PIDD interaction with KEAP1 as a new mutation-independent mechanism to promote NRF2 stabilization and chemoresistance in NSCLC

Chemotherapy resistance is a major problem in non-small cell lung cancer (NSCLC) treatment. A major mechanism of chemoresistance involves stabilization of the NRF2 transcription factor. NRF2 levels are normally tightly regulated through interaction with KEAP1, an adaptor that targets NRF2 to the CUL3 E3 ubiquitin ligase for proteolysis. In NSCLC, aberrant NRF2 stabilization is best understood through mutations in NRF2, KEAP1, or CUL3 that disrupt their interaction. Biochemical studies, however, have revealed that NRF2 can also be stabilized through expression of KEAP1-interacting proteins that competitively sequester KEAP1 away from NRF2. Here, we have identified PIDD, as a novel KEAP1-interactor in NSCLC that regulates NRF2. We show that this interaction allows PIDD to reduce NRF2 ubiquitination and increase its stability. We also demonstrate that PIDD promotes chemoresistance in NSCLC cells both in vitro and in vivo, and that this effect is dependent on NRF2. Finally, we report that NRF2 protein expression in a NSCLC cohort exceeds the typical incidence of combined NRF2, KEAP1, and CUL3 mutations, and that NRF2 expression in this cohort is correlated with PIDD levels. Our data identify PIDD as a new NRF2 regulator, and suggest that variations in PIDD levels contribute to differential chemosensitivities among NSCLC patients.

against oxidative stress and harmful environmental toxicants and carcinogens. Consistent with these molecular functions, high levels of NRF2 in NSCLC are associated with resistance to chemotherapy and poor prognosis 9 . In functional studies, the loss of Kelch-like ECH-associated protein 1 (KEAP1), a negative regulator of NRF2, increases nuclear accumulation and activation of NRF2 in NSCLC cell lines, and promotes growth and chemoresistance 10 . In addition, silencing of NRF2 through RNAi increases sensitivity to cisplatin in NSCLC cells [11][12][13] . Thus, the development of therapeutic strategies that effectively interfere with NRF2 activity offer great promise for overcoming chemoresistance.
NRF2 activity is highly regulated. It shifts from essentially non-existent under healthy conditions to highly active during stress. This regulation primarily occurs through modulation of NRF2 proteolysis. Under non-stressed conditions, NRF2 is constitutively degraded through interaction with its binding partner KEAP1, which recruits the CUL3-E3 ubiquitin ligase to stimulate proteasomal degradation of NRF2 14 . Upon exposure to oxidative stress or chemopreventive compounds like sulforaphane and curcumin, modification of the cysteine residues on KEAP1 imposes a conformational change that disrupts the binding between KEAP1 and NRF2, resulting in diminished NRF2 ubiquitination and increased NRF2 protein levels and target gene expression 15,16 . The critical importance of this form of regulation to NRF2 activity is supported by the discovery of gain-of-function mutations in NRF2 that disrupt regulation by KEAP1-CUL3, and loss-of-function mutations in KEAP1 and CUL3 in several cancer types 17 . A growing body of evidence suggests that KEAP1 could be a focal point for additional mechanisms of NRF2 regulation. In tumors, NRF2 levels are sometimes increased in the absence of genomic alterations in the NRF2, KEAP1, and CUL3 genes. Several studies suggest that this might occur through elevated expression of proteins that compete with NRF2 for binding to KEAP1, and thus, sequester KEAP1 away from NRF2, preventing its ubiquitination. This list of proteins currently includes p62 (SQSTM1), WTX, PALB2 and DPP3 [18][19][20][21] .
To study potential regulation of NRF2 activity by KEAP1-interacting proteins in NSCLC, we used KEAP1 immunoprecipitation-mass spectrometry (IP-MS) to identify KEAP1-interactors in a NSCLC cell line. Using this approach, we identified P53-induced protein with a death domain (PIDD) as a novel binding partner for KEAP1. We also provide evidence for PIDD being a new clinically relevant regulator of NRF2 and NSCLC malignancy and chemoresistance, and suggest that its further study may yield insight into novel treatment options for NSCLC.

NRF2 regulatory pathway alterations in NSCLC and other cancers.
To better understand the extent of involvement of NRF2 activation to chemotherapy resistance in NSCLC, we surveyed NRF2 pathway genomic alterations in NSCLC relative to other major cancers. Using the comprehensive TCGA data available through the cBioportal database (www.cBioportal.org) 22 , we examined DNA amplifications, deletions, and mutations in NRF2, KEAP1, and CUL3. When considering the cumulative alteration frequencies of all three genes as an overall assessment of pathway activation, the top two cancers of 31 surveyed malignancies were lung squamous cell carcinoma (35.96%) and lung adenocarcinoma (23.48%), the two main types of NSCLC, which are shown in Supplementary Fig. S1 (Panel A). The detailed alteration frequencies for individual genes are presented in Supplementary Fig. S1 (Panel B-D) and Supplementary Table S1. These data support an especially prominent role for NRF2 pathway activation in NSCLC and suggest that additional mechanisms may promote NRF2 activation in cases where DNA alterations were not detected.

