Identification of SARS-CoV2-mediated suppression of NRF2 signaling reveals a potent antiviral and anti-inflammatory activity of 4-octyl-itaconate and dimethyl fumarate

Antiviral strategies to inhibit Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV2) and the pathogenic consequences of COVID-19 are urgently required. Here we demonstrate that the NRF2 anti-oxidant gene expression pathway is suppressed in biopsies obtained from COVID-19 patients. Further, we uncover that NRF2 agonists 4-octyl-itaconate (4-OI) and the clinically approved dimethyl fumarate (DMF) induce a cellular anti-viral program, which potently inhibits replication of SARS-CoV2 across cell lines. The anti-viral program extended to inhibit the replication of several other pathogenic viruses including Herpes Simplex Virus-1 and-2, Vaccinia virus, and Zika virus through a type I interferon (IFN)-independent mechanism. In addition, induction of NRF2 by 4-OI and DMF limited host inflammatory responses to SARS-CoV2 infection associated with airway COVID-19 pathology. In conclusion, NRF2 agonists 4-OI and DMF induce a distinct IFN-independent antiviral program that is broadly effective in limiting virus replication and suppressing the pro-inflammatory responses of human pathogenic viruses, including SARS-CoV2. One Sentence Summary NRF2 agonists 4-octyl-itaconate (4-OI) and dimethyl fumarate inhibited SARS-CoV2 replication and virus-induced inflammatory responses, as well as replication of other human pathogenic viruses.

The 2020 SARS-CoV2 pandemic emphasizes the urgent need to identify cellular factors and 72 pathways that can be targeted by new broad-spectrum anti-viral therapies. Viral infections 73 usually cause disease in humans through both direct cytopathogenic effects and through 74 excessive inflammatory responses of the infected host. This also seems to be the case with 75 SARS-CoV2 as COVID-19 patients develop cytokine storms that are very likely to contribute to, 76 if not drive, immunopathology and severe disease(1, 2). For these reasons, anti-viral therapies 77 must aim to not only inhibit viral replication but also to limit inflammatory responses of the host. 78 Nuclear factor (erythroid-derived 2) -like 2 (NRF2) functions as a cap´n´collar basic leucine 79 zipper family of transcription factors characterized structurally by the presence of NRF2-ECH 80 homology domains (3). At homeostasis, NRF2 is kept inactive in the cytosol by its inhibitor 81 protein KEAP1 (Kelch-like ECH-associated protein 1), which targets NRF2 for proteasomal 82 degradation(4). In response to oxidative stress, KEAP1 is inactivated and NRF2 is released to 83 induce NRF2-responsive genes. In general, the genes under the control of NRF2 protect against 84 stress-induced cell death and NRF2 has thus been suggested as the master regulator of tissue 85 damage during infection (5). Importantly, NRF2 is now demonstrated as an important regulator of 86 the inflammatory response (6,7) and functions as a transcriptional repressor of inflammatory 87 genes, most notably interleukin (IL-) 1b, in murine macrophages (8). Recent reports have now 88 demonstrated that NRF2 is induced by several cell derived metabolites including itaconate and 89 fumarate, to limit inflammatory responses to stimulation of TLR signaling with 90 lipopolysaccharide stimulation (9). The chemically synthesized and cell-permeable derivative of 91 itaconate, 4-octyl-itaconate (4-OI) was then demonstrated to be a very potent NRF2 inducer (9). 92 Of special interest is the derivative of fumarate, dimethyl fumarate (DMF), a US Food and Drug 93 Administration (FDA) approved drug, which is used as an anti-inflammatory therapeutic in 94 multiple sclerosis (MS) and demonstrated, at least in animal models, a potent capacity to 95 suppress pathogenic inflammation through a Nrf2-dependent mechanism (10,11). 