Correction

MEDICAL SCIENCES Correction for ‘‘Cancer related mutations in NRF2 impair its 13573; first published August 29, 2008; 10.1073/ recognition by Keap1-Cul3 E3 ligase and promote malignancy,’’ pnas.0806268105). by Tatsuhiro Shibata, Tsutomu Ohta, Kit I. Tong, Akiko The authors note that an incorrect amino acid abbreviation Kokubu, Reiko Odogawa, Koji Tsuta, Hisao Asamura, Masayuki was used in Table 1 and in Figs. 1 and 3. ‘‘T80K’’ should read Yamamoto, and Setsuo Hirohashi, which appeared in issue 36, ‘‘T80R’’. This error does not affect the conclusions of the article. September 9, 2008, of Proc Natl Acad Sci USA (105:13568– The corrected table, figures, and legends appear below.

Table 1. NRF2 mutation in cancer cell lines and cancer patients Sample name Nucleotide mutation Amino acid change Tumor histology Gender Smoking status Age

LK-1 239 C–G (heterozygous) T80R SQ Male Current 68 LK-8 101 G–A (heterozygous) R34Q SQ Male Current 57 LK-9 72 G–C (heterozygous) W24C SQ Female Former 73 LK-17 235 G–A (heterozygous) E79K SQ Male Former 73 LK-29 245 A–G (heterozygous) E82G SQ Male Former 70 LK-31 235 G–A (heterozygous) E79K SQ Male Current 71 LK-39 95 T–G (heterozygous) V32T SQ Male Former 74 LK-40 88 C–T (homozygous) L30F SQ Male Current 70 LK-47 83 T–C (heterozygous) I28T SQ Female Never 59 LK-72 235 G–C (heterozygous) E79Q AD Male Former 75 LK-102 72 G–C (heterozygous) W24C LCNEC Male Current 77 HN-3 225 A–C (heterozygous) Q75H HN Female Former 51 HN-5 239 C–T (heterozygous) T80I HN Male Never 58 HN-10 86 A–G (heterozygous) D29G HN Male Current 74 EBC-1 230 A–T (heterozygous) D77V NSCLC NA NA NA LK2 235 G–A (homozygous) E79K NSCLC NA NA NA HO1-U1 246 A–T (heterozygous) E82D OC NA NA NA

Fig. 1. NRF2 mutations in human cancer. (A) Functional domains of Nrf2 showing distribution and types of NRF2 mutations. SQ, lung squamous cell carci- noma; AD, adenocarcinoma; LCNEC, large cell neuroendocrine carcinoma; HN, head and neck cancer; cell, lung cancer cell lines. (B) Representative DNA chromatograph of NRF2 mutations in human cancer cell lines with corresponding normal sequences in the bottom panel. (C) Copy number alter- ation map [shown by log2 ratio (tumor/ normal)] of 2 in the tumors with mutated NRF2 and having lost the wild-type allele, detected by array-based comparative genomic hybridization. The arrow indicates the chromosomal location of NRF2.(D) Kaplan–Meier plot showing disease-free survival (DFS) of squamous cell lung carcinoma patients segregated ac- cording to the presence or absence of NRF2 mutations.

10392–10393 ͉ PNAS ͉ June 23, 2009 ͉ vol. 106 ͉ no. 25 www.pnas.org Downloaded by guest on September 24, 2021 CORRECTION

Fig. 3. NRF2 mutations disturb proper Nrf2-Keap1 binding, inhibit Keap1-mediated degradation, promote transcriptional activity, and enhance nuclear localization of Nrf2 in vivo.(A) Expression level of the transfected Flag-tagged Keap1 and Myc-tagged wild-type or mutant Nrf2 (E79K, T80R, and L30F) in 293T cells. (B) Keap1-Nrf2 association was evaluated by immunoprecipitation of transfected 293T cells using anti-Flag antibody and immunoblotted by anti-Myc antibody. (C) Ubiquitination efficiency of wild-type or mutant Myc-tagged Nrf2 in vivo.(D) Immunoblot (left) showing degradation rate of Nrf2 protein from 15 to 60 min after cycloheximide chase. Intensities (%) relative to time 0 are plotted (right). Data represent mean Ϯ SD of three independent runs. Dotted line indicates 50% expression of control. (E) Transactivation activities of Nrf2 mutants. Data represent mean Ϯ SD of three independent runs (*, P Ͻ 0.001). (F) Rescue analysis in KEAP1-orNRF2-mutated cancer cells. H1650, A549, LK2, and EBC-1 cells bearing both wild-type KEAP1 and NRF2 or mutations in one of the are indicated. Relative expressions of Nrf2 target genes peroxiredoxin1 (PRDX1), multidrug resistance protein 3 (MRP3), and NAD(P)H quinine oxidoreductase-1 (NQO1)to␤-ACTIN with or without cotransfected KEAP1 plasmid are shown. (G) Nuclear accumulation of mutant Nrf2 when coexpressed with Keap1 (left). Ratio of cytoplasmic/nuclear localization of Nrf2 proteins was shown as bar chart (right).

www.pnas.org/cgi/doi/10.1073/pnas.0905305106

PNAS ͉ June 23, 2009 ͉ vol. 106 ͉ no. 25 ͉ 10393 Downloaded by guest on September 24, 2021 Cancer related mutations in NRF2 impair its recognition by Keap1-Cul3 E3 ligase and promote malignancy

