Stress-sensing mechanisms and the physiological roles of the Keap1–Nrf2 system during cellular stress

Transcription factor Nrf2 (NF-E2-related factor 2) is a master regulator of cellular responses against environmental stresses. Nrf2 induces the expression of detoxification and antioxidant enzymes and suppresses the induction of pro-inflammatory cytokine genes. Keap1 (Kelch-like ECH-associated protein 1) is an adaptor subunit of Cullin 3-based E3 ubiquitin ligase. Keap1 regulates the activity of Nrf2 and acts as a sensor for oxidative and electrophilic stresses. In this review, we discuss the molecular mechanisms by which the Keap1–Nrf2 system senses and regulates the cellular response to environmental stresses. In particular, we focus on the multiple stress-sensing mechanisms of Keap1 and novel regulatory functions of Nrf2.

such that cellular Nrf2 activity is constitutively suppressed. Upon exposure to oxidative or electrophilic stresses, Keap1 loses its ability to ubiquitinate Nrf2, allowing Nrf2 to accumulate in the nucleus and activate its target genes.
Recent studies expanded our knowledge on the targets of the Keap1-Nrf2 system, and the molecular mechanisms underpinning how this system senses a variety of environmental stresses. Keap1 primarily regulates Nrf2 in the cytoplasm; however, a Keap1-independent mechanism utilizing ␤-TrCP (␤-transducin repeat-containing protein) in the nucleus also operates (11).
In addition to protecting against oxidative and xenobiotic insults, Nrf2 has also been known to attenuate inflammation (19). Nrf2 deficiency exacerbates inflammation, such as sepsis, pleurisy, and emphysema, in a variety of murine models (20 -22). In human clinical studies, the Nrf2 inducer Tecfidera (dimethyl fumarate) has been approved for the treatment of multiple sclerosis (23,24)-at least in part based on its antiinflammatory function. Thus, Nrf2 is essential in controlling inflammation.
Recent microarray and ChIP-sequence analyses have revealed that Nrf2 binds in the promoter proximal region of proinflammatory cytokine genes, including IL-6 and IL-1b, and inhibits the lipopolysaccharide-induced expression of these genes ( Fig. 1) (17). Nrf2 activation disrupts the recruitment of RNA polymerase II to IL-6 and IL-1b loci and relieves inflammatory phenotypes. This Nrf2-mediated transcriptional interference appears to be independent of the level of reactive oxygen species. Although the conventional hypothesis is that Nrf2 alleviates inflammation as a secondary consequence of its antireactive oxygen species and anti-oxidation function (25), these results suggest that Nrf2 inhibits the induction of pro-inflammatory cytokine gene transcription. Thus, Nrf2 appears to be the key regulator of two important cytoprotective pathways, anti-inflammation and anti-oxidation. The precise mechanism by which Nrf2 impedes pro-inflammatory cytokine gene transcription requires further study.
It is plausible that Nrf2 inducers would be beneficial in the treatment of a variety of diseases that are aggravated by IL-6/ IL-1 signaling (26). In fact, neutralizing antibodies and receptor antagonists that block IL-6/IL-1 signaling have been developed and approved for the treatment of rheumatoid arthritis and other inflammatory diseases (27). Nrf2 inducers are small molecules that can be administered orally with a lower burden on patients-an advantage over antibody-based drugs that require subcutaneous or intravenous administration.
The stoichiometry of Keap1 and Nrf2 within the Keap1-Nrf2 complex is 2:1, as shown by isothermal calorimetry analysis (31), nuclear magnetic resonance (NMR) titration (31), and an analytical centrifugation experiment (32). A single Nrf2 protein binds to a Keap1 homodimer using a high-affinity ETGE motif and a low-affinity DLG motif ( Fig. 2A). This two-site recognition/binding mechanism of the Nrf2 Neh2 domain and Keap1 homodimer is essential for the efficient ubiquitination of Nrf2 (31,33,34).
