAK1-mediated NLRP1 signaling. Using the PP2A activator ATUX-1215 in 300 ZAKα KO A549 ASC-GFP/NLRP1 reporter cells transfected with dsRNA, we observed that 301 PP2A activation strongly reduced ASC speck formation during dsRNA stimulation (Fig. S4C). 302 To the contrary, PP2A activation had no effect on VbP-induced NLRP1 activation (Fig. S4C). 303 Hence, these results suggest that PP2A activity might specifically increase the activation 304 threshold of the TAK1-dependent NLRP1 pathway upon dsRNA exposure. 305 Altogether, our findings show that PP1/PP2A-targeting toxins and dsRNA converge on TAK1 306 activation, which in turn drives both p38-dependent and p38-independent activation of the 307 hNLRP1 inflammasome in keratinocytes. 308 309 TAK1 and P38 kinases play a major role in hNLRP1 inflammasome response in Human 310 native skin models of toxin exposure 311 The identification of TAK1 as a major contributor to PP1/PP2A-targeting toxin–induced 312 hNLRP1 inflammasome activation prompted us to examine the relevance of this pathway in a 313 more physiologically complex setting. To this end, we established a model of human native skin 314 explants exposed to the dermatitis-inducing toxins cantharidin, dinophysistoxin, and okadaic acid 315 (34). These explants, obtained from healthy donors, contain the full pattern of skin-resident cell 316 types including keratinocytes, fibroblasts, melanocytes, mast cells, and Langerhans cells, thus 317 providing a biologically representative environment (34). 318 In this model, all three toxins induced pronounced intraepidermal adhesion loss and extensive 319 cell death (Fig. 5A). Notably, pretreatment with the TAK1 inhibitor HS-276 or the pan-p38 320 inhibitor doramapimod markedly preserved intradermal adhesion and overall tissue architecture 321 in explants exposed to dinophysistoxin or okadaic acid (Fig. 5A-C). This finding suggests that 322 both TAK1 and p38 kinases are key mediators of PP1/PP2A-targeting toxin–induced skin 323 damage. In contrast, the protective effect of these inhibitors against cantharidin-induced injury 324 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 13 was less pronounced (Fig. 5A), suggesting that cantharidin may elicit skin pathology through 325 both PP1/PP2A-dependent and -independent mechanisms (35). 326 We next assessed whether these toxins could trigger an inflammasome response within the 327 explants. Longitudinal analysis of IL-1 β release revealed robust cytokine production beginning 328 24 hours after exposure, consistent with an inflammasome activation (Fig. S5A). 329 Immunohistochemistry further demonstrated abundant ASC specks in keratin 14-positive 330 keratinocytes (Fig. 5B), and proximity ligation assays showed close spatial association between 331 ASC and NLRP1 (Fig. S5B). Together, these data suggest that cantharidin, dinophysistoxin, and 332 okadaic acid all activate the NLRP1 inflammasome in human skin. Finally, to define the 333 contribution of TAK1 and p38 kinases to this response, we pretreated skin explants with HS-276 334 or doramapimod prior to toxin exposure. Both inhibitors markedly reduced IL-1 β release and 335 ASC speck formation across all three toxins (Fig. 5B, C). 336 All in one, these results demonstrate that TAK1 and p38 kinases are not only critical for NLRP1 337 inflammasome activation in response to PP1/PP2A-inhibiting toxins but also for the associated 338 tissue damage. 339 340

341 Because of its strong sensitivity to phosphorylation-driven activation, we investigated the ability 342 of cellular phosphatases to modulate the NLRP1 inflammasome response. By performing a 343 phosphatase screening approach, we identified PP1 and PP2A phosphatases as critical inhibitors 344 of the NLRP1 inflammasome response. Specifically, genetic inactivation of PP1/PP2A catalytic 345 subunits or exposure to PP1/PP2A-inhibiting toxins that can cause skin dermatitis, such as 346 Cantharidin from the beetles’ blister or Okadaic acid and Dinophysis toxins from Dinophysis 347 algae, all triggered the MAP3K7 TAK1, which promoted a phosphorylation cascade that 348 culminated with the NLRP1 inflammasome activation in human keratinocytes. Given their 349 presence in various living organisms such as viruses (e.g. small antigen T) ( 36, 37), insects (29), 350 algal species ( 26), PP1/PP2A-inhibiting toxins may represent a broad, yet poorly characterized, 351 family of triggers of NLRP1 activation with potential dermatologic, gastrointestinal, or 352 respiratory manifestations. However, in a coevolution point of view, it appears unlikely that the 353 NLRP1 inflammasome has been selected by the existence of various PP1/PP2A-targeting toxins, 354 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 14 given their presence in very distinct and specific geographical areas ( 26, 27, 29). To this regard, 355 our findings along the ones from two companion studies from Chua et al and Corley et al., found 356 that TAK1 kinase also contributes to the NLRP1 inflammasome response upon double stranded 357 RNA (dsRNA) exposure or viral infections driven by the oncolytic virus VSV-M51 and Sindbis 358 viruses. Of importance, although toxin-inactivated PP1/PP2A solely promotes a TAK1-359 dependent NLRP1 response, dsRNA and viral infections engage multiple activating pathways 360 including but not restricted to 1/ NLRP1 binding, 2/ Ribotoxic stress response-driven ZAK α 361 activation and 3/ TAK1 hyperactivation, suggesting that yet to be discovered and characterized 362 effectors/signaling pathway might be at play in cells exposed to viruses/dsRNA (19, 20). 363 Our findings that PP1 and PP2A phosphatase regulate the human NLRP1 inflammasome through 364 TAK1 also underscore the versatility of NLRP1 as a sensor of diverse threats. In this context, an 365 emerging picture suggests that NLRP1 is positioned as a central integrator of stress signals 366 originating from infectious agents, environmental toxins, and chemical disruptors that all 367 converge on MAPK signaling and regulation (38). To this regard, we hypothesize that PP1/PP2A 368 but also additional phosphatases could be acting as counterbalances against multiple MAPK 369 activity under homeostatic or stress conditions (39), hence increasing the NLRP1 inflammasome 370 activation threshold, as observed in our study during viral infections and dsRNA exposure. An 371 additional hypothesis of work is that phosphatases of the host, including PP1/PP2A, might also 372 be involved at directly dephosphorylating NLRP1 disordered region (DR) in addition to 373 inactivating kinases, hence fostering the cells to avoid NLRP1 spontaneous activation. Although 374 we did not address specifically this question in our study, this is currently under investigations in 375 our group. This is of particular relevance for understanding how keratinocytes, the primary cells 376 of the epidermis, maintain immune homeostasis while being exposed to a constant barrage of 377 environmental toxins and pathogens. 378 While our study provides significant insights into the role of PP1 and PP2A phosphatases in the 379 regulation of NLRP1 inflammasome activation in keratinocytes, several important questions 380 remain unanswered, and further research is necessary to fully understand the mechanistic details 381 and broader implications of these findings. To this regard, although we demonstrated the critical 382 involvement of several of the catalytic subunits of PP1 and PP2A in regulating NLRP1 383 inflammasome activation, the specific adaptor proteins and subunits that mediate this effect 384 remain unclear. Indeed, PP1 and PP2A consist of multiple isoforms with distinct tissue-specific 385 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 15 functions associated with a multitude of different adaptor proteins that confer to those 386 holoenzymes their specificity of action ( 23–25, 37). Future studies should focus on 387 characterizing the individual subunits of PP1 and PP2A that mediate their inhibitory function on 388 TAK1-mediated NLRP1 inflammasome regulation, as well as the potential role of regulatory 389 subunits in fine-tuning inflammasome dephosphorylation. 390 Furthermore, the strong importance of PP1/PP2A at restricting kinase-driven NLRP1 391 phosphorylation and activation pushes us to speculate that multiple viruses, bacteria, fungi or 392 parasites might have set up phosphatase or phosphatase-like virulence factors that in theory 393 could dampen the NLRP1 inflammasome response (40), either by targeting key MAPK and/or by 394 directly dephosphorylating NLRP1 (41), which will constitute an exciting future area of research. 