Keap1 regulates both cytoplasmic-nuclear shuttling and degradation of Nrf2 in response to electrophiles


  • Ken Itoh,

    1. Exploratory Research for Advanced Technology, Japan Science and Technology Corporation, and
    2. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author
  • Nobunao Wakabayashi,

    1. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author
  • Yasutake Katoh,

    1. Exploratory Research for Advanced Technology, Japan Science and Technology Corporation, and
    2. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author
  • Tetsuro Ishii,

    1. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author
  • Tania O'Connor,

    1. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author
  • Masayuki Yamamoto

    Corresponding author
    1. Exploratory Research for Advanced Technology, Japan Science and Technology Corporation, and
    2. Center for Tsukuba Advanced Research Alliance and Institute of Basic Medical Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8577, Japan
    Search for more papers by this author

  • Communicated by: Shunsuke Ishii



Background: Transcription factor Nrf2 regulates the expression of a set of detoxifying and anti-oxidant enzyme genes. Several lines of evidence suggest that electrophiles and reactive oxygen species liberate Nrf2 from its cytoplasmic repressor Keap1 and provoke the accumulation of Nrf2 in the nucleus. To elucidate the molecular mechanisms as to how Nrf2 is activated by inducers, we examined the cytoplasmic-nuclear shuttling and turnover of Nrf2.

Results: We found that Nrf2 is rapidly degraded through the proteasome pathway, while electrophiles cause Nrf2 nuclear translocation with concomitant stabilization. Crucial to the inducible accumulation of Nrf2 is the enfeebling of the Nrf2–Keap1 interaction by electrophiles. Exploiting mice which have the LacZ reporter gene knocked into the nrf2 locus, we revealed that the inducible accumulation of Nrf2 protein by electrophiles in macrophages and intestinal epithelia could be recapitulated by the Nrf2 N-terminal region in combination with a nuclear localization signal. We also found constitutive Nrf2 nuclear accumulation in Keap1-deficient mouse macrophages.

Conclusions: Our results highlight the fact that Nrf2 protein turnover is regulated by Keap1 mediated subcellular compartmentalization.


To counteract the damage that can be provoked by electrophiles and reactive oxygen species (ROS), higher animals have developed elaborate defence mechanisms (Prestera et al. 1993; Primiano et al. 1997). A battery of genes encoding detoxifying and anti-oxidative stress enzymes/proteins is coordinately induced following exposure to electrophiles and ROS (Beutler et al. 1995; Hayes & Pulford 1995). This coordinated response is regulated through a cis-acting element called the anti-oxidant responsive element (ARE) or electrophile responsive element (EpRE) within the regulatory region of target genes (Rushmore et al. 1991; Friling et al. 1990). Genes encoding a subset of drug metabolizing enzymes, such as glutathione S-transferases (GSTs) (Friling et al. 1990) and NAD(P)H-quinone oxidoreductase 1 (NQO1) (Rushmore et al. 1991), have been shown to be under ARE/EpRE regulation, along with a subset of anti-oxidant genes, such as haem oxygenase-1 (HO-1) (Alam et al. 1995), the subunits of γ-glutamylcysteine synthetase (γ-GCS) (Mulcahy et al. 1997) and thioredoxin (Kim et al. 2001).

Nrf2/ECH (NF-E2-Related Factor 2 (Moi et al. 1994) or chicken Erythroid-derived CNC-Homology factor (Itoh et al. 1995)) was recently identified as the major regulator of ARE-mediated gene expression (reviewed in Itoh et al. 1999b; Ishii et al. 2002). Nrf2/ECH belongs to the Cap-N-Collar (CNC) family of transcription factors that share a highly conserved basic region-leucine zipper (bZip) structure (reviewed in Motohashi et al. 1997). Nrf2 requires a member of the small Maf proteins as an obligatory partner molecule for binding to their cognate DNA sequence (Itoh et al. 1995; Marini et al. 1997). Through nrf2 gene targeting analysis, we demonstrated that Nrf2 regulates a battery of genes encoding drug metabolizing enzymes and anti-oxidant proteins (Itoh et al. 1997; Ishii et al. 2000). Subsequently, we identified Keap1 (Kelch-like ECH Associating Protein 1) as a direct binding partner of Nrf2 (Itoh et al. 1999a). The central role attributed to the Nrf2-Keap1 system is the regulation of cellular defence against a variety of environmental insults; for example, in the electrophile counterattack response (Itoh et al. 1999b; Ishii et al. 2000), in acetaminophen intoxication (Enomoto et al. 2001), in chemical carcinogenesis (Ramos-Gomez et al. 2001), and in diesel exhaust inhalation (Aoki et al. 2001).

Several key features have emerged from an extensive study of the molecular mechanisms of Nrf2 activation by electrophiles and ROS. In peritoneal macrophages, nuclear Nrf2 accumulation in response to electrophiles or ROS appears to be an important step in Nrf2 mediated gene induction (Ishii et al. 2000). Forced over-expression of Nrf2 gave rise to ARE-reporter gene transcription in a reporter co-transfection/transactivation assay (Itoh et al. 1999a) and switched on endogenous target genes in zebrafish (Kobayashi et al. 2002) without any additional electrophilic or oxidative signals. Therefore, increasing the amount of Nrf2 circumvents the normal activation pathway of electrophiles and ROS and actuates cytoprotective gene transcription. In the co-transfection/transactivation model, the concomitant expression of Keap1 sequesters Nrf2 in the cytoplasm and represses Nrf2 transactivation activity (Itoh et al. 1999a). Treatment of the cells with electrophiles liberates Nrf2 from Keap1. Nrf2 subsequently translocates to the nucleus and activates the transcription of cytoprotective enzyme genes through the ARE.

