Inhibition of the inflammatory response to stress by targeting interaction between PKR and its cellular activator PACT

PKR is a cellular kinase involved in the regulation of the integrative stress response (ISR) and pro-inflammatory pathways. Two N-terminal dsRNA Binding Domains (DRBD) are required for activation of PKR, by interaction with either dsRNA or PACT, another cellular DRBD-containing protein. A role for PKR and PACT in inflammatory processes linked to neurodegenerative diseases has been proposed and raised interest for pharmacological PKR inhibitors. However, the role of PKR in inflammation is subject to controversy. We identified the flavonoid luteolin as an inhibitor of the PKR/PACT interaction at the level of their DRBDs using high-throughput screening of chemical libraries by homogeneous time-resolved fluorescence. This was further validated using NanoLuc-Based Protein Complementation Assay. Luteolin inhibits PKR phosphorylation, the ISR and the induction of pro-inflammatory cytokines in human THP1 macrophages submitted to oxidative stress and toll-like receptor (TLR) agonist. Similarly, luteolin inhibits induction of pro-inflammatory cytokines in murine microglial macrophages. In contrast, luteolin increased activation of the inflammasome, in a PKR-independent manner. Collectively, these data delineate the importance of PKR in the inflammation process to the ISR and induction of pro-inflammatory cytokines. Pharmacological inhibitors of PKR should be used in combination with drugs targeting directly the inflammasome.

PKR (Protein Kinase dsRNA-dependent) is one of the four eIF2α kinases which controls general protein translation and concomitantly triggers the integrative stress response through the eIF2α-independent enhanced translation of transcription factors such as ATF4 1 . In addition, PKR participates in the NF-κB signaling pathways leading to induction of pro-inflammatory cytokines. For this activation, PKR may act through its kinase activity or also through protein/protein interaction [2][3][4][5][6][7][8] . A link between PKR and the inflammasome was also reported but here, the situation is less clear as PKR has been proposed to participate in the assembly of the inflammasome, dependent 4 or not of its kinase activity 6 , to have no effect 8 or to diminish inflammasome activity through its control on translation 5 . Understanding the role of PKR in the inflammation process is of particular interest in view of studies indicating its participation in neurodegenerative diseases and other human pathologies related to inflammation. For instance, following a study showing that phosphorylation of eIF-2α was impairing memory formation 9 , cognitive studies with PKR deficient mice revealed that suppression of PKR promotes network excitability and enhanced cognition 10 .
The N-terminus of PKR contains two basic helical domains referred to as dsRNA Binding Domains (DRBD) through which PKR binds to dsRNA or to other DRBD-containing proteins. One of these, the cellular PACT protein (PKR Activator) interacts with PKR in response to a variety of cellular stresses, such as those resulting from perturbations of the endoplasmic reticulum or the oxidative phosphorylation function of the mitochondria. PACT has been demonstrated to activate PKR in vitro as well as in vivo after induction by an oxidative stress [11][12][13][14][15][16] . Indeed, such a stress prevents PACT to be sequestered as an inactive heterodimer with the TAR RNA Binding Protein (TRBP) and releases its PKR activation ability 17,18 . Colocalisation of PACT with phosphorylated PKR was observed by immunohistochemistry in the cytoplasm of hippocampal neurons of post-mortem brains of patients whith Alzheimer's disease, in line with a possible role for PKR in cognitive disorders 19 . Furthermore, oxidative stress can increase, in a PKR-dependent manner, the translation of BACE1 (beta-site APP cleaving enzyme 1), the rate-limiting enzyme involved in the generation of amyloid β (Aβ)-peptide 20 . In the brain, Aβ is known to bind to the microglial receptor complex CD36/TLR4-6 and trigger induction of pro-inflammatory cytokines, such as IL-8, IL-6 and IL1-β, similar to the action of microbial effectors, such as LPS 21 . While IL-8 and IL6 are directly released from the cells under their active form, production of IL1-β requires activation of the inflammasome for its cleavage by caspase-1 from the pro-IL1-β form. Formation of the NLRP3 inflammasome complex 22 can occur following Aβ phagocytosis and subsequent lysosomal damage which activates an oxidative stress through the plasma membrane-localized NADPH oxidase (Nox2) 23,24 . It is possible that PKR could be involved both in the generation of Aβ through its eIF-2α kinase activity and in the action of Aβ through NF-κB signaling and regulation of the inflammasome, thus raising interest to generate PKR inhibitors in order to be able to deal with neurodegenerative pathologies.