96 Besides limiting the inflammatory response to LPS, induction of NRF2 by 4-OI also inhibits the 97 Stimulator of Interferon Genes (STING) antiviral pathway along with interferon (IFN) 98 stimulated gene expression (12). In opposition to this anti-viral effect of NRF2 on the IFN-99 response a recent single-cell RNA-seq analysis has demonstrated that NRF2 gene expression 100 signatures correlated negatively with susceptibility to HSV1 infection (13). If NRF2 agonists can 101 be used to limit viral replication of SARS-CoV2 or other pathogenic viruses is, however, not 102 known. 103 Here we demonstrate that expression of NRF2-dependent genes is suppressed in biopsies from 104 COVID-19 patients and that treatment of cells with NRF2 agonists 4-OI and DMF induces a 105 strong anti-viral program that limits SARS-CoV2 replication. The anti-viral effect of activating 106 NRF2 extended to other pathogenic viruses including Herpes Simplex Virus-1 and-2 (HSV-1 and 107 HSV-2), Vaccinia Virus (VACV), and Zika Virus (ZIKV). Further, 4-OI and DMF limited the 108 release of pro-inflammatory cytokines in response to SARS-CoV2 infection and to virus-derived 109 ligands through a mechanism that limits IRF3 dimerization. In summary, we demonstrate that 110 NRF2 agonists are plausible broad-spectrum anti-viral and anti-inflammatory agents and we 111 suggest a repurposing of the already clinically approved DMF for the treatment of SARS-CoV2. 112 113 114

NRF2 dependent anti-oxidant response is suppressed in COVID-19 patient biopsies 117
To identify host factors or pathways that are important for controlling SARS-CoV2 infection, 118 publicly available transcriptome data sets including transcriptome analysis of lung biopsies from 119 COVID-19 patients were analyzed using differential expression analysis (14). Here, genes linked 120 with inflammatory and anti-viral pathways, including RIG-I receptor and Toll-like receptor 121 signaling, were highly enriched in COVID-19 lung patient samples, whereas genes associated 122 with the NRF2 dependent anti-oxidant response were highly suppressed ( Fig. 1a-c). That NRF2-123 induced genes are repressed during SARS-CoV2 infections was supported by re-analysis of 124 another data-set builing on transcriptome analysis of lung autopsies obtained from five individual 125 COVID-19 patients (Desai et al., 2020) (Fig. 1d). Further, that the NRF2-pathway is repressed 126 during infection with SARS-CoV2 was also supported by in vitro experiments where the 127 expression of NRF2-inducible proteins Heme Oxygenase 1 (HO-1) and NAD(P)H quinone 128 oxydoreducatse 1 (NqO1) was repressed in SARS-CoV2 infected Vero hTMPRSS2 cells while 129 the protein levels of canonical anti-viral transcription factors such as STAT1 and IRF3 seemed 130 unaffected (Fig. S1). These data indicate that SARS-CoV2 targets the anti-oxidant NRF2 131 pathway and thus suggests that the NRF2 pathway restricts SARS-CoV2 replication. 132     Nrf2 agonist 4-OI (Fig. 2h). The effect of 4-OI was also retained in the lung cancer cell line 169 Calu-3, where SARS-CoV2 RNA levels were reduced by >2-logs (Fig. 2i), while release of 170 progeny virus was reduced by > 6-logs based on TCID50 analysis of cell supernatants (Fig. 8 low compared to what we could observe in Calu3 and Vero cells but 4-OI treatment still reduced 173 SARS-CoV2 RNA levels and release of progeny virus (Fig. 2l+m). We further tested the anti-174 viral effect towards SARS-CoV2 in primary human airway epithelial (HAE) cultures (Fig. 2n). 175 Here, 4-OI treatment also significantly reduced viral RNA levels (Fig. 2o). Interestingly, when 176 treating Calu3 cells with DMF, another known NRF2 inducer and a clinically approved drug in 177 the first-line-of treatment of multiple sclerosis, we could also observe an anti-viral effect toward 178 SARS-CoV2 replication similar in magnitude as what we had observed with 4-OI (Fig 2p-q) as 179 well as a reduced but significant effect when using Vero cells (Fig. 2r). To further evaluate if the 180 NRF2/KEAP1 axis controls an anti-viral pathway effective in inhibiting SARS-CoV2 181 replication, we used genetic activation of NRF2 by silencing of KEAP1. This approach 182 supported the 4-OI data as silencing of KEAP1 led to suppressed replication of SARS-CoV2 in 183 Calu3 cells by qPCR analysis, immunoblotting, and TCID50 analysis (Fig. 2s-u). Finally, the 184 anti-viral effect of SARS-CoV2 was not isolated to this particular isolate as the effect of 4-OI 185 was reproduced using a different SARS-CoV2 isolate obtained in Japan(16) (Fig. 2v+x). These 186 data demonstrate that NRF2 inducers 4-OI and DMF induce potent anti-viral responses that 187 efficiently inhibit SARS-CoV2 replication across multiple cellular systems. 