While these results clearly demonstrate that electrophiles or ROS induce a subset of cytoprotective enzyme genes by counteracting Keap1 repression of Nrf2 activity, many questions still remain regarding the manner by which electrophilic signals dissociate Nrf2 from Keap1, and Nrf2 translocates from the cytoplasm to the nucleus. To address these questions, we investigated the accumulation, stability and nuclear-cytoplasmic shuttling of Nrf2 in mouse peritoneal macrophages. Our results indicate a rapid turnover of Nrf2 in the cytoplasm under non-stress conditions, whereas Nrf2 protein stabilizes in the presence of electrophiles and accumulates in the nucleus. We found that the N-terminal region of Nrf2, in combination with the Nuclear Localization Signal (NLS), is sufficient for this electrophilic induced nuclear accumulation, and that Keap1 plays important roles in this process by interacting with the Nrf2 N-terminal Neh2 domain. Our data suggest that the N-terminal region of Nrf2 initiates rapid proteolysis within the cytoplasm, but mediates turnover via a slower pathway in the nucleus.


Electrophiles accumulate Nrf2 protein in peritoneal macrophages

To elucidate the molecular basis of the Nrf2 activation mechanism in response to electrophiles, we examined the relationship between Nrf2 activation and Nrf2 target gene induction in primary cultures of peritoneal macrophages. We executed a time course study of Nrf2 target gene induction (Fig. 1A) and Nrf2 accumulation in the nucleus (Fig. 1B) using diethylmaleate (DEM), an electrophile classified as a Michael reaction acceptor. The inducible expression of Peroxiredoxin I (PrxI)/MSP23, one of the major Nrf2 target genes in peritoneal macrophages (Ishii et al. 2000), occurred as early as 2 h after DEM treatment (Fig. 1A, top panel). In contrast, the mRNA level of Nrf2 was slightly reduced by DEM treatment (Figs 1A and 2A, middle panels), indicating that DEM was inducing the expression of the Nrf2 target gene through some post-transcriptional mechanism.

Figure 1.

Nrf2 activation by DEM accompanies the accumulation of Nrf2 in whole cell lysates. (A) RNA blot analysis of Nrf2 and PrxI. Peritoneal macrophages were cultured in the presence (lanes 3, 5 and 7–9) or absence of 100 µm DEM (lanes 1, 2, 4 and 6), and total RNAs were prepared at the various time points indicated in the Figure. (B) DEM caused Nrf2 to accumulate in the nucleus. Nuclear extracts from macrophages treated with (lanes 3, 5 and 7) or without (lanes 1, 2, 4 and 6) 100 µm DEM for the indicated times were immunoblotted with either anti-Nrf2 antibody (N) or anti-Lamin B antibody (L). (C) Nrf2 accumulated in total cell extracts after treatment with DEM. Total cell extracts from macrophages treated with (lanes 3, 5 and 7) or without (lanes 1, 2, 4 and 6) 100 µm DEM for the indicated times were immunoblotted with either anti-Nrf2 antibody (N) or anti-Lamin B antibody (L). Over-expressed recombinant Nrf2 in 293T cells (R. Nrf2) was also loaded as a control (lane 8). (D) Nrf2 was accumulated in total cell lysates by various electrophiles and ROS producing agents. Total cell extracts prepared from macrophages treated with DMSO (lane 1), 2.5 µm menadione (lane 2), 100 µm DEM (lane 3), 10 µm CDNB (lane 4), or 10 µm sulforaphane (lane 5) for 4 h were immunoblotted with anti-Nrf2 antibody (N) or anti-Lamin B antibody (L).

Figure 2.

Both the induction of target genes and the activation of Nrf2 are sensitive to treatment with cycloheximide (CHX). (A) The effect of CHX on the induction of PrxI. Total RNAs prepared from macrophages that were untreated (lanes 1) or treated with either 100 µm DEM (lane 2), 1 µg/mL CHX (lane 4) or both (lane 3) for 4.5 h were blot hybridized with the cDNA probes indicated in the figure. (B) Macrophages were either untreated (lanes 1 and 2) or treated with 100 µm DEM (lanes 3 and 4) or both DEM and CHX (lanes 5 and 6) for 4 h and 0.5% Triton X soluble (S) (lanes 1, 3 and 5) or insoluble (I) (lanes 2, 4 and 6) fractions were prepared. Fractions were probed with anti-Nrf2 antibody (N), anti-lamin B antibody (L), or anti-β-actin antibody (A). Over-expressed recombinant Nrf2 in 293T cells (R, Nrf2) was also loaded as a control (lane 7).

The Nrf2 protein level in the nuclear fraction was examined by immunoblot analysis using anti-Nrf2 antibody. Nrf2 migrated as 110 kDa in SDS-PAGE, which was the same as over-expressed recombinant Nrf2 in 293T cells (Fig. 1C, lane 8; Kwak et al. 2002). A significant increase in Nrf2 was observed as early as 0.5 h following DEM treatment (Fig. 1B). We also found that Nrf2 binding activity to the ARE sequence was markedly increased and showed a similar induction profile (data not shown). This is consistent with our previous analysis in that, although Nrf2 target genes are significantly induced, Nrf2 mRNA is not increased after the treatment of macrophages with DEM or similar agents (Ishii et al. 2000). These results demonstrate that the nuclear accumulation of Nrf2 precedes induction of target gene expression and is thus a key event in the inducible expression of phase II and anti-oxidant enzyme genes.