A limited number of PKR inhibitors have been previously described. Screening 26 different ATP-binding site inhibitors to target the catalytic activity of PKR led to the isolation of the oxindole/imidazole derivative C16 25 . Inhibiting PKR activation at the level of its N-terminus was demonstrated by using a cell penetrating peptide, referred to as PRI, which contains the 21-aa peptide corresponding to the first DRBD of PKR 26 . A different approach by high-throughput screening aimed at identifying molecules that protect macrophages from anthrax lethal toxin-induced cell death through NLRP1 inflammasome activation, led to the identification of a compound (7-desacetoxy-6,7-dehydrogedunin (7DG)) that binds to the C-terminus of PKR but does not interfere with the PKR kinase activity 6 .
Here, we have performed a high-throughput screening of chemical libraries to isolate molecules that can interfere with the interaction of the N-terminus of PKR with its cellular activator PACT, in order to identify novel inhibitors of PKR and to better understand the mechanism of action of PKR on the signaling pathways linked to inflammation.

Results
Set up of in vitro PKR/PACT interaction by HTRF and screening of libraries of chemical compounds. We set up an in vitro Homogeneous Time-Resolved Fluorescence (HTRF) interaction assay between PKR and PACT, using purified preparations of His-PACT and GST-PKR-Nter (Fig. 1A,B). The two proteins were incubated in the presence of the acceptor XL665 -labeled anti-GST antibody which binds to the GST moiety of the PKR construct and the donor Lumi4 Tb labeled anti-6His antibody, which binds to the His tag of His-PACT (Fig. 1C). Interaction between the two proteins resulted in transfer of energy and fluorescence which can be expressed as percentage of ΔF (see Materials and Methods). Maximum fluorescence signal was obtained after 6 hrs and did not vary over 24 hrs. Experiments reproducibly showed a ΔF value around 300-400 (Fig. 1D). Addition of an excess of the PRI peptide corresponding to the sequence located in the first DRBD of PKR 26 strongly inhibited fluorescence, thus validating our assay (Fig. 1E). DMSO, at concentration up to 2% did not significantly affect the interaction (Fig. 1F), allowing to conveniently dilute the chemical compounds at a final concentration of 1% DMSO for the screening procedure. The PKR/PACT HTRF interaction assay was then used to screen the Prestwick Chemical Library ® (1,200 compounds with high chemical and pharmacological diversity) and a total collection of 38,173 compounds, coming from the ≪ Chimiothèque Nationale ≫ and CHEM-X-INFINITY (see Materials and Methods). The screening procedure yielded 113 hits that could inhibit the transfer of fluorescence higher than 95%. Two successive HTRF assays were then performed on these hits to determine the specificity of the inhibition as well as their IC50 for efficiency of inhibition and for cytotoxicity (Supplementary Table 2) This resulted in the selection of 2 compounds from the Prestwick library (myricetin and quercetin) and 14 compounds from the other libraries, among which luteolin, a natural polyphenol, member of the flavonoid family and similar to myricetin and quercetin. Interestingly, a number of studies have reported anti-inflammatory effects for these flavonoids and interest in therapy 27 . The next step was to evaluate the effect of the selected compounds on the PKR/PACT interaction in a cellular context.