Endogenous Keap1 and Cul3 mainly localize in the cytoplasm (35). Oxidative and/or electrophilic stresses provoke the nuclear accumulation of Nrf2 without altering the cytoplasmic localization of Keap1 (35) or triggering dissociation of the Keap1-Nrf2 complex (36). Keap1-based E3 ubiquitin ligase activity is inhibited by oxidative and electrophilic stresses. This reduces Nrf2 ubiquitination and leads to the stabilization and nuclear accumulation of de novo synthesized Nrf2 (36). One key premise is that this stress-induced disruption of Keap1-Cul3 E3 ubiquitin ligase activity sanctions the constitutive and active synthesis of Nrf2 molecules. This pool of Nrf2 then floods the cytoplasm, and Nrf2 rapidly translocates to the nucleus.
The absolute numbers of Nrf2, Keap1, and Cul3 proteins within cells were quantitatively analyzed using five different cell lines (32). In the basal state, Nrf2 protein was maintained at a level significantly lower than the levels of Keap1 and Cul3 proteins. When challenged with the electrophilic Nrf2 inducer diethyl maleate (DEM), the abundance and localization of Keap1 and Cul3 proteins did not change, whereas Nrf2 rose to a level considerably higher than the levels of Keap1 and Cul3 proteins (32).
These results support our contention that in the basal state the Keap1-Cul3 ubiquitin ligase complex acts as a "floodgate" and degrades Nrf2 efficiently through the proteasome system (32). In response to oxidative/electrophilic stimuli (or inducers), the Keap1-Cul3 complex loses ubiquitin ligase activity (or the floodgate function), and Nrf2 accumulates in the nucleus and activates target genes (Fig. 2B).
Conflicting reports exist regarding the stoichiometry of Keap1 and Cul3. Although one study reports that one Cul3 protein binds to a homodimer of Keap1 (37), another report shows that two Cul3 proteins bind to a Keap1 homodimer (38). Recent analytical centrifugation experiments verified that the stoichiometry of the Keap1-Cul3 complex is actually two Cul3 monomers bound to one Keap1 homodimer in a 2:2 ratio (32).
The Keap1 Cys-151 residue is essential for the cellular response to a number of Nrf2-inducing electrophiles (39 -45). Cys-151 is located in the BTB domain of Keap1, which is

Figure 2. Molecular dynamics of the Nrf2-Keap1-Cul3 complex in cells.
A, two-site binding model of the Nrf2 Neh2 domain and Keap1 homodimer. Note that the ETGE and DLG motifs in the Neh2 domain are high-affinity and low-affinity motifs, respectively. B, in the basal state, a portion of the ubiquitin (Ub) ligase Keap1-Cul3 complex acts as a floodgate and degrades Nrf2 in the proteasome system. In response to oxidative and electrophilic stimuli, Nrf2 accumulates significantly in the nucleus. These stimuli do not affect the abundance and subcellular localization of Keap1 and Cul3 or the interaction of Keap1 with Cul3 and Nrf2. Pooled Nrf2 dimerizes with one of the small Maf proteins (sMaf) to form an Nrf2-sMaf heterodimer that recognizes the ARE and activates target gene expression. MINIREVIEW: Stress response by the Keap1-Nrf2 system responsible for the interaction of Keap1 with Cul3. Consistent with these observations, several reports have shown that modification of Cys-151 inhibits the Keap1-Cul3 interface and prevents ubiquitination of Nrf2 (30,41). It should be noted that there are a couple of reports that disagree with this model (46,47). For instance, the Keap1-Cul3 complex does not dissociate upon exposure to 5,6-dihydrocyclopenta-1,2-dithiole-3-thione or sulforaphane (46). Similarly, a live-cell imaging analysis showed no dissociation of the Keap1-Cul3 complex upon exposure to 1-[2-cyano-3,12-dioxooleana-1,9(11)-dien-28-oyl] (CDDO), sulforaphane, sulfoxythiocarbamate alkyne, or hydrogen peroxide (47).
In this regard, pulldown and analytical centrifugation analyses show that many of the Cys-151-targeting electrophilic inducers do not provoke dissociation of the Keap1-Cul3 complex, except in the case of CDDO-imidazole (CDDO-Im) (32). This special effect of CDDO-Im on the Keap1-Cul3 association concurs with co-crystal analysis of CDDO and the BTB domain of Keap1, which shows that Cys-151 modification by CDDO is capable of inhibiting the binding of Cul3 to Keap1 (48).