Tatsuhiro Shibata*†, Tsutomu Ohta‡, Kit I. Tong§, Akiko Kokubu*, Reiko Odogawa‡, Koji Tsuta¶, Hisao Asamuraʈ, Masayuki Yamamoto†§**, and Setsuo Hirohashi*

*Cancer Genomics Project, Pathology Division, ‡Center for Medical Genomics, National Cancer Center Research Institute, 5-1-1, Tsukiji, Chuo-ku, Tokyo, 104-0045, Japan; §Graduate School of Comprehensive Human Sciences, Center for Tsukuba Advanced Research Alliance, and Environmental Response Project, Exploratory Research for Advanced Technology-Japan Science and Technology Corporation, University of Tsukuba, 1-1-1, Tennoudai, Tsukuba, 305-8577, Japan; ¶Clinical Laboratory Division, ʈThoracic Surgery Division, National Cancer Center Hospital, 5-1-1, Tsukiji, Chuo-ku, Tokyo, 104-0045, Japan; and **Department of Medical Biochemistry, Tohoku University Graduate School of Medicine, 2-1 Seiryo-cho, Aoba-ku, Sendai, 980-8575, Japan

Communicated by Paul Talalay, Johns Hopkins University School of Medicine, Baltimore, MD, June 29, 2008 (received for review January 15, 2008) The nuclear factor E2-related factor 2 (Nrf2) is a master transcrip- the nucleus where it stimulates activation (7–9). Hence, tional activator of genes encoding numerous cytoprotective en- experimental strategies to target Keap1, including siRNA knock- zymes that are induced in response to environmental and endog- down and small molecule modulation, have also been used as a enously derived oxidative/electrophilic agents. Under normal, means to activate Nrf2 (10, 11). nonstressed circumstances, low cellular concentrations of Nrf2 are Enzymes that are encoded by Nrf2 target genes, such as maintained by proteasomal degradation through a Keap1-Cul3- NAD(P)H quinine oxidoreductase-1 (NQO1) and peroxiredox- Roc1-dependent mechanism. A model for Nrf2 activation has been ins (PRDX), are cytoprotective under normal conditions. Par- proposed in which two amino-terminal motifs, DLG and ETGE, adoxically, these enzymes are increased in many cancers, thereby promote efficient ubiquitination and rapid turnover; known as the affording these cells with protective capabilities to environmen- two-site substrate recognition/hinge and latch model. Here, we tal stresses, which is a hallmark of malignancy (12, 13). This show that in human cancer, somatic mutations occur in the coding observation implicates aberrant Nrf2 activation in the process of region of NRF2, especially among patients with a history of human cancer development. In this regard, recent studies have smoking or suffering from squamous cell carcinoma; in the latter revealed genetic alterations in KEAP1 that can lead to uncon- case, this leads to poor prognosis. These mutations specifically trolled Nrf2 activation in cancers of the lung (14–16). Although alter amino acids in the DLG or ETGE motifs, resulting in aberrant polymorphisms have been identified in the promoter region of cellular accumulation of Nrf2. Mutant Nrf2 cells display constitutive the Nrf2 gene (17, 18), which are thought to predispose indi- induction of cytoprotective enzymes and drug efflux pumps, which viduals to lung injury and oxidative related diseases, no conclu- are insensitive to Keap1-mediated regulation. Suppression of Nrf2 sive findings on the occurrence of genetic variations in the coding protein levels by siRNA knockdown sensitized cancer cells to region have yet been published. Given that the intermolecular and chemotherapeutic reagents. Our results association between Nrf2 and Keap1 is mandatory for Nrf2 strongly support the contention that constitutive Nrf2 activation degradation and repression, we hypothesized that subtle amino affords cancer cells with undue protection from their inherently acid changes in the coding region of Nrf2 might also affect stressed microenvironment and anti-cancer treatments. Hence, proper Keap1-Nrf2 interaction and hence Nrf2 turnover in inactivation of the Nrf2 pathway may represent a therapeutic cancer. strategy to reinforce current treatments for malignancy. Congru- ously, the present study also provides in vivo validation of the Results two-site substrate recognition model for Nrf2 activation by the NRF2 Mutation in Human Cancer. We sequenced the entire coding Keap1-Cul3-based E3 ligase. region of NRF2 in 85 cancer cell lines and found mutations in two lung cancer cell lines (LK2 and EBC1) and one oral cancer cell cancer cell microenvironment ͉ multidrug resistant-associated protein ͉ line (HO1-U1) [Fig. 1 A and B and supporting information (SI) oxidative stress ͉ somatic mutation ͉ -proteasome system Table S1]. We also examined 103 primary lung cancers of various histological subtypes and 12 primary head and neck (HN) xidative and mutagenic damage to DNA, proteins, and cancers, and observed mutations in 11 lung tumors (10.7%) and Olipids occurs continuously as a result of exposure to envi- 3 HN tumors (25%). All mutations (14/115, 12.2%) in primary ronmental pollutants, ionizing radiation, endogenous reactive tumors were missense amino acid substitutions and determined oxygen species (ROS), electrophilic intermediates of xenobiotic to be of somatic origin (Fig. 1A, Fig. S1 and Table S1), whereas metabolism, and inflammation. Damage from these exogenous no synonymous somatic alteration was detected. These muta- and endogenous insults, if not repaired, can contribute to various tions are conspicuously high in patients with a history of smoking pathologies including cancer (1, 2). The transcription factor Nrf2 (12/14) and occur frequently in squamous cell carcinoma (12/60, mediates gene induction of protective phase-2 detoxifying en- zymes and its activation by thiol-active compounds, found in a variety of natural foods and dietary supplements, has been Author contributions: T.S., T.O., K.I.T., M.Y., and S.H. designed research; T.S., T.O., K.I.T., A.K., and R.O. performed research; T.S., T.O., K.I.T., K.T., and H.A. contributed new advocated as a means to prevent cancer (3, 4). By contrast, reagents/analytic tools; T.S., T.O., K.I.T., A.K., R.O., M.Y., and S.H. analyzed data; and T.S., Nrf2-deficient mice are highly susceptible to chemically induced T.O., K.I.T., A.K., R.O., K.T., H.A., M.Y., and S.H. wrote the paper. carcinogenesis of multiple organs (5, 6). Under normal, non- The authors declare no conflict of interest. stressed circumstances, low cellular concentrations of Nrf2 are †To whom correspondence may be addressed. E-mail: [email protected] or masi@ maintained by proteasomal degradation, through a Keap1-Cullin mail.tains.tohoku.ac.jp. 3-Roc1-dependent mechanism in which Keap1 serves as the This article contains supporting information online at www.pnas.org/cgi/content/full/ substrate adaptor subunit in the E3 holoenzyme. Activation of 0806268105/DCSupplemental. Nrf2 allows it to escape and to rapidly accumulate in © 2008 by The National Academy of Sciences of the USA