Flavonoids inhibit the PKR/PACT interaction at the level of their DRBDs. Based on a previous
experience with a split-luciferase protein-protein interaction detection system 28 , we designed a split-Nanoluc luciferase complementation assay (NPCA) to examine the effect of the compounds on the interaction between PKR and PACT in a cellular context, using HEK-293T cells. For the NPCA assay, we constructed plasmids expressing the N-terminus of PKR (1-265 aa) or the full length PACT with the two complementary nanoluc moieties (referred to as 1 and 2) inserted in all available combinations at the N-or C-terminus of PKR or PACT 29 . The highest efficiency of interaction, revealed after cotransfection of the different constructs, was obtained when these moieties are placed at the N-terminus of PKR and PACT, giving the combination PKR-N2/PACT-N1 ( Fig. 2A). The cells were then cotransfected with this pair of plasmids and 4 hrs later, incubated in the presence of different concentrations of the compounds. Luciferase activity was measured after 24 hrs. The results are shown for the Luteolin strongly inhibits PKR phosphorylation in a cellular stress assay. We then used the human THP1 macrophages as a convenient cellular model for subsequent analysis of the effects of the compounds on PKR activation as well as on induction of pro-inflammatory cytokines and activation of inflammasome. PKR activation was correctly detected upon treatment of the cells with sodium arsenite, inducing oxidative stress known to trigger PKR activation through PACT in accord with previous reports 15,31 . Kinetics experiment showed an increase in PKR phosphorylation with the maximum effect at 8 hrs post treatment (Fig. 3A) and a dose response experiment showed maximum phosphorylation for 25 μM of sodium arsenite (Fig. 3B).The cells were then incubated in the presence of luteolin, myricetin and quercetin for 30 min before incubation with 25 μM of sodium arsenite for 8 hr, and the cell extracts were analysed for PKR activation. Treatment of cells with 400 nM of C16, a small molecule inhibitor of the catalytic activity of PKR 25 , was included in the assay as control and the degree of efficiency of the compounds to inhibit PKR was estimated by comparison with the action of C16 (Fig. 3C). Luteolin presented the best ability to inhibit phosphorylation of PKR, with inhibition starting at 4 µM and being maximal around 20 µM. A ten-fold higher concentration (200 µM) was required for quercetin, while myricetin had no effect (Fig. 3C). Some decrease in the expression levels of PACT and PKR was observed for the highest concentrations of luteolin. This may indicate their enhanced susceptibility to degradation in response to stress when they are not longer in association. Altogether, these data show that, out of the three flavonoids that were identified as disrupting the PKR/PACT interaction by both HTRF and NPCA, luteolin was the most efficient as a specific inhibitor of PKR phosphorylation in response to oxidative stress.

Activation of the integrative stress response (ISR) and induction of pro-inflammatory cytokines in THP1 cells upon treatment with sodium arsenite or thapsigargin and with an LPS analog.
In the next assay, we submitted THP1 cells to oxidative stress to trigger activation of PKR as a kinase and to treatment with a TLR agonist to activate the NF-κB signaling pathway, in which PKR can participate as a scaffolding partner 2 . This allowed to examine the effect of luteolin on these pathways. Oxidative stress was triggered upon treatment with either sodium arsenite or thapsigargin, another inducer of oxidative stress known to trigger PKR activation through PACT 31 . As TLR agonist, we used Kdo 2 -Lipid A (KLA), a defined LPS from E. coli which triggers NF-κB activation through TLR4 32 . Induction of the ATF4-dependent GADD34 gene and of the pro-inflammatory cytokines IL-8 and pro-IL-1β were chosen as markers of the integrative response to stress and NF-κB activation, respectively. We first showed that similar to sodium arsenite, thapsigargin triggers PKR phosphorylation. In contrast, KLA did not trigger PKR phosphorylation, as expected since activation of NF-κB does not require the kinase function of PKR, and did not interfere with the effect of thapsigargin and sodium arsenite on this phosphorylation (Fig. 4A). THP1 macrophages were then treated for 4 and 8 hrs with either sodium arsenite (Fig. 4B) or thapsigargin (Fig. 4C). The results showed that GADD34 was induced by sodium arsenite or thapsigargin with a more sustained induction over time by thapsigargin treatment and that it was not significantly affected by KLA treatment (Fig. 4B,C; left). Conversely, induction of the NF-κB-dependent genes IL8 and pro-IL1β was responding almost exclusively to KLA (Fig. 4B,C; middle and right). In addition, we observed that Na Arsenate or thapsigargin treatment interfered with induction of pro-inflammatory cytokines by KLA. Oxidative stress is also known to activate the Nrf2 pathway, independently of PKR or PACT and this has been shown to occur in response to arsenic 33 or thapsigargin 34 . Therefore, it is possible that this observed inhibition might be the result of the well known cross-talk between the NF-κB and Nrf2 response pathways through competition at the transcriptional level, for binding to the transcriptional co-activator CBP (CREB-binding)-p300 complex 35 . However, thapsigargin proved to be less inhibitory that Na Arsenate and thapsigargin/KLA was therefore chosen over sodium arsenite/KLA to be able to correctly evaluate the effect of the PKR inhibitor on the following events of the inflammatory response in which PKR is thought to participate: the integrative stress response, the NF-κB activation pathway and the inflammasome.
