Antioxidant and food additive BHA prevents TNF cytotoxicity by acting as a direct RIPK1 inhibitor

BHA specifically protects cells from RIPK1 kinase-dependent cell death

TNF triggers necroptosis in mouse L929 cells upon single exposure (Fig. 1A), in MEFs and dermal fibroblasts (MDFs) by the additional presence of the pan-caspase inhibitor zVAD-fmk (Fig. 1B, C), and in human HT-29 cells by the further inhibition of IKKα/β (IKKi) (Fig. 1D). In all four cellular systems, the requirement of RIPK1 kinase activity for necroptosis induction can be demonstrated by the use of the RIPK1 inhibitor Nec-1s, which completely protects the cells from TNF cytotoxicity. In line with previous studies, we found that BHA pretreatment also greatly protected these cells from TNF-induced necroptosis (Fig. 1A–D). However, we found that BHA did not provide protection against necroptosis that does not rely on the kinase activity of RIPK1, as observed following infection of MEFs by Herpes simplex virus 1 (HSV1) [38] (Fig. 1E, F). In these experiments, the cells are pretreated for 24 h with IFNβ to induce expression of ZBP1, and are subsequently infected with a mutated strain of HSV1 (ICP6 RHIM mutant) to specifically induce ZBP1/RIPK3-dependent necroptosis [39]. As shown in Fig. 1E, F, Nec-1s and BHA similarly and marginally protected these cells from death, whereas pharmacological (Fig. 1E) and genetic (Fig. 1F) inhibition of RIPK3 completely prevented necroptosis induction. Of note, zVAD-fmk was added to the experiment making use of the RIPK3 inhibitor GSK’872 to prevent spontaneous induction of apoptosis [40].

A–D, G–K L929 cells (A), MEFs (B, G–K), MDFs (C), and HT-29 cells (D) were pretreated for 30 min with indicated compounds (100 µM BHA, 10 µM Nec-1s, 500 ng/ml CHX) before stimulation with 20 ng/ml hTNF (A–D, G–I), 10 µM etoposide (J) or 2 µM staurosporine (K) for the indicated duration. E–F MEFs of indicated genotypes were pretreated for 24 h with 100 U/ml IFNβ followed by pretreatment for 30 min with indicated compounds and infected with ICP6 RHIM mutant HSV1 fmutRHIM with an MOI of 3. Cell death was measured over time by Sytox Green (SG+) positivity. Cell death assays are presented as mean ± SEM of three independent experiments (n = 3). Statistical analysis on kinetic cell death assays with more than one timepoint is detailed in the Methods section. Significance between samples is indicated in the figures as follows: *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Full size image

Interestingly, we found that the correlation between the anti-death potential of BHA and the dependency on RIPK1 kinase activity was also true in the context of apoptosis. TNF triggers RIPK1 kinase-dependent apoptosis when combined with IKKα/β [35, 41] (IKKi) (Fig. 1G) or IKKε/TBK1 [42, 43] (TBK1i) inhibition (Fig. 1H), and RIPK1-independent apoptosis in the presence of the translational inhibitor CHX [44] (Fig. 1I). Remarkably, we observed perfect overlap in the protection provided by Nec-1s and BHA in cells undergoing TNF-induced RIPK1 kinase-dependent apoptosis (Fig. 1G, H). In addition, BHA had no effect against RIPK1-independent apoptosis induced by TNF in a combination of CHX (Fig. 1I), and by etoposide (Fig. 1J) or staurosporine (Fig. 1K) treatment.

Together, these results demonstrated that the protective effect of BHA is not specific to necroptosis but rather to RIPK1 cytotoxicity, which suggested a role for ROS in RIPK1 activation, as previously reported [26].

The protective role of BHA does not originate from ROS scavenging

In order to evaluate whether ROS has an essential role during RIPK1 kinase-dependent death, we tested the effect of other ROS scavengers on TNF-induced RIPK1 cytotoxicity, starting with the structurally related synthetic compound BHT. Interestingly, and in contrast to BHA, pretreatment with BHT did not protect MEFs, MDFs, or L929 cells from TNF-induced necroptosis (Fig. 2A, SFig. 1A–B), nor from TNF-induced RIPK1 kinase-dependent apoptosis (Fig. 2B, C). The absence of protection was also noticed in the presence of five additional antioxidants, including the hydrophilic ROS scavenger NAC (Fig. 2D–F), the ubiquinone analog and membrane-targeted ROS scavenger DeCylQ, the vitamin E isomer, and natural antioxidant α-tocopherol, its water-soluble counterpart Trolox and finally the lipid ROS scavenger Ferrostatin-1 (Fig. 2G–I). Importantly, all these antioxidants were in contrast equally efficient at preventing ferroptosis, a ROS-dependent cell death modality, induced by GPX4 inhibition via ML162 treatment or by system xc-inhibition following erastin exposure [45, 46] (Fig. 2J, K). These results demonstrated that ROS is dispensable for TNF-induced RIPK1 kinase-dependent apoptosis and necroptosis, and thereby questioned the origin of the protective effect of BHA against RIPK1 cytotoxicity.

