Activation of the Keap1/Nrf2 pathway for neuroprotection by electrophillic phase II inducers

Abstract

Electrophilic neurite outgrowth-promoting prostaglandin (NEPP) compounds protect neurons from oxidative insults. At least part of the neuroprotective action of NEPPs lies in induction of hemeoxygenase-1 (HO-1), which, along with other phase II enzymes, serve as a defense system against oxidative stress. Here, we found that, by using fluorescent tags and immunoprecipitation assays, NEPPs are taken up preferentially into neurons and bind in a thiol-dependent manner to Keap1, a negative regulator of the transcription factor Nrf2. By binding to Keap1, NEPPs prevent Keap1-mediated inactivation of Nrf2 and, thus, enhance Nrf2 translocation into the nucleus of cultured neuronal cells. In turn, Nrf2 binds to antioxidant/electrophile-responsive elements of the HO-1 promoter to induce HO-1 expression. Consistent with this notion, NEPP induction of an HO-1 reporter construct is prevented if the antioxidant-responsive elements are mutated. We show that NEPPs are neuroprotective both in vitro from glutamate-related excitotoxicity and in vivo in a model of cerebral ischemia/reperfusion injury (stroke). Our results suggest that NEPPs prevent excitotoxicity by activating the Keap1/Nrf2/HO-1 pathway. Because NEPPs accumulate preferentially in neurons, they may provide a category of neuroprotective compounds, distinct from other electrophilic compounds such as tert-butylhydroquinone, which activates the antioxidant-responsive element in astrocytes. NEPPs thus represent a therapeutic approach for stroke and neurodegenerative disorders.

Cellular antioxidants are crucial for reducing oxidative stress and preventing neuronal death. A recently elucidated pathway to induce antioxidant enzymes involves transcriptional activation through the antioxidant-responsive element (ARE) (1, 2). In this case, electrophilic agents induce a set of genes encoding “phase-II” enzymes, including hemeoxygenase-1 (HO-1), NADPH quinone oxidoreductase 1, and γ-glutamyl cysteine ligase (γ-GCL). These enzymes provide efficient cytoprotection, in part, by regulating the intracellular redox state (1, 2).

The ARE is a cis-acting element essential for transcriptional activation of phase-II genes by electrophiles (1, 2). The transcription factor Nfr2 complexes with Maf family proteins to transactivate the ARE. Under basal conditions, the cytosolic regulatory protein Keap1 binds tightly to Nrf2, retaining it in the cytoplasm (1, 2). In this regard, the action of Keap1 is analogous to that of IκB, preventing activation and translocation of the transcription factor NF-κB (1, 2). In the case of Keap1, electrophiles make a Michael adduct with critical cysteine residues in this regulatory protein, causing the liberation of Nrf2 and allowing it to translocate into the nucleus (1, 2).

Among phase-II enzymes, HO-1 has attracted special attention because of its therapeutic effects against neurodegenerative diseases (3, 4). HO-1 oxidatively cleaves heme to biliverdin, forms CO, and releases the chelated Fe²⁺ (3). Bilirubin (a reduction product of biliverdin) serves as a potent radical scavenger (4) and protects neuronal cells against oxidative stress at nanomolar concentrations (5). Studies using gene-knockout and transgenic mice have confirmed the biological significance of HO-1 as a cellular antioxidant (6). HO-1 has been proposed to play an obligatory role in endogenous defense against oxidative stress, because cells from HO-1^–/– mice are highly susceptible to oxidative insults (6). The significance of HO-1 in terms of drug development against neurodegenerative diseases is based on two facts: (i) HO produces several antioxidative compounds, including biliverdin and bilirubin (5), and (ii) the induction of HO-1 can be regulated by various compounds (7). Thus, our group, in addition to those of S. H. Snyder and M. D. Maines (3, 5, 7), has proposed that an inducer of HO-1 in neurons could represent an efficient neuroprotective compound.

A recent study showed that activation of the Keap1/Nrf2/ARE pathway mediates HO-1 induction by electrophiles (1, 2). Thus, we focused on this pathway as a possible mechanism by which neurite outgrowth-promoting prostaglandins (NEPPs) promote HO-1 induction and neuroprotection. Here, we present data showing that NEPPs protect cortical neurons both in vitro and in vivo against neuronal degeneration by acting as electrophiles to activate the Keap1/Nrf2 pathway.