We then examined whether Nrf2 accumulation can be detected in total cell lysate (Fig. 1C). Importantly, before DEM treatment, Nrf2 was scarcely detectable in total cell lysate, indicating that the size of the cytoplasmic Nrf2 pool is very small in non-induced cells. Following DEM treatment, however, Nrf2 accumulated in total cell lysate, albeit the immunoblot signal was very weak compared to that with nuclear extracts. Electrophiles other than DEM, such as menadione (lane 2), chlorodinitrobenzene (CDNB) (lane 4), and sulforaphane (lane 5) provoked a similar accumulation of Nrf2 in total cell lysate (Fig. 1D), showing that cellular Nrf2 accumulation is a mechanism which is common to several Nrf2 inducers. Since the Nrf2 protein signal was scarcely detectable in immunoblot analysis of the cytosolic fraction (data not shown), we assume that Nrf2 accumulation in the nucleus mainly occurs through new protein synthesis and not through the translocation of pre-existing Nrf2 protein from the cytosol to the nucleus.

De novo protein synthesis is required for nuclear accumulation of Nrf2

In order to test whether de novo protein synthesis is required for nuclear accumulation of Nrf2, we treated peritoneal macrophages with DEM in either the presence or absence of the protein synthesis inhibitor cycloheximide (CHX). The induction of PrxI mRNA by DEM was largely abolished in the presence of CHX (1 µg/mL; Fig. 2A). Similarly, the induction of HO-1 mRNA by DEM was significantly affected by CHX treatment (data not shown). The major regulator of inducible defence gene expression may therefore have a short degradation half-life and rapid turnover.

Since Nrf2 mRNA levels did not significantly increase following treatment by DEM and/or CHX, the possibility exists that Nrf2 protein actually turns over rapidly and its level decreases in response to CHX, hence giving rise to the decrease in Nrf2 target gene expression. We therefore examined the effect of CHX on the levels of nuclear Nrf2. The build up of Nrf2 in the 0.5% Triton X insoluble nuclear fraction can be monitored biochemically (Fig. 2B). Nrf2 accumulated in the Triton X insoluble fraction, but not in the soluble fraction, of peritoneal macrophages from control, DEM-treated and DEM plus CHX-treated mice. Although DEM markedly induced the Nrf2 level (compare lanes 2 and 4), CHX largely cancelled this induction (lane 6). These results thus indicate that Nrf2 turns over rapidly in macrophages and that new protein synthesis is required for the nuclear accumulation of Nrf2.

Proteasome inhibitors also induce Nrf2 accumulation

In order to gain insight into the rapid turnover rate of Nrf2, we carried out two further experiments. Firstly, we examined the effect of various protease inhibitors on the level of Nrf2 protein. Immunoblot analysis using Nrf2 antibody revealed that proteasome inhibitors, such as clasto-Lactacystin β-Lactone, MG132 and MG115, led to the accumulation of Nrf2 in the Triton X insoluble fraction, but that a calpain inhibitor failed to accumulate Nrf2 (Fig. 3A, top panel). Our results suggest the degradation of Nrf2 specifically by the proteasome system in non-induced cells.

Figure 3.

DEM inhibits the degradation of Nrf2 by proteasomes. (A) Nrf2 protein was accumulated in the 0.5% Triton X insoluble fraction by specific proteasome inhibitors. Macrophages were treated with DMSO (lanes 1 and 2), 10 µm MG132 (lanes 3 and 4), 10 µm clasto-Lactacystin β-Lactone (lanes 5 and 6), 50 µm MG115 (lanes 7 and 8), or 10 µm calpain inhibitor (lanes 9 and 10) for 4.5 h. The 0.5% Triton soluble (lanes 2, 4, 6, 8 and 10) or insoluble (lanes 1, 3, 5, 7 and 9) fractions were prepared and immunoblotted with the indicated antibody. (B) Total cell lysates from macrophages treated with 1 µg/mL CHX in the presence or absence of 100 µm DEM for the indicated times were subjected to immunoblot analysis with anti-Nrf2 antibody or anti-Lamin B antibody. (C) The band intensities in (B) were measured by densitometric analysis. The relative Nrf2 band intensities (Nrf2/Lamin B) are shown against a semilog plot. (D) The cells were treated with 10 µm MG132 alone (lane 1) or in combination with 50 (lane 2), 100 (lane 3) or 200 µm (lane 4) DEM for 4 h. Total cell lysates were subjected to immunoblot analysis with anti-Nrf2 antibody or anti-Lamin B antibody.