395 Our study further delves into the kinase networks that drive NLRP1 inflammasome activation 396 following PP1/PP2A inhibition. Specifically, we found that TAK1 could mediate both P38 397 kinase-dependent and P38-independent activation of the NLRP1 inflammasome. To this regard, 398 it is worth mentioning that among the P38 kinases activated by PP1/PP2A-targeting toxins were 399 the multiple P38 isoforms (α , β , and δ ), which suggests a strong redundancy among those kinases 400 at promoting NLRP1 phosphorylation and activation. Those findings join the recent study from 401 Jenster et al., that also demonstrated a key role of all P38 kinase isoforms on NLRP1 402 phosphorylation and activation (20). In line with this, Chua et al. (companion manuscript) found 403 a minor role for P38 α /β isoforms in Cantharidin-induced NLRP1 inflammasome response. We 404 hypothesize that, given the redundant impact of the three P38 isoforms on the NLRP1 405 inflammasome response, P38α /β KO might not be sufficient to measure the contribution of P38 406 or that under certain conditions TAK1 might overwhelm P38 kinase requirement. 407 Regarding the relevance of our findings in complex settings, our Human native skin explant 408 approaches highlighted the strong contribution of both TAK1 and P38 kinases on the NLRP1 409 inflammasome activation in response to PP1/PP2A-targeting toxins. Importantly, pretreatment 410 with TAK1 or P38 inhibitors also significantly reduced skin damage, thus suggesting that these 411 kinases contribute not only to inflammasome signaling but also to the inflammatory skin 412 pathology observed in response to toxin exposure. Whether additional physiological or 413 pathological contexts involve dysregulated P38/TAK1 kinases such as psoriasis, hypomorphic 414 PP1/PP2A conditions ( 42) or genetic deficiency in RNA regulation (e.g. ADAR1 alterations) 415 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 16 (43) will warrant for further studies and testing preclinical P38/TAK1 inhibitors in a topical way. 416 To the contrary, such formulation of PP1/PP2A inhibition, when added locally holds strong 417 potential to act as a curing agent. To this regard, Cantharidin is the main compound of the 418 YCanth cure, commonly used to eliminate local warts induced by the molluscum virus in the 419 skin (44). It is tempting to speculate that such process might involve NLRP1-driven pyroptosis 420 and inflammation in treated cells. 421

of the study 422 Although our study identifies PP1/PP2A phosphatases as critical regulators of the NLRP1 423 inflammasome, we did not address the involvement of the multiple adaptors that drive the 424 specificity of those holoenzymes, which is currently under extensive study in the lab. 425 Furthermore, despite the use of human native skin explants from healthy donors, our study lacks 426 the complete physiology of an organism. Given the lack of conservation between rodents and 427 humans on the NLRP1 phosphorylation, such approach is at this time not available as we have 428 not yet succeeded at generating robust humanized models. 429 430

431 432 Study design 433 This study aimed to define how phosphorylation controls human NLRP1 inflammasome 434 activation. We first screened phosphatases inhibitors and environmental phosphatase-targeting 435 toxins in reporter cells. Anisomycin and Val-boro-Pro, two known activators of NLRP1 were 436 used as a treatment of reference. The involvement of the Ribotoxic Stress Response and ZAK α 437 was examined in A549, HEK293T and N/TERT keratinocytes through pharmacological 438 inhibition and CRISPR-Cas9 depletion. Candidates’ kinases were investigated by kinase activity 439 assay, inhibitor profiling, genetic knock-out, and in-vitro kinase and lambda protein phosphatase 440 assays coupled to PhosTag SDS-PAGE. The requirement for TAK1 upstream kinase in NLRP1 441 activation by dsRNA and viral infection was evaluated in keratinocytes stimulated with poly(I:C) 442 or infected with various viral strains, including VSV or VSV-M51, or Sindbis virus. Finally, 443 toxin-induced NLRP1 activation and kinase dependency were validated in human skin explants 444 using pharmacological inhibitors and histology, immunofluorescence, proximity ligation assays 445 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 17 and cytokine measurements. Details on sample size and statistical analyses are provided in figure 446 legends. 447 448 Human Skin Tissue Models 449 HypoSkin 22-mm-diameter human skin explants (GenoSkin, Toulouse, France) derived from 450 abdominal tissue were obtained from two healthy adult Caucasian donors (female, 35-year-old; 451 male, 43-year-old), both of whom had no documented dermatological conditions and gave 452 written informed consent. This biological material complies fully with the Declaration of 453 Helsinki and all relevant regulations. The study is not classified as human subject research, and 454 Institutional Review Board approval was not required. Each donor batch underwent the 455 manufacturer’s internal quality control prior to shipment, including histological validation of 456 tissue integrity and virological screening for HIV-1 and -2 and Hepatitis B and C. 457 Explants were maintained in the recommended culture medium (GenoSkin, NSMED2) and 458 standard culture conditions (37°C, 5% CO 2). Culture medium was replaced daily. When 459 indicated, inhibitors (Doramapimod, 20µM, MedChem Express, HY-10320; TAK-1 inhibitor 460 HS-276 (20 µM, MedChem Express, HY-147141) were added to fresh medium and applied as a 461 24-hour pre-treatment before injection. Each explant received a 55µL intradermal injection 462 delivered just beneath the epidermis. Injection solutions contained either phosphatase-targeting 463 toxins (Dinophysistoxin, 500nM, Bertin Technologies, WP-10011497; Okadaic acid, 2400nM, 464 Bertin Technologies, WP-10011490; Cantharidin, 25µM, Sigma-Aldrich, C7632-25MG) diluted 465 in PBS, PBS, or toxin/inhibitor mixtures. Explants were collected 72 hours later, fixed in 466 paraformaldehyde (PFA) for histological staining or snap-frozen in liquid nitrogen for 467 immunohistochemical analyses. All experiments were performed in two independent runs, each 468 including three biological replicates per condition. 469 470 Cell culture 471 472 The complete description of all cells and cell lines used in this study is provided in Table S1. 473 Human Embyonic Kidney (HEK) 293T and human alveolar basal epithelial (A549) cells were 474 cultured in Dulbecco’s Eagme’s Medium (DMEM, high-glucose, Gibco, 41965039) 475 supplemented with 10% Fetal Bovine Serum (FBS, Cytiva, SV30160.03IR). 476 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 18 Immortalised human keratinocytes (N/TERT) were provided by J. Rheinwald (Material Transfer 477 Agreements to FL. Zhong and E. Meunier). WT, ZAK α KO and NLRP1 KO cells were 478 previously described by ( 21, 22) and WT cells stably expressing NLRP1 Disordered Region 479 (DR) construct (aa 86-275) expressing a SNAP tag (here referred to as N/TERT + 86-275-480 SNAP) were generated in a previous study (21, 22). N/TERT cells were cultivated in 481 Keratinocyte Serum Free Media (SFM, Gibco, 17005042) supplemented with final concentration 482 of 25mg/L Bovine Pituitary Extract (Gibco, 13028-014), 294.4 ng/L human recombinant EGF 483 (Gibco, 10450-013) and 300µM of CaCl2 (PromoCell, C-34006). Human primary keratinocytes 484 (pHEKs, PromoCell, C-12007) were cultivated in Keratinocyte Growth Medium 2 (PromoCell, 485 C-20011). 486 N/TERT keratinocytes, HEK293T cells, A549 cells and pHEKs were cultured under standard 487 conditions, at 37°C, 5% CO2. 488 489 490 491 CRISPR/Cas 9 KO generation with electroporation 492 CRISPR/Cas9 ribonucleoprotein (RNP) complexes were prepared by mixing recombinant Cas9 493 protein (Invitrogen, A36498) with custom-designed sgRNA (benchling.com, SYNTHEGO) at 494 a 1 :1 molar ratio in Genome Editing Buffer (Invitrogen, A5430001). Complexes were incubated 495 at room temperature for 25 minutes. Cell pellets were resuspended in Genome Editing Buffer 496 (7µL buffer per 1.5 x 10 5 cells), supplemented with GFP mRNA (1µg, OZ Biosciences, 497 MRNA15-100) to assess electroporation efficiency, and combined with RNPs. Electroporation 498 was performed according to the manufacturer’s protocol using 1700 V, 20 ms, single-pulse 499 setting. Cells were immediately transferred into pre-warmed (37°C) 12-well plates and cultured 500 for 48 hours before selection. Knock-out efficiency was confirmed by immunoblotting and 501 further experiments were performed after one week of recovery. 502 The following sites were targeted: MAPK14 (p38α ) (5′ -TTGTGTCAAAAGCAGCACTA-3′ / 5′ -503 AAAGAACCTACAGAGAACTG-3′ ); MAPK11 (p38 β ) (5 ′ -GCAGAACGTACCGGGAGCTG-504 3′ / 5 ′ -CCGGGCGTCGTAGGCCGAAC-3′ ); MAPK13 (p38 δ ) (5 ′ -505 CTCCCCTGACCGCTTGTCGA-3′ / 5 ′ -CAGCAGCAGCAGCTCCCGGT-3′ ); PPP1CA (5 ′ -506 GGCTCACCGCAGATCTTGAG-3′ / 5 ′ -GCGCCCCAGTGCAGGGCTCG-3′ ); PPP1CB (5 ′ -507 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 19 ATTTATCGTTTGTCAGTACG-3′ / 5 ′ -TACATACCACAAATTTTCAG-3′ ); PPP1CC (5 ′ -508 GAAGCACCACTCAAAATATG-3′ / 5 ′ -ATTTTTCTCTTGCAGTGAGA-3′ ); PPP2CA (5 ′ -509 ATAAACAAGTAATTTGTATC-3′ / 5 ′ -ACATCGAACCTCTTGCACGT-3′ ); PPP2CB (5 ′ -510 GGATACAAACTACTTATTCA-3′ / 5′ -AAAAGAATCAAATGTGCAAG-3′ ). 