188   Virus (HSV) type 1 and 2 we could observe that treatment with 4-OI reduced both the release of 225 progeny virus, the cellular content of virus RNA determined by RNA sequence analysis, and 226 viral protein as determined by both immunoblotting and flow cytometry (Fig. 3a-e and Fig. S3). 227 By contrast, the expression of NRF2-inducible HO-1, NqO1, and Sequestosome 1 (SQSTM1) 228 was highly increased in response to 4-OI treatment (Fig. 2c and Fig. S3). The anti-viral effect of 229 4-OI was at least partially dependent on NRF2 as silencing hereof by siRNA clearly reduced the 230 suppression of HSV1 infection by 4-OI (Fig. 2f-g). 231 Vaccinia virus (VACV) belongs to the family of human pathogenic poxviruses. We used HaCaT 232 cells, but also bone marrow derived dendritic cells (BMDCs), to test if the anti-viral effect of 4-233 OI extended to these viruses. Here we could observe that both HaCaT cells and BMDCs became 234 highly resistant to infection with VACV when these were pre-treated with 4-OI as measured by 235 plaque assay and flow cytometry (Fig. 2h-n). This seemed also to be the case for another 236 poxvirus Ectromelia virus (ECTV) as assesses by confocal imaging (Fig. 2k-l). For both HSV1 237 and VACV the anti-viral effect of 4-OI was extended to other cell type including murine cancer 238 cell line 4T1 and human renal carcinoma 786-O cells (Fig. S4). Interestingly, the anti-viral effect 239 of 4-OI was not extended to infection with vesicular stomatitis virus (VSVd51M) emphasizing 240 that the anti-viral program induced by 4-OI effectively inhibits replication of many, but not all, 241 viruses (Fig. S4). The anti-viral effect of 4-OI relied on intracellular restriction of replication, 242 since viral entry was not affected by 4-OI treatment -if anything it seemed to be slightly 243 increased (Fig. S5). 244 To determine if the anti-viral effect of 4-OI extended to an in vivo model of viral pathogenesis, 245 female C57BL6J mice were treated with 4-OI prior to vaginal inoculation with HSV; pre-treatment 246 with 4-OI decreased disease progression (Fig. S6), an effect that was enhanced in mice deficient 247 in STING (TMEM173 -/-) most likely to due to the pro-viral effect 4-OI has on the STING signaling 248 pathway (12,17), which is eliminated in these mice. 249 Finally, we tested the efficacy of 4-OI on Zika virus, an important human pathogenic virus 250 causing mild symptoms in the competent adult but severe disease when transmitted in utero (18). 251 Here, we could demonstrate that the anti-viral program induced by 4-OI reduced replication of 252 Zika virus in the human lung cancer cell line A549 and in the human liver cell line Huh-7 (Fig.  253   2o-p). Given that Vero cells are deficient in type I IFN (19), this suggested that the inhibitory 254 effect of 4-OI was actually independent of type I IFN signaling. To address this possibility, we 255 used either HaCaT cells deficient in IFN alpha receptor 2 (IFNAR2), Signal Transducer and 256 Activator of Transcription 1 (STAT1), both of which is necessary for type I IFN-signaling(20); 257 or deficient in STING, which is central to type I IFN-response to DNA viruses. Here, cells were 258 treated with 4-OI, followed by infection with HSV1 and VACV. Replication of both viruses was 259 inhibited by 4-OI in STAT1 KO cells and for HSV1 also in STING KO cells as measured by 260 plaque assay and expression of viral proteins by immunoblotting and flow cytometry (Fig. S7). 261 In conclusion, 4-OI induces an anti-viral program that operates independently of IFN signaling 262 (Fig. S7).