Second, we analysed the effect of electrophiles on Nrf2 protein turnover. Since the expression level of Nrf2 in the total cell lysate of peritoneal macrophages was too low to measure quantitatively, we induced the intracellular level of Nrf2 protein by MG132, a reversible inhibitor of proteasome, for 2.5 h (Lee & Goldberg 1998). After the thorough removal of MG132 and treatment of the peritoneal macrophages with CHX and/or DEM, the Nrf2 protein level in the total cell lysate was measured by immunoblot analysis with anti-Nrf2 antibody. Nrf2 was rapidly degraded with a half-life of 18.5 min, whereas DEM stabilized the Nrf2 protein and prolonged its intracellular half-life to 31.7 min (Fig. 3B,C). To clarify the mechanism of Nrf2 stabilization by DEM, we examined the effect of DEM in the presence of a saturating amount of MG132. If DEM had worked through a different pathway than MG132 to accumulate Nrf2, they should synergistically accumulate Nrf2. As shown in Fig. 3D, DEM did not affect the protein level of Nrf2, suggesting that DEM attenuates the pathway leading to the proteasomal degradation of Nrf2. From this, we postulate that Nrf2 protein is stabilized by electrophiles through the inhibition of proteasomal Nrf2 degradation.

The N-terminus of Nrf2 and the NLS are sufficient for the electrophilic response

We generated a germ line nrf2 mutant mouse by replacing exon V of the nrf2 gene, which encodes the DNA binding and dimerization domains, with an NLS-LacZ-neo (nuclear localization signal-LacZ-neomycin resistant gene) cassette (Itoh et al. 1997). This genetically engineered mouse expressed the Nrf2-LacZ fusion gene, giving rise to a protein product containing the N-terminal portion of Nrf2 and complete β-galactosidase. The structure of the fusion protein is shown in Fig. 4A. However, although we are confident that the knock-in targeting was executed in a flawless manner, it puzzled us that we could not detect β-galactosidase activity or its protein in these mutant mice. This was particularly strange since mRNA encoding the Nrf2-LacZ fusion protein was expressed at a level comparable to endogenous Nrf2 mRNA (Itoh et al. 1997). The finding that Nrf2 is rapidly degraded in vivo prompted us to examine whether the β-galactosidase protein might also be rendered unstable due to its fusion with the N-terminal region of Nrf2.

Figure 4.

The N-terminal region of Nrf2 is sufficient for the response to electrophiles. (A) Schematic representation of the Nrf2 knock-in strategy. NLS-LacZ was knocked into exon V of the nrf2 gene locus (top panel). The lower structure represents the resultant fusion protein. (B) Effect of DEM and MG132 on the Nrf2 LacZ fusion proteins. nrf2−/− peritoneal macrophages were untreated (lane 1) or treated with 100 µm DEM (lane 2) or 10 µm MG132 (lane 3) for 4 h and total cell lysates were subjected to immunoblot analysis using anti-LacZ antibody. (C) and (D) Immunohistochemical staining of BHA treated or untreated mice by anti-LacZ antibody. nrf2−/− mice were fed on a diet supplemented with (C) or without (D) 0.7% BHA for 3 days, and their tissue sections were analysed by anti-lacZ antibody. (E) RNA blot analysis of Nrf2-LacZ. nrf2−/− mice were fed in duplicate with a BHA-containing diet for the indicated periods and total RNAs were analysed by Nrf2 cDNA probe or β-actin probe.

To test the effect of MG132 and DEM on the level of Nrf2-lacZ protein, we carried out several series of experiments exploiting the homozygous Nrf2-LacZ knock-in mouse (we will refer to this mouse as nrf2–/–) and peritoneal macrophages derived from this mouse. We first determined the Nrf2-LacZ protein level in peritoneal macrophages from the nrf2−/− mouse treated with either MG132 or DEM using anti-LacZ and anti-Nrf2 antibodies. Showing very good agreement with our previous analysis, the Nrf2-LacZ protein was not detectable in peritoneal macrophages in primary culture (Fig. 4B, lane 1). In contrast, accumulation of Nrf2-LacZ protein, which migrates to approximately the 180 kDa-size position, was clearly identified in the macrophages treated with either DEM (lane 2) or MG132 (lane 3) by both anti-LacZ antibody (Fig. 4B) and anti-Nrf2 antibody that specifically recognizes the N-terminal region of Nrf2 (data not shown). These results thus revealed that the N-terminus of Nrf2 is responsible for the accumulation of Nrf2 in response to DEM.

We previously found that Nrf2 is essential for the BHA induction of phase II drug metabolizing enzymes and anti-oxidant stress enzymes/proteins in the intestine and liver (Itoh et al. 1997; Ishii et al. 2000). Based on this, we next examined the effect of BHA on the level of Nrf2-LacZ protein in vivo. After feeding the nrf2−/− mice with BHA for 3 days, they were dissected and tissue sections of the intestine were examined immunohistochemically with anti-LacZ antibody. Consistent with our expectation, positive nuclear staining was observed specifically in the intestinal epithelium of BHA-treated nrf2−/− mice (Fig. 4C), but not in the sections of nrf2−/− mice fed on a normal diet (Fig. 4D). Thus, Nrf2-LacZ protein also turns over rapidly in vivo, with DEM and BHA preventing rapid degradation of the protein. Importantly, the in vivo level of Nrf2-LacZ mRNA remained constant during BHA induction (Fig. 4E). In conclusion, the N-terminus of Nrf2, in combination with the NLS, is sufficient for the response to the proteasome inhibitor MG132, the electrophile DEM, and the dietary anti-oxidant BHA.