511 512 Plasmid and nucleic acid transfection 513 For plasmid expression, DNA (1 µg/µL) were diluted 1:10 in nuclease-free water and mixed 514 with LyoVec (1:100 ratio, InvivoGen, lyec-1). The DNA-LyoVec solution was gently mixed by 515 pipetting and incubated at room temperature for 20 minutes. The transfection mixture was added 516 dropwise to cells in antibiotic-free medium, and incubated at 37°C, 5% CO 2. In all, 10 ng of 517 previously described NLRP1 plasmid (pSelect2B-hNLRP1 WT, pSelect2B-hNLRP1 Ser107A, 518 pSelect2B- hNLRP1 112TST114/112AAA114Ser107Pro, pSelect2B-hNLRP1 178TST180/178AAA180) 519 (18, 21) were transfected per well. Cells were analysed 48 hours after transfection. 520 For Poly(I:C) (Invivogen, tlrl-picw) stimulation, keratinocytes were transfected with 521 Lipofectamine LTX (Invitrogen, 15338030). Briefly, poly(I:C) was diluted in OPTI-MEM 522 (1:100) and Lipofectamine LTX (1:2). The RNA-Lipofectamine solution incubated at room 523 temperature for 25 minutes, then added dropwise to cells in OPTI-MEM and incubated at 37°C, 524 5% CO2. 525 526 Viral infections 527 Vesicular Stomatitis Indiana Virus (VSV) and mutant for the M protein (VSV-M51) viruses 528 were provided by David Olagnier ( 45). If not otherwise indicated, cells were infected with 529 viruses at a Multiplicities Of Infection (MOI) of 10 and incubated for indicated time at 37°C, 5% 530 CO2. 531 532 Cell stimulation in 2D cultures 533 If not otherwise indicated, cells were plated at 3 x 10 4, 5 x 10 4, 7.5 x 10 4, 1 x 10 5 cells/well for 534 96-, 24-, 12-, and 6-well plates respectively, in complete culture medium. Next day, when 535 indicated, cells were pre-incubated for 1 hour in OPTI-MEM (Gibco, 51985034) with inhibitors 536 before the addition of toxins or control stimuli. 537 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 20 If not otherwise indicated, stimuli and inhibitors were used at the following concentrations : 538 PLX-4720 (10 µM, MedChem Express, HY-51424); SB203580 (10 µM, MedChem Express, 539 HY-10256); Doramapimod (10 µM, MedChem Express, HY-10320); BX795 (1 µM, Invivogen, 540 Tlrl-bx7); Abemaciclib (25 nM, MedChem Express, HY-16297A); Ribociclib (40 nM, 541 MedChem Express, HY-15777); Palbociclib (25 nM, MedChem Express, HY-50767); 542 Fadraciclib (25 nM, MedChem Express, HY-101212); Roniciclib (25 nM, MedChem Express, 543 HY-13914); SCH772984 (1 µM, MedChem Express, HY-50846); SP600125 (10 µM, MedChem 544 Express, HY-12041); TAK-1 inhib HS-276 (10-20 µM, MedChem Express, HY-147141); TAO 545 Kinase inhibitor 1 (5 µM, MedChem Express, HY-112136); Selonsertib (10 µM, MedChem 546 Express, HY-18938); 5z-7-Oxozeaenol (5 µM, MedChem Express, HY-12686); GW806742X (1 547 µM, MedChem Express, HY-112292A); DN1289 (1 µM, MedChem Express, HY-152142); 548 Gossypetin (40 µM, MedChem Express, HY-119917); BSJ-04-122 (10 µM, MedChem Express, 549 HY-152185); GSK-872 (10 µM, MedChem Express, HY-101872); Bortezomib (1 µM, Selleck, 550 SE-S1013-5MG); MLN9424 (1 µM, Tocris Bioscience, 6499); DT-061 (1-20 µM; MedChem 551 Express, HY-112929); Dinophysistoxin (250 nM, Bertin Technologies); Okadaic acid (100-600 552 nM, Bertin Technologies, WP-10011490); Cantharidin (5-25 µM, Sigma-Aldrich, C7632-553 25MG); Calyculin-A (1 µM, Bertin Technologies, WP-19246); LB-100 (50 µM, Bertin 554 Technologies, 29105); Raphin-1 (50 µM, Tocris Bioscience, 6760/10); Sephin-1 (100 µM, 555 Tocris Bioscience, 5553/10); Anisomycin (1-5 µM, Selleck, SE-S7409-10MG); Valboro-Pro/ 556 Talabostat (10 µM, Selleck, SE-S8455-5MG); Phosphatase Inhibitor Library (117 items) (10 557 µM, MedChem Express, HY-L081). 558 559 ASC specks imaging and quantification 560 ASC specks formation in A549 and HEK293 cells was monitored using EVOS 7000 561 fluorescence microscope using x 10 or x 20 objectives. Specks were manually quantified as the 562 proportion of ASC aggregates over total nuclei number (staining with Hoechst 33342, 563 Invitrogen, H3570) using Fiji software (ImageJ). At least three wells and ≥ 500 cells per well 564 were analysed in a blinded manner. 565 566 Immunoblotting 567 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 21 Whole-cell lysates were prepared by lysing cells in RIP-A buffer (150 mM NaCl, 50 mM Tris-568 HCl, 1% Triton X-100, 0.5% Na-deoxycholate) supplemented with protease inhibitor cocktail 569 (Roche, PHOSS-RO). For phosphorylated proteins samples, phosphatase inhibitor cocktail 570 (Roche, 04693159001) was added to the lysis buffer. Before addition of 4X Laemmli buffer 571 (Bio-Rad, 1610747) (with a 3 : 1 ratio), protein concentration was measured with BCA protein 572 assay (PIERCE, 23227). Lysates were then boiled at 95°C for 5 minutes. 573 Cell culture supernatants were precipitated by addition of TriChloroacetic Acid (TCA, 100% 574 w/v, 4°C) to a final concentration of 10%, incubated on ice for 1 hour and centrifuged for 1 hour 575 (15,000 rpm, 4°C). Pellets were washed with ice-cold acetone and centrifuged again (15,000 576 rpm, 4°C). The acetone was removed, proteins were resuspended in TE buffer and then boiled at 577 95°C for 5 minutes. 578 Samples were separated by home-made 12% denaturing SDS-PAGE, blotted onto 0.2-µm PVDF 579 membranes (Bio-Rad, 12023927), and blocked in 5% milk diluted in TBS-T (Tris 10 mM, pH 8, 580 NaCl 150 mM, Tween 20 0.05%). Membranes were incubated with indicated primary antibody 581 overnight at 4°C and with corresponding secondary HRP-conjugated antibodies for 1 hour at 582 room temperature. Chemiluminescent signals were detected with ECL revelation kit (Advansta, 583 K-12043-D10, K-12042-D20) and recorded with C-DiGit Imaging System (Li-cor). 584 Antibodies used in this study were anti-Gasdermin D (E8G3F) Monoclonal antibody (1 :1000, 585 Cell signaling, 97558S) ; anti-Caspase 1 (p20) (human) mAb (Bally-1) antibody (1 :500, 586 Adipogen, AG-20B-0048) ; anti-DFNA5/Gasdermin E (EPR19859) Monoclonal antibody 587 (1 :1000,Abcam, ab215191) ; anti-Caspase 3 antibody (1 :500, Cell signaling, 9662S) ; anti-588 Human IL-1 β /IL-1F2 Polyclonal antibody (1 :500, R&D systems, AF-201-NA) ; anti-Cleaved 589 IL-1β (Asp116) Monoclonal antibody (1 :750, Cell signaling, 83186S) ; anti-Tubulin- α antibody 590 (1 :1000, Abcam, ab4074) ; anti-PP1alpha Polyclonal antibody (1 :1000, Invitrogen, PA5-591 119781) ; anti-PP1CB-Specific antibody (1 :1000, ProteinTech, PR-55136-AP-150) ; anti- 592 PP1CC antibody (1 :1000, ProteinTech, PR-11082-1-AP-150) ; PP2A C Subunit antibody 593 (1 :500, Cell signaling, 2038S) ; anti- Purified NLRP1 (N-terminal) antibody (1 :300, Biolegend, 594 679802) ; anti- NLRP1 (C-terminal) Polyclonal antibody (1 :500,Abcam, ab36852) ; anti- GFP 595 antibody (1 :1000, Abcam, ab6673) ; anti- SNAP/CLIP-tag Monoclonal antibody (1 :1000, 596 ProteinTech, 6F9-100) ; anti-ZAK α Polyclonal antibody (1 :1000, Bethyl Laboratories, A301-597 993A) ; anti- P38 MAPK antibody (1 :1000, Cell signaling, 9212S) ; anti- Phospho-p38 598 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 22 (Thr180/Tyr182) (D3F9) Monoclonal antibody (1 :1000, Cell signaling, 4511S) ; anti- 599 Puromycine antibody Clone 12D10 (1 :1000, Sigma-Aldrich, MABE343) ; anti- TAK1 antibody 600 (1 :1000, Cell signaling, 4505S) ; anti- Phospho-TAK1 (Ser412) antibody (1 :1000, Cell 601 signaling, 9339S) ; anti- Phospho-TAK1 (Thr184/187) 90C7 Monoclonal antibody (1 :1000, Cell 602 signaling, 4508S) ; anti- P38 alpha/MAPK14 antibody (E229) (1 :1000, Abcam, ab170099) ; 603 anti- P38 MAPK beta Monoclonal antibody (1 :1000, Invitrogen, MA514-950) ; anti- P38 beta 604 MAPK (C28C2) Monoclonal antibody (1 :1000, Cell Signaling, 2339S) ; anti- P38 605 gamma/MAPK12 antibody (EPR6528N) (1 :1000, Abcam, ab205926) ; anti- P38 gamma MAPK 606 antibody (1 :1000, Cell Signaling, 2307S) ; anti- Human P38 delta antibody (1 :1000, R&D 607 system, AF1519) ; anti- P38 δ MAPK (10A8) mAb (1 :1000, Cell Signaling, 2308S) ; anti- 608 GAPDH antibody (1 :1000, GeneTex, GTX100118) ; Goat anti-rabbit HRP secondary antibody 609 (1 :5000, Advansta R-05072-500) ; Goat anti-mouse HRP secondary antibody (1 :5000, 610 Advansta, R-05071-500) ; Goat anti-rat IgG H&L (HRP) (1 :5000, Abcam, ab97057) ; Goat IgG 611 HRP-conjugated Antibody (1 :5000, Biotechne, HAF109). 612 613 PhosTag SDS-PAGE 614 Cells were lysed in RIP-A buffer supplemented with protease inhibitor and phosphatase inhibitor 615 cocktails, and processed as described above. Phosphorylated proteins were separated by home-616 made 10% SDS-PAGE gel, with addition of Phos-tag Acrylamide (Wako Chemicals, AAL-107) 617 to a final concentration of 30 µM and Manganese Chloride (II) (Sigma-Aldrich, 63535) to 60 618 µM. Gel were washed in Transfer buffer containing EDTA (10 mM), blotted onto 0.45-µm 619 PVDF membranes (Invitrogen, LV2005), and blocked in 5%. Membranes were incubated with 620 indicated primary antibody and corresponding secondary HRP-conjugated antibodies. 