Chemical inducers produce conformational changes in Keap1, as shown by a hydrophobicity probe (49). We surmise that cysteine modification elicits structural alterations in Keap1 by affecting the complex status of Keap1 and Cul3 without causing their dissociation. This structural alteration in Keap1 prevents Nrf2 ubiquitination. It is possible that Cys-151 modification affects the angle of orientation of the Keap1-Cul3 association, thereby significantly altering the distance between ubiquitin and the target lysine residues in the Neh2 domain.

Nrf2 chemical inducers interacting with Cys-151
A variety of Nrf2 chemical inducers have been reported, and the majority are electrophilic and react with nucleophilic thiols, including cysteine sulfhydryl groups (50). Specific patterns of Keap1 cysteine modifications by individual Nrf2 chemical inducers have been identified by mass spectrometry analyses (23,42,51,52), and these modifications affect the ubiquitin E3 ligase activity of the Keap1-Cul3 complex.
The functional significance of the cysteine residues in Keap1 has been examined using site-directed mutagenesis (39,40,44,45). The results suggest that the chemicals triggering Nrf2 activation are associated with Keap1 cysteine modification and show distinct profiles. This unique use of cysteine residues as sensors has led to the "cysteine code" concept. The cysteine code supports the uniqueness of the Keap1-Nrf2 system that responds to a diverse array of chemicals and oxidative insults (18). Indeed, our recent results revealed that the three major cysteine residues Cys-151, Cys-273, and Cys-288 act individually and/or redundantly as sensors of various electrophiles (45). These studies unequivocally demonstrate that Keap1 employs multiple cysteine residues as sensors enabling Nrf2-mediated cytoprotective responses. The three major cysteine residues in Keap1 act specifically and/or collaboratively as sensors (45). Based on these analyses, Nrf2 inducers have been classified into at least five classes (Fig. 3).
The importance of Cys-151 as a sensor has been verified (39 -45). Knock-in or transgenic mice expressing a Keap1 mutant where Cys-151 has been substituted with serine are referred to as Keap1 C151S . In Keap1 C151S mouse embryonic fibroblasts (MEFs) and peritoneal macrophages, the Cys-151 residue was found to be indispensable for the accumulation of Nrf2 in response to a set of electrophilic chemicals. These Nrf2 inducers include DEM, tert-butylhydroquinone, dimethyl fumarate, nitric oxide (NO), TFM-735, and CDDO-Im (44,45,53) and are categorized into Class I as Cys-151-preferring inducers (Fig. 3).

The other Nrf2 inducers
It has been assumed that other cysteine residues, especially Cys-273 and Cys-288, might also contribute to the stress-sensor function of Keap1. It has been suggested that Cys-273 and Cys-288 residues react to 15-deoxy-⌬ 12,14 -prostaglandin J 2 (15d-PGJ 2 ) (42). However, substitution of Cys-273 and Cys-288 with serine or alanine failed to repress Nrf2 activity in both a reporter co-transfection transactivation assay (39,54) and in transgenic complementation rescue mouse experiments (40). Thus, a technical difficulty is precluding a simple validation of this notion.
Comparison of Cys-273 and Cys-288 among Kelch-like (KLHL) family members suggests that KLHL proteins, including Keap1, likely possess a similar structure contiguous with Cys-273 and Cys-288 for functioning as a ubiquitin ligase (45). In light of this knowledge, we infer that electrophilic modification of Cys-273 and Cys-288 readily disrupts the conformation of Keap1 that is critical for ubiquitin ligase activity.
To this end, we systematically introduced amino acid substitutions into Cys-273 and Cys-288, and we identified amino acids that do not affect the ability of Keap1 to repress Nrf2 accumulation (45). Keap1 mutants in which Cys-273 was replaced with hydrophobic residues (methionine or tryptophan) and Cys-288 was replaced with hydrophilic residues (glu- MINIREVIEW: Stress response by the Keap1-Nrf2 system tamate, asparagine, or arginine) (45) retained the ability to repress Nrf2 accumulation. In unstressed conditions, it seems that Cys-273 and Cys-288 are kept in hydrophobic and hydrophilic conditions, respectively. These characteristics of Cys-273 and Cys-288 might be critical for the structural integrity of Keap1 for maintaining ubiquitin ligase activity. However, conclusive insights into these structural requirements await elucidation of the structure surrounding the intervening region that incorporates Cys-273 and Cys-288.