A–K MEFs were pretreated for 30 min with indicated compounds (100 µM BHT, 5 mM NAC, 100 µM BHA, 10 µM DecylQ, 100 µM α-tocopherol, 100 µM Trolox, 500 nM Ferrostatin-1, 10 µM Nec-1s) before stimulation with 20 ng/ml hTNF (A–I), 5 µM ML162 (J), or 10 µM Erastin (K) for the indicated duration. Cell death was measured over time by Sytox Green (SG+) positivity. Cell death assays are presented as mean ± SEM of three independent experiments (n = 3). A–F Statistical analysis on kinetic cell death assays with more than one timepoint is detailed in the Methods section. G–K Statistical significance was determined via ordinary one-way ANOVA followed by a Tukey post hoc test. Significance between samples is indicated in the figures as follows: *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Full size image

BHA prevents cellular activation of RIPK1

A previous study reported that the anti-necroptotic potential of BHA was partially due to its ability to inhibit the activity of mitochondrial complex I and of lipoxygenases, which was supported by the protection against TNF-induced necroptosis obtained by their respective inhibition with rotenone and NDGA [34]. We found that the protection conferred by rotenone and NDGA likely originates from defective TNFR1 complex I assembly, as monitored by defective TRADD and RIPK1 recruitment to TNFR1 (SFig. 2A), which consequently affects global signaling from the receptor, as exemplified by defective IκBα degradation (SFig. 2A). The fact that BHA does not affect complex I assembly, therefore, indicates that BHA protects cells from TNF-mediated RIPK1 cytotoxicity through a distinct mechanism (SFig. 2B).

We, and others, previously reported an early boost in RIPK1 activity, monitored by autophosphorylation on S166/T169 [35, 42, 47, 48], at the level of TNFR1 complex I under conditions of RIPK1 kinase-dependent apoptosis. In these experiments, TNFR1 complex I is incubated with USP21 to remove the ubiquitin chains attached to RIPK1, allowing proper visualization of RIPK1 phosphorylation by immunoblot. Interestingly, we found that BHA pretreatment of MEFs and human HT-29 and BT549 cells not only prevented the massive activation of RIPK1 resulting from IKKα/β or IKKε/TBK1 inhibition, but also the basal activity detected at the receptor complex upon single TNF stimulation (Fig. 3A, B, SFig. 2C, D). We confirmed that this effect was specific to BHA as pretreatment with the panel of antioxidants did not alter RIPK1 activation in TNFR1 Complex I (Fig. 3C). Such an early boost in RIPK1 activity was not observed at the receptor complex upon stimulation of cells with the necroptotic trigger TNF + zVAD-fmk (Fig. 3D), but was instead detected, at around the same time, in a pool of cytosolic RIPK1 (Fig. 3E). In these experiments, immunoprecipitated pS166/pT169 RIPK1 was preventively incubated with USP21 and λ-phosphatase to limit the potential complexity of its migration profile by immunoblot. Remarkably, both BHA and Nec-1s again prevented RIPK1 activation under these necroptotic conditions (Fig. 3E, F), which was not observed with any of the other antioxidants (Fig. 3F). As the catalytic activity of RIPK1 is required for the formation of the secondary cytosolic complex IIb/necrosome that triggers RIPK1 kinase-dependent apoptosis and necroptosis, we finally confirmed that the early inhibition of RIPK1 by BHA translated into defective complex IIb/necrosome assembly. As shown in Fig. 3G, BHA and Nec-1s similarly prevented the RIPK1 kinase-dependent association between RIPK1, FADD, and Caspase-8 in response to TNF + IKKi+zVAD-fmk, an inhibition that was not observed by BHT or α-tocopherol pretreatment (Fig. 3G). Together, these results demonstrated that BHA protects cells from TNF-induced RIPK1 kinase-dependent cell death by inhibiting cellular RIPK1 enzymatic activity.