Luteolin inhibits the integrative stress response and induction of pro-inflammatory cytokines in THP1 and in murine primary macrophages. We then submitted the THP1 cells to the combined treatment thapsigargin/KLA in the presence of different concentrations of luteolin. The results showed that luteolin inhibited induction of GADD34, IL8 and pro-IL1β (Fig. 5A-C). We noticed a residual induction of GADD34, even for the highest concentration of luteolin used. This residual induction could result from the activation of PERK, another eIF2α kinase, which can also be activated by thapsigargin, through Endoplasmic Reticulum (ER) stress 36 . To address this, we compared the effect of luteolin to that of inhibitors of the catalytic activity of PKR (C16) or PERK (PERKi). The effect of luteolin was found to be similar to that of C16 and both were less effective than the PERK inhibitor (Fig. 5D). These data show that, in conditions where PKR and PERK are both activated to trigger the integrative stress response, treatment with luteolin is able to significantly attenuate this effect, but not to inhibit it completely.
Because of the possible interference by the Nrf2/ NF-kB cross-talk in our assay, it was possible that the observed reduced induction of pro-inflammatory cytokines by KLA in the presence of thapsigargin might mask a role of the PKR/PACT association in the early events of NF-kB activation. To examine this, we analysed NF-kB activation 60 min post-treatment with the drugs through translocation of NF-kB p65 protein in the nucleus by confocal microscopy. The results show that only KLA triggers the translocation of NF-kB, whether alone or in the presence of thapsigargin and that luteolin can inhibit this translocation. Therefore, the ability of luteolin to attenuate the induction of the pro-inflammatory cytokines is independent from its effect on the PKR/PACT association (Fig. 5E).
The ability of luteolin to inhibit induction of pro-inflammatory cytokines was then assayed on primary macrophages, using microglia isolated from the hippocampal regions of murine embryos. In this experiment, the microglial cells were submitted to 4 hrs of treatment with 5 μM thapsigargin and 100 ng/ml KLA, in the absence or presence of different concentrations of luteolin (1, 5 and 50 μM). The murine microglial cells proved to respond Data represents means ± SD from three experiments; *P < 0.05, **P < 0.005, ***P < 0.0005, ****P < 0.00005, ns: non significant, Tukey's multiple comparison test in ANOVA.
well to the stress situation by induction of significant induction of the murine pro-inflammatory markers IL6 (6-fold) or IL1β (15-fold). Luteolin was found to inhibit these inductions in a concentration manner, with maximum effect for 50 µM and an already significant effect at 5 µM (Fig. 6). We noted that the oxidative stress response (induction of GADD34) was not as sensitive to luteolin as induction of pro-inflammatory cytokines; only a slight but significant inhibition was detected at 50 μM. Altogether, these data show that luteolin can interfere with the cellular response to stress by limiting the induction of pro-inflammatory cytokines in cells of two different mammalian species.

The PKR inhibitor luteolin increases the activation of inflammasome.
We then examined the effect of luteolin on the activation of NLRP3 inflammasome. Such activation can take place following an increase in the NLRP3 expression through NF-κB (priming event) and its assembly as a complex. The latter can be triggered in response to several different stimuli, in particular via oxidative stress. Once activated, the inflammasome triggers activation of caspase-1, subsequent cleavage of pro-IL-1β and secretion of the resulting IL-1β as well as secretion of caspase-1. Activation of NF-κB by KLA and activation of oxidative stress by thapsigargin can here be conveniently used to study activation of the inflammasome and, accordingly, we showed that treatment of THP1 cells with thapsigargin/ KLA could trigger the expression of caspase-1 and IL-1β as measured by ELISA (Fig. 7). We found that treatment with luteolin further increased the expression of Caspase-1 (Fig. 7A). These data could agree with a previous report 5 showing that PKR negatively controls the inflammasome. By preventing PKR to negatively control translation, luteolin would then favour the translation of effectors required for the formation of the inflammasome complex and subsequently its activation. Although no increase in the expression of IL-1β was observed after luteolin treatment (Fig. 7B), this could be explained because of the ability of luteolin to first inhibit its induction through NF-κB (see Fig. 5). To determine whether the hypothesis that PKR controls the inflammasome through translation was correct, we compared the action of luteolin to that of Salubrinal 37 which can strongly inhibit translation by preventing the dephosphorylation of eIF2α and therefore acts downstream of PKR. If luteolin restores translation by inhibiting PKR, then its positive effect would be lost in the presence of salubrinal. The results showed that, on the contrary, luteolin kept its ability to increase the expression of Caspase-1 or IL-1β in the presence of Salubrinal (Fig. 7A,B). Therefore, these data are arguing against a role for PKR in controlling the inflammasome through translation.