263
To examine what general pathways are affected by 4-OI treatment either alone or during 264 infection that could predict the antiviral mode of action itaconate, we treated HaCaT cells with 4-265 OI before infection with HSV-1. RNA was then collected and analysed by RNA sequencing. 266 Pathway analysis was then used to compare untreated cells to 4-OI treated cells either with or 267 without infection with HSV-1. This analysis identified several pathways that were either induced 268 or repressed by 4-OI treatment while re-confirming that IFN-signaling pathways was repressed 269 by 4-OI (Fig. S8). Amongst the top up-regulated genes induced by 4-OI was the heme 270 oxygensase-1, an enzyme canonicaly involved in stress detoxification, also reported to have 271 antiviral activity against amongst others Zika, Dengue and Ebola viruses (21)(22)(23). To assess 272 whether, HO-1 had any antiviral activity in our cellular system, Vero hTMPRSS2 and Calu-3 273 cells were either transfected with an overexpression plasmid encoding HO-1 or genetically 274 silenced for KEAP1 and HO-1 by siRNA, respectively before infection with SARS-CoV-2. None 275 of the treatments (HO-1 overexpression or silencing) really altered SARS-CoV-2 276 infection/replication suggesting of an HO-1-independent antiviral program induced by NRF2 277 (Fig. S9). 278 In an attempt to pin-point the anti-viral mode of action of 4-OI, we also used microscope-based 279 analysis of morphology by cell-paint technology (Fig. S10). With this analysis we are able to 280 compare morphological changes in cells treated with 4-OI to cells treated with compounds that 281 have known cellular targets and with cells treated with other compounds with reported anti-viral 282 activity towards SARS-CoV2 including Remdesivir and Hydroxychloroquine (24,25). In this 283 analysis, 4-OI was determined to have an low but significant morphological activity whitout loss 284 of cell viability. Interestingly, the activity of 4-OI did not seem to overlap with other compound 285 with known perturbation in cell morphology including Rapamycin, Bafilomycin, Tunicamycin, 286 Cyclohexamide, Emetine, Mitomycin, or Doxorubicin. Interestingly, there was also no 287 observable overlab with the activity profile of Remdesivir or hydroxychloroquine indicating that 288 the anti-viral mode of action of 4-OI is distinct from known anti-viral mechanisms.  Student's t-test to determine statistical significance where **p< 0.01, ***p<0.001, and ****p<0.0001.

4-OI and DMF suppress the inflammatory response to SARS-CoV2 309
In COVID-19, an uncontrolled pro-inflammatory cytokine storm contributes to disease 310 pathogenesis and lung damage (26). For this reason, we investigated if 4-OI and DMF could 311 inhibit expression of pro-inflammatory cytokines induced by SARS-CoV2. In Calu-3 cells, 312 infection with SARS-CoV2 increased the expression of IFNB1, C-X-C motif chemokine 10 313 (CXCL10), Tumor Necrosis Factor alpha (TNFA), IL-1B and C-C chemokine ligand 5 (CCL5). 314 Interestingly, this was abolished by pretreatment with 4-OI thus severely reducing the pro-315 inflammatory response to SARS-CoV2 (Fig. 4a-b). By contrast, expression of the NRF2 316 inducible gene HMOX1 was highly increased in response to 4-OI treatment (Fig. 4c). The 317 potential anti-inflammatory effect of 4-OI in this context was supported when using HAE 318 cultures. Here, treatment with 4-OI also reduced the expression of IFNB1, CXCL10, TNFa, and 319 CCL5 in the context of SARS-CoV2 infection (Fig. 4d-e), while increasing the expression of the 320 NRF2 inducible gene HMOX1 (Fig. 4f). A similar pattern was seen in experiments where Calu3 321 cells were treated with DMF before SARS-CoV2 infection. Here, IFNB1, CXCL10 and CCL5 322 mRNA levels were highly reduced in DMF treated cells while TNFA mRNA levels seemed 323 unaffected (Fig. 4g+h). By contrast, treatment with DMF increased the mRNA expression levels 324 of NRF2 inducible gene HMOX1 (Fig. 4i). As inflammatory responses often stem from immune 325 cells we also tested the effect of 4-OI on Peripheral Blood Mononuclear Cells (PBMCs) 326 harvested from healthy donors. Although stimulation of PBMCs with SARS-CoV2 yielded a 327 very weak