A–G MEFs were pretreated for 30 min with the indicated compounds (100 µM BHA, 100 µM BHT, 5 mM NAC, 10 µM DecylQ, 100 µM α-tocopherol, 100 µM Trolox, 500 nM Ferrostatin-1, 10 µM Nec-1s, 5 µM TPCA-1 (IKKi), 10 µM GSK8612 (TBK1i), 50 µM zVAD-fmk) before stimulation with 1 µg/ml FLAG-hTNF (A–D), 1 µg/ml hTNF (E–F) or 20 ng/ml hTNF (G) for the indicated duration. A–D TNFR1 complex I was FLAG-immunoprecipitated and the IPs were treated with USP21 before analysis by immunoblot. The signal for pRIPK1 refers to active RIPK1 autophosphorylated on residue S166/T169. E–F Autophosphorylated active RIPK1 (pRIPK1) was immunoprecipitated using the specific anti-pS166/T169 RIPK1 antibody and the IPs were treated with USP21 and λ phosphatase before analysis by immunoblot. G Complex IIb/necrosome was pulled down by immunoprecipitation of caspase-8 and analyzed by immunoblot. Immunoblots are representative of at least two independent experiments.

Full size image

BHA acts as a direct RIPK1 inhibitor

We next tested the possibility that BHA would directly inhibit RIPK1 by performing in vitro kinase assays using recombinant RIPK1. Remarkably, we observed a dose-dependent inhibitory effect of BHA on RIPK1 (Fig. 4A), reaching a 50% reduction in RIPK1 enzymatic activity at 100 µM of BHA. Importantly, BHT had no effect on RIPK1 activity (Fig. 4A), and the inhibition by BHA was not originating from interference with the biochemical principle of the ADP-GLOTM kinase assay, the conversion of ADP to light (Fig. 4B). In addition, the inhibitory effect of BHA showed specificity to RIPK1, as the enzymatic activity of the close family member RIPK3 was not affected by BHA, while completely repressed by the RIPK3 inhibitor Dabrafenib [49] (Fig. 4C). As previously observed for other RIPK1 inhibitors [50], the inhibitory potential of BHA on RIPK1 was better in cells than in vitro. Indeed, BHA completely prevented cellular RIPK1 activity and cytotoxicity when used at 100 µM (Fig. 4D, E), a concentration that is most commonly found in the literature and that was used in our cellular assays (Figs. 1–4). This cellular inhibitory effect was lost at 10 µM, but still substantial at 50 µM (Fig. 4D, E). These results identified BHA, and not BHT, as a direct RIPK1 inhibitor.

A, C Quantitative RIPK1 (AA 1–479) (A) or Ripk3 (AA 1–439) (C) enzymatic activities were measured by ATP consumption using ADP-Glo kinase assays. The compounds were used at indicated concentrations and ‘D’ is short for DMSO. B The ADP-Glo reaction was performed in the absence of active kinase to assess the possible interference of BHA (100 µM) with the luminescence reaction. Results are presented as a percentage relative to the activity of the RIPK1/3 in absence of inhibitors (A, C), or as a percentage relative to the ADP-Glo luminescence reaction (B) and are the mean ± SEM of three independent kinase assays (n = 3). Statistical significance was determined by one-way ANOVA followed by a Tukey post hoc test (A–B) or by a two-tailed paired t test (C). D MEFs were pretreated for 30 min with the indicated compounds (5 µM TPCA-1) before stimulation with 1 µg/ml FLAG-hTNF for the indicated duration. TNFR1 complex I was then FLAG-immunoprecipitated and the IPs were treated with USP21 before analysis by immunoblot, where pRIPK1 refers to autophosphorylation of RIPK1 on S166/T169. The results are representative of at least two independent experiments. E MEFs were pretreated for 30 min with the indicated compounds before stimulation with 20 ng/ml hTNF for the indicated duration. Cell death was measured over time by Sytox Green (SG+) positivity, and the results are presented as mean ± SEM of three independent experiments (n = 3). Statistical analysis on kinetic cell death assays with more than one timepoint is detailed in the Methods section (E). Significance between samples is indicated in the figures as follows: *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Full size image