The Neh2 domain mediates proteasomal degradation

To determine which portion of the Nrf2 N-terminal region is recognized by proteasome, the Neh2, Neh4 and Neh5 domains of Nrf2 were independently fused to enhanced green fluorescent protein (EGFP) (Fig. 5A) and transfected into NIH3T3 mouse fibroblasts. Neh4-EGFP was excluded from the analysis, as its expression, even in the presence of MG132, could not be detected. Neomycin-resistant cells were selected and multiple mass stable transformants were established for each construct. Cells were treated with MG132 for 9 h in order to amass the fusion proteins and EGFP. After thoroughly washing MG132 from the cells, the degradation half-life of each EGFP fusion protein was determined in the presence of the protein synthesis inhibitor CHX by following a procedure similar to that used for Fig. 3. Intriguingly, the Neh2-EGFP protein level declined rapidly after the addition of CHX, with a degradation half-life of 3 h, whereas the levels of EGFP alone and Neh5-EGFP remained high, even 10 h after CHX treatment (Fig. 5B,C). The rapid degradation of the Neh2 domain was proteasome dependent, because treatment with MG132 along with CHX inhibited Neh2 degradation (data not shown).

Figure 5.

Neh2 mediates the rapid degradation of Nrf2. (A) Schematic representation of Nrf2-EGFP fusion proteins. ‘Neh domain’ stands for highly conserved regions of the Nrf2 protein within various species. Neh2 and Neh5 correspond to amino acid sequences 1–99 and 153–227, respectively. (B) Neh2 enhances the degradation of the reporter gene EGFP in the mouse fibroblast cell line NIH 3T3. Fusion proteins and EGFP were first accumulated by 10 µm MG132 for 9 h. The protein levels of EGFP (top panel), Neh2-EGFP (middle panel), or Neh5-EGFP (bottom panel) were determined using anti-EGFP antibody after treating the cells with 20 µg/mL CHX for the indicated times. (C) Determination of the degradation half-life of Neh2-EGFP. The band intensities in (B) were measured by densitometry. The relative EGFP values (EGFP/Lamin B) are shown against a semilog plot. (D–F) Localization of Nrf2-EGFP fusion proteins. NIH3T3 cells were transiently transfected with EGFP (D), Neh2-EGFP (E) or Neh5-EGFP (F).

The accumulation or rapid degradation of EGFP and its fusion proteins was visualized by green fluorescence. Since EGFP, and the Neh2 and Neh5 EGFP fusion proteins did not retain a canonical NLS, they dispersed throughout the cells. The important finding here is that, although cells transfected with EGFP or Neh5-EGFP showed bright green fluorescence (Fig. 5D,F), those expressing Neh2-EGFP showed only marginal fluorescence (Fig. 5E), reflecting the rapid turnover and low accumulation level of the Neh2-EGFP fusion protein. Thus, these results clearly support our hypothesis that Neh2 is the core domain within the Nrf2 N-terminus which is responsible for mediating the rapid degradation of Nrf2.

The Nrf2 protein level was unaffected by electrophiles in Keap1-deficient peritoneal macrophages owing to its constitutive nuclear accumulation

Our revelation that Nrf2 is rapidly degraded by the proteasome system, and that such degradation requires the N-terminal region of Nrf2, led us to hypothesize that Keap1 may be an important regulator of Nrf2 stability in vivo. Therefore, we explored the roles of Keap1 in the process of Nrf2 degradation using mice constitutive for keap1, heterologous for keap1, or deficient in the keap1 gene.

The juvenile death of keap1 knockout mice due to squamous cell hyperkeratinization provoked by constitutive Nrf2 activation prevented us from obtaining sufficient peritoneal macrophages (manuscript submitted for publication). To circumvent this problem, we exploited keap1–/–::nrf2+/– mice, which retain only one copy of the Nrf2 gene and survive to the adult stage. Thus, the protein level of Nrf2 in peritoneal macrophages was examined against a genetic background of heterozygous nrf2 in the presence of two, one or zero alleles of the keap1 gene.

The basal level of Nrf2 protein was markedly higher in keap1 knockout mouse peritoneal macrophages compared to those of keap1 heterozygous mutant or wild-type mice (Fig. 6A, compare lanes 1, 4 and 7). It is important to note that the constitutive expression level of Nrf2 protein negatively correlated with the number of intact keap1 alleles. The DEM mediated induction of Nrf2 was completely abolished in keap1−/− macrophages, because of the high constitutive expression of Nrf2 (lanes 2, 5 and 8). These results strongly support our contention that Keap1 sequestration to the cytosol enhances Nrf2 degradation. Based on these findings, we envisage that Keap1 maintains a minimal basal level of Nrf2 by directing the rapid degradation of Nrf2, thus allowing a sharp response to a sudden onslaught by electrophiles.

Figure 6.

Nrf2 is constitutively accumulated in the nucleus and lacks the response to electrophiles in keap1−/−nrf2+/– peritoneal macrophages. (A) Nrf2 immunoblot analysis in keap1 mutant macrophages. Peritoneal macrophages from wild-type (lanes 1–3), keap1 heterozygous (lanes 4–6), and keap1 homozygous (lanes 7–9) mice were either untreated (lanes 1, 4 and 7) or treated with 100 µm DEM (lanes 2, 5 and 8) or 10 µm MG132 (lanes 3, 6 and 9) for 3 h. Total cell extracts were subjected to immunoblot analysis using anti-Nrf2 antibody (upper panel). Anti-lamin B antibody was used as a loading control (lower panel). (B) Immunoblot analysis of nuclear extracts from keap1–/– macrophages. Macrophages from keap1−/− (lanes 3 and 4) or keap1+/+ mice (lanes 1 and 2) were treated (lanes 2 and 4) or untreated (lanes 1 and 3) with 100 µm DEM for 3 h and nuclear extracts were subjected to immunoblot analysis using anti-Nrf2 antibody (upper panel). Anti-lamin B antibody was used as a loading control (lower panel). (C) Model of Nrf2 degradation. In the cytoplasm, Nrf2 is bound by Keap1 and subjected to rapid proteasomal degradation. Keap1 enhances proteasomal Nrf2 degradation by localizing Nrf2 near the proteasome. In the nucleus, Nrf2 is relatively stabilized by the lack of Keap1.

