Activation of NRF2 by Nitrosative Agents and H₂O₂ Involves KEAP1 Disulfide Formation^*

Cell Culture and Reagents

HeLa cells were grown at 37 °C, 5% CO₂, in Dulbecco's modified Eagle's medium containing 1 g/liter of glucose, 110 mg/ml sodium pyruvate, 4 mm GlutaMAX (Invitrogen), complemented with 10% fetal calf serum (Sigma). For plasmid transfection, ∼5.5 10⁵ cells were incubated 5 h with 3 μg of DNA and Lipofectamine 2000 (Invitrogen) following the supplier's recommendations, washed, and incubated in fresh medium. The cells were treated as described in the text 24 h after transfection. For the TRxR1 knockdown, ∼2.5 10⁵ cells were transfected with 3 μg of DNA, by the same protocol. After 24 h, hygromycin B was added (250 μg/ml) to the culture medium twice every 2 days. The cells were then expanded and kept in the presence of hygromycin B (125 μg/ml). The selection was withdrawn before experiments. N-Ethylmaleimide (NEM), t-BHQ, H₂O₂, and cycloheximide were purchased from Sigma, and SpNO was from Cayman Chemical. S-nitrosocysteine (Cys-NO) was prepared by mixing stoichiometric amounts of l-cysteine and sodium nitrite at pH 4, followed by the measure of its concentration by recording the absorbance at 334 nm.

Plasmids

Plasmid pcDNA3-HA-mKEAP1 was a gift from Dr. M. Yamamoto, pCI-HA-mNRF2 from was J. A. Diehl, and peYFP-N1 was purchased from Clontech. Plasmid pcDNA3- Myc-His-mKEAP1 was constructed by PCR-mediated replacement of the HA tag sequence of pcDNA3-HA-mKEAP1 with three Myc tag sequences followed by a stretch of eight His codons. Mutagenesis was done with the Stratagene QuikChange multi kit following the manufacturer's instructions. For the TrxR1 knockdown, we used TrxR1-specific small hairpin RNA (shRNA) that targeted the following sequences within the open reading frame: GGATTAAGGCAACAAATAA (sh1), GCATCAAGCAGCTTTGTTA (sh2), and GCAAGACTCTCGAAATTAT (sh3). shRNA sequences were designed with the DSIR program that also operates an exact similarity search algorithm for potential off target detection (34). These shRNA were expressed under the control of the H1 promoter from the Epstein-Barr virus-based replicative plasmid that contains an oriP, the EBNA open reading frame, and a hygromycin B selection cassette (35). Control lines expressed a nonfunctional shRNA as reported (35).

Redox Western

The cells were washed on ice with 40 mm NEM in phosphate-buffered saline and lysed in lysis buffer (0.1 m Tris-HCl, pH 8.0, 120 mm NaCl, 0.2% deoxycholic acid, 5% Nonidet P-40, 0.2 mm NaF, 0.2 mm EGTA, 0.1 mm phenylmethylsulfonyl fluoride, Roche-Complete mini protease inhibitor mixture, 40 mm NEM). Centrifuged-cleared lysates were diluted with 2 volumes of 3× loading buffer (0.2 m Tris-HCl, pH 6.8, 45% glycerol, 6% SDS, 0.03% bromphenol blue). Half of the samples were reduced by the addition of β-mercaptoethanol (6% v/v). After heat denaturation, the proteins were separated by SDS-PAGE, transferred onto a nitrocellulose membrane, and immunostained with anti-NRF2 (H300; Santa Cruz), anti-HA (HA11; Covance), anti-Myc (9E10; a kind gift from G. Clément and C. Créminon, France), anti-TrxR1 (a kind gift from A Holmgren, Sweden), or anti-YFP (JL8; Clontech) specific antibodies. Detection was performed after chromophore-coupled secondary antibody staining, using the LICOR Odyssey infrared imager.