Results

Thiols as Targets of Electrophilic NEPPs. We generated NEPP-related compounds based on the chemical structures of cyclopentenone prostaglandins and found that NEPP6 and -11 protected neurons against oxidative stress (7, 8) and that induction of HO-1 played an essential role in these neuroprotective effects (7). NEPP11 afforded more potent neuroprotection than did NEPP6, probably because NEPP11 is more lipophilic, allowing for better CNS permeability (8). The cross-conjugated dienone structure of NEPPs is critical for their biological effects (8) and underlies the electrophilicity of carbon #11 and, thus, its high chemical reactivity with thiols (8).

Through its single free thiol group (on cysteine residue #34), BSA has been used as an in vitro example for adduct formation by electrophilic compounds. We tested whether the free cysteine of BSA could form an adduct with NEPP compounds. If carbon #11 on the NEPP compounds binds to this free cysteine of BSA, then pretreatment with N-ethylmaleimide (NEM), an irreversible thiol alkylating agent, should abolish this binding. In an experiment to test this idea, we synthesized NEPP6-biotin (8). Biotin was conjugated to the C1 carbonic acid site on NEPP6 by using a chemical linker. With streptavidin as a probe, we could then detect proteins bound to NEPP6-biotin. BSA was mixed with vehicle or NEPP6-biotin in PBS at room temperature for 5 h, electrophoresed, and probed with streptavidin (Fig. 1). After exposure to NEPP6-biotin, a single band (68 kDa) corresponding to BSA/NEPP6-biotin was detected (lane 3). Pretreatment with NEM depressed this signal in a dose-dependent manner, although levels of protein were virtually the same, as judged from Coomassie brilliant blue staining of the gel. NEM also abolished the binding of NEPP6-biotin to lysates of HT22 cells or brain (data not shown). These results suggest that cysteine thiols are the target of NEPP compounds binding to cellular proteins.

Fig. 1.

Thiols as targets of NEPP binding. BSA (1 μg per lane) was incubated with various concentrations of NEM for 30 min at room temperature. Vehicle (lanes 1 and 2) or NEPP6-biotin (10 μM, lanes 3–8) was then added, and the samples were incubated for5hat room temperature. (Upper) Proteins were subjected to SDS/PAGE and probed with peroxidase-conjugated streptavidin. (Lower) The gel was also stained with Coomassie brilliant blue.

NEPPs Activate HO-1 Transcription Through the ARE. We found that NEPP11 induced HO-1 protein levels at the same concentration that prevented neuronal cell death (Fig. 2A) (7). Next, we studied transcriptional activation of the HO-1 gene by NEPP11 in HT22 cells, a differentiated neuronal cell line, transfected with pHO15luc, a luciferase reporter construct under the control of a 15-kb mouse HO-1 promoter fragment (Fig. 2B). Based on luciferase activity, NEPP11 stimulated the expression of HO-1 transcription >5-fold in a dose-dependent manner. Similar responses were obtained with NEPP6 (data not shown). Gong et al. (2) had identified two enhancer regions, E1 and E2, located upstream of the transcription initiation site by ≈4 and 10 kb that contained the elements responsible for activation by 15-deoxy-Δ^12,14-PGJ₂ (2). To determine the role of the E1 and E2 enhancers, we expressed a mutant promoter construct lacking both enhancer sites [pHO15lucΔ(E1+E2)]. As shown in Fig. 2B, the mutant was only minimally responsive to NEPP11 (1.8-fold induction). Similar results were obtained with NEPP6 (data not shown).

Fig. 2.