The addition of MG132 significantly increased the amount of Nrf2 protein (Fig. 6A, lanes 3, 6 and 9), even in keap1−/− macrophages where the level actually exceeded that brought about by DEM (compare lanes 8 and 9). This observation suggests that Nrf2 degradation in keap1 (–/–) macrophages and DEM-treated macrophages is partially attenuated, with MG132 being the more specific inhibitor of the proteasomal degradation pathway. The examination of Nrf2 expression in the nucleus presented us with a clearer resolution of Nrf2 accumulation compared to that in total cells. The constitutive accumulation of Nrf2 was observed in nuclear extracts from Keap1-deficient macrophages (Fig. 6B, compare lanes 1 and 3), consistent with that observed in total cell extracts (Fig. 6A). This further suggested that Nrf2 was liberated from Keap1, hence the rapid degradation machinery, and efficiently accumulated in the nucleus. It should be noted that DEM elicited an increase in the nuclear Nrf2 level which was comparable to, but not greater than, the constitutive Nrf2 level in the Keap1 deficient animals (Fig. 6B, lanes 3 and 4), suggesting that DEM acts to increase the Nrf2 protein level by repressing Keap1 activity.

In summary, while the combination of Keap1 and proteasome rapidly turns over Nrf2 in non-induced cells, Nrf2 dissociates from Keap1 following exposure to electrophiles, such that Nrf2 enters the nucleus in a more stable form. Our study has highlighted the rapid degradation of Nrf2 occurring in the cytosol compared to the slower degradation in the nucleus.


In this study, we investigated the mechanism which underlies activation of the transcription factor Nrf2 by electrophiles leading to the expression of cellular defence genes against xenobiotic and oxidative insults. The present study demonstrated that nuclear accumulation of Nrf2 is an indispensable step for and strictly coincides with the induction of cellular defence genes. We also found that, in mouse peritoneal macrophages, the nuclear accumulation of Nrf2 requires Nrf2 protein to be newly synthesized.

Our current hypothesis for the accumulation of Nrf2 by electrophiles is depicted in Fig. 7. Whereas Nrf2 is rapidly turned over by the proteasome protein degradation system, electrophiles attenuate the degradation process by weakening the Nrf2–Keap1 interaction. Two lines of evidence support this contention. Firstly, the Nrf2-LacZ fusion gene knock-in mouse analysis provides convincing evidence that the N-terminal portion of Nrf2 is essential for the accumulation of Nrf2 in response to electrophiles in vivo. Secondly, Nrf2 is accumulated constitutively in the nuclei of keap1−/− macrophages at a level that did not increase further in response to electrophiles. We surmise that, in the absence of Keap1, Nrf2 protein escapes from the rapid cytosolic degradation process and becomes relatively stabilized in the nuclei of keap1−/− cells. Thus, electrophiles stabilize Nrf2 by repressing the effect of Keap1 on Nrf2.

Figure 7.

Hypothesis of Nrf2 activation. Keap1 and proteasome co-localize in the actin cytoskeleton. Keap1 localizes Nrf2 in the vicinity of the proteasome, thereby enhancing its degradation in the cytoplasm of quiescent cells. Sulfhydril reagents interact with the reactive cysteine(s) of Keap1, thus liberating Nrf2 from Keap1 repression. The dissociation of Nrf2 from Keap1 leads to the nuclear localization of Nrf2 with concomitant stabilization.

The results of this study support our contention that the accumulation of Nrf2 mainly occurs by exploiting a post-transcriptional mechanism. In fact, Nrf2 accumulation does not accompany an increase in Nrf2 mRNA. On the contrary, our previous analyses showed that transcription of the nrf2 gene is induced by electrophiles, both in PE keratinocytes and in vivo (Kwak et al. 2001, 2002). Thus, the accumulation of Nrf2 can be achieved via multiple pathways, including transcriptional and post-transcriptional mechanisms.

Amongst the possible post-transcriptional mechanisms, it appears that Nrf2 accumulates through the stabilization of Nrf2 protein in peritoneal macrophages. Two lines of evidence are noteworthy here. First, in addition to proteasome inhibitors, electrophiles markedly prolonged the degradation half-life of Nrf2. Second, electrophiles, such as DEM, and proteasome inhibitors, did not show any additive effect on the accumulation of Nrf2 (Fig. 3D). These results support our assertion that electrophiles act as inducers through the stabilization of Nrf2 protein by displacing it from proteasome-mediated degradation.