Protein Pulldown Assays

For pulling down His tag-containing polypeptides, the cells (10⁶/sample) were washed in ice-cold phosphate-buffered saline containing NEM (40 mm) and lysed in precipitation buffer (0.1 m Tris-HCl, pH 8, 1% Nonidet P-40, 2% glycerol, 0.3 m NaCl, 0.2 mm phenylmethylsulfonyl fluoride, Roche-Complete mini protease Inhibitor mixture, 40 mm NEM). 10 mm imidazole was then added, and 90% of the sample (250 μl) was incubated with 50 μl of a nickel-nitrilotriacetic acid resin (Qiagen) for 1 h. After extensive washing with wash buffer (0.1 m Tris-HCl, pH 8, 1% Nonidet P-40, 2% glycerol, 0.3 m NaCl, 0.2 mm phenylmethylsulfonyl fluoride, 10 mm NEM, 10 mm imidazole), the proteins were eluted with 2× protein-loading buffer containing 10 mm NEM and separated by SDS-PAGE.

The Biotin-switch Technique

The biotin-switch method was performed as described in Ref. 36, with the difference that cells were directly lysed in HEN buffer (250 mm HEPES, pH 7.7, 1 mm EDTA, 0.1 mm neocuproine) containing SDS (2.5%) and the thiol-reactive compound S-methyl methanethiosulfonate (0.1%) to block thiol modification during and after the lysis. Briefly, the proteins were precipitated and washed twice with acetone, then the pellets were dissolved in HEN buffer containing SDS (1%), and reduction and labeling of S-nitrosothiols were achieved by the addition of sodium ascorbate (100 μm) and biotin-HPDP (0.25 mg/ml) upon incubation in the dark at room temperature. For the specific detection of S-nitroso-KEAP1, the samples were precipitated with acetone, washed with acetone, dissolved in HEN buffer 0.1× containing SDS (1%). 3 volumes of (v/v) neutralization buffer (25 mm HEPES, pH 7.5, 100 mm NaCl, 1 mm ETDA, 1% Triton X-100) was added before incubation with a streptavidin-agarose slurry. Elution was performed in 0.1× HEN buffer containing β-mercaptoethanol (1%). Western blots were then performed using anti-Myc antibodies.

Oxidation of KEAP1 in Cells Exposed to H₂O₂

We evaluated whether H₂O₂ could oxidize KEAP1 at Cys residues by analyzing the redox state of HA-KEAP1 ectopically expressed in HeLa cells. To prevent Cys residue oxidation, free sulfhydryls were blocked with NEM during sample preparation. HA-KEAP1 from untreated cell lysates that had been reduced by β-mercaptoethanol migrated in SDS-PAGE as a single band with an apparent molecular mass of ∼70 kDa (Fig. 1A, lower panel). When reduction was omitted, HA-KEAP1 from the same lysates still migrated as a major band of same molecular mass (denoted Red for reduced), but now a second less intense band of faster mobility was observed (denoted OxIM for oxidized intramolecular; see below) (Fig. 1A, upper panel). A 5-min exposure of cells to H₂O₂ (200 μm) further altered migration of KEAP1 under nonreducing conditions (Fig. 1A, lane 2); the reduced KEAP1 70-kDa band decreased in intensity, whereas OxIM increased and two new bands of slower migration appeared (denoted OxIR1 and OxIR2 for oxidized intermolecular; see below; lane 2). OxIM, OxIR1, and OxIR2 were absent under reducing conditions, indicating that they probably result from disulfide formation. OxIM might correspond to an intramolecular disulfide (see below), the formation of which is predicted to increase SDS-PAGE mobility because of a decrease in the hydrodynamic radius of the SDS-bound polypeptide, especially if its two constitutive Cys residues are far apart in the primary sequence. OxIR1/2 could correspond to intermolecular disulfides between KEAP1 and itself or with another polypeptide. Such oxidative KEAP1 modifications were not seen upon cell exposure to t-BHQ (supplemental Fig. S1A). Time course analysis showed that KEAP1 was maximally oxidized 5 min after exposure to H₂O₂ and then returned to the redox state of untreated cells 40 min after this treatment (Fig. 1A, lanes 2–7). Such kinetics presumably reflects KEAP1 reduction by endogenous reductases and not de novo protein synthesis, because it was similar in cells treated with the protein synthesis inhibitor cycloheximide (supplemental Fig. S1B). H₂O₂ dose-response analysis (Fig. 1B) indicated that the basal KEAP1 oxidation seen in lysates from untreated cells started to increase at a concentration of 100 μm and was maximal at 200 μm. Higher doses were not tested because of potential toxicity.