3-BHA docks into RIPK1 in a DLG-out/Glu-out conformation

Commercial BHA is supplied as an isomeric mixture of 2- and 3-BHA (respectively, representing ~15% and 85%). Interestingly, we found that 3-BHA, but not 2-BHA or BHT, docks into an existing crystal structure of RIPK1 (PDB: 4ITH) in which Nec-1s is bound to RIPK1 in a DLG-out/Glu-out conformation, a typical feature of RIPK1 type III inhibitors but, to our knowledge, that has not been reported yet for other kinases (Fig. 5A) [51, 52]. Similarly, as Nec-1s, 3-BHA is capable of hydrogen bonding with V76 and S161 of RIPK1 in the DLG-out conformation (Fig. 5A). In addition, both Nec-1s and BHA engage in hydrophobic interactions with the hydrophobic residues M67, L70, and M92 of RIPK1, which is possible for BHA thanks to the tert-butyl group in the 3-position (Fig. 5A). The fact that such hydrophobic interactions would not be possible with the tert-butyl group in the 2-position may explain why 2-BHA could not dock into this crystal structure of RIPK1. In line with these predictions, we confirmed the superior inhibitory capacity of 3-BHA over 2-BHA, both in kinase assays (SFig. 3) and in cells, using RIPK1 activation in TNFR1 complex I (Fig. 5B) and RIPK1 kinase-dependent apoptosis (Fig. 5C) and necroptosis (Fig. 5D) as readouts. In contrast, the two BHA isomers showed similar potency in inhibiting erastin-induced ferroptosis, confirming their comparable ROS-scavenging capacities (Fig. 5E).

A Docking of Nec-1s (purple) and 3-BHA (yellow) into RIPK1 (PDB: 4ITH). Hydrogen bonds (2.7–3.0 Å) with residues V76 and S161 of RIPK1 are shown as red dashes. Hydrophobic residues M67, L70, and M92 of RIPK1 engaged in hydrophobic interaction are also indicated. B, H MEFs were pretreated for 30 min with indicated compounds before stimulation with 1 µg/ml FLAG-hTNF for the indicated duration. TNFR1 complex I was FLAG-immunoprecipitated and the IPs were treated with USP21 post-IP when indicated. The results are representative of at least two independent experiments. pRIPK1 refers to autophosphorylation of RIPK1 on S166/T169. C–E, I–J MEFs were pretreated for 30 min with indicated compounds (100 µM BHA, 100 µM 2-BHA, 100 µM 3-BHA, 100 µM TBHQ, 10 µM Nec-1s, 10 µM GSK8612 (TBK1i), 50 µM zVAD-fmk) before stimulation with 20 ng/ml hTNF (C–D, I–J) or 10 µM erastin (E) for the indicated duration. Cell death was measured over time by Sytox Green (SG+) positivity, and the results are presented as mean ± SEM of three independent experiments (n = 3). Statistical analysis on kinetic cell death assays with more than one timepoint is detailed in the Methods section. F Molecular structure models are presented for 2-, 3-BHA, BHT, and THBQ. G Kinase activity was quantitatively measured by ATP consumption using the ADP-Glo kinase assay, compounds were used at indicated concentrations. D is short for DMSO. Results are presented as a percentage relative to the activity of the kinase in absence of inhibitor and are the mean ± SEM of three independent kinase assays (n = 3). Statistical significance was determined via ordinary one-way ANOVA followed by a Tukey post hoc test. Significance between samples is indicated in the figures as follows: *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Full size image

The structural insights on the binding of 3-BHA to RIPK1 also provided an explanation for the lack of inhibition of RIPK1 by BHT. Antioxidant BHT lacks the oxygen in the 4-position of the benzene ring that is required for hydrogen bonding with S161 of RIPK1 (Fig. 5A, Fig. 5F). Furthermore, the extra tert-butyl group in the 5-position makes BHT significantly more sterically bulky than BHA, resulting in an inability for BHT to bind to RIPK1. In contrast, the structural similarities between 3-BHA and TBHQ (oxygens in the 1- and 4-position for the respective hydrogen bonding with S161 and V76 of RIPK1, and tert-butyl group in the 3-position for hydrophobic interactions with RIPK1) suggested that TBHQ could also function as a direct inhibitor for RIPK1 (Fig. 5F). This prediction was confirmed by in vitro kinase assays using recombinant RIPK1, where ~60% reduction in RIPK1 enzymatic activity was observed at 100 µM of TBHQ (Fig. 5G). Inhibition of RIPK1 by TBHQ was further validated in cells, with TBHQ pretreatment completely preventing activation of RIPK1 in TNFR1 complex I (Fig. 5H), and TNF-mediated RIPK1 kinase-dependent apoptosis (Fig. 5I) and necroptosis (Fig. 5J).

Together, these results provided a structural basis for the direct inhibition of RIPK1 by 3-BHA, which led to the identification of the structurally related antioxidant and food additive TBHQ as an additional type III RIPK1 inhibitor. The fact that 2-BHA partially affected RIPK1 kinase activity despite its inability to bind RIPK1 in silico reveals some flexibility in the kinase domain of RIPK1 that still allows 2-BHA docking, albeit less efficiently than 3-BHA.