Most Nrf2 inducers are known to interact with the cysteine residues of Keap1, thereby inactivating Keap1 E3 ligase activity. Classic inducers of Nrf2 interact with the cysteine residues of Keap1 by virtue of their electrophilic nature and can therefore inherently react with glutathione or thiol in proteins. Because an overdose of such thiol-reactive chemicals could cause electrophilic damage in cells, chemicals that directly inhibit the protein-protein interaction (PPI) of Keap1 and Nrf2 are emerging as attractive novel Nrf2 inducers (Class V in Fig. 3) (57-61).
It is interesting to note that Astex Pharmaceuticals and GlaxoSmithKline Pharmaceuticals disclosed a novel phenylpropanoic acid-based Keap1-Nrf2 PPI inhibitor through the fragment-based drug design method (59). This compound has an IC 50 (inhibitory concentration of 50%) value of 15 nM and showed a high potency in cellular and in vivo models (59). To our knowledge, this compound is one of the most potent Keap1-Nrf2-based inhibitors reported to date, and other candidates also exist.

Novel function of Nrf2
Keap1 is the critical negative regulator of Nrf2 in vivo, as widely validated through experiments using mouse genetics. Direct Keap1 knock-out mice (Keap1 Ϫ/Ϫ ) die around the time of weaning due to hyperkeratosis of the upper digestive tract, which leads to feeding problems. Nonetheless, Nrf2 accumulates in the nucleus, and cytoprotective genes are constitutively up-regulated in various tissues in these global Keap1 knock-out mice (62). Hyperkeratosis and the massive induction in cytoprotective gene expression are reversed in Keap1 Ϫ/Ϫ ::Nrf2 Ϫ/Ϫ mice (62), indicating that the lethal phenotype of Keap1 Ϫ/Ϫ mice is due to constitutive stabilization of Nrf2 (Table 1).
The Keap1 gene is deleted in squamous epithelium in Kera-tin5-Cre mice (Keap1 flox/flox ::K5-Cre) that express the Keratin5 promoter-regulated Cre recombinase in squamous epithelium (63) and results in lethality at the weaning age (64). Keap1 deficiency in the esophagus is the primary cause of this lethality due to malnutrition from obstructive lesions in the upper digestive tract.
We found that a loxP addition to the Keap1 locus generates knockdown of the Keap1 gene, which leads to constitutive accumulation of Nrf2 throughout the whole body without lethality (64). These Keap1 knockdown mice (Keap1 KD/KD or Keap1 KD/Ϫ ) prove quite useful for examining the effects of Nrf2 activation in vivo (Table 1) (64 -68). Importantly, compared with global Keap1-null mice, Nrf2 activation is weaker in Keap1 knockdown mice due to partial retention of Keap1 gene expression.
Analyses of NEKO mice led to the discovery of a novel phenotype in the kidney, which is attributable to full activation of Nrf2 by complete deletion of Keap1. NEKO mice display polyuria with low osmolality, an impaired response to dehydration, Table 1 Summary of Keap1 gene-modified mice and their phenotypes MINIREVIEW: Stress response by the Keap1-Nrf2 system and consequent renal structural damage like hydronephrosis (69). This phenotype might be caused by reduced aquaporin 2 (AQP2) protein in the kidney, because AQP2 is a well-known water channel responsible for the reabsorption of water in the kidney (70,71).
These renal phenotypes are recapitulated by renal tubularspecific Keap1 gene deletion during development, but not in adulthood (69,72), indicating that Nrf2 activation in renal tissue at an early stage is responsible for the polyuria and kidney damage observed in NEKO mice (Table 1). Thus, full Nrf2 activation during renal development leads to nephrogenic diabetes insipidus (NDI).