Activation of HO-1 promoter and ARE by NEPP. (A) Induction of HO-1 protein by NEPP11. Various concentrations of NEPP11 were added to HT22 cells for 24 h. Thereafter, cell lysates (10 μg per lane) were subjected to SDS/PAGE and probed with anti-HO-1 (Upper) or anti-β-actin (Lower). (B) Activation of the HO-1 promoter by NEPP11. HT22 cells were transfected with reporter cDNAs (1 μg per well), and various concentrations of NEPP11 were added to the cultures. (Upper) After a 24-h incubation, cell lysates were subjected to luciferase reporter assays. Data represent mean ± SD; *, P < 0.01 by paired ANOVA between pHO15luc and pHO15lucΔ(E1+E2) at each concentration of NEPP11. (C) Activation of ARE core element by NEPP11. cDNA constructs represent wild-type pAREluc (5′-CTCAGCCTTCCAAATCGCAGTCACAGTG ACTCAGCAGAATC-3′) and mutant pGC-AREluc (5′-CTCAGCCTTCCAAATCGCAGT CACAGTGACTCAATAGAATC-3′) (12). (D) Effect of NEPP11 (2 μM) on the activity of various transcriptional elements. (E) EMSA to assess binding of nuclear lysates to the ARE. HT22 cells were incubated for 8 h with vehicle (lane 2) or NEPP11 (l μM, lanes 3–5). EMSAs were performed by using 10 μg of nuclear lysate per lane and a biotin-labeled ARE probe. In lane 4, α-Nrf2 antibody (100×) was added to supershift Nrf2 protein binding to ARE probe. In lane 5, an excess amount of nonlabeled probe (100 nM) was added to the sample to compete out the labeled ARE probe from binding to proteins in the nuclear lysates.

The E1 and E2 enhancer sites in the HO-1 promoter each contain an ARE. Recent evidence has suggested that electrophiles can activate the ARE via the Keap1/Nrf2 pathway (1, 2). Because NEPP compounds have an electrophilic carbon at position #11 and can bind cysteine residues of cellular proteins, we studied this pathway of ARE activation (Fig. 2C). To provide direct evidence that NEPP11 could activate the HO-1 enhancer through an ARE, we studied transcriptional activation of a wild-type ARE core element (pAREluc) and a mutated form (pGC-AREluc) (9). NEPP11 stimulated the wild-type transcriptional activity up to 7-fold in a dose-dependent manner, whereas the mutant was unaffected. Similar responses were obtained with NEPP6 (data not shown). These results strongly support the notion that NEPP compounds activate the HO-1 enhancer by activating an ARE.

Because the E1 and E2 regions also contain other enhancer elements, we tested the effects of NEPP11 (2 μM) on the expression of luciferase activity derived from these elements, including the cAMP-responsive element (CRE; p3xCREluc), AP1-binding site (p3xAP1luc), NF-κB binding site (p3xNF-κBluc), NFAT-binding site (p3xNFATluc), ETS-binding site (p3xETSluc), and MEF2-binding site (p3xMEF2luc) (Fig. 2D). The plasmid pHO15luc, used for comparison, showed >5-fold induction of reporter activity. In contrast, NEPP11 (2 μM) had no effect on the activity of p3xMEF2luc, slightly depressed the activities of the p3xCREluc, p3xAP1luc, p3xNFATlu, and p3xETSluc constructs, and minimally activated p3xNF-κBluc expression. These results suggest that activation of these transcriptional elements plays a minor role in the activation of the HO-1 promoter by NEPP11.

To confirm that NEPP11 leads to an increase in transcription factors, including Nrf2, that bind to the ARE, we performed EMSAs (Fig. 2E). The band representing the labeled ARE probe was shifted in the presence of control cell lysates (Fig. 2E, lane 2), indicating binding of endogenous transcription factors to the ARE. Lysates prepared after exposure to NEPP11 (1 μM) manifest a significant increase in the intensity of this band (Fig. 2E, lane 3). In contrast, the band was totally abrogated in the presence of excess unlabeled probe (Fig. 2E, lane 5). Importantly, addition of anti-Nrf2 antibody produced a supershifted band (Fig. 2E, lane 4), consistent with the notion that one of the transcription factors binding to the ARE under these conditions represented Nrf2.

Activation of the Keap1/Nrf2 Pathway by NEPP Compounds in a Neuronal Cell Line. Next, we examined the localization of an Nrf2-GFP fusion protein (10) in transfected HT22 cells. Under basal conditions, the Nrf2-GFP fusion protein was predominantly localized in the cytoplasm but translocated into the nucleus upon exposure to 2 μM NEPP11 (Fig. 3A). In general, Nrf2 is rapidly ubiquitinated and degraded by the proteasome pathway in the cytoplasm but becomes stable when translocated into the nucleus (1, 2). Thus, we hypothesized that nuclear levels of Nrf2 protein should increase when cells are exposed to an electrophile such as NEPP, which promotes nuclear translocation. Indeed, we found this to be the case (Fig. 3B).