KEAP1 Disulfide Bond Formation Induced by NO Derivatives and HOCl

We were interested to see whether other oxidants known to activate NRF2, such as NO and derivatives or the potent thiol oxidant hypochlorous acid (HOCl) could also lead to KEAP1 Cys residue oxidation. We used the NO donor SpNO that decomposes in culture medium into two NO equivalents with an approximate half-life of 2 h (not shown). In lysates from HA-KEAP1-expressing HeLa cells exposed for 15 min to SpNO (1 mm), KEAP1 appeared oxidized in nonreduced SDS-PAGE, with a migration very similar to that of H₂O₂-oxidized KEAP1 (Fig. 1C, compare lanes 1–3). Oxidation was not as important but was maintained up to 3 h versus 40 min with H₂O₂ (Fig. 1A). Such a prolonged response probably reflects the 2-h half-life of SpNO-released NO, in contrast to a bolus of H₂O₂ that is rapidly degraded by cellular consumption. KEAP1 also became oxidized upon cell exposure to HOCl at the low dose of 1 mm, with the formation of oxidized species similar to those seen upon exposure to H₂O₂ and SpNO (supplemental Fig. S2A). NO can cause protein S-nitrosylation, but we could not detect any modification of KEAP1 by S-nitrosylation when cells were exposed to SpNO (2 mm) for 45 min or to H₂O₂ for 5 min (Fig. 1D). In contrast, and as a positive control, a 30-min cell exposure to the trans-nitrosylating agent Cys-NO (0.5 mm) led to potent KEAP1 S-nitrosylation. Interestingly, Cys-NO also led to KEAP1 disulfide bond formation (supplemental Fig. S2B).

Identification of the KEAP1 Cys Residues Engaged in Disulfide Bonds

We used a mutagenesis approach to identify the KEAP1 Cys residues whose oxidation caused alteration of protein migration. 23 of the 25 Cys residues of murine KEAP1 (Fig. 2A) were substituted by a serine, generating 19 single and two double-Cys residues mutants (C513S/C518S and C622S/C624S). Cys³⁶⁸ and Cys⁴⁸⁹ were not tested because, as predicted by a KEAP1 Kelch crystallographic structure (37), they are unlikely to participate in redox regulation, as not being solvent-exposed and lacking a proximal Cys residue. These mutants were each expressed in HeLa cells and tested for their oxidation upon a 5-min exposure to H₂O₂ (200 μm) or to SpNO (2 mm) (Fig. 2B and supplemental Fig. S3). Except KEAP1 Cys¹⁵¹, Cys²²⁶, and Cys⁶¹³, all of the mutants had nonreducing SDS-PAGE migrations similar to that of wild type HA-KEAP1.

KEAP^C151S still formed OxIM, but not OxIR1 and 2 (Fig. 2B, compare lanes 5 and 6 with lanes 9 and 10). A faint and fuzzy band just below OxIR1 was present instead. Note also here a β-mercaptoethanol-insensitive KEAP1 band just above OxIR1 already present in lysates from untreated cells that also disappeared in KEAP^C151S (Fig. 2B) (denoted cov for covalent). This KEAP1 band, which varied in intensity between experiments, was resistant to all reducing agents tested (not shown) and therefore corresponds to a covalent modification of unknown nature engaging Cys¹⁵¹. Disappearance of OxIR1 and 2 in KEAP^C151S indicates that KEAP1 engages Cys¹⁵¹ in an intermolecular disulfide with itself and/or with another protein to account for the two slow migrating β-mercaptoethanol-sensitive species. In the absence of Cys¹⁵¹, illegitimate unstable disulfides might form to account for the faint slow migrating band seen with KEAP1^C151S.

Substituting Cys²²⁶ or Cys⁶¹³ had the same effect on KEAP1 migration (Fig. 2B, lanes 3, 4, 7, 8, 11, and 12). These mutants were both unable to form OxIM but still formed the OxIR1 and 2 bands, although their migration was slightly up-shifted. These data indicate that Cys²²⁶ and Cys⁶¹³ are engaged in an intramolecular disulfide linkage spanning across the Kelch domain (Fig. 2A). The presence of this disulfide in the wild type KEAP1 OxIR1/2 complexes would then explain the migration up-shift of these two bands when either Cys²²⁶ or Cys⁶¹³ are mutated. Similarly, the higher intensity of OxIM in KEAP1^Cys151 in oxidant-treated cells reflects the fact that all of the KEAP1 species that carry the intramolecular disulfide are now shifted down to the size of this band.