Vasopressin-mediated trafficking of AQP2 to the apical membrane is known to elicit urinary excretion of AQP2 (73). Compared with control mice, the AQP2 protein level in NEKO mouse urine is significantly greater, and the glycosylated form of AQP2 protein is strikingly reduced in NEKO kidneys (69). Therefore, Nrf2 seems to regulate glycosylation-related gene expression involved in AQP2 trafficking. Indeed, we found upregulated expression of Clec4d and Clec4n genes (69), which encode members of the C-type lectin family (74). This suggests that Clec4d and Clec4n enhance the transport of AQP2 to the apical membrane, thereby enhancing the urinary excretion of AQP2 (69).
Murine phenotypes of Nrf2 hyperactivation mediated by the loss of Keap1 function include esophageal hyperkeratosis and a urinary concentrating defect (62,69). This implies that Nrf2 regulates not only cytoprotective genes but also other genes involved in cell-fate determination and the maintenance of organismal homeostasis. Importantly, although elevated Nrf2 levels have been identified in various types of human cancers (75,76), spontaneous tumorigenesis has not been observed in NEKO mice (69), indicating that mere hyperactivation of Nrf2 is not sufficient to elicit cancer per se.
An intriguing discussion here is that dioxin has been known to induce hydronephrosis in childhood without anatomical obstruction of the ureter (77). Dioxin-induced hydronephrosis is elicited by an elevated production of prostaglandin E 2 (PGE 2 ), which is a potent electrophilic inducer of Nrf2 (78). Of note, the development of hydronephrosis is induced by exposure to dioxin in the neonatal stage but not during adulthood (79). This developmental stage-specific effect of dioxin (or window to dioxin) is akin to the observation that inducible and renal tubular-specific Keap1 knock-out mice develop hydronephrosis in childhood but not in adulthood (69). These similarities imply that dioxin-induced hydronephrosis could be explained by the constitutive activation of Nrf2 by chronic elevation of PGE 2 in the developing kidney. This advocates that chronic exposure to Nrf2 activators, such as environmental pollutants, prostaglandins, and oxidative stress, during kidney development may cause NDI and hydronephrosis.
In light of this observation, the timing at which Nrf2-activation treatment is given needs to be carefully considered. The administration of Nrf2-activating chemicals during adulthood might not cause strong adverse effects in the kidney (69). Even if treatment includes chronic administration of potent Nrf2 inducers, the body does not seem to be impaired. Indeed, many previous studies revealed that pharmacological Nrf2 activation has a protective effect on a variety of stress-mediated diseases, including kidney injury (80,81).
In addition to the renal phenotype, mild anemia, growth retardation, and poor survival have been observed in NEKO mice (69). The anemia, growth retardation, and poor survival cannot be recapitulated in renal tubular-specific Keap1-deficient mice (69), indicating that these phenotypes are independent of Nrf2 activation in renal tubules. Loss of Keap1 in hematopoietic cells suppresses differentiation toward the erythroid lineage (82), so the anemia in NEKO mice is likely due to a lack of Keap1 in hematopoietic cells. Growth retardation in NEKO mice might be caused by Nrf2 activation in skeletal muscle and adipose tissue, because Keap1 deletion in skeletal muscle reduces body weight (83), and loss of Keap1 represses the differentiation of adipose cells (84).
Because NEKO mice likely have other phenotypes that have not yet been studied, further analysis of NEKO mice will provide new insights for a better understanding of the physiological functions of Nrf2.

Future perspectives
Accumulating lines of evidence have proven that Keap1 senses a wide range of Nrf2-inducing chemicals (44,45). Nonetheless, how Keap1 senses a variety of chemicals utilizing multiple and distinct sets of cysteine residues remains to be clarified. Additional studies on the sensing mechanisms will advance our understanding of the environmental stress response and hopefully lead to the eradication of environmental stressrelated disorders.
Identification of the unexpected adverse effects of Nrf2 hyperactivation, such as NDI (69) and reduced placental angiogenesis in preeclampsia (85), are important. We believe that wide-ranging clinical applications of Nrf2 activators will come of age. Therefore, the adverse effects of Nrf2 activators need to be rigorously explored to use Nrf2 activators in a safe and appropriate manner. We believe that the unfavorable effects of Nrf2 activation can be overcome with accurate and detailed knowledge of the Keap1-Nrf2 system.