Fig. 3.

Activation of the Keap1/Nrf2 pathway by NEPP11. (A) Nuclear translocation of Nrf2 protein induced by NEPP11. HT22 cells transfected with pNrf2-GFP were treated with vehicle or NEPP11 (2 μM) for 24 h and observed under epifluorescence microscopy. (Scale bar, 25 μm.) (B)(Upper) Increase in nuclear Nrf2 protein in response to NEPP11. Cytosolic and nuclear fractions (200 μg protein) from cells that had been treated with vehicle or 2 μM NEPP11 for 24 h were precipitated and probed with anti-Nrf2 antibody. (Lower) The fractions (10 μg per lane) were probed with anti-NeuN, a neuronal-specific nuclear protein. (C) Induction of HO-1 protein by NEPP6-biotin. Lysates (10 μg per lane) of HT22 cells treated with various concentrations of NEPP6-biotin were probed with anti-HO-1 (Upper) or anti-β-actin (Lower). (D and E) Binding of NEPP6-biotin to Keap1. Vehicle, (–), lane 1 or NEPP6-biotin, (+), lane 2 was added to the cultures, which were then incubated for 24 h. Subsequently, lysates were prepared and subjected to precipitation with either anti-Keap1 or streptavidin. The precipitates were electrophoresed and probed. The precipitates were also probed with anti-Keap1 as a control to ensure that equal amounts of Keap1 were pulled down under each condition (D Lower). (F) Inhibition of the HO-1 promoter by Keap1(C151S) protein. The ordinate shows the fold induction by NEPP11. Values are mean ± SD; *, P < 0.01 by ANOVA.

To elucidate the mechanism of electrophile action in this regard, we examined whether NEPP compounds bind Keap1 (Fig. 3 C–E). Similar to NEPP6 and -11, we found that the conjugated product of NEPP6 and biotin (NEPP6-biotin) is neuroprotective (8). Additionally, at concentrations of 1–10 μM, NEPP6-biotin induced the expression of HO-1 protein (Fig. 3C), confirming that NEPP6-biotin retains its biological effect. We then performed immunoprecipitation experiments to provide direct evidence for adduct formation between NEPP compounds and Keap1. HT22 cells were treated with vehicle or 10 μM NEPP6-biotin and lysed. Cell lysates were immunoprecipitated with anti-Keap1 antibody and probed with streptavidin (Fig. 3D). A 73-kDa protein, corresponding to Keap1 that had bound to NEPP6-biotin, was observed in cells treated with NEPP6-biotin but not with vehicle. The precipitates were also probed with anti-Keap1 to confirm that the amount of precipitated Keap1 was the same between vehicle- and NEPP6-biotin-treated cells. Next, the lysates were treated in reverse fashion, i.e., precipitated with streptavidin and then probed with anti-Keap1 antibody (Fig. 3E). Again, precipitated protein, corresponding to Keap1 that had bound to NEPP6-biotin, was detected only in the cells treated with NEPP6-biotin, but not with vehicle. Taken together, these results are consistent with the notion that NEPP compounds bind to Keap1 in cells.

To demonstrate that binding of NEPP compounds to Keap1 protein is involved in the biological actions of NEPPs, we examined the effects of a Keap1 cysteine mutant (C151S in which cysteine residue #151 is replaced by serine) on HO-1 transcriptional activation. Cysteine residue #151 has been reported to be essential for activation of Nrf2-mediated transcription by electrophiles (11). Overexpression of this mutant should abolish the functional link between Keap1 and Nrf2 proteins and, thus, reduce sensitivity to electrophiles, such as NEPP11. Along these lines, we found that activation of the HO-1 promoter (pHO151luc) by NEPP11 was significantly depressed by cotransfection of pKeap1(C151S), but not wild-type pKeap1 (Fig. 3F).