The down-regulation of key regulatory proteins is essential in many biological processes, including cell cycle control and signal transduction. In such processes, the down-regulation is often brought about by specific proteolysis mediated by the ubiquitin-proteasome system (Hochstrasser 1996; Varshavsky 1997). Alternatively, specific inhibitory proteins are known to bind and inhibit important regulatory proteins. One prominent example for the latter mechanism is the NF-κB/IκB system, in which IκB binds to NF-κB and sequesters it stably in the cytoplasm in the absence of inducing signals. Since Nrf2 is activated by electrophiles through its dissociation from Keap1 (Itoh et al. 1999a,b), Jaiswal and colleagues recently proposed that an analogous activation mechanism exists between Nrf2 and NF-κB (Dhakshinamoorthy & Jaiswal 2001). However, the finding that Nrf2 is rapidly degraded by proteasomes argues against such a claim. The Nrf2 activation mechanism rather shares a common feature with the hypoxia inducible transcription factors, hypoxia inducible factor-1α (HIF-1α) and HLF (HIF-1-like factor or EPAS-1). HIF-1α and HLF are rapidly degraded by proteasomes during normoxia through the hydroxylation of specific proline and/or aspartate residues, but stabilized under conditions of hypoxia (reviewed in Semenza 2001). However, the rapid turnover of Nrf2 is prevented by electrophiles through dissociation from Keap1, and this mechanism is completely different from that of the hypoxia inducible transcription factors. Thus, these results demonstrate that the activation mechanism of Nrf2 is a unique biological system.

Both keap1 gene ablation and DEM result in the freedom of Nrf2 to translocate to the nucleus and accumulate, and these events appear to play an important role in the stabilization of Nrf2. It is becoming evident that the fast and efficient proteolysis of a short-lived protein not only depends on the presence of a degradation signal, but also requires a specific localization of the substrate (Lenk & Sommer 2000). This most probably reflects the fact that components of the ubiquitin proteasome system are not evenly distributed throughout the cell, but localize specifically to certain subcellular compartments (Hirsch & Ploegh 2000). Endoplasmic reticulum-associated protein degradation (ERAD) provides a prime example of compartmentalization of the ubiquitin proteasome system (Sommer & Wolf 1997). ERAD components localize on the cytosolic surface of the ER membrane and are responsible for the degradation, not only of misfolded proteins in the ER lumen, but of cytosolic proteins such as the transcriptional repressor Matα2 (Swanson et al. 2001).

Keap1 binds to the actin cytoskeleton and localizes in the perinuclear space where the ER network is enriched (our unpublished observation). Since most chemicals are transformed to electrophiles on the cytosolic surface of the ER membrane (Guengerich 1990), it is reasonable to hypothesize that the direct modification of Keap1 by reactive electrophiles takes place near the cytosolic surface of the ER membrane. Similarly, as the 20S proteasome is located at the cytoskeleton of intermediate and actin filaments (Arcangeletti et al. 2000), an alternative possibility is that Keap1 mediated sequestration of Nrf2 to the actin cytoskeleton might be essential for rapid Nrf2 degradation (Figs 6C and 7). Since Nrf2 protein in the nucleus is still sensitive to proteasomal degradation in keap1−/− cells (Fig. 6A), the Nrf2 degradation activity, albeit weaker, must also exist in the nucleus as well as in the cytosol. In the nucleus, considering the absence of Keap1, it is unlikely that Keap1 is directly involved in the recognition of Nrf2 by proteasome. In the case of the cytoplasm, however, Keap1 is likely to determine the localization of Nrf2 to the proteasome, thereby enhancing Nrf2 degradation.

We demonstrated, by its linkage to EGFP, that the Neh2 domain is responsible for Nrf2 degradation, at least in part. The Neh2 domain can be divided into three subdomains based on interspecies amino acid conservation. These subdomains comprise an N-terminal amphipathic helix subdomain, which is conserved in Nrf1 and nematode SKN-1 (Itoh et al. 1999b), a central hydrophilic region, and a Keap1 binding ETGE motif at the C-terminus (Itoh et al. 1999b; Kobayashi et al. 2002). It is interesting to note that the hydrophobic surface of the amphipathic helix present in the Matα2 N-terminal region is the major structural feature of the Matα2 degradation signal (Johnson et al. 1998). The Ubc6p-Ubc7p ubiquitin-conjugating enzyme pair, which is a component of ERAD, mediates this type of degradation signal (Gilon et al. 2000; Sadis et al. 1995). Thus, the hydrophobic surface of the Neh2 amphipathic helix might be important for Nrf2 degradation.

Quite recently, it was reported that Nrf2 is degraded through the ubiquitin–proteasome pathway, and that phase II inducers stabilize Nrf2 against degradation (Nguyen et al. 2002; Sekhar et al. 2002; Stewart et al. 2002). The mechanisms as to how Nrf2 is stabilized, however, remain to be clarified. Furthermore, it should be tested carefully through in vivo systems whether recognition of Nrf2 by the proteasome system requires ubiquitination and whether such Nrf2 ubiquitination is regulated by electrophiles. Interestingly, Nrf2 is stabilized by cadmium without any apparent changes in ubiquitination status (Stewart et al. 2002), suggesting that any changes in Nrf2 ubiquitination may not be associated with its stabilization. On the contrary, phosphorylation of Nrf2 has been shown to be important for Nrf2 activation (Huang et al. 2002) and it was reported that Nrf2 stabilization by tert-butylhydroquinone (tBHQ) was inhibited by MAP kinase inhibitors, but not by protein kinase C (PKC) inhibitors (Nguyen et al. 2002). Thus, the MAPK pathway may be responsible for the tBHQ-mediated accumulation of Nrf2, although future studies will be absolutely essential for shedding light on the relationship between the function of Keap1 and the modification of Nrf2 by the kinase pathways or ubiquitination pathway in the Nrf2 degradation process.