621 622 Plasma membrane permeabilization monitoring 623 N/TERT cells were seeded at 2 x 10 4 cells/well in Black/Clear 96-well plates (Greiner, 655090) 624 in complete culture medium. Twenty-four hours later, culture medium was changed for OPTI-625 MEM supplemented with SYTOX-Green Nucleic Acid Stain (500 nM, Invitrogen, S7020) and 626 treated with the indicated stimulations. Real-time fluorescence was measured using Clariostar 627 (BMG Labtech) plate reader equipped with a 37°C incubator. Maximal permeabilization was 628 defined using 1% Triton X-100. 629 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 23 630 Cell lysis assay 631 Cell lysis was evaluated by the quantification of LDH release into the cell supernatant, 632 employing the LDH-Blue kit (rep-ldh-1) from Invivogen. Keratinocytes were seeded in 96-well 633 plates at 2 /i9 ×/i9 104 cells per well in Keratinocyte Growth Medium 2 (PromoCell Inc). The 634 following day, cells were stimulated and 50 /i9 µL of cell supernatant was mixed with an equal 635 volume of LDH substrate and left to incubate for 30 /i9 min at room temperature, protected from 636 light. The enzymatic reaction was stopped by adding 50 /i9 µL of stop solution. Maximal cell 637 death was determined with whole cell lysates from unstimulated cells incubated with 1% Triton 638 X-100. 639 640 Kinase activity assay 641 Kinase activity profiling was performed using PamChip 96-microarrays (PamGene, 86311). 642 N/TERT keratinocytes were stimulated with dinophysistoxin or not for 3 hours, washed in ice-643 cold PBS, and lysed in manufacturer-recommended lysis buffer. Protein concentrations were 644 normalised using BCA assay and equal amounts of lysates were loaded per array, according to 645 the supplier’s protocol. Fluorescent signals were acquired and quantified in BioNavigator 646 software. Technical replicates passing automated quality control were averaged to generate one 647 kinase activity profile per biological replicate (n = 3 per condition). Differential kinase activity 648 was determined by comparing dinophysistoxin-treated samples to matched controls. 649 650 Cytokine analysis/ quantification/ Soluble mediator release analysis 651 To measure hIL-1 β levels, cell supernatants were recovered 24 hours after stimulation and 652 human IL-1β Enzyme Linked Immunosorbent Assay (ELISA) kit (Invitrogen, 88-7261-77) was 653 used according to the supplier’s protocol. 654 For the Human IL-1 Family Cytokine Array (Tebu-bio, AAH-IL1F-1-8), N/TERT were seeded 655 in T-75 flask at 5 x 10 5 cells per wells in complete culture medium. Supernatants were collected 656 24 hours after stimulation, cleared by centrifugation (15 000 rpm, 10 minutes, 4°C), concentrated 657 with Amicon Ultra 4 – 10K (4,000 g, 15 minutes, 4°C, Sigma-Aldrich, UFC801024), and 658 cytokine array was performed according to the supplier’s protocol. 659 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 24 To measure secreted cytokine and soluble mediators’ levels for human skin models, culture 660 mediums were collected, cleared by centrifugation (15 000 rpm, 10 minutes, 4°C), and 661 LEGENDPlexTM Human cytokine panel 2 (Biolegend, 741378) was used according to the 662 supplier’s protocol. Samples were processed with BD Fortessa x20 Nothern Lights C6 L2 663 (Genotoul Tri, IPBS, toulouse) and analysed using the LEGENDplex Data Analysis Software 664 Suite (Biolegend). The following targets were analysed: TSLP, IL-1 α , IL-1β , GM-CSF, IFN-α 2, 665 IL-23, IL-12p40, IL-12p70, IL-15, IL-18, IL-11, IL-27, and IL-33 666 667 Histology 668 Human skin tissue samples were fixed in a 10% formalin bath for 48 hours at room temperature, 669 dehydrated automatically and embedded in paraffin (Epredia HistoStar, Epredia). Paraffin blocks 670 were cut into 5 µm-thick sections and stained with Hematoxylin and Eosin (H&E) according to 671 standard protocols. Images were acquired using a Zeiss Axio Imager M2 (Zeiss) and 20x/0.8 EC 672 Plan-neofluar Zeiss objective (Zeiss). Images were acquired and processed using Zeiss Axiocam 673 503 color camera and the Zeiss Zen software (Zeiss). 674 675 Immunohistology 676 Human skin tissue samples were snap-frozen in liquid nitrogen and embedded in Tissue Freezing 677 Medium (TFM) (Microm Microtech France, TFM-5). Frozen blocks were cut into 7 µm-thick 678 sections (Phi plateform, I2MC, Toulouse), fixed in acetone 100% for 20 minutes at room 679 temperature and rehydrated in PBS for 15 minutes. Sections were then blocked for 1 hour at 680 room temperature with 3% Bovine Serum Albumin in PBS-T (PBS 0.1% Tween 20) and 681 incubated overnight at 4°C in a humidity chamber with primary antibody (Cytokeratin K-682 14,1 :400, Abcam, Ab7800; ASC,1 :250, AdipoGen, AG-25B-0006) diluted in staining buffer 683 (1% BSA, PBS 0.05% Tween). Slides were then incubated with Dylight 550 (1 :500, 684 ImmunoReagents Inc., GtxMu-003-D550NHSX), Dylight 488 (1 :500, ImmunoReagents Inc., 685 GtxRb-003-0488NHSX) -Conjugated secondary antibodies in staining buffer for 1 hour at room 686 temperature. Hoechst (1:1000) was added to the last wash for 10 minutes at room temperature. 687 Wild-field images were acquired using EVOS 7000 fluorescence microscope with x 20 688

and confocal imaging was performed on a Spinning disk Andor (Olympus) with 689 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 25 60XO/1.35 ULSAPO (WD 0.15mm) objective at the Genotoul Imagerie plateform (IPBS, 690 Toulouse). Images were processed using IQ3 software (Olympus). 691 692 Proximity Ligation assay 693 ASC-NLRP1 interactions were visualised on frozen sections (as described above) using the 694 NaveniFlex Tissue MR (Navinci, 60027) kit, according to manufacturer’s protocol. Briefly, 695 sections were blocked for 1 hour with Blocking buffer, incubated overnight at 4°C with primary 696 antibody NLRP1 (1 :200, Biolegend, 679802), ASC (1 :200, AdipoGen, AG-25B-0006) and 697 incubated with Navenibody M1 and Navenibody R2 diluted in Diluent buffer (1 :40) for 1 hour. 698 Slides were incubated with Enzyme 1 diluted in 1X Buffer 1 (1 :40) for 30 minutes and with 699 Enzyme 2 (1 :40) for 90 minutes. For visualisation, slides were incubated in Detection buffer for 700 30 minutes, incubated in PBS supplemented with Hoechst (1 :1000) for 5 minutes and mounted 701 in mounting medium. Otherwise specified, every incubation step was made at 37°C in a 702 humidity chamber. Images were acquired using Confocal and Multiphoton Zeiss 710 NLO 703 Spectral (Zeiss) and 20X/0.8 PLAN APO AIR objective. Images were processed using Zeiss Zen 704 software (Zeiss). 705 706 Quantification of epithelial damage and ASC specks in 3D skin models 707 Intradermal disruptions were quantified as the ratio of disruption area to total epidermis area. 708 Regions of interest were manually delineated in Fiji using H&E sections from three technical 709 replicates per treatment within the same experiment. 710 ASC specks and Proximity Ligation Assay (PLA)-positive puncta in tissue sections were 711 quantified as the proportion of ASC aggregates or PLA signals per nucleus using Fiji software. 712 Quantifications were performed on images from three technical replicates per treatment within 713 the same experiment. 714 715 Proteins expression and purification 716 HEK 293T cells were seeded at 3 x 10 5 cells per well on a 6-well plate and transfected with a 717 plasmid coding for NLRP1 Disordered Region (DR) construct (aa 86-254) expressing a GFP tag 718 (22). Transfection was made using Lipofectamine 2000 (Invitrogen, 11668027) and 2 µg of 719 DNA per well, according to manufacturer’s protocol. Cells were lysed in RIP-A buffer and 720 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 26 cleared by centrifugation (17 000 g, 10 minutes, 4°C). Immunoprecipitation (IP) was conducted 721 using GFP-Trap Agarose (Chromotek, gta) kit, according to supplier’s protocol. Briefly, diluted 722 lysates were added to equilibrated beads (in dilution buffer) and incubated for 1 hour at 4°C on a 723 rotating wheel. Subsequently, beads were washed four times with Wash buffer before elution in 724 homemade acidic elution buffer (0.1 M Glycine pH 2.7) and neutralisation with 1 M Tris pH 8.5. 725 Purified NLRP1-DR-GFP was quantified using Bradford/ BCA assay and used for lambda 726 phosphatase dephosphorylation assay. 727 728 Lambda Phosphatase dephosphorylation assay 729 Purified hNLRP-GFP-DR was adjusted to a final volume of 40 µL with distilled water, yielding 730 at 40 µg protein per reaction. Samples were supplemented with 10X NEBuffer for Protein 731 MetalloPhosphatases (PMP, 5 µL), MnCl 2 (10 mM, 5 µL), and Lambda Protein Phosphatases (1 732 µL, New England Biolab, P0753S). Reactions were incubated at 30°C for 30 minutes. Following 733 dephosphorylation, samples were processed for SDS-PAGE phostag immunoblotting as 734 described previously in the Phostag SDS-PAGE section. 