Oral administration of BHA protects mice from TNF-induced lethal shock

As BHA and TBHQ are commonly used food additives, we next examined the effect of the oral administration of BHA on inflammatory conditions originating from TNF-mediated RIPK1 kinase-dependent cell death. We first used a chronic model of disease caused by SHARPIN deficiency in mice. These animals develop a severe multi-organ inflammatory condition, known as chronic proliferative dermatitis in mice (cpdm), resulting from tissue-specific induction of TNF/TNFR1-mediated RIPK1 kinase-dependent apoptosis or necroptosis [29]. MDFs isolated from the Sharpincpdm/cpdm mice succumb by RIPK1 kinase-dependent apoptosis upon single TNF stimulation, and we found that BHA, but not BHT, could protect these cells from death (Fig. 6A). We therefore next tested the effects of feeding the Sharpincpdm/cpdm mice with a BHA-containing diet for a period of 5 weeks, starting at 4-weeks of age, before the onset of symptoms. Interestingly, we observed that the mice fed with BHA looked macroscopically better than the ones receiving a standard diet or a diet enriched with BHT (Fig. 6B). For instance, the BHA-fed mice did not show signs of skin lesion characteristic of the Sharpincpdm mutation (Fig. 6B). Nevertheless, closer histological analysis of several organs, including not wounded skin, did not reveal a statistically significant reduction in the number of tissue-associated dead cells or immune infiltrates (Fig. 6C and data not shown). Also, the reduction in the levels of LDH and IL-6 detected in the serum of the BHA-fed mice did not reach statistical significance (Fig. 6D, E). So, while inhibition of RIPK1 by BHA could prevent TNF cytotoxicity in Sharpincpdm/cpdm MDFs, the oral administration of BHA to the Sharpincpdm/cpdm mice only marginally protected them from cell death-driven inflammation.

A Sharpincpdm/cpdm MDF cells were pretreated for 30 min with indicated compounds (Nec-1s 10 µM, BHA 100 µM, BHT 100 µM) before stimulation with 1 ng/ml hTNF for the indicated duration. Cell death was measured over time by Sytox Green (SG + ) positivity, and the results are presented as mean ± SEM of three independent experiments (n = 3). Statistical significance was determined via one-way ANOVA followed by a Tukey post hoc test. B–E 4-weeks old Sharpin+/+ and Sharpincpdm/cpdm littermate mice were fed for 5 weeks on a control diet or on a diet enriched in BHA or BHT. B A picture of the mice was taken at the end of the feeding period. C TUNEL quantification was performed on liver sections from 3–4 mice per diet condition. D–E Serum levels of lactate dehydrogenase (LDH) (Sharpin+/+: control diet n = 8, BHA diet n = 8, BHT diet n = 8; Sharpincpdm/cpdm: control diet n = 9, BHA diet n = 9, BHT diet n = 9) (D) and IL-6 (Sharpin+/+: control diet n = 8, BHA diet n = 9, BHT diet n = 8; Sharpincpdm/cpdm: control diet n = 9, BHA diet n = 9, BHT diet n = 9) (E) were determined at the end of the feeding period. C–E Statistical significance between Sharpincpdm/cpdm mice was determined via ordinary one-way ANOVA followed by a Tukey post hoc. F–G C57BL/6 J female mice were administered 100 µl pure corn oil or 100 µl corn oil containing 625 mg/kg BHA or BHT via oral gavage after 16 h of starvation. Food was re-introduced 30 min after gavage and hTNF was injected 1 h after the oral gavage at 15 µg hTNF per 20 g of body weight. Body temperature (F) and cumulative survival rates (G) were determined over time. The number of mice used in each condition is indicated. The temperature results are represented as mean ± SEM. Statistical significance of the temperature curves was determined using two-way ANOVA followed by a Tukey post hoc test. Survival curves were compared using the log-rank Mantel–Cox test. Significance between samples is indicated in the figures as follows: *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Full size image

Next, we evaluated the potential protective effect of BHA in the acute model of systemic inflammatory response syndrome (SIRS) caused by intravenous injection of TNF, a lethal shock model previously demonstrated to originate from RIPK1 kinase-dependent cell death [53, 54]. Remarkably, oral administration of BHA significantly protected the mice from TNF-induced hypothermia and lethality (Fig. 6F, G). Importantly, the protection provided by BHA resulted from RIPK1 inhibition, and not ROS scavenging, since BHT had no effect in this acute model of disease (Fig. 6F, G).