The rapid cytosolic degradation of Nrf2 seems to have a physiological relevance. When the cell encounters highly toxic electrophiles, a state of emergency arises in which it is vital that potent transactivators such as Nrf2 (Katoh et al. 2001) can rapidly and effectively provide cytoprotection. One drawback is that potent transcriptional activators may place cells under the possible danger of deregulated activation. Indeed, mice with constitutively activated Nrf2 due to the absence of Keap1 die within 3 weeks of birth, most probably due to excessive Nrf2 mediated transcription (manuscript submitted for publication). The rapid and irreversible proteolysis of Nrf2 in the cytosol might provide a solid basis for the tight control of Nrf2 activity.

One question remaining is whether the stabilization per se triggers Nrf2 activation or not. Since Nrf2 over-expression can cause the constitutive expression of an ARE reporter gene in cell culture (Itoh et al. 1999a) and an endogenous Nrf2 target gene in vivo (Kobayashi et al. 2002), it is plausible that the saturation of Keap1 repression can lead to Nrf2 activation. Proteasome inhibition can result in the accumulation of Nrf2 in the nucleus (Fig. 3A) and the subsequent induction of target gene expression (our unpublished observations; Sekhar et al. 2000); thus, protein stabilization might at least enhance the nuclear translocation of Nrf2 in response to inducers (Fig. 7). Further analysis, however, is required to elucidate the degradation mechanisms of Nrf2.

Experimental procedures

Plasmid construction

Construction of the Neh2-EGFP expression plasmid was previously described (Kobayashi et al. 2002). To generate the Neh5-EGFP expression plasmid, a cDNA fragment encoding the Neh5 region (amino acids 153–227) of Nrf2 was amplified by PCR and subcloned into the KpnI and AgeI sites of pcDNA3-EGFP (Kobayashi et al. 2002).

Cell culture and treatment

Mouse peritoneal macrophages were cultured as previously described (Ishii et al. 2000). Cells were treated with 10 µm MG132 (Peptide Institute Inc.), 50 µm MG115 (Peptide Institute Inc.), 10 µm clasto-Lactacystin β-Lactone (Calbiochem), or 10 µm calpain inhibitor (Peptide Institute Inc.). The cells were also treated with 2.5 µm menadione, 100 µm DEM, 10 µm 1-chloro-2-4-dinitrobenzene (CDNB) or 10 µm sulforaphane. To remove the MG132 from the culture, the cells were washed three times with medium, with incubations of 5 min each. NIH 3T3 fibroblast cells were maintained in DMEM supplemented with 10% FBS and stable clones were selected in the presence of 0.5 mg/mL G418.


Total cell extracts or fractionated extracts were separated by SDS–polyacrylamide gel electrophoresis in the presence of 2-mercaptoethanol and electro-transferred on to Immobilon membrane (Millipore, Bedford, MA). The membrane was blocked in 3% skimmed milk and 2% goat serum overnight at 4 °C and subsequently incubated with anti-Nrf2 antibody (Ishii et al. 2000) overnight at 4 °C. As a loading control, the membrane was also hybridized with anti-β-actin antibody or anti-lamin B antibody. To detect immunoreactive proteins, we used horseradish peroxidase-conjugated anti-rabbit IgG and ECL blotting reagents (Amersham Japan, Tokyo).

RNA blot analysis

Total cellular RNA was extracted from macrophages by RNAzol™ B (Tel-Test Inc., Friendswood, TX). The RNA samples (10 µg) were electrophoresed and transferred on to Zeta-Probe GT membranes (Bio-Rad Japan, Tokyo). The membranes were probed with [32P]-labelled cDNA probes, as indicated in the figures. β-Actin cDNA was used as a positive control.

Cell fractionation

Nuclear extracts from macrophages were prepared as previously described (Ishii et al. 2000). Briefly, 7.5 × 106 peritoneal macrophages were suspended in hypotonic buffer and vortexed for 15 s and the nuclear fraction was pelleted at 7700 g for 1 min. The nuclei were resuspended in SDS sample loading buffer (without dye or 2-mercaptoethanol) and boiled for 5 min. Protein concentrations were estimated by BCA protein assay (Pierce, Rockford, IL). 0.5% Triton X soluble and insoluble fractions were prepared as previously described (Fey et al. 1984).


Livers or intestines were fixed in ice-cold 10% formalin in phosphate-buffered saline for 2 h, dehydrated with ethanol, embedded in paraffin, and cut into 3 µm sections. The sections were de-waxed and incubated for 20 min with anti-β-galactosidase antibody. They were then incubated with biotin-conjugated goat anti-rabbit IgG and avidin-DAB.


We are grateful to Drs Thomas Kensler, Mi-Kyoung Kwak, Satoru Takahashi and Kazuhiko Igarashi for critical discussions and advice. We also thank Ms N. Kaneko for technical support in histological analysis and R. Kawai for mouse breeding. This work was supported by grants from the Japan Science and Technology Corporation-ERATO project (M.Y.), Ministry of Education, Culture, Sports, Science and Technology of Japan (K.I., I.T. and M.Y.), Japanese Society for Promotion of Sciences (JSPS)-RFTF (M.Y.), and Probrain (K.I.). N.W. was a JSPS postdoctoral fellow.