735 736 In vitro kinase assay 737 Kinase reactions were carried out by combining 500 ng of recombinant NLRP1 (Clinisciences, 738 P316481) with 100 ng of either recombinant p38 α (Abcam, ab271606) or TAK1 (Abcam, 739 ab89692) in homemade reaction buffer (HEPES-KOH pH 7.0, 5 mM MgCl 2, 2 mM ATP, 0.3 M 740 DTT) for 1 hour at 30°C. When indicated, kinase products were subsequently exposed to 741 Lambda Protein Phosphatase in 1X NEBuffer for PMP and MnCl 2 (1 mM). Mixtures were 742 incubated for 30 minutes at 30°C. Final reaction products were processed for SDS-PAGE 743 Phostag immunoblotting as described previously in the Phostag SDS-PAGE section. 744 745 Statistical analysis 746 Statistical analyses were performed using GraphPad Prism (version 10, GraphPad Software, 747 Inc.). The exact test, sample size (n), and correction methods are provided in the corresponding 748 figure legend. Unless otherwise specified, data are presented as mean ± SEM. Comparisons 749 between two or more groups were analysed using one-way ANOVA or two-way ANOVA 750 (respectively) with the appropriate multiple-comparison correction. P values are reported in the 751 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 27 figures, with the following significance thresholds: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001; ns 752 indicates no significant difference. 753 754

755 1. K. S. Robinson, D. Boucher, Inflammasomes in epithelial innate immunity: front line 756 warriors. FEBS Lett 598, 1335–1353 (2024). 757 2. R. Medzhitov, The spectrum of inflammatory responses. Science 374, 1070–1075 (2021). 758 3. B. Sundaram, R. E. Tweedell, S. Prasanth Kumar, T.-D. Kanneganti, The NLR family of 759 innate immune and cell death sensors. Immunity 57, 674–699 (2024). 760 4. N. Kayagaki, O. S. Kornfeld, B. L. Lee, I. B. Stowe, K. O’Rourke, Q. Li, W. Sandoval, D. 761 Yan, J. Kang, M. Xu, J. Zhang, W. P. Lee, B. S. McKenzie, G. Ulas, J. Payandeh, M. 762 Roose-Girma, Z. Modrusan, R. Reja, M. Sagolla, J. D. Webster, V. Cho, T. D. Andrews, L. 763 X. Morris, L. A. Miosge, C. C. Goodnow, E. M. Bertram, V. M. Dixit, NINJ1 mediates 764 plasma membrane rupture during lytic cell death. Nature 591, 131–136 (2021). 765 5. K. Newton, V. M. Dixit, N. Kayagaki, Dying cells fan the flames of inflammation. Science 766 374, 1076–1080 (2021). 767 6. J. Shi, Y. Zhao, K. Wang, X. Shi, Y. Wang, H. Huang, Y. Zhuang, T. Cai, F. Wang, F. 768 Shao, Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death. 769 Nature 526, 660–665 (2015). 770 7. N. Kayagaki, I. B. Stowe, B. L. Lee, K. O’Rourke, K. Anderson, S. Warming, T. Cuellar, 771 B. Haley, M. Roose-Girma, Q. T. Phung, P. S. Liu, J. R. Lill, H. Li, J. Wu, S. Kummerfeld, 772 J. Zhang, W. P. Lee, S. J. Snipas, G. S. Salvesen, L. X. Morris, L. Fitzgerald, Y. Zhang, E. 773 M. Bertram, C. C. Goodnow, V. M. Dixit, Caspase-11 cleaves gasdermin D for non-774 canonical inflammasome signalling. Nature 526, 666–671 (2015). 775 8. F. Martinon, K. Burns, J. Tschopp, The inflammasome: a molecular platform triggering 776 activation of inflammatory caspases and processing of proIL-beta. Mol Cell 10, 417–426 777 (2002). 778 9. G. Fenini, T. Karakaya, P. Hennig, M. Di Filippo, M. Slaufova, H.-D. Beer, NLRP1 779 Inflammasome Activation in Keratinocytes: Increasing Evidence of Important Roles in 780 Inflammatory Skin Diseases and Immunity. J Invest Dermatol 142, 2313–2322 (2022). 781 10. A. Sandstrom, P. S. Mitchell, L. Goers, E. W. Mu, C. F. Lesser, R. E. Vance, Functional 782 degradation: a mechanism of NLRP1 inflammasome activation by diverse pathogen 783 enzymes. Science 364, eaau1330 (2019). 784 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 28 11. Q. Wang, J. C. Hsiao, N. Yardeny, H.-C. Huang, C. M. O’Mara, E. L. Orth-He, D. P. Ball, 785 Z. Zhang, D. A. Bachovchin, The NLRP1 and CARD8 inflammasomes detect reductive 786 stress. Cell Rep 42, 111966 (2023). 787 12. M. B. Geeson, J. C. Hsiao, L. P. Tsamouri, D. P. Ball, D. A. Bachovchin, The interaction 788 between NLRP1 and oxidized TRX1 involves a transient disulfide bond. Cell Chemical 789 Biology 31, 955-961.e4 (2024). 790 13. E. L. Orth-He, H.-C. Huang, S. D. Rao, Q. Wang, Q. Chen, C. M. O’Mara, A. J. Chui, M. 791 Saoi, A. R. Griswold, A. Bhattacharjee, D. P. Ball, J. R. Cross, D. A. Bachovchin, Protein 792 folding stress potentiates NLRP1 and CARD8 inflammasome activation. Cell Rep 42, 793 111965 (2023). 794 14. R. Planès, M. Pinilla, K. Santoni, A. Hessel, C. Passemar, K. Lay, P. Paillette, A.-L. C. 795 Valadão, K. S. Robinson, P. Bastard, N. Lam, R. Fadrique, I. Rossi, D. Pericat, S. 796 Bagayoko, S. A. Leon-Icaza, Y. Rombouts, E. Perouzel, M. Tiraby, COVID Human 797 Genetic Effort, Q. Zhang, P. Cicuta, E. Jouanguy, O. Neyrolles, C. E. Bryant, A. R. Floto, 798 C. Goujon, F. Z. Lei, G. Martin-Blondel, S. Silva, J.-L. Casanova, C. Cougoule, B. 799 Reversade, J. Marcoux, E. Ravet, E. Meunier, Human NLRP1 is a sensor of pathogenic 800 coronavirus 3CL proteases in lung epithelial cells. Mol Cell 82, 2385-2400.e9 (2022). 801 15. K. S. Robinson, D. E. T. Teo, K. S. Tan, G. A. Toh, H. H. Ong, C. K. Lim, K. Lay, B. V. 802 Au, T. S. Lew, J. J. H. Chu, V. T. K. Chow, D. Y. Wang, F. L. Zhong, B. Reversade, 803 Enteroviral 3C protease activates the human NLRP1 inflammasome in airway epithelia. 804 Science 370, eaay2002 (2020). 805 16. B. V. Tsu, C. Beierschmitt, A. P. Ryan, R. Agarwal, P. S. Mitchell, M. D. Daugherty, 806 Diverse viral proteases activate the NLRP1 inflammasome. Elife 10, e60609 (2021). 807 17. K. S. Robinson, G. A. Toh, M. J. Firdaus, K. C. Tham, P. Rozario, C. K. Lim, Y. X. Toh, Z. 808 H. Lau, S. C. Binder, J. Mayer, C. Bonnard, F. I. Schmidt, J. E. A. Common, F. L. Zhong, 809 Diphtheria toxin activates ribotoxic stress and NLRP1 inflammasome-driven pyroptosis. J 810 Exp Med 220, e20230105 (2023). 811 18. M. Pinilla, R. Mazars, R. Vergé, L. Gorse, M. Paradis, B. Suire, K. Santoni, K. S. 812 Robinson, G. A. Toh, L. Prouvensier, S. A. Leon-Icaza, A. Hessel, D. Péricat, M. Murris, 813 H. Guet-Revillet, A. Henras, J. Buyck, E. Ravet, F. L. Zhong, C. Cougoule, R. Planès, E. 814 Meunier, EEF2-inactivating toxins engage the NLRP1 inflammasome and promote 815 epithelial barrier disruption. J Exp Med 220, e20230104 (2023). 816 19. S. Bauernfried, M. J. Scherr, A. Pichlmair, K. E. Duderstadt, V. Hornung, Human NLRP1 817 is a sensor for double-stranded RNA. Science 371, eabd0811 (2021). 818 20. L.-M. Jenster, K.-E. Lange, S. Normann, A. vom Hemdt, J. D. Wuerth, L. D. J. Schiffelers, 819 Y. M. Tesfamariam, F. N. Gohr, L. Klein, I. H. Kaltheuner, S. Ebner, D. J. Lapp, J. Mayer, 820 J. Moecking, H. L. Ploegh, E. Latz, F. Meissner, M. Geyer, B. M. Kümmerer, F. I. Schmidt, 821 P38 kinases mediate NLRP1 inflammasome activation after ribotoxic stress response and 822 virus infection. J Exp Med 220, e20220837 (2023). 823 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 29 21. L. Gorse, L. Plessis, S. Wearne, M. Paradis, M. Pinilla, R. Chua, S. S. Lim, E. Pelluz, G.-A. 824 TOH, R. Mazars, C. Bomfim, F. Hervé, K. Lhaute, D. Réveillon, B. Suire, L. Ravon-825 Katossky, T. Benoist, L. Fromont, D. Péricat, K. Neil Mertens, A. Derrien, A. Terre-826 Terrillon, N. Chomérat, G. Bilien, V. Séchet, L. Carpentier, M. Fall, A. Sonko, H. Hakim, 827 N. Sadio, J. Bourdeaux, C. Cougoule, A. K. Henras, A. B. Perez-Oliva, P. Brehmer, F. J. 828 Roca, F. L. Zhong, J. Common, E. Meunier, P. Hess, Portimine A toxin causes skin 829 inflammation through ZAKα -dependent NLRP1 inflammasome activation. EMBO Mol Med 830 17, 535–562 (2025). 831 22. ZAK α -driven ribotoxic stress response activates the human NLRP1 inflammasome | 832 Science. https://www.science.org/doi/10.1126/science.abl6324. 833 23. D. L. Brautigan, S. Shenolikar, Protein Serine/Threonine Phosphatases: Keys to Unlocking 834 Regulators and Substrates. Annu Rev Biochem 87, 921–964 (2018). 835 24. V. Janssens, J. Goris, Protein phosphatase 2A: a highly regulated family of serine/threonine 836 phosphatases implicated in cell growth and signalling. Biochem J 353, 417–439 (2001). 837 25. D. Kerk, C. White-Gloria, J. J. Johnson, G. B. Moorhead, Eukaryotic-like phosphoprotein 838 phosphatase (PPP) enzyme evolution: interactions with environmental toxins and regulatory 839 proteins. Biosci Rep 43, BSR20230378 (2023). 840 26. R. E. Honkanan, B. A. Codispoti, K. Tse, A. L. Boynton, Characterization of natural toxins 841 with inhibitory activity against serine/threonine protein phosphatases. Toxicon 32, 339–350 842 (1994). 843 27. Cantharidin, a potent and selective PP2A inhibitor, induces an oxidative 844 stress/i9 independent growth inhibition of pancreatic cancer cells through G2/M cell/i9 cycle 845 arrest and apoptosis. 