PKR and PACT are not involved in the activation of the inflammasome.
We have shown that luteolin disrupts the PKR/PACT interaction and behaves as an inhibitor of integrative stress response. We have shown also that luteolin inhibits the induction of pro-inflammatory cytokines but that it could enhance activation of the inflammasome. We next checked whether PKR or PACT could be responsible for this effect, in a kinase-independent manner. For instance, their dissociation in the presence of luteolin might have favoured association of PKR or PACT as monomers with some components of the inflammasome. The THP1 cells were silenced for either PKR or PACT using siRNAs and submitted to thapsigargin/KLA treatment in the presence or not of luteolin. An immunoblot analysis first confirmed the ability of luteolin to inhibit PKR phosphorylation in Figure 6. Luteolin inhibits the integrative stress response and induction of pro-inflammatory cytokines in murine primary macrophages. Microglia from hippocampal tissues of C57/BL6 embryos was prepared as described in Materials and Methods. The cells, in Poly L-ornithine-coated dishes, were incubated in the presence of 1, 5 or 50 μM of luteolin for 30 min and either treated or not with 5 μM of thapsigargin and 100 ng/ ml KLA for 4 hrs. Supernatants were collected for ELISA analysis of cytokines. Total RNA was isolated from harvested cells and levels of GADD34 (A), IL-6 (B) and IL-1ß (C) were determined by RT-qPCR. Data are expressed as RNA relative expression. Data represents means ± SD from three independent dishes; *P < 0.05, **P < 0.005, ****P < 0.00005, ns: non significant, Tukey's multiple comparison test in ANOVA. response to thapsigargin/KLA (Fig. 8A; compare lane 13 to 9). In addition, these data revealed that silencing of PACT also inhibits activation of PKR by thapsigargin/KLA and that this effect is aggravated in the presence of luteolin ( Fig. 8A; compare lane 10 to 9 and 14 to 13). This confirms the role of PACT and of its interaction with PKR to trigger PKR activation, in response to stress. Silencing of either PKR or PACT were indicated by significant inhibition of expression levels of these proteins (Supplemental Fig. 8A). In regards to the effect of the depletion of PKR or PACT on the activation of the inflammasome, we found that PKR or PACT, either silenced separately or together, had no influence on the activation of caspase 1 by the thapsigargin/KLA treatment, (Fig. 8B-D). Furthermore, the increase in the activation of caspase1 by luteolin was not affected by silencing PKR or PACT alone, or PKR and PACT together. Only a slight inhibition was observed by PACT silencing (Fig. 8C). Altogether, these data show that activation of inflammasome by luteolin may involve interaction with partners independent on its action on PKR and PACT and that PKR or PACT are not participating in activation of the inflammasome.

Discussion
In this study, we performed a high-throughput screening to identify compounds that target the stress kinase PKR at the level of its activation by PACT, rather than at its catalytic domain. Our screening procedure was based on in vitro HTRF assay between tagged purified preparations of the DRBD-containing N-terminus PKR and full length PACT. The selected hits were further assayed for their ability to interfere with the interaction between PKR and PACT in a cellular context, using a split-Nanoluc protein complementation assay. This led to the identification of three members of the flavonoid family: myricetin, quercetin and luteolin. Using luteolin as a representative of this chemical series, we then demonstrated that inhibition occurs at the level of interaction between the DRBDs, whether in the homo-or hetero-dimerization configuration. In other words, those compounds are able to inhibit the PKR/PKR or the PACT/PACT interaction in addition to inhibit the PKR/PACT interaction. These data reveal for the first time that the PKR/PACT association can be specifically disrupted by members of the flavonoid family. Luteolin proved to be more efficient than the two other flavonoids in a subsequent PKR phosphorylation assay following an oxidative stress and its mechanism of action was further evaluated.