Activation of the Keap1/Nrf2 Pathway in Primary Cortical Cultures. To begin to test the effects of NEPP compounds on primary cerebrocortical neurons, we treated mixed neuronal/glial cortical cultures with NEPP6-biotin and then fixed and stained with rhodamine conjugated to streptavidin to determine the site of NEPP accumulation. We found that MAP-2-positive neurons stained strongly for NEPP6-biotin-streptavidin, suggesting that NEPP compounds accumulate in neurons (Fig. 4A). Incubation in biotin by itself did not result in neuronal accumulation of biotin–streptavidin, and incubation in NEPP6-biotin plus excess free biotin (4 mM) did not affect the degree of neuronal accumulation, indicating that NEPP6-biotin accumulation was not facilitated by the presence of biotin but rather by NEPP itself.

Fig. 4.

Activation of the Keap1/Nrf2 pathway in neurons. (A) Accumulation of NEPP6-biotin in neurons. Cortical cultures treated with NEPP6-biotin (10 μM) were stained with anti-MAP-2 monoclonal (green) and rhodamine-conjugated streptavidin (red) antibodies and with Hoechst 33,258 (blue, 5 μg/ml). Arrows indicate neurons. (Scale bar, 25 μm.) (B) Induction of HO-1 protein in neurons by NEPP. Cortical cultures (E17 and DIV14–21) treated with vehicle or NEPP11 (0.7 μM) were stained with anti-MAP-2 (green) and anti-HO-1 (red) and with Hoechst dye (blue). After exposure to NEPP11, HO-1 increased substantially in neurons. (Scale bar, 25 μm.) (C) Induction of HO-1 protein after exposure to NEPP shown in immunoblots. Lysates (10 μg per lane) of primary cortical cultures treated with vehicle or NEPP11 for 24 h were probed with anti-HO-1 (Upper) or anti-β-actin (Lower). The fold induction of HO-1 by NEPP11 was 2.2 ± 0.25 (for 0.5 μM NEPP11) and 3.5 ± 0.25 (for 1.0 μM NEPP11), as assessed with densitometry. (Scale bar, 50 μm.) (D) Activation of HO-1 promoter and ARE by NEPP11. Cortical cultures were transfected with reporter cDNAs (1 μg per well), and 0.7 μM NEPP11 was added to the cultures. After a 24-h incubation, cell lysates were subjected to luciferase reporter assays. The ordinate shows the fold induction by NEPP11. (E) Inhibition of the HO-1 promoter by Nrf2(S40A) protein. Values are mean ± SD; *, P < 0.01 by ANOVA.

We reasoned that, if, unlike other electrophiles (12), NEPPs accumulate predominantly in neurons, then HO-1 might be induced preferentially in this cell type. To check this possibility, we performed immunofluorescence studies with anti-HO-1 antibody after exposure to a potent NEPP compound (NEPP11, 0.7 μM). In control cultures, nonneuronal cells expressed relatively more HO-1 at baseline than did neurons. After addition of NEPP11, HO-1 immunofluorescence increased mainly in neurons, in both the cytosol and nucleus (Fig. 4B). NEPP11 also increased total HO-1 protein in the cultures, as detected by immunoblotting (Fig. 4C).

Next, transcriptional activation of the HO-1 promoter by NEPP was examined by reporter gene assays in primary neurons. In transfected cortical cultures, NEPP11 (0.7 μM) significantly increased activity of the HO-1 promoter and ARE core element, an effect that was abrogated by mutation of the HO-1 enhancer sites or the ARE site, respectively (Fig. 4D). These results suggest that NEPP11 induced HO-1 protein in primary cortical neurons through activation of the ARE elements in the HO-1 promoter. If the Keap1/Nrf2 pathway, indeed, mediates activation of the HO-1 promoter by NEPP11, mutant Nrf2 protein should inhibit this activation. For this purpose, we used the mutated construct pNrf2(S40A)-GFP (in which the serine residue at position #40 is replaced by an alanine); the encoded protein does not activate the ARE, because it cannot translocate into the nucleus (10). NEPP11 significantly activated the HO-1 promoter, and cotransfection with pNrf2-GFP did not affect activation (Fig. 4E). In contrast, cotransfection with pNf2(S40A)-GFP almost completely knocked down activation by NEPP11 (additionally, the basal level of HO-1 promoter activity was reduced). Taken together, our results suggest that NEPP11 activates the Keap1/Nrf2 pathway selectively in neurons. Moreover, the selective activation in neurons may explain the relatively small amplitude of total HO-1 activation seen in this mixed neuronal/glial culture system in which glia predominate.