846 28. H. Fujiki, E. Sueoka, T. Watanabe, M. Suganuma, The concept of the okadaic acid class of 847 tumor promoters is revived in endogenous protein inhibitors of protein phosphatase 2A, 848 SET and CIP2A, in human cancers. J Cancer Res Clin Oncol 144, 2339–2349 (2018). 849 29. L. Moed, T. A. Shwayder, M. W. Chang, Cantharidin Revisited: A Blistering Defense of an 850 Ancient Medicine. Arch Dermatol 137, 1357–1360 (2001). 851 30. J. Regan, A. Capolino, P. F. Cirillo, T. Gilmore, A. G. Graham, E. Hickey, R. R. Kroe, J. 852 Madwed, M. Moriak, R. Nelson, C. A. Pargellis, A. Swinamer, C. Torcellini, M. Tsang, N. 853 Moss, Structure-activity relationships of the p38alpha MAP kinase inhibitor 1-(5-tert-butyl-854 2-p-tolyl-2H-pyrazol-3-yl)-3-[4-(2-morpholin-4-yl-ethoxy)naph- thalen-1-yl]urea (BIRB 855 796). J Med Chem 46, 4676–4686 (2003). 856 31. S. Scarneo, P. Hughes, R. Freeze, K. Yang, J. Totzke, T. Haystead, Development and 857 Efficacy of an Orally Bioavailable Selective TAK1 Inhibitor for the Treatment of 858 Inflammatory Arthritis. ACS Chem. Biol. 17, 536–544 (2022). 859 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 30 32. K. Huang, Q. Zhang, H. Wan, X.-X. Ban, X.-Y. Chen, X.-X. Wan, R. Lu, Y. He, K. Xiong, 860 TAK1 at the crossroads of multiple regulated cell death pathways: from molecular 861 mechanisms to human diseases. FEBS J 292, 3849–3877 (2025). 862 33. N. S. Benedik, M. Proj, C. Steinebach, M. Sova, I. Sosi č , Targeting TAK1: Evolution of 863 inhibitors, challenges, and future directions. Pharmacol Ther 267, 108810 (2025). 864 34. M. Scholaert, M. Peries, E. Braun, J. Martin, N. Serhan, A. Loste, A. Bruner, L. Basso, B. 865 Chaput, E. Merle, P. Descargues, E. Pagès, N. Gaudenzio, Multimodal profiling of 866 biostabilized human skin modules reveals a coordinated ecosystem response to injected 867 mRNA-1273 COVID-19 vaccine. Allergy 79, 3341–3359 (2024). 868 35. J.-Y. Liu, X.-E. Chen, Y.-L. Zhang, Insights into the key interactions between human 869 protein phosphatase 5 and cantharidin using molecular dynamics and site-directed 870 mutagenesis bioassays. Sci Rep 5, 12359 (2015). 871 36. P. O. Corda, M. Bollen, D. Ribeiro, M. Fardilha, Emerging roles of the Protein Phosphatase 872 1 (PP1) in the context of viral infections. Cell Commun Signal 22, 65 (2024). 873 37. R. Gohmann, D. Mackey, Protein phosphatase 2A: a high-value target of virulence factors. 874 Trends in Parasitology 39, 803–805 (2023). 875 38. A. Vervaeke, M. Lamkanfi, MAP Kinase Signaling at the Crossroads of Inflammasome 876 Activation. Immunol Rev 329, e13436 (2025). 877 39. S. I. Kim, J. H. Kwak, L. Wang, M. E. Choi, Protein Phosphatase 2A Is a Negative 878 Regulator of Transforming Growth Factor-β 1-induced TAK1 Activation in Mesangial 879 Cells. J Biol Chem 283, 10753–10763 (2008). 880 40. P. Parameswaran, L. Payne, J. Powers, M. Rashighi, M. H. Orzalli, A viral E3 ubiquitin 881 ligase produced by herpes simplex virus 1 inhibits the NLRP1 inflammasome. J Exp Med 882 221, e20231518 (2024). 883 41. A. Sajid, G. Arora, A. Singhal, V. C. Kalia, Y. Singh, Protein Phosphatases of Pathogenic 884 Bacteria: Role in Physiology and Virulence. Annu Rev Microbiol 69, 527–547 (2015). 885 42. P. Sandal, C. J. Jong, R. A. Merrill, G. J. Kollman, A. H. Paden, E. G. Bend, J. Sullivan, R. 886 C. Spillmann, V. Shashi, A. T. Vulto-van Silfhout, R. Pfundt, B. B. A. de Vries, P. P. Li, L. 887 S. Bicknell, S. Strack, De novo missense variants in the PP2A regulatory subunit PPP2R2B 888 in a neurodevelopmental syndrome: potential links to mitochondrial dynamics and 889 spinocerebellar ataxias. Hum Mol Genet 34, 193–203 (2025). 890 43. J. Rehwinkel, P. Mehdipour, ADAR1: from basic mechanisms to inhibitors. Trends in Cell 891 Biology 35, 59–73 (2025). 892 44. P. Venkatesan, First treatment approved for molluscum contagiosum. The Lancet Infectious 893 Diseases 23, e402 (2023). 894 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 31 45. N. Kurmasheva, A. Said, B. Wong, P. Kinderman, X. Han, A. H. F. Rahimic, A. Kress, M. 895 E. Carter-Timofte, E. Holm, D. van der Horst, C. F. Kollmann, Z. Liu, C. Wang, H.-D. 896 Hoang, E. Kovalenko, M. Chrysopoulou, K. S. Twayana, R. N. Ottosen, E. B. Svenningsen, 897 F. Begnini, A. E. Kiib, F. E. H. Kromm, H. J. Weiss, D. Di Carlo, M. Muscolini, M. 898 Higgins, M. van der Heijden, R. Arulanandam, A. Bardoul, T. Tong, A. Ozsvar, W.-H. 899 Hou, V. R. Schack, C. K. Holm, Y. Zheng, M. Ruzek, J. Kalucka, L. de la Vega, W. A. M. 900 Elgaher, A. R. Korshoej, R. Lin, J. Hiscott, T. B. Poulsen, L. A. O’Neill, D. G. Roy, M. M. 901 Rinschen, N. van Montfoort, J.-S. Diallo, H. F. Farin, T. Alain, D. Olagnier, Octyl itaconate 902 enhances VSVΔ 51 oncolytic virotherapy by multitarget inhibition of antiviral and 903 inflammatory pathways. Nat Commun 15, 4096 (2024). 904 46. K. S. Robinson, G. A. Toh, P. Rozario, R. Chua, S. Bauernfried, Z. Sun, M. J. Firdaus, S. 905 Bayat, R. Nadkarni, Z. S. Poh, K. C. Tham, C. R. Harapas, C. K. Lim, W. Chu, C. W. S. 906 Tay, K. Y. Tan, T. Zhao, C. Bonnard, R. Sobota, J. E. Connolly, J. Common, S. L. Masters, 907 K. W. Chen, L. Ho, B. Wu, V. Hornung, F. L. Zhong, ZAKα -driven ribotoxic stress 908 response activates the human NLRP1 inflammasome. Science 377, 328–335 (2022). 909 910 Acknowledgments 911 The authors acknowledge all lab members as well as members from FL Zhong’s and P. 912 Mitchell’s labs from fruitful discussions. for their critical advices and support on this project. 913 The authors also acknowledge the IPBS microscopy, cytometry and histology platforms from 914 IPBS and I2MC institutes. 915 916 Fundings 917 Facility TRI-IPBS received financial support from ITMO Cancer Aviesan (Alliance Nationale 918 Pour les Sciences de la Vie et de la Santé, National Alliance for Life Science and Health) within 919 the framework of the Cancer Plan. This project was supported by the Agence Nationale de la 920 Recherche (ANR-PSICOPAK, ANR-INFLAMATOX, ANR-INTOX) and the European 921 Research Council (StG INFLAME 804249) to E. Meunier, an ANR-COMETH- to C. Cougoule, 922 an Invivogen-Conventions industrielles de formation par la recherche (CIFRE) PhD grant to L. 923 Ravon-Kattowsky, and a PhD fellowship from the foundation Air Liquide and the Region 924 Occitanie to A. Gomes. 925 926 Author contribution 927 Conceptualization: MP, EM 928 Methodology: MP, LG, NG, FLZ, DO, EM 929 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 32 Investigation: MP, LG, LSDC, RC, GA, AM, LB, TB, AG, CB, LRK, BS, LF, RSO, DP, EM 930 Visualization: MP, LG, DO, EM, FLZ, NG, VS, LB, CC 931 Funding acquisition: EM, CC, DO 932 Project administration: EM 933 Supervision: EM, RM, CC, DO 934 Writing – original draft: MP, LG, EM 935 Writing – review & editing: MP, LG, EM, DO, RM 936 937 Competing interests 938 Authors declare that they have no competing interests. 939 All data are available in the main text or the supplementary materials. 940 All tools generated in this study are available upon request to [email protected] 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 33 957 958 959 960 Figures 961 Figure 1 962 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 34 963 964 Figure 1. A pharmacological screen identifies PP1/PP2A phosphatase inhibitors and toxins965 as inducers of the human NLRP1 inflammasome 966 967 ns .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 35 A, B. Screening methodology, associated quantifications of ASC-GFP specks in HEK293T cells 968 individually expressing or not NLRP1 and IL-1 β release in NTERT-keratinocytes (WT and 969 NLRP1KO) exposed to 1µM /i9 of compounds from the phosphatase screening library for 8 /i9 h. 970 ASC-GFP (green) pictures were taken with an EVOS7000 after adding Hoechst (nuclei staining) 971 directly in medium. The percentage of ASC complex was performed by determining the ratios 972 between cells positive for ASC speckles and the total of cell nuclei (Hoechst) by automatic 973 fluorescence microscopy. At least ten fields from each experiment were analyzed. Values are 974 expressed as mean/i9 ±/i9 SEM. ***P/i9≤/i9 0.0001, one-way ANOVA. Graphs show one experiment 975 performed in triplicates at least three times. 976 977 C. Immunoblotting of GSDMD and analysis of the subsequent IL-1β release in WT and NLRP1-978 deficient NTERT-keratinocytes after 8 /i9 h exposure to Dinophysis toxin (Dino. toxin, 100nM), 979 Okadaic acid (Ok. Acid, 250nM) and Cantharidin (Canth. 5µM). Immunoblots show combined 980 lysates+supernatants from one experiment performed at least three times. For cytokine release, 981 ***P/i9≤/i9 0.0001, two-way ANOVA with multiple comparisons. Values are expressed as 982 mean/i9 ±/i9 SEM. Graphs show one experiment performed in triplicates at least three times. 