Luteolin belongs to the group of natural flavonoids, or polyphenols, that is abundant in vegetables, fruits and medicinal herbs. The plants rich in luteolin, such as Chrysanthemum flowers have been long known in Chinese traditional medecine for their anti-inflammatory effects and anti-cancer effects 27 . However, luteolin has pleotropic effects and its exact mechanism of action remains ill-defined. Luteolin is generally referred to as an anti-oxidant, or ROS scavenger, either through its own oxidation or through the inhibition of ROS-generating oxidases or by protecting endogenous anti-oxidant enzymes. In addition to this direct anti-oxidant properties, luteolin was also reported to inhibit LPS-induced activation of NF-κB and MAPK pathways, but its mode of action was not understood 27 . Luteolin is blood-brain barrier permeable and was reported to have a neuroprotective effect in drug-induced Alzheimer's rat model 38 . These properties would be therefore in line with a role for PKR in impairment of cognition through inflammation. At present, there are only few clinical assays to assess the beneficial effects of members of the flavonoid family. A combination of luteolin and quercetin proved effective in reducing symptoms of autism spectrum disorders (ASD), which might be associated with inflammation in brain regions related to cognitive function 39 . Our study reveals that luteolin can inhibit two of the known functions of PKR: triggering the integrative stress response and its participation in the induction of NF-κB-dependent pro-inflammatory cytokines but, in contrast, was increasing the activity of the inflammasome, in a PKR (and PACT)-independent manner. A schematic representation of the inflammatory pathways at play in response to oxidative stress, and TLR activation is shown in Fig. 8E and the role of luteolin on these pathways is shown on Fig. 8F. The clear ability of luteolin to abrogate PKR/PACT association and PKR/PKR dimerization provided therefore a different approach to appreciate the role of PKR in mechanisms leading to inflammation, particularly in view of previous contradictory reports. For instance, it has been shown that murine macrophages either deficient for PKR due to a deletion in its catalytic domain 40 , or treated with 2-aminopurine (2AP) or the C16 compound presented inhibition of caspase 1 and production of IL-1β in response to DAMPs (Danger Associated Molecular Patterns) 4 . PKRwt, but not a catalytically inactive PKR point mutant was then shown to physically interact with NLRP3 which led to the conclusion that PKR interacts with NLRP3 and that its activity were required to activate the inflammasome 4 . In another study using macrophages from either the same PKR deficient mice or other PKR deficient mice through deletion of the first two exons of PKR 41 , no difference was observed in the activation of caspase 1 and the production of IL-1β after treatment of cells with LPS and DAMPs, leading to the opposite conclusion that PKR was not involved in the regulation of the inflammasome 8 . Compound 7DG isolated from a systematic search for inhibitors of the lethal anthrax toxin-activation of inflammasome by library screening, was reported to inhibit events leading to activation of caspase1 as well as interacting with the C-terminal half of PKR, without affecting its catalytic activity. Assembly of the inflammasome was inhibited in the presence of 7DG but not in the presence of 2AP or C16. The conclusion of the authors was that PKR was involved in inflammasome assembly but ruled out its role as a kinase in this process 6 . Recently, a novel model of PKR inactive mice has been generated in which PKR kinase activity was knocked-in by point mutation. The use of peritoneal macrophages either from these mice or from PKR deficient mice showed an increase in the formation of inflammasome (ASC nucleation), in the activity of caspase1 and secretion of IL-1β and IL-18 after cotreatment with LPS and DAMPs. The authors concluded that PKR exerts a negative control on the inflammasome at the level of protein translation, allowing to repress the induction of factors that are critical for the activity of the cryopyrin inflammasome 5 . NLRP3, as component of the inflammasome, is an important sensor of altered homeostasis and its deregulation is linked to a number of metabolic diseases. In particular, NLRP3 appears to be involved in obesity-induced inflammation and Type 2 diabetes. For instance, NLRP3-deficient mice are insulin hypersensitive when submitted to high fat diet (reviewed in 42 ). Some studies have indicated a role for PKR in obesity-induced inflammation and Type 2 diabetes 43 but again this has been questionned 44 .  