13568–13573 ͉ PNAS ͉ September 9, 2008 ͉ vol. 105 ͉ no. 36 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0806268105 Fig. 1. NRF2 mutations in human cancer. (A) Functional domains of Nrf2 protein showing distribution and types of NRF2 mutations. SQ, lung squamous cell carcinoma; AD, adenocarcinoma; LCNEC, large cell neuroendocrine carcinoma; HN, head and neck cancer; cell, lung cancer cell lines. (B) Representative DNA chromatograph of NRF2 mutations in human cancer cell lines with corresponding normal sequences in the bottom panel. (C) Copy number alteration map [shown by log2 ratio (tumor/normal)] of chromosome 2 in the tumors with mutated NRF2 and having lost the wild-type allele, detected by array-based comparative genomic hybridization. The arrow indicates the chromosomal location of NRF2.(D) Kaplan–Meier plot showing disease-free survival (DFS) of squamous cell lung carcinoma patients segregated according to the presence or absence of NRF2 mutations.

20%). In lung cancer, no tumor harbored both NRF2 and KEAP1 NRF2 Mutations Impair the Two-site Substrate Recognition Mecha- mutations concurrently, and these alterations occur in tumors nism of Keap1. The double glycine repeat and the C-terminal without EGFR mutation (Fig. 1A, Table 1, and Table S2). A region of Keap1 (hereafter refer as to Keap1-DC) recognize homozygous Nrf2 mutant allele was observed in LK2 cells and both the DLG and ETGE motifs of Nrf2 (20, 21). To evaluate one primary tumor (LK-40) and an increase in the gene copy the effect that mutations in the Nrf2 protein may have on its number of NRF2 (3 copies) was also seen in LK2 cells (Fig. 1C). physical association with Keap1, we used isothermal calorimetry Prognostic analysis of lung squamous cell carcinomas carrying a (ITC) to probe the interaction between purified recombinant mutation in NRF2 revealed a significant correlation with poor Neh2 protein mutants (either ⌬1–33Neh2/ETGE mutants or full survival (P ϭ 0.0247 by Log-rank analysis; Fig. 1D). Strikingly, length Neh2 DLG mutants), and the Keap1-DC or the full- the mutations were clustered in exon 2 and coded for amino acid length Keap1 protein. All of the mutations that occur in the changes in either the DLG or ETGE motifs of the regulatory ETGE motif (77DxETGE82), such as D77V, E79Q or T80K (Fig. Neh2 (Nrf2-ECH homology 2) domain (Fig. 1A) (19). 2 A–C), compromised the association with Keap1-DC. In par-

Table 1. NRF2 mutation in cancer cell lines and cancer patients Sample name Nucleotide mutation Amino acid change Tumor histology Gender Smoking status Age

LK-1 239 C–G (heterozygous) T80K SQ Male Current 68 LK-8 101 G–A (heterozygous) R34Q SQ Male Current 57 LK-9 72 G–C (heterozygous) W24C SQ Female Former 73 LK-17 235 G–A (heterozygous) E79K SQ Male Former 73 LK-29 245 A–G (heterozygous) E82G SQ Male Former 70 LK-31 235 G–A (heterozygous) E79K SQ Male Current 71 LK-39 95 T–G (heterozygous) V32T SQ Male Former 74 LK-40 88 C–T (homozygous) L30F SQ Male Current 70 LK-47 83 T–C (heterozygous) I28T SQ Female Never 59 LK-72 235 G–C (heterozygous) E79Q AD Male Former 75 LK-102 72 G–C (heterozygous) W24C LCNEC Male Current 77 HN-3 225 A-C (heterozygous) Q75H HN Female Former 51 HN-5 239 C-T (heterozygous) T80I HN Male Never 58 HN-10 86 A-G (heterozygous) D29G HN Male Current 74 EBC-1 230 A–T (heterozygous) D77V NSCLC NA NA NA LK2 235 G–A (homozygous) E79K NSCLC NA NA NA