983 984 D. Immunoblotting of PP1 and PP2A catalytic subunits (PP1C α or PP1C γ and PP2Ac α or 985 PP2Acβ ) and analysis of the subsequent IL-1 β release in WT and NLRP1-deficient NTERT-986 keratinocytes 24 hours after CRISPR-Cas9 (RNP)-mediated PP1/PP2A catalytic subunit 987 invalidation. For cytokine release, ***P /i9≤/i9 0.0001, two-way ANOVA with multiple 988 comparisons. Values are expressed as mean /i9 ±/i9 SEM. Graphs show one experiment performed 989 in triplicates at least three times. 990 991 992 993 994 995 Figure 2 996 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 36 997 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 37 Figure 2. PP1/PP2 A-targeting toxins activate the NLRP1 inflammasome in a ZAK α -998 independent manner 999 1000 A. Schematic representation of the mechanism of ZAK α /P38 stress kinases activation upon 1001 induction of Ribotoxic Stress Response (RSR). 1002 1003 B. Phosphotag blotting of phosphorylated NLRP1 disordered Region (DR) in HEK293T 1004 expressing the NLRP1 DR construct (aa 86-275-GFP (described in A)) and exposed to all 1005 PP1/PP2A-targeting compounds identified in Fig. 1A/B ) or to the known RSR inducer 1006 Anisomycin (1 /i9 µg/mL) for an hour. Tubulin- α was used as internal protein loading controls. 1007 Immunoblots show lysates from one experiment performed at least two times. 1008 1009 C. Phosphotag blotting of phosphorylated full length NLRP1 in primary human keratinocytes 1010 exposed Dinophysis toxin (100 nM) for various time. Tubulin- α was used as internal protein 1011 loading controls. Immunoblots show lysates from one experiment performed at least two times. 1012 1013 D. Phosphotag blotting of phosphorylated ZAK α and NLRP1 disordered Region (DR) in WT or 1014 ZAK KO NTERT NLRP1 KO /i9 +/i9 86-275-SNAP keratinocytes exposed to Dinophysis toxin 1015 (100nM), Cantharidin (5%M) or Val-boro-Pro (VbP, 10µM) for an hour. Tubulin- α was used as 1016 internal protein loading controls. Immunoblots show lysates from one experiment performed at 1017 least three times. 1018 1019 E. Plasma membrane permeabilization (SYTOX Green incorporation, 16 /i9 h) and IL-1 β release 1020 evaluation (10 /i9 h) in WT, ZAK α or NLRP1 KO NTERT keratinocytes after exposure to 1021 Dinophysis toxin (100nM), Cantharidin (5%M), Okadaic acid (250nM) or Anisomycin (1µM). 1022 ***P/i9≤/i9 0.0001, one-way ANOVA. Values are expressed as mean/i9 ±/i9 SEM. Graphs show one 1023 experiment performed in triplicates at least three times. 1024 1025 F. Immunoblotting and clonal selection (clone 3, red), fluorescence microscopy and associated 1026 quantifications of ASC-GFP specks in WT or ZAK α KO HEK293T ASC-GFP/NLRP1 reporter cells 1027 exposed to Dinophysis toxin (100nM) or Anisomycin (1µM) for 5 hours. ASC-GFP (green) 1028 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 38 pictures were directly taken in dish after adding Hoechst (nuclei staining). Images shown are 1029 from one experiment and are representative of three independent experiments; scale bars, 1030 50/i9 µm. ASC complex percentage was performed by determining the ratios of cells positive for 1031 ASC speckles on the total nuclei (Hoechst). At least ten fields from each experiment were 1032 analyzed. Values are expressed as mean/i9 ±/i9 SEM. ***P/i9≤/i9 0.0001, one-way ANOVA. 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 39 Figure 3 1060 1061 1062 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 40 Figure 3. Multiple P38 MAPkinases contribute to NLRP1 inflammasome activation 1063 1064 A. Pamgene analysis of activated Serine/threonine kinases in primary human keratinocytes 1065 exposed to Dinophysis toxin (100nM) for 2 hours and subsequent determination of IL-1β release 1066 in WT NTERT-keratinocytes after 8 /i9 h exposure to Dinophysis toxin (Dino. toxin, 100nM), 1067 Okadaic acid (Ok. Acid, 250nM), Cantharidin (Canth. 5µM) or Anisomycin (1µM) in 1068 presence/absence of inhibitors of identified kinases in. For all kinases, inhibitors were used at 1069 10µM. ***P /i9≤/i9 0.0001, one-way ANOVA. Values are expressed as mean /i9 ±/i9 SEM. Graphs 1070 show one experiment performed in triplicates at least two times. 1071 1072 B. Phosphotag blotting of phosphorylated P38 kinase isoforms in NTERT keratinocytes exposed 1073 to Dinophysis toxin (Dino. toxin, 100nM), Okadaic acid (Ok. Acid, 250nM), Cantharidin (Canth. 1074 5µM) for 1 hour in presence/absence of the Pan P38 inhibitor Doramapimod (10µM). Tubulin- α 1075 was used as internal protein loading controls. Immunoblots show lysates from one experiment 1076 performed at least three times. 1077 1078 C. Immunoblotting of P38, ZAK α , NLRP1 and phosphorylated P38 kinases in WT, ZAK α KO 1079 or NLRP1 KO NTERT keratinocytes exposed to Dinophysis toxin (Dino. toxin, 100nM), 1080 Okadaic acid (Ok. Acid, 250nM), Cantharidin (Canth. 5µM) for 1 hour. Tubulin- α was used as 1081 internal protein loading controls. Immunoblots show lysates from one experiment performed at 1082 least three times. 1083 1084 D. Immunoblotting characterization of the P38 isoform genetic knockdown (CRISPR-Cas9) and 1085 of the subsequent IL-1 β release in WT, P38 δ KO, P38 α /β d K O , o r P 3 8α /β /δ TKO NTERT 1086 keratinocytes exposed or not to Dinophysis toxin (Dino. toxin, 100nM), Okadaic acid (Ok. Acid, 1087 250nM), Cantharidin (Canth. 5µM) or Anisomycin (1µM) for 8 hours. ***P /i9≤/i9 0.0001, one-1088 way ANOVA. Values are expressed as mean/i9 ±/i9 SEM. Graphs show one experiment performed 1089 in triplicates at least three times. 1090 1091 E. Western blot showing NLRP1 (anti-NLRP1 N-terminal antibody (aa 1–323)) and associated 1092 fluorescence microscopy/quantifications of ASC-GFP specks in HEK293 ASC-GFP reporter cells 1093 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 41 reconstituted with hNLRP1 or hNLRP1 plasmid constructs mutated for important 38 1094 phosphorylation sites (S107A, TST112-114AAA and TST178-180AAA) after 10/i9 h of exposure 1095 to Dinophysis toxin (100nM), Okadaic acid (250nM), Cantharidin (5µM) or Val-boro-Pro (VbP, 1096 10µM). ASC-GFP (green) pictures were taken in the dish after toxin exposure. Images shown are 1097 from one experiment and are representative of three independent experiments; scale bars, 1098 10/i9 µm. ASC complex percentage was performed by determining the ratios of cells positive for 1099 ASC speckles (green, GFP) on the total nuclei (Hoechst). At least ten fields from three 1100 independent experiments were analyzed. Values are expressed as mean /i9 ±/i9 SEM. 1101 ***P/i9≤/i9 0.0001, two-way ANOVA with multiple comparisons. Graphs show one experiment 1102 performed in triplicate at least three times. 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 42 Figure 4 1125 1126 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 43 Figure 4. TAK1 apical MAP kinase both activates P38 kinases and directly contributes to 1127 triggering of the hNLRP1 inflammasome 1128 A. Quantifications of ASC-GFP specks in HEK293 ASC-GFP/NLRP1 reporter cells exposed to 1129 Dinophysis toxin (100nM) or not for 6 /i9 h in presence or absence of various MAPK inhibitors 1130 (10µM). TAK1 inhibitor; HS-276, ZAK α inhibitor; PLX4720, TAOK inhibitor; CP-43, MLKL 1131 inhibitor; necro sulfonamide, ASK1 inhibitor; GS-444217, DLK/LZK inhibitor; DN-1289, 1132 RIPK3 inhibitor; GSK-872. ASC-GFP (green) pictures were directly taken in dish after adding 1133 Hoechst (nuclei staining). Images shown are from one experiment and are representative of three 1134 independent experiments; scale bars, 10 /i9 µm. ASC complex percentage was performed by 1135 determining the ratios of cells positive for ASC speckles on the total nuclei (Hoechst). At least 1136 ten fields from each experiment were analyzed. Values are expressed as mean /i9 ±/i9 SEM. 1137 ***P/i9≤/i9 0.0001, one-way ANOVA. 1138 1139 B. Determination of the IL-1 β release in WT, P38 α /β /δ TKO, TAK1 KO and NLRP1 KO 1140 NTERT keratinocytes exposed or not to Dinophysis toxin (Dino. toxin, 100nM), Okadaic acid 1141 (Ok. Acid, 250nM), Cantharidin (Canth. 5µM), Anisomycin (1µM) and VbP (10µM) for 10 1142 hours. ***P /i9≤/i9 0.0001, one-way ANOVA. Values are expressed as mean /i9 ±/i9 SEM. Graphs 1143 show one experiment performed in triplicates at least three times. 1144 1145 C. Immunoblotting of P38, TAK1, cleaved GSDMD and IL-1 β and of the subsequent IL-1 β 1146 release in WT, P38 α /β /δ TKO, TAK1 KO, P38 α /β /δ TKO/TAK1 KO and NLRP1 KO NTERT 1147 keratinocytes exposed or not to Dinophysis toxin (Dino. toxin, 100nM), Okadaic acid (Ok. Acid, 1148 250nM), Cantharidin (Canth. 5µM), Anisomycin (1µM) and VbP (10µM) for 10 hours. 1149 ***P/i9≤/i9 0.0001, one-way ANOVA. Values are expressed as mean/i9 ±/i9 SEM. Graphs show one 1150 experiment performed in triplicates at least three times. 1151 1152 D. Phosphotag blotting of phosphorylated P38, TAK1 and NLRP1-DR-SNAP in WT, P38 α /β /δ 1153 TKO, TAK1 KO or P38 α /β /δ TKO/TAK1 KO NTERT keratinocytes exposed to Dinophysis 1154 toxin (Dino. toxin, 100nM), Okadaic acid (Ok. Acid, 250nM), Cantharidin (Canth. 