Neuroprotection by NEPP11. Neuron-selective activation of the Keap1/Nrf2/HO-1 pathway by NEPP compounds should provide neuroprotection. Hence, we examined the action of NEPP11 both in vitro and in vivo in excitotoxic paradigms, first in culture as a protectant from NMDA-receptor-mediated insults and then after MCAO by using the intraluminal filament model of transient focal ischemia/reperfusion in mice.

Exposure of primary cortical cultures to relatively mild insults, such as low concentrations of NMDA (50 μM) for short durations (15 min), is known to cause delayed and predominantly apoptotic neuronal cell death (13). We stained the cultures with both anti-MAP-2 and anti-NeuN monoclonal antibodies to label neuronal dendrites and nuclei, respectively. Apoptotic nuclei were identified by morphological changes seen with Hoechst staining. In this system, NEPP11 significantly decreased the number of apoptotic neurons (Fig. 5 A and B), suggesting that NEPP11 protected neurons against excitotoxicity in vitro. Moreover, the relatively specific HO-1 antagonist, zinc protoporphyrin, abrogated the neuroprotective effect of NEPP11 in these cerebrocortical cultures. This result is consistent with the notion that NEPP11 protects primary cortical neurons against excitotoxicity, at least in part, through induction of HO-1. If this antioxidant pathway is important for NEPP action, then downstream events should also be affected. Along these lines, we found that NEPP11 (1 μM) inhibited NMDA-induced caspase-3 activation (see Fig. 7, which is published as supporting information on the PNAS web site). In contrast, if this is the predominant pathway, then other known anti-apoptotic genes, e.g., bcl-xL and bcl-2, might not be induced by NEPP compounds (see Fig. 8, which is published as supporting information on the PNAS web site). Indeed, we found this to be the case.

Fig. 5.

Neuroprotection by NEPP11 in vitro and in vivo. (A) Inhibition of excitotoxicity by NEPP11 in vitro. Vehicle or NEPP11 (0.7 μM) was added to cerebrocortical cultures (E17 and DIV14–21) 60 min before treatment with NMDA (50 μM) for 15 min. The cultures were then incubated for 20 h and subsequently stained with anti-MAP-2 and anti-NeuN (red) and with Hoechst dye (blue). (B) Statistical analysis of the neuroprotective effect afforded by NEPP11 and inhibition of the effect by zinc protoporphyrin (ZnPP). The number of apoptotic neurons was assessed by determining the apoptotic index [(No. of condensed nuclei in MAP-2 or NeuN-positive cells)/(No. of total MAP-2 or NeuN-positive cells) × 100%], as reported in ref. 13. ZnPP (10 μM) was added simultaneously with NEPP11. Values are mean ± SD; *, P < 0.01 by ANOVA. (C) Effect of NEPP11 and NEM on cellular levels of total GSH in primary cortical cultures. NEPP11 or NEM was added at t = 0 and levels of total GSH were measured at the indicated times. GSH content of control cortical cultures (set arbitrarily at 100%) was 46.8 ± 4.5 nmol/mg protein. Unlike NEPP11, NEM did not produce neuroprotection in these cultures and, in fact, resulted in neuronal death after 24 h. (D) Neuroprotective effect of NEPP11 after MCAO/reperfusion. Representative coronal brain sections stained with 2,3,5-triphenyltetrazolium chloride from vehicle- and NEPP11-treated mice. (E) Statistical analysis of the neuroprotective effect of NEPP11 on MCAO/reperfusion. Values are means ± SEM, n = 4 for each group; *, P < 0.01 by paired ANOVA, with corresponding vehicle-treated value. (F and G) Induction of HO-1 protein in the mouse brain by NEPP11 detected by Western blotting (F) and immunostaining (G). Vehicle or NEPP11 was injected in the same manner as in the MCAO experiments. (F) Brain lysates (10 μg per lane) were extracted and subjected to Western blot with anti-HO-1 (Upper) and anti-β-actin antibody (Lower). (G) Coronal sections of mouse brain were stained with anti-MAP-2 (green) and anti-HO-1 (red) antibodies. Neurons in layer V of the parietal region of the cerebral cortex from vehicle-injected (Left) and NEPP11-injected (Right) mice are shown. (Scale bar, 25 μm.)