HO1-U1 246 A–T (heterozygous) E82D OC NA NA NA MEDICAL SCIENCES

Shibata et al. PNAS ͉ September 9, 2008 ͉ vol. 105 ͉ no. 36 ͉ 13569 Fig. 2. NRF2 mutations impair the two-site substrate recognition mechanism of Keap1. ITC titration profiles of Keap1-DC with Neh2[⌬1–33,D77V] (A), Neh2[⌬1–33,E79Q] (B), Neh2[⌬1–33,T80K] (C), and of Keap1 with Neh2 (D), Neh2[W24C] (E), Neh2[L30F] (F), raw thermograms (upper graph) and fitted binding isotherms (lower graph). Stoichiometry (n) and association constant (Ka) are as indicated. (G–I), binding site of the human Keap1-DC and wild-type ETGE peptide structure (G) and modeled structures of Keap1-DC with E79Q (H) or E82D (I) mutant ETGE peptide. Stick representation showing side chains of interacting residues from Keap1 (aqua) and ETGE peptide (pink). Dotted lines indicate predicted hydrogen bonds. Figures were drawn with PyMOL (www.pymol.org). ticular, the T80K mutation displayed no observable affinity. The the ability to associate with Keap1 to a similar degree as wild wild-type Neh2 protein exhibited a biphasic-binding curve with type, presumably because of the presence of an intact ETGE Keap1 (Fig. 2D), which has been shown to be a dimer in solution motif, whereas the weaker binding DLG motif alone could not (20), indicating that two binding sites (DLG and ETGE) in the sufficiently support a stable Nrf2/Keap1 interaction. Neh2 domain are capable of interacting with Keap1. However, Because Keap1 plays a central role in ubiquitination and mutants bearing the missense alterations W24C or L30F within proteasomal degradation of Nrf2, we proceeded to analyze the the DLG motif (23LxxQDxDLG31) regress to monophasic bind- ubiquitination status of mutant Nrf2. Wild-type and mutant ing patterns (Fig. 2 E and F). The residual binding affinity most Nrf2 were coexpressed with Keap1 and hemagglutinin (HA)- likely is contributed by interaction with the ETGE motif alone. tagged ubiquitin, and the levels of Nrf2 ubiquitination were These data unequivocally demonstrate that Nrf2 mutations determined after treatment with MG132, a proteasomal inhib- occurring in either the DLG or ETGE motif can affect binding itor. Wild-type Nrf2 was efficiently polyubiquitinated whereas to the Keap1 dimer. mutant Nrf2 proteins were only weakly polyubiquitinated (Fig. Using the x-ray crystal structure of human Keap1-DC in 3C). In particular, the L30F mutant was strongly resistant to association with the ETGE peptide as template (PDB code ubiquitination. We also investigated whether mutant Nrf2 pro- 2FLU) (22) and by means of the MODELLER 9v1 program (23), teins are more stable than wild-type Nrf2. Wild-type and mutant we modeled some of the amino acid substitutions found in cancer Nrf2 were expressed in 293T cells and treated with cyclohexi- patients or cell lines to access their impact on the structure- mide to prevent new protein synthesis. Wild-type Nrf2 protein

1 ϭ function relationship of Keap1 with Nrf2. The model detailing decreased rapidly (t ͞2 16 min) as reported (7). By contrast, the structure of the E79Q mutation (Figs. 1A and 2H) suggests mutant Nrf2 proteins were degraded more slowly, having half-

1 ϭ that compared with wild-type protein, there is a loss of electro- lives of approximately twice that of wild type (t ͞2 28–35 min; static interaction with R483 and S508 of Keap1 (Fig. 2G), which Fig. 3D). likely accounts for the 1,000-fold reduction in binding affinity with Keap1 seen earlier (Fig. 2B). Similarly, the E82D mutation NRF2 Mutations Promote Transcriptional Activity and Nuclear Local- (Fig. 1A and Fig. 2I), despite both being acidic residues, seems ization of Nrf2. We next determined the transcriptional activation to eliminate the contact with S363, R380, and S508 of Keap1, ability of mutant Nrf2 by analyzing luciferase activity from a possibly because of the shorter side chain of Asp. reporter plasmid carrying a promoter region - responsive element (ARE), the canonical Nrf2 recognition NRF2 Mutations Disturb Proper Nrf2-Keap1 Binding and Inhibit Keap1- motif. Mutant Nrf2 proteins were significantly more active (P Ͻ Mediated Degradation of Nrf2 In Vivo. We next examined the 0.001 by unpaired t test) than wild-type Nrf2 (Fig. 3E). Coex- interaction between mutant Nrf2 and Keap1 in vivo. Three of the pression of Keap1 markedly decreased luciferase activity in- Nrf2 mutations that were detected in cancer biopsies were duced by wild-type Nrf2 whereas the activity of mutant Nrf2 selected (E79K, T80K, and L30F) and the proteins expressed in proteins was resistant to Keap1 inhibition (Fig. 3E). We also 293T cells, together with Keap1. All three mutant proteins tested this reporter construct in cancer cell lines that carry showed enhanced doublet bands (Fig. 3A, lanes 4–6). Immuno- genetic alterations in the Nrf2-Keap1 pathway (A549, LK2, and precipitation analysis revealed that the E79K and T80K muta- EBC1) and found that endogenous Nrf2 mutants also exhibited tions both had substantially reduced ability to interact with transactivation activity that was resistant to suppression by Keap1 (Fig. 3B). Protein harboring the L30F mutation retained Keap1 coexpression, especially in LK2 cells, which carry a