Thus, a fraction of KEAP1 carries an intramolecular Cys²²⁶–Cys⁶¹³ disulfide. H₂O₂ and SpNO further increase this disulfide, also causing formation of an intermolecular KEAP1 disulfide(s). These two disulfides are independent, because one can form in the absence of the other(s).

KEAP1 Homodimerizes through a Cys¹⁵¹-dependent Disulfide Bond

We tested whether either one or both of the two KEAP1 intermolecular disulfide-linked complexes (OxIR1/2) could be the result of KEAP1 dimerization by co-immunoprecipitation assays, using two versions of KEAP1 that differed by both their size and tag. We thus co-expressed in HeLa cells HA-KEAP1 (molecular mass, 71 kDa) and a version of KEAP1 containing a three-Myc epitope tag fused to an eight-His tag (Myc-His-KEAP1) (molecular mass, 76 kDa).

When individually expressed, HA-KEAP1 from cells exposed to H₂O₂ displayed the characteristic migration of oxidized protein (Fig. 3, lane 5). Note here that reduced KEAP1 was absent, indicating almost full oxidation. When HA-KEAP1 was co-expressed with Myc-His-KEAP1, two new HA-immunoreactive faint bands could now be seen that migrated just above the HA-KEAP1 OxIR1/2 bands (Fig. 3, lane 6, see red arrows). In these lysates anti-Myc antibodies revealed Myc-His-KEAP1, which had the characteristic migration of oxidized protein but was slightly slower than HA-KEAP1 because of its extra 5 kDa (lane 2). However, migration of Myc-His-KEAP1 was not affected by co-expressed HA-KEAP1 in contrast to the latter, probably because of a higher level of expression of the former. Western blots were then performed with eluates of a nickel column on which lysates of cells co-expressing the two KEAP1 derivatives had been adsorbed to pull down Myc-His-KEAP1. In these eluates, the anti-Myc antibody revealed enriched oxidized Myc-His-KEAP1 (lane 4), the anti-HA antibody HA-KEAP1 OxIM, and two slow migrating bands of the size of the extra HA-immunoreactive faint bands seen in crude extracts (Fig. 3, lane 8, see the red arrows). When the same experiment was performed with the Cys¹⁵¹-substituted versions of the two KEAP1 derivatives, HA-KEAP1 OxIM1 was still pulled down with Myc-His-KEAP1, but no slower oxidized band could be seen (supplemental Fig. S4). We conclude that these two HA-immunoreactive bands pulled down with Myc-His-KEAP1, each corresponding to a Cys¹⁵¹-Cys¹⁵¹ mixed disulfide between one molecule of HA-KEAP1 and one of Myc-His-KEAP1. Such a conclusion is compatible with both the sizes of these two bands, which are intermediate between those of the KEAP1-derivatives OxIR1/2 bands, and with the ∼140-kDa apparent molecular mass of OxIR1 that equals two KEAP1 molecules. The faster OxIR2 band, always much fainter, might be redox conformation-dependent but is not the result of a N-terminal protein cleavage (not shown). Importantly, whereas the monomeric intramolecular disulfide form of HA-KEAP1 (HA-OxIM) co-precipitated with Myc-His-KEAP1 (see Fig. 3, lane 8 and supplemental Fig. S4), most probably by virtue of noncovalent dimerization imparted by the BTB domain (4, 5, 38), the intermolecular HA-KEAP1-HA-KEAP1 disulfide was not co-precipitated, which suggests that dimerization of KEAP1 through the BTB and through the intermolecular disulfide might exclude each other.

Functional Consequences of KEAP1 Disulfide Bond Formation

Stress signals activate NRF2 by inhibiting KEAP1-dependent NRF2 degradation. To verify the functional significance of KEAP1 oxidation and of identified Cys residues, we monitored NRF2 protein abundance as readout of KEAP1 activity in HeLa cells co-expressing HA-NRF2, Myc-His-KEAP1 or its Cys mutant derivatives, and YFP as transfection control.