induction of CXCL10 compared to sendai virus (SeV) infection, and no detectable 328 induction of other cytokines, 4-OI treatment also reduced CXCL10 mRNA levels in this context 329 (Fig. 4j). Further, when using PBMCs harvested from four individual patients with severe 330 COVID-19 and admitted to hospital Intensive Care Units (ICUs), we could conclude that in three 331 out of four patients, expression levels of CXCL10 were increased when compared to healthy 332 controls; and that in all four patients, these levels were strongly reduced to or below normal 333 when treating the PBMCs with 4-OI (Fig. 4k) indicating that 4-OI is able to relieve this RIG-I agonist M8 (Fig 4l-m), through an effect linked to the inhibition of Interferon 344 Regulatory Factor 3 (IRF3) dimerization but not of upstream phosphorylation of Tank Binding 345 Kinase 1 (TBK1) or of IRF3 expression itself (Fig 4n). Importantly, NRF2 expression itself was 346 closely associated with the inhibition of IRF3 dimerization and host antiviral gene expression, 347 since NRF2 silencing by siRNA was sufficient to restore IRF3 dimerization and limit the inhibitory 348 effect of 4-OI (Fig. 4n-o). When using the constitutively active form of IRF3, IRF3(5D) (28), 4-349 OI was still able to block IRF3 dimerization, and again, this effect was eliminated when NRF2 350 expression was suppressed by siRNA (Fig 4p-q). These data indicate that an NRF2 inducible and 351 dependent mechanism targets the induction of IFN by inhibition of IRF3 dimerization. This 352 phenomenon is likely to add to the inhibition of SARS-CoV2 induced cytokine release we could 353 observe when using NRF2 agonists. We have previously reported that 4-OI inhibits the expression 354 of STING, which is important for the induction of the IFN-response in cells stimulated with 355 cytosolic DNA (12). In line, 4-OI inhibited the IFN-response to HSV1 infection and to stimulation 356 with STING agonists dsDNA and cGAMP (Fig. 4R-T).

378
and HaCaT(q) cells were transfected with indicated plasmids before treatment with 4-OI at 125µM. In (q), HaCaT 379 cells were lipofected with siRNAs for 72h before plasmid transfection. Cells were then collected for analysis by qPCR 380 (p) and immunoblotting (q). (r-t) HaCaT cells were treated with 4-OI at 125µM before infection with HSV1 at MOI 381 0.01 or transfection with dsDNA (4 µg.mL -1 ). Cell pellets were collected for qPCR and immunoblotting at 6 and 3 382 hours respectively. In (n-t), data display data from one experiment representative of at least to independent 383 experiments. All statistical analysis were performed using a two-tailed Student's t-test to determine statistical 384 significance where **p< 0.01, ***p<0.001, and ****p<0.0001.

Discussion 387
Altogether, this study demonstrated that the expression of NRF2 dependent anti-oxidant genes was 388 significantly inhibited in COVID-19 patients, and that the NRF2 agonists 4-OI and DMF inhibited 389 both SARS-CoV2 replication, as well as the expression of associated inflammatory markers. The 390 ability of these NRF2 inducers to also reduce potentially pathogenic IFN-and inflammatory 391 responses while retaining their anti-viral properties is unique to these compounds and promotes 392 their applicability to prevent virus-induced pathology. That NRF2 might be a natural regulator of 393 IFN-responses in the airway epithelium is supported by a recent report demonstrating that NRF2 394 activity is high while IFN activity is low in the bronchial epithelium (29). As DMF is currently 395 used as an anti-inflammatory drug in relapsing-remitting MS, this drug could be easily repurposed 396 and tested in clinical trials to test its ability to limit SARS-CoV2 replication and inflammation-397 induced pathology in COVID-19 patients. Our observation that 4-OI strongly inhibits the IFN-398 response to both cytosolic DNA and d RNA, which are canonical anti-viral pathways, but still 399 retain its ability to block viral replication also suggests a spectrum of unidentified cellular 400 programs that are inducible through NRF2 and efficiently restrict viral replication independently 401 of IFNs. This is supported by already mentioned negative correlation between expression of