One caveat to the mechanism of NEPP compounds acting at the level of Keap1 to induce HO-1 transcription is that NEPP could potentially react indiscriminately with other thiol-containing compounds in cells. To approach this question, we assessed cellular GSH levels to determine whether NEPP11 would affect this abundant antioxidant thiol. We found that NEPP11 did not deplete GSH levels in cortical cultures, unlike many other electrophiles that have this effect (e.g., N-ethylmaleimide) (Fig. 5C). In fact, GSH levels transiently increased after exposure to NEPP11. This increase in GSH may have occurred by induction γ-GCL (14), the rate-limiting enzyme in GSH biosynthesis, and could also contribute to cytoprotection. In fact, this scenario seems likely, because the Keap1/Nrf2 pathway regulates the expression of γ-GCL in addition to HO-1 (1). Besides GSH, cells have another major reduction pathway representing the thioredoxin–glutaredoxin system. However, we found that NEPP11 (1 μM) did not significantly affect the expression of thioredoxin or glutaredoxin under our conditions (see Supporting Results in Supporting Text, which is published as supporting information on the PNAS web site).

Next, we tested whether NEPP11 could decrease the size of brain infarcts after MCAO/reperfusion injury (Fig. 5 D and E). NEPP11 or vehicle was injected i.p. 1 h before and 4 h after MCAO. The area of brain infarction (corrected for possible edema) was assessed on coronal sections stained with 2.5% 2,3,5-triphenyltetrazolium chloride 24 h after the onset of reperfusion. As described in ref. 11, we monitored physiological variables, including arterial pressure, blood gases and glucose, core body temperature, and regional cerebral blood flow; these parameters did not differ between the control and NEPP11-treated groups (data not shown). NEPP11 significantly reduced the infarct area in coronal sections, suggesting that NEPP11 is neuroprotective in vivo. We did not examine the effects of NEPP11 administered postinfarct in this study, because the drug requires several hours to exert its neuroprotective effect by transcriptional activation and therefore requires pretreatment (7, 8).

To test the hypothesis that NEPP protection against brain ischemia is associated with HO-1 expression, we examined HO-1 induction by Western blotting and immunostaining. NEPP11 or vehicle was injected i.p. 12 h before the animals were killed. We found that the same concentration of NEPP11 that prevented neuronal cell death during brain ischemia increased the level of HO-1 protein in the brain (Fig. 5F). Induction of HO-1 protein was observed in neuronal soma and dendrites (Fig. 5G).

Discussion

This study provides evidence that electrophilic drugs can afford neuroprotection through activation of the Keap1/Nrf2 pathway and consequent up-regulation of HO-1 and possibly other class II enzymes. It was known that up-regulation of HO-1 decreased stroke size, as assessed in HO-1 transgenic mice (3). Here, we develop and characterize a set of small-molecule electrophiles that activate HO-1 transcription in neurons, showing that this pathway may represent a druggable target in the brain. The key element in successful neuroprotection by NEPP compounds is that they can activate the Keap1/Nrf2 pathway at nontoxic concentrations. Many other electrophilic molecules cause systemic side effects and are not neuroprotective, probably because they also deplete critical reducing substances in the cell, such as GSH, but this is not the case with the NEPP drugs (Fig. 5C).

NEPP compounds are lipophilic, an important characteristic for their accumulation in neurons, as demonstrated in this study with labeled NEPP (NEPP-biotin). Kraft et al. (12), however, reported that another electrophile, tert-butylhydroquinone (TBHQ), activates the ARE in astrocytes, a fact that may appear inconsistent with our observations. Nevertheless, it should be noted that the chemical structures of electrophiles, such as NEPP and TBHQ, vary widely and may affect their cellular uptake (12). Hence, one electrophile may very well localize to astrocytes, whereas another predominates in neurons, as observed here for NEPP compounds. NEPP compounds (Δ⁷-prostaglandinA₁ analogues) have been molecularly designed based on the chemical structure of Δ¹²-prostaglandinJ₂, and these molecules share many chemical and biological properties (15). Δ¹²-prostaglandinJ₂ is reportedly transported into cells by active transport through the cell membrane (16). We speculate, therefore, that neurons may have a more active transport system for NEPP compounds than do glia, because we observed that NEPP compounds accumulate preferentially in neurons. In contrast, the electrophile TBHQ may simply diffuse into cells and, thus, affect glia, which greatly outnumber neurons.