13570 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0806268105 Shibata et al. Fig. 3. NRF2 mutations disturb proper Nrf2-Keap1 binding, inhibit Keap1-mediated degradation, promote transcriptional activity, and enhance nuclear localization of Nrf2 in vivo.(A) Expression level of the transfected Flag-tagged Keap1 and Myc-tagged wild-type or mutant Nrf2 (E79K, T80K, and L30F) in 293T cells. (B) Keap1-Nrf2 association was evaluated by immunoprecipitation of transfected 293T cells using anti-Flag antibody and immunoblotted by anti-Myc antibody. (C) Ubiquitination efficiency of wild-type or mutant Myc-tagged Nrf2 in vivo.(D) Immunoblot (left) showing degradation rate of Nrf2 protein from 15 to 60 min after cycloheximide chase. Intensities (%) relative to time 0 are plotted (right). Data represent mean Ϯ SD of three independent runs. Dotted line indicates 50% expression of control. (E) Transactivation activities of Nrf2 mutants. Data represent mean Ϯ SD. of three independent runs (*, P Ͻ 0.001). (F) Rescue analysis in KEAP1-orNRF2-mutated cancer cells. H1650, A549, LK2, and EBC-1 cells bearing both wild-type KEAP1 and NRF2 or mutations in one of the genes are indicated. Relative expressions of Nrf2 target genes peroxiredoxin1 (PRDX1), multidrug resistance protein 3 (MRP3), and NAD(P)H quinine oxidoreductase-1 (NQO1)to␤-ACTIN with or without cotransfected KEAP1 plasmid are shown. (G) Nuclear accumulation of mutant Nrf2 proteins when coexpressed with Keap1 (left). Ratio of cytoplasmic/nuclear localization of Nrf2 proteins was shown as bar chart (right). homozygous L79K mutation (Fig. S2). We further analyzed the cancer cells frequently encounter various oxidative stresses in endogenous expression of various Nrf2 target genes (NQO1, their microenvironment (24), we examined whether activating PRDX1 and MRP3) (15). The expression of these target genes mutations to Nrf2 affords these cells with resistance to oxidative was increased in cells with either altered KEAP1 (A549) or NRF2 stress. NRF2- or KEAP1-mutated cancer cells exhibited resis- (LK2 and EBC1), but remained at low basal levels in a cell line tance to hydrogen peroxide treatment, and Nrf2 down- Ϸ that had functional Nrf2 regulation (H1650). Introduction of regulation in these cells significantly reduced cell viability ( 1.6- Keap1 cDNA decreased expression of these genes in KEAP1 fold decrease in both cells) to a level similar to what was observed mutated A549 cells but failed to accomplish a similar down- in Nrf2-pathway intact cancer cells (H1650) (Fig. 4B). Nrf2 downstream targets include enzymes for metabolizing (such as regulation in cancer cells with mutated NRF2 (Fig. 3F). Immu- PRDX1) and exporting (MRP3) anti-cancer agents. Therefore, nofluorescent analysis also revealed that mutant Nrf2 proteins we also tested the effect of Nrf2-siRNA on drug sensitivity. We preferentially accumulated in the nucleus compared with wild chose to examine the cytotoxic effect of cisplatin, a drug that is type when coexpressed with Keap1 (Fig. 3G). routinely used for treating lung cancers. Nrf2 down-regulation, in cancer cells harboring genetic alterations in the Nrf2-Keap1 Down-regulation of NRF2 Restores Sensitivity to Oxidative Stress and pathway, significantly increased sensitivity to this compound Chemotherapeutic Agent. Finally, we examined the effects of (Fig. 4C). down-regulating Nrf2 in cancer cell lines where the Nrf2- pathway had become deregulated. Suppression of Nrf2 protein Discussion production by short interfering RNA (siRNA) reduced the Nrf2 degradation is governed by the Cullin3-Keap1 system, expression of Nrf2-target genes (Fig. 4A). Because invasive which is unique in terms of the adaptor-substrate recognition MEDICAL SCIENCES