D). Activation of Caspase1 in the supernatants was determined by ELISA and the values were normalised to the amount of total protein for each condition. Data represents means ± SD from three independent supernatants. (E,F) Graphical model Activation of ROS (Reactive Oxidative Stress) in response to a stress leads to dimerization of PACT, its association with PKR, dimerization and activation of PKR as a kinase (PACT and PKR are represented here only as monomers for clarity). PKR inhibits general protein translation (GT), while turning on an integrative stress response (ISR), in which the favoured translation of transcription factor ATF4, allows restoration of homeostasis, through induction of specific genes, such as GADD34. Activation of the NF-κB signalling pathway, in response to specific TLR agonists (Priming), triggers induction of pro-inflammatory cytokines such as IL-8 (not shown here) and pro-IL-1ß. PKR can participate in this NF-κB activation as a scaffold protein, through binding to TRAF6 via its C-terminus and binding to the IKK complex via its DRBDs, independently of its interaction with PACT. The conjugation of both priming and ROS-induced events trigger the inflammasome (complex NLRP3/ASC), leading to cleavage of pro-Caspase1 and pro-IL-1ß Our results show that activation of the inflammasome is increased in the presence of luteolin that affects the homo-or heterodimerization of PKR and PACT but that this effect can not be abrogated when the expression of PKR or PACT is inhibited through silencing. Therefore, our data support the notion that PKR would not play an essential role in the control of the inflammasome. The use of luteolin, following its identification as disruptor of the PKR/PACT interaction and as a novel inhibitor of PKR allowed us also to show its ability to inhibit the induction of pro-inflammatory cytokines and we suggest that this might be, at least in part, to its ability to bind the DRBDs of PKR and interfere with the interaction of PKR with the NF-κB pathway. Our data support the notion that it could be beneficial for the organism to inhibit activation of PKR as a kinase in response to stress. Indeed, this would attenuate unwanted translations of mRNAs, as part of the integrative stress response, such as translation of BACE1 with subsequent secretion of amyloid-beta. PKR is also known to trigger activation of c-Jun N-terminal kinase (JNK) and both kinases are activated in the brains in the case of Alzheimer disease and are involved in Aβ production, neuroinflammation, and neuronal death. Recently, we have shown that a dual inhibition of PKR and JNK could nearly abolish Aβ toxicity in primary cultured neurons and would therefore mediate neuroprotection 45 . Similarly, thiamine deficiency, which leads to neuronal death, activates the PKR-eIF2α pathway and increases the BACE1 expression levels of Aβ in specific thalamus nuclei. This effect could be reversed by PKR downregulation 46 .
In conclusion, the anti-oxidant luteolin has been identified here as a novel inhibitor of PKR, by preventing PKR activation through its association with PACT and preventing also homodimerization of PKR. Luteolin can inhibit both the integrative stress response and induction of pro-inflammatory cytokines, two effects related to kinase and non-kinase functions of PKR, but can also increase activation of the inflammasome in a PKR-independent manner. This indicates that the use of compounds like luteolin in inflammatory-related diseases might be taken with caution and should be associated with direct anti-NLRP3 agents, such as the recently described MCC950 47 , to have the safest and more efficient outcome in patients.
Expression vectors and antibodies. The pGEX-4T vector expressing the PKR 1-265 fragment in fusion with GST at its N-terminus (GST-PKR Nter) and the pET15b vector expressing the PACT protein have been previously described 2,15 . The pENTR ™ -SD⁄D-TOPO ® vector and the pDEST24 were purchased from Invitrogen.
Immunofluorescence. Cells were fixed in PBS containing 4% paraformaldehyde (PFA) for 20 min in room temperature and permeabilized with PBS containing 10% FBS and 0.3% Triton X-100. Non-specific antibody sites were blocked with PBS containing 2.5% BSA, 10% FBS for 30 min at room temperature. For staining, cells were incubated with antibodies diluted in this blocking solution. Primary antibodies were added for 1 hr and secondary antibodies were added for 30 min in room temperature. Nuclei were visualized by incubation with DAPI for 5 min at room temperature. and their secretion (E). Luteolin dissociates the PKR/PACT interaction, inhibits PKR phosphorylation and the IRS (since ROS activates also the IRS through PERK, the latter inhibition is not complete). Luteolin also inhibits induction of pro-inflammatory cytokines, at least in part through its ability to interfere with the DRBDs of PKR. In contrast, luteolin activates the inflammasome (through a still unknown mechanism,) in a PKR and PACTindependent manner, which rules out a role for PKR in the control of the inflammasome (F).