Our findings suggest the neuroprotective mechanism of NEPP action shown schematically in Fig. 6. Within cells, these drugs bind to the cytosolic regulator protein Keap1, which, in turn, liberates Nrf2. Nrf2 is then translocated into the nucleus, where it activates AREs on the HO-1 promoter (1, 2). Transcription of HO-1 is thus activated in neurons, and an increase in HO-1 protein leads to degradation of heme molecules, producing biliverdin and bilirubin (1, 2). The accumulation of bilirubin, a potent antioxidant molecule, is responsible, at least part, for the neuroprotective effects of HO-1 and, thus, of NEPP compounds (7, 14). Additionally, we found that inhibition of HO-1 by zinc protoporphyrin prevented the protective effect of NEPP, consistent with the notion that the therapeutic action of these drugs is mediated predominantly by this pathway.

Fig. 6.

Proposed mechanism of neuroprotective effects afforded by NEPP compounds through Keap1/Nrf2 transcriptional activation of HO-1 and subsequent antioxidant action.

Recently, decreased Nrf2 transcriptional activity was also reported to cause age-related loss of GSH synthesis (17). Low-molecular-weight compounds can induce γ-GCL through activation of the ARE to increase GSH levels. Thus, compounds that regulate the Keap1/Nrf2 pathway may be promising candidates for neuroprotection against free radical stress through induction of γ-GCL as well as HO-1, both of which help prevent accumulation of reactive oxygen species.

In summary, we found that modulation of the Keap1/Nrf2 pathway by NEPP compounds leads to activation of the HO-1 promoter by Nrf2. Induction of HO-1 protein is known to play an important neuroprotective role against excitotoxicity and brain ischemia (3–8). How can clinically useful drugs be developed based on the chemical structures of cyclopentenone prostaglandins like the NEPPs? One possible approach is to synthesize drugs of “moderate” electrophilicity like the NEPPs. Strongly electrophilic compounds are known to deplete the cell of critical thiol-containing compounds like GSH and, hence, contribute to cell death. In contrast, NEPP compounds and their congeners interact with Keap1 without depleting GSH. The present results suggest that selective activators of the Keap1/Nrf2 pathway may be promising candidates as neuroprotective agents that act through induction of phase II genes, including HO-1.

Materials and Methods

Cell Cultures, Transfection, and Glutathione (GSH) Measurement. HT22 cells (7, 8, 14) and primary cortical neurons (13) were cultured as described. Transfection was performed with Lipofectamine 2000 (Invitrogen). In the reporter gene assays, firefly luciferase activity in cell lysates was measured with a luminometer (Promega). Total GSH (reduced and oxidized) was determined as described in ref. 9.

Immunoprecipitation, Western Blots, and Immunofluorescence. These assays were performed as described in ref. 18 by using the following antibodies: anti-HO-1 (SPA895, 1:1,000, Stressgen Biotechnologies, Victoria, Canada), anti-Nrf2 (1:100, Santa Cruz Biotechnology), anti-Keap1 (1:100, Santa Cruz Biotechnology), or anti-actin (1:5,000, Oncogene Research Products, San Diego).

EMSAs. Double-stranded AREs were labeled by using a biotin 3′-end DNA labeling kit (Pierce). Nuclear lysates were incubated with the labeled probe for 20 min at room temperature, resolved on an 8% native polyacrylamide gel, and transferred to Hybond-N⁺ (Amersham Pharmacia). Signals were visualized with peroxidase-conjugated streptavidin (Pierce).

Focal Cerebral Ischemia and Reperfusion. The filament model of middle cerebral artery occlusion (MCAO)/reperfusion was used as described in ref. 18 (see Supporting Methods in Supporting Text). NEPP11 was injected at 100 μg/ml in a 7.5% solution of DMSO in PBS; controls received the diluent alone. The investigator was blinded to treatment group.

Statistical Analysis. Experiments presented were repeated at least three times with four samples. The data are presented as mean ± SD (for in vitro experiments) or SEM (for in vivo experiments).