Shibata et al. PNAS ͉ September 9, 2008 ͉ vol. 105 ͉ no. 36 ͉ 13571 Keap1-DC-DLG peptide and Keap1-DC-ETGE-peptide com- plexes have depicted that the side chains of Q26, D27, and D29 of the DLG motif (23LxxQDxDLG31) and E79 and E82 of the ETGE motif (77DxETGE82) have direct interactions with the Keap1-DC. Whereas T80 has both inter- and intramolecular interactions, Q75 has intramolecular contact (14, 21, 22). Can- cer-related mutations within the ETGE motif are primarily substitutions of these essential residues for Keap1 recognition whereas mutations in the DLG motif are either changes in the conserved (D29 and L30) or nonconserved residues (W24, I28, V32, and R34). These findings suggest that the weaker Keap1- binding DLG region is more vulnerable to structural changes and fits into its anticipated role of the ‘‘latch’’ for the Nrf2-repression gate in the stress response. Collectively the NRF2 somatic mutations found in human cancers indubitably validate the necessity for simultaneous recognition of these two N-terminal degrons in Nrf2 for its timely proteolysis and regulation in vivo. The correlation of mutation in the ubiquitin-proteasome pathway with cancer growth has long been recognized (27, 28). Altered degradation pattern of transcription factors and tumor suppressor proteins are found in cancers of different pathologies and origin, where oncogenic changes often arise from mutations in the subunits of E3 ligases. For example, von Hippel-Lindau tumor suppressor protein (pVHL) and BRAC1, together with BARD1 (BRAC1-associated RING domain 1), are components of E3 ubiquitin ligases (29–31). VHL ubiquitin ligase complex facilitates protein turnover of HIF␣ for angiogenesis or hypoxic response regulation, whereas BRAC1/BARD1 system targets a variety of substrates that involve in regulation of transcription, chromatin remodeling, DNA repair, and centrosome dynamics (32, 33). Mutations of pVHL or BRAC1 predispose tumorigen- esis in patients probably by either enhancing adaptive capability of cancer cells to their microenvironment or by jeopardizing proper transcriptional regulations, DNA repair response, and genome stability, respectively (27, 28, 33). Similarly, recent studies have shown different somatic mutations of Keap1 (14– 16), the substrate binding module of a Cullin3-based E3 ligase, in lung cancer patients that led to aberrant expression of ARE Fig. 4. Down-regulation of NRF2 restores sensitivity to oxidative stress and genes. Cancer-related somatic mutations in the degradation chemotherapeutic agent. (A) Down-regulation of Nrf2 expression by siRNA. signals of Nrf2, presented in this study, for E3 ligase recognition Nonsilencing control or NRF2 siRNA was transfected into cancer cells with again reinforce the importance of functional ubiquitin- either KEAP1 (A549) or NRF2 (EBC-1) mutation. Cell lysates were electropho- proteasome system. ␤ resed and immunoblotted with anti-Nrf2 or anti- -actin (loading control) Additionally our results demonstrated that gain-of-function antibody (top). Relative expressions of Nrf2 target genes (PRDX1, MRP3, and NQO1)to␤-ACTIN are shown in the histogram (bottom). (B) Cell viability of mutations of NRF2 were observed in approximately 11% of lung siRNA treated A549, EBC-1, and H1650 cells exposed to oxidative stress. Data cancer. Together with previous reports (14–16), constitutive present percentage of cell viability after a 5-h treatment with H2O2 relative to activation of Nrf2 (caused by KEAP1 or NRF2 mutations) occurs nontreated control (mean Ϯ SD of three independent runs) (C) A549 and EBC-1 in Ͼ20% of lung cancer cases. Our analysis concurrently re- cells were treated with control or NRF2 siRNA and various concentrations of vealed frequent NRF2 mutation in head and neck cancers, cisplatin. Data present percentage of cell viability after 48 h relative to vehicle suggesting perturbations of the Nrf2 pathway in a broad range (DMSO)-treated control (mean Ϯ SD of three independent runs). (*, P Ͻ 0.05; of cancers, particularly those arising in tissues exposed to oxygen. **, P Ͻ 0.01; ***, P Ͻ 0.001) Intriguingly KEAP1-deficient mice, which show constitutive activation of Nrf2 in the whole body, develop postnatal hyper- keratosis without accompanying by increased cell proliferation whereby the two Neh2 recognition motifs (DLG and ETGE) of in skin and upper digestive tract (34). Thus, preferential NRF2 a Nrf2 monomer bind to a Keap1 homodimer with different mutations among patients diagnosed with squamous cell carci- affinities; known as the two-site substrate recognition/hinge and noma and a history of smoking could be explained by stress- latch model (Fig. S3). Unlike some other substrates of protea- ⌲ related Nrf2-dependent intervention to squamous cell lineage somal degradation system, such as inhibitory kappa B (I B) in differentiation. However, malignant development induced by inflammatory response and hypoxia-inducible transcription fac- deviated Nrf2-mediated profile appears to be tor (HIF), which have to be post-translationally modified before more complicated because a number of genes involved in cell recruiting to their respective E3 ligase complexes (25), Nrf2 is proliferation were found to be transcriptionally modulated by being constitutively recognized by Keap1 through negative- Nrf2 through glutathione as the effector (35) and the growth rate charged residues of two conserved degradation signal sequences of immortalized Keap1 null MEF cells also increased on culture in the substrate. Therefore, the structural integrity of both the dishes (16). Nevertheless, the existence of two tumors with both Hinge (ETGE) and Latch (DLG) within the Neh2 domain is NRF2 alleles mutated and the association of poor prognosis crucial for maintaining contemporaneous two-site binding for suggest a direct role of Nrf2 mutation in tumor progression. In proper Nrf2 turnover and the narrow window of Nrf2-ARE terms of stress cellular defense, temporary induction of cyto- mediated gene expression (21, 26). X-ray crystal structures of the protective enzymes by Nrf2 activation is beneficial to the repair