H₂O₂ induced a slight but reproducible and transient stabilization of NRF2 that closely paralleled KEAP1 oxidation (Fig. 4A). Such a marginal effect of H₂O₂ might relate to the transient nature of KEAP1 modification induced by this very short-lived inducer that denied NRF2 build up by neo-synthesis. HOCl, as a short-lived oxidant, also slightly stabilized NRF2, with kinetics that paralleled KEAP1 oxidation (supplemental Fig. S2A), whereas Cys-NO, which had a very potent and extended effect on KEAP1 oxidation, stabilized NRF2 up to 3 h (supplemental Fig. S2B). SpNO was also kinetically different, because stabilization of NRF2 was not detected early in the kinetics (supplemental Fig. S2C) but was significant when measured at 5 h (Fig. 4B, lane 8), consistent with the long half-life of SpNO causing lower but prolonged KEAP1 oxidation (Fig. 1C) and therefore steady NRF2 accumulation with time.

We took advantage of the potent stabilization of NRF2 by SpNO to evaluate the effect of mutations that affect KEAP1 oxidation. As shown in Fig. 4B, NRF2 levels were elevated in the absence of KEAP1 and decreased upon KEAP1 co-expression (compare lanes 1 and 2). Under basal conditions, substitution of neither Cys¹⁵¹ nor Cys²²⁶ altered KEAP1 influence on NRF2 levels (Fig. 4B, lanes 3 and 4). As a control experiment, we checked the response to t-BHQ, which as previously shown relieved KEAP1-mediated NRF2 repression in the presence of wild type KEAP1 but not of KEAP1^Cys151S (Fig. 4B, lanes 5 and 6) (6, 11). However, KEAP1^Cys226S still responded to t-BHQ (Fig. 4B, lane 7). Checking the response to SpNO showed that it was similar to t-BHQ; the potent repression relief caused by SpNO at 5 h in the presence of wild type KEAP1 was not observed with KEAP1^Cys151S (Fig. 4B, lanes 8 and 9); however, KEAP1^Cys226S had a wild type response to SpNO-induced derepression of NRF2 (Fig. 4B, lane 10). Therefore, intermolecular disulfide formation via Cys¹⁵¹ is important for relieving KEAP1-mediated NRF2 degradation.

The Effect of Inactivating Thiol Redox Control on KEAP1 Oxidation and Function

The transient nature of KEAP1 oxidation by H₂O₂ or SpNO (Figs. 1 and 4) suggests that cellular thiol reductases recycle the protein back to its reduced form. We sought to evaluate the redox state and activity of KEAP1 upon inactivating either one or both of the two thiol-redox control system. To inactivate the GSH pathway, we used buthionine sulfoximine, a specific inhibitor of the GSH rate-limiting biosynthetic enzyme γ-glutamyl cysteine synthase. To inactivate the thioredoxin pathway, we stably expressed in HeLa cells shRNAs targeting the TrxR1 mRNA. Two of the three tested TrxR1 shRNA caused protein level reduction of ∼90%, as compared with a TrxR1 unrelated small nonhairpin RNA (Fig. 5A). Buthionine sulfoximine treatment (0.1 mm for 24 h) lead to an 80% decrease of the total GSH cellular content as measured by the 5,5′-dithiobis (2-nitrobenzoic acid)-GSSG reductase recycling assay (not shown) but altered neither KEAP1 oxidation nor NRF2 stabilization, only slightly delaying protein reduction (Fig. 5B, compare lanes 1–5 with lanes 6–10). Knockdown of TrxR1 moderately increased the effect of H₂O₂ on KEAP1 oxidation and on NRF2 stabilization but did not delay reduction (compare lanes 1–5 with lanes 11–15). In contrast, buthionine sulfoximine treatment of TrxR1 knockdown cells caused constitutive KEAP1 oxidation and major NRF2 stabilization. In these cells, H₂O₂ further increased KEAP1 oxidation up to shifting all of the protein to the OxIR1 and 2 bands but did not further stabilize NRF2 because its levels might have already reached a plateau. Surprisingly 30 min after H₂O₂ treatment, some reduction of KEAP1 started to occur back to the levels of untreated cells, which indicate the presence of a remnant thiol reductase activity in these cells.

Thus, upon inactivating both thiol redox pathways, low endogenous reactive oxygen species levels cause a build up of KEAP1 oxidation and NRF2 levels with time, further indicating a correlation between KEAP1 oxidation and NRF2 stabilization. These data could also indicate redundancy of the two pathways in KEAP1 reduction, but partial inactivation of these pathways, as indicated by the residual KEAP1 reduction, does not allow such a conclusion.