5µM) for 1 1155 hour. Tubulin-α was used as internal protein loading controls. Immunoblots show lysates from 1156 one experiment performed at least three times. 1157 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 44 1158 E. Phosphotag immunoblotting of phosphorylated recombinant NLRP1 full-length protein or 1159 immunoprecipitated GFP-tagged NLRP1-DR incubated with recombinant TAK1-TAB1 fusion 1160 or P38α kinases for 60 minutes in presence/absence of lambda phosphatase. Immunoblots show 1161 proteins from one experiment performed at least three times. 1162 1163 F. Fluorescence microscopy quantifications of ASC-GFP specks in WT, P38 α /β /δ TKO or in 1164 P38α /β /δ TKO/TAK1 KO HEK293 ASC-GFP/NLRP1 reporter cells after 6 /i9 h of exposure to 1165 Dinophysis toxin (100nM) or Val-boro-Pro (VbP, 10µM). ASC complex percentage was 1166 performed by determining the ratios of cells positive for ASC speckles (green, GFP) on the total 1167 nuclei (Hoechst). At least ten fields from three independent experiments were analyzed. Values 1168 are expressed as mean /i9 ±/i9 SEM. ***P /i9≤/i9 0.0001, two-way ANOVA with multiple 1169 comparisons. Graphs show one experiment performed in triplicate at least three times. 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 45 Figure 5 1189 1190 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 46 Figure 5. TAK1 and P38 kinases play a major role in hNLRP1 inflammasome response in 1191 Human native skin models of toxin exposure 1192 A-C. Hematoxylin (H) & Eosin (E) or ASC immunobiological staining showing P38 and TAK1-1193 dependent histological/inflammasome changes caused by Dinophysis toxin (250 nM), Okadaic 1194 acid (600 nM) and Cantharidin (10µM) exposure for 24 hours. When specified, pan P38 kinase 1195 inhibitor Doramapimod (10µM) and TAK1 inhibitor HS-276 (20µM) were used. Associated 1196 quantification of IL-1 β release, the dermal–epidermal layer detachment and the percentage of 1197 ASC specks in the Human skin explants. P values indicated in figure, one-way ANOVA. Images 1198 are representative of two (A) and three (B) biological replicates. Scale bar/i9 =/i9 50/i9 µm. 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 47 Supplemental information 1222 1223 Supplemental Figures 1224 1225 1226 1227 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 48 Supplemental Figure 1 (Refers to Figure 1). Multiple PP1/PP2A inhibitors trigger the 1228 human NLRP1 inflammasome activation 1229 A. Fluorescence microscopy of ASC-GFP specks in WT HEK293T ASC-GFP/NLRP1 reporter cells 1230 exposed to selected phosphatase inhibitory compounds for 8 hours. ASC-GFP (green) pictures 1231 were directly taken in dish after adding Hoechst (nuclei staining). Images shown are from one 1232 experiment and are representative of three independent experiments; scale bars, 50 /i9 µm. ASC 1233 complex percentage was performed by determining the ratios of cells positive for ASC speckles 1234 on the total nuclei (Hoechst). 1235 1236 B. Fluorescence microscopy of ASC-GFP specks in multiple HEK293T ASC-GFP reporter cells 1237 expressing NLRP1, NLRP10, PYRIN or NLRP3 inflammasome-forming sensors and exposed to 1238 selected Dinophysistoxin (100nM) compounds for 8 hours. ASC-GFP (green) pictures were 1239 directly taken in dish after adding Hoechst (nuclei staining). Images shown are from one 1240 experiment and are representative of three independent experiments; scale bars, 50 /i9 µm. ASC 1241 complex percentage was performed by determining the ratios of cells positive for ASC speckles 1242 on the total nuclei (Hoechst). 1243 1244 C. Cytokine analysis 8 /i9 h after exposure of WT and NLRP1KO TERT keratinocytes to 1245 Dinophysistoxin (100nM), Okadaic acid (250nM) and Cantharidin (5µM). Representative 1246 experiment of three independent replicates. 1247 1248 D. Determination of IL-1 β release in WT NTERT-keratinocytes after 8 /i9 h exposure to 1249 Dinophysis toxin (Dino. toxin, 100nM) in presence/absence of bortezomib (1µM). 1250 ***P/i9≤/i9 0.0001, one-way ANOVA. Values are expressed as mean/i9 ±/i9 SEM. Graphs show one 1251 experiment performed in triplicates at least three times. 1252 1253 1254 1255 1256 1257 1258 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 49 1259 1260 1261 1262 1263 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 50 Supplemental Figure 2 (Refers to Figure 4). TAK1-driven NLRP1 inflammasome 1264 activation expands to dsRNA and viral infections 1265 1266 A. Determination of IL-1β release in WT, P38 TKO, TAK1 KO, NRLP1 KO, ZAKα KO, TAK1 1267 KO/ P38 TKO and ZAK α KO/ TAK1 KO NTERT-keratinocytes after 24 /i9 h infection Sindbis, 1268 VSV and VSVM51 viruses (MOI 10) or after Poly:IC (5µg/mL) transfection. ***P /i9≤/i9 0.0001, 1269 one-way ANOVA. Comparisons of each treatment in various genotype to it respective WT 1270 condition. Values are expressed as mean /i9 ±/i9 SEM. Graphs show one experiment performed in 1271 triplicates at least three times. 1272 1273 B. Immunoblotting of P38, TAK1, cleaved GSDMD and IL-1 β in WT, P38 α /β /δ TKO, TAK1 1274 KO, P38α /β /δ TKO/TAK1 KO in NTERT keratinocytes infected or not with Sindbis virus (MOI 1275 10) for 30 hours or transfected with poly:IC (5µg/mL). Pictures show one experiment performed 1276 at least two times. 1277 1278 C. Fluorescence microscopy and associated quantifications of ASC-GFP specks in ZAK α KO 1279 A549ASC-GFP/NLRP1 reporter cells exposed to VbP (5µM) or transfected with poly:IC (5µg/mL) for 1280 6 hours in presence/absence of the PP2A activator ATUX-1215 (15µM). ASC-GFP (green) 1281 pictures were directly taken in dish after adding Hoechst (nuclei staining). Images shown are 1282 from one experiment and are representative of three independent experiments; scale bars, 1283 50/i9 µm. ASC complex percentage was performed by determining the ratios of cells positive for 1284 ASC speckles on the total nuclei (Hoechst). 1285 1286 1287 1288 1289 1290 1291 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 51 1292 1293 1294 1295 1296 1297 1298 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 52 Supplemental Figure 3 (Refers to Figure 5). Extended results regarding Human native 1299 skins 1300 A. Evaluation of IL-1β release upon exposure of Human skin explants to Dinophysistoxin (250 1301 nM), Okadaic acid (600 nM) and Cantharidin (10µM) for various time. ***P /i9≤/i9 0.0001, one-1302 way ANOVA compared to their respective PBS controls. Values are expressed as 1303 mean/i9 ±/i9 SEM. Graphs show one experiment performed in duplicate at least two times. 1304 B. Hematoxylin (H) & Eosin (E), ASC-NLRP1 Proximity Ligation Assay (PLA) staining and 1305 associated quantifications showing ASC and NLRP1 in close proximity after exposure of Human 1306 native skins to Dinophysis toxin (250 nM), Okadaic acid (600 nM) and Cantharidin (10µM) for 1307 24 hours. P values indicated in figure, one-way ANOVA. Images are representative of three 1308 biological replicates. Scale bar/i9 =/i9 50/i9 µm. 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 53 Supplemental material 1330 Table S1. List of all cell lines used in this study is provided in Table S1. 1331 Cell line Source Catalog number A549 ASC-GFP Invivogen a549-ascg A549 ASC-GFP NLRP1 Invivogen A549-ascgnlrp1 HEK ASC-GFP Previous study (21) Previous study (21) MTA-Invivogen- E.Meunier HEK ASC-GFP NLRP1 Previous study (21) Previous study (21) MTA-Invivogen- E.Meunier HEK ASC-GFP NLRP3 Previous study (21) Previous study (21) MTA-Invivogen- E.Meunier HEK ASC-GFP NLRP10 Previous study (21) Previous study (21) MTA-Invivogen- E.Meunier HEK ASC-GFP PYRIN This study ([email protected]) This study (Etienne.meunier@ip bs.fr ) MTA-Invivogen-E. Meunier Normal Human Epidermal Keratinocytes (NHEK) juvenile foreskin, pooled PromoCell C-12007 k-NTERT WT Previous study Previous study .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint 54 (46) (46) k-NTERT NLRP1 KO Previous study (46) Previous study (46) k-N/TERT ZAKα KO Previous study (46) Previous study (46) k-NTERT RNP Control + 86-275 DR-SNAP Previous study (21) Previous study (21) k-NTERT ZAKα KO + 86-275 DR- SNAP Previous study (21) Previous study (21) k-NTERT TAK1 KO + 86-275 DR- SNAP This study ([email protected] g) This study ([email protected] du.sg) k-NTERT TAK1, ZAKα KO + 86- 275 DR-SNAP This study ([email protected] g) This study ([email protected] du.sg) k-NTERT p38α , p38β KO + 86-275 DR-SNAP This study ([email protected]) This study k-NTERT p38δ KO + 86-275 DR- SNAP This study ([email protected]) This study k-NTERT p38 total (α , β , δ ) KO + 86-275 DR-SNAP This study ([email protected]) This study k-NTERT p38 total (α , β , δ ), TAK1 KO + 86-275 DR-SNAP This study ([email protected]) This study 1332 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 25, 2026. ; https://doi.org/10.64898/2026.01.23.701233doi: bioRxiv preprint