13572 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0806268105 Shibata et al. of oxidative damage and cancer prevention before the accumu- Isothermal Calorimetry. ITC titration experiments were performed at 25°C on lation of any genetic lesions at the early stage of carcinogenesis a VP-ITC system (MicroCal Inc.). All proteins were prepared as described in ref. (36, 37). Although constitutive expression in a tumorigenic 21. All runs were performed in triplicate. Binding data were analyzed using the computer program Origin version 5.0, supplied by MicroCal Inc. Data situation could be devastating because it could provide a survival processing was carried out as reported (21). advantage to invasive and metastatic cancer cells, such as the adaptation to microenvironment and the evolvement of che- Molecular and Cell Biology Analysis. Anti-Myc (Cell signaling), anti-Flag moresistance in cancer cells that are known to occur in tumor (Sigma), and anti-HA (MBL) antibodies, and MG132 (Calbiochem), cyclohexi- hypoxia (38, 39). In addition, cross-talk between tumor hypoxia mide (Nacalai), and cisplatin (WAKO) were purchased, whereas anti-Nrf2 and induction of Nrf2 has also been suggested recently (40). It polyclonal antibody was produced by immunization with Nrf2 peptide. Im- is also noteworthy that inhibition of the Nrf2 pathway in these munoprecipitation, immunoblotting, and immunofluorescent analyses were performed as described (7, 8) with slight modifications. cancer cells enhanced sensitivity to the anti-cancer reagent cisplatin. Taken together, our results set a new paradigm for Statistical Analysis. The Kaplan–Meier method was used to estimate the prob- cancer therapy in which the inhibition of Nrf2 could be used to ability of disease-free survival. Log-rank analysis was used to assess the signifi- enhance chemotherapeutic sensitivity. They further suggest that cance of the differences between tumors with and without NRF2 mutation. mutation in the Keap1-Nrf2 system may represent a novel oncogenic mechanism leading to malignancy (Fig. S3). SI. Further description of experimental procedures can be found in SI Materials and Methods, which is published as SI on the PNAS web site. Materials and Methods ACKNOWLEDGMENTS. We thank Dr. K. Nakayama (Kyushu University, DNA Sample Collection. DNA was extracted from cell lines using a DNA Fukuoka, Japan) for HA-tagged ubiquitin vector and Dr. V.P. Kelly (Trinity extraction kit (Qiagen). All tumor samples were subjected to laser-capture College, Dublin) for critical reading of the manuscript. This work was sup- microdissection using LM200 (Arcuturus). ported in part by a Grant-in-Aid for Cancer Research from the Ministry of Health, Labor and Welfare of Japan; a grant for the program for promotion PCR Amplification and Genetic Analysis of NRF2. All coding exons of NRF2 were of Fundamental Studies in Health Sciences of the National Institute of Bio- medical Innovation (NiBio); and a grant for the program for promotion of amplified by PCR using High Fidelity Taq polymerase (Roche) and appropriate Fundamental Studies in Health Sciences of the Organization for Pharmaceu- primers (Table S3). All PCR products were purified and analyzed by sequenc- tical Safety and Research; the Japanese Ministry of Education, Culture, Sports, ing. A high-density oligonucleotide array (Human genomic CGH microarray Science and Technology, and Exploratory Research for Advanced Technology- 44B, Agilent Technologies) was used. Japan Science and Technology Corporation.

1. Kasai H, Kawai K (2006) Oxidative DNA damage: Mechanisms and significance in health 21. Tong KI, et al. (2007) Different electrostatic potentials define ETGE and DLG motifs as and disease. Antioxid Redox Signal 8:981–983. hinge and latch in oxidative stress response. Mol Cell Biol 21:7511–7521. 2. Motohashi H, Yamamoto M (2007) Carcinogenesis and transcriptional regulation 22. Lo SC, Li X, Henzl MT, Beamer LJ, Hannink M (2006) Structure of the Keap1: Nrf2 through Maf recognition elements. Cancer Sci 98:135–139. interface provides mechanistic insight into Nrf2 signaling. EMBO J 25:3605–3617. 3. Fahey JW, Kensler TW (2007) Role of Dietary Supplements/Nutraceuticals in Chemo- 23. Marti-Renom MA, et al. (2000) Comparative protein structure modeling of genes and prevention through Induction of Cytoprotective Enzymes. Chem Res Toxicol 20:572– genomes. Annu Rev Biophys Biomol Struct 29:291–325. 576. 24. Mori K, Shibanuma M, Nose K (2004) Invasive potential induced under long-term 4. Kensler TW, Wakabayashi N, Biswal S (2007) Cell survival responses to environmental oxidative stress in mammary epithelial cells. Cancer Res 64:7464–7672. stresses via the Keap1-Nrf2-ARE pathway. Annu Rev Pharmacol Toxicol 47:89–116. 25. Petroski MD, Desharies RJ (2005) Function and regulation of cullin-RING ubiquitin 5. Ramos-Gomez M, et al. (2001) Sensitivity to carcinogenesis is increased and chemo- ligases. Nat Rev Mol Cell Biol 6:9–15. protective efficacy of enzyme inducers is lost in nrf2 transcription factor-deficient mice. 26. Tong KI, Kobayashi A, Katsuoka F, Yamamoto M (2006) Two-site substrate recognition Proc Natl Acad Sci USA 98:3410–3415. model for the Keap1-Nrf2 system: A hinge and latch mechanism. Biol Chem 387:1311– 6. Iida K, et al. (2004) Nrf2 is essential for the chemopreventive efficacy of oltipraz against 1320. urinary bladder carcinogenesis. Cancer Res 64:6424–6431. 27. Kim WY, Kaelin WG (2004) Role of VHL gene mutation in human cancer. J Clin Oncol 7. Kobayashi A, et al. (2004) Oxidative stress sensor Keap1 functions as an adaptor for Cul3-based E3 ligase to regulate proteasomal degradation of Nrf2. Mol Cell Biol 22:4991–5004. 24:7130–7139. 28. Mani A, Gelmann EP (2005) The ubiquitin-proteasome pathway and its role in cancer. 8. Kobayashi A, et al. (2006) Oxidative and electrophilic stresses activate Nrf2 through J Clin Oncol 23:4776–4789. inhibition of ubiquitination activity of Keap1. Mol Cell Biol 26:221–229. 29. Wu LC, et al. (1996) Identification of a RING protein that can interact in vivo with the 9. Zhang DD, Lo SC, Cross JV, Templeton DJ, Hannink M (2004) Keap1 is a redox-regulated BRCA1 gene product. Nat Genet 14:430–440. substrate adaptor protein for a Cul3-dependent ubiquitin ligase complex. Mol Cell Biol 30. Pause A, et al. (1997) The von Hippel-Lindau tumor-suppressor gene product forms a 24:10941–10953. stable complex with human CUL-2, a member of the Cdc53 family of proteins. Proc Natl 10. Devling TW, Lindsay CD, McLellan LI, McMahon M, Hayes JD (2005) Utility of siRNA Acad Sci USA 94:2156–2161. against Keap1 as a strategy to stimulate a cancer chemopreventive phenotype. Proc 31. Lisztwan J, Imbert G, Wirbelauer C, Gstaiger M, Krek W (1999) The von Hippel-Lindau Natl Acad Sci USA 102:7280–7285. tumor suppressor protein is a component of an E3 ubiquitin-protein ligase activity. 11. Kern JT, Hannink M, Hess JF (2007) Disruption of the keap1-containing ubiquitination Genes Dev 13:1822–1833. complex as an antioxidant therapy. Curr Top Med Chem 7:977–983. 32. Iliopoulos O, Levy AP, Jiang C, Kaelin WG Jr., Goldberg MA (1996) Negative regulation 12. Jaiswal AK (1994) Jun and Fos regulation of NAD (P)H: Quinone oxidoreductase gene of hypoxia-inducible genes by the von Hippel-Lindau protein. Proc Natl Acad Sci USA expression. Pharmacogenetics 4:1–10. 93:10595–10599. 13. Gupta GP, Massague J (2006) Cancer metastasis: Building a framework. Cell 127:679– 33. Starita LM, Parvin JD (2006) Substrates of the BRCA1-dependent ubiquitin ligase. 695. Cancer Biol Ther 5:137–141. 14. Padmanabhan B, et al. (2006) Structural basis for defects of Keap1 activity provoked by 34. Wakabayashi N, et al. (2003) Keap1-null mutation leads to postnatal lethality due to its point mutations in lung cancer. Mol Cell 21:689–700. constitutive Nrf2 activation. Nat Genet 35:238–245. 15. Singh A, et al. (2006) Dysfunctional KEAP1-NRF2 interaction in non-small-cell lung 35. Reddy NM, et al. (2007) Genetic dissection of the Nrf2-dependent redox signaling cancer. PLoS Med 3:e420. regulated transcriptional programs of cell proliferation and cytoprotection. Physiol 16. Ohta T, et al. (2008) Loss of Keap1 function activates Nrf2 and provides advantages for Genomics 32:74–81. lung cancer cell growth. Cancer Res 68:1303–1309. 36. Bindra RS, Glazer PM (2005) Genetic instability and the tumor microenvironment: 17. Yamamoto T, et al. (2004) Identification of polymorphisms in the promoter region of Towards the concept of microenvironment-induced mutagenesis. Mutat Res 569:75– the human NRF2 gene. Biochem Biophys Res Commun 321:72–79. 18. Marzec JM, et al. (2007) Functional polymorphisms in the transcription factor NRF2 in 85. humans increase the risk of acute lung injury. FASEB J 21:2237–2246. 37. Adriana A, Sporn MC (2007) The tumor microenvironment as a target for chemopre- 19. Itoh K, et al. (1999) Keap1 represses nuclear activation of antioxidant responsive vention. Nat Rev Cancer 7:139–147. elements by Nrf2 through binding to the amino-terminal Neh2 domain. Genes Dev 38. Semenza GL (2003) Targeting HIF-1 for cancer therapy. Nat Rev Cancer 3:721–732. 13:76–86. 39. Zhou J, Schmid T, Schnitzer S, Bru¨ ne B (2006) Tumor hypoxia and cancer progression. 20. Tong KI, et al. (2006) Keap1 recruits Neh2 through binding to ETGE and DLG motifs: Cancer Lett 237:10–21. Characterization of the two-site molecular recognition model. Mol Cell Biol 26:2887– 40. Kim YJ, et al. (2007) Human prx1 gene is a target of Nrf2 and is up-regulated by 2900. hypoxia/reoxygenation: Implication to tumor biology. Cancer Res 67:546–554. MEDICAL SCIENCES

Shibata et al. PNAS ͉ September 9, 2008 ͉ vol. 105 ͉ no. 36 ͉ 13573