PIDD interacts with KEAP1.
To investigate other potential mechanisms for NRF2 activation, we focused on characterizing novel KEAP1-interacting proteins that might compete with NRF2, and hence, indirectly promote NRF2 stabilization. In pilot discovery experiments, we used immunoprecipitation-mass spectrometry (IP-MS) and H1299 NSCLC cells, which have wild-type KEAP1 (determined from publicly available data from the Cancer Cell Line Encyclopedia), to identify 58 putative KEAP1-interactors, including P53-induced protein with a death domain (PIDD). We decided to focus on PIDD since it, like NRF2, has been implicated in chemoresistance 23 , which suggested a possible functional connection to NRF2 regulation. To validate the interaction of KEAP1 with PIDD, we first used epitope-tagged constructs and anti-tag antibodies to look for co-immunoprecipitation in transiently transfected HEK293T cells. After transfection of HEK293T cells with FLAG-KEAP1 and HA-PIDD, HA-PIDD was found in anti-FLAG-KEAP1 immunoprecipitates (Fig. 1A), which migrated in the range of ~100 kDa, as detected with the anti-HA antibody. Conversely, FLAG-KEAP1 was found in the reciprocal anti-HA-PIDD immunoprecipitates (Fig. 1B), which consistently migrated as a ~55 kDa doublet, as detected with the anti-FLAG antibody. The Western bands of both PIDD and KEAP1 were consistent with previous papers [24][25][26] . To determine if the KEAP1 and PIDD interaction could also be detected among endogenously expressed proteins in NSCLC, we also performed co-immunoprecipitation using antibodies directed towards the native proteins. Consistent with the transfection studies, endogenous KEAP1 and PIDD could be co-immunoprecipitated from H1299 NSCLC cells (Fig. 1C,D).
PIDD reduces the amount of ubiquitinated NRF2. Since some KEAP1-interacting proteins can indirectly regulate NRF2 levels by sequestering KEAP1and the CUL3-dependent ubiquitylating machinery away from NRF2 26-28 , we next investigated whether PIDD could also regulate NRF2 ubiquitination. HEK293T cells were transfected with increasing doses of a PIDD expression vector and treated with a proteasome inhibitor, MG132, for 4 hours to prevent ubiquitinated NRF2 from being degraded. Endogenous NRF2 was then immunoprecipitated from the lysates and blotted with an anti-ubiquitin antibody. Consistent with our hypothesis, as PIDD levels increased, NRF2 ubiquitination decreased, indicating that PIDD can regulate NRF2 ubiquitination (Fig. 2). www.nature.com/scientificreports www.nature.com/scientificreports/

PIDD promotes chemoresistance in NSCLC cells both in vitro and in vivo.
To determine whether PIDD might promote chemoresistance in NSCLC, we used a lentivirus vector to raise PIDD expression in H1299 cells where we found PIDD interacts with KEAP1. In LV-PIDD-infected cells, PIDD expression was increased ~3-fold, which was sufficient to cause a more than 2-fold increase in NRF2 levels, and is consistent with PIDD's ability to reduce NRF2 ubiquitination (Fig. 3A,B). These changes in PIDD and NRF2 expression in the LV-PIDD-infected cells translated to increased resistance to cisplatin (p < 0.05, Fig. 3C).
To determine whether the increased resistance to cisplatin is also manifested in vivo, we generated LV-PIDD H1299 xenografts. Equivalent numbers of LV-PIDD-infected and LV-mock-infected H1299 cells were injected into nude mice and after 10 days, cisplatin treatment was initiated. At the endpoint of the therapeutic regimen, the sizes, volumes, and weights of the LV-PIDD tumors were all significantly larger than the mock group (p < 0.05, Fig. 3D-F), demonstrating increased resistance to cisplatin upon elevation of PIDD levels.

PIDD-mediated chemoresistance is dependent on NRF2.
To explore the mechanism of PIDD-induced chemoresistance in NSCLC, we focused on NRF2, whose stabilization is promoted by PIDD. Using control MGH7 cells, a lung squamous cell carcinoma cell line with high levels of NRF2 29,30 (Fig. 4A), we The upper and lower panels were from the same gel. The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. Figure 2. PIDD decreases the amount of ubiquitinated NRF2. HEK293T cells were transfected with different amounts of PIDD expression plasmid in the presence of 10 µM MG132, which blocks degradation of ubiquitinated proteins. NRF2 was then immunoprecipitated (IP) from cell lysates and its level of ubiquitination was quantified by anti-ubiquitin immunoblotting (IB). The upper and lower panels were from the same gel. The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. The upper and lower panels were from the same gel. The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. (B) Quantification of higher levels of PIDD and NRF2 in lentivirus infected H1299 cells. Data are presented as the densitometric means of three immunoblotting replicates and are normalized to expression of GAPDH. Error bars represent standard error of the mean (SEM) (Two tailed Student's t-test, **P = 0.0011; ***P < 0.0001). (C) LV-PIDD infected H1299 cells showed increased viability as compared to the LV-mock H1299 cells after treatment with different doses of cisplatin for 24 hours. Viability was quantified by the MTT assay. Data are mean ± SEM for three experiments (Two way ANOVA; **P < 0.01; ***P < 0.001; ns, P > 0.05). www.nature.com/scientificreports www.nature.com/scientificreports/ identified a lentiviral shRNA construct (shRNA2) that was especially effective at reducing protein levels of NRF2 and the NRF2 target gene, NQO1 (Fig. 4A,B). MGH7 cells stably infected with NRF2 shRNA2 also had a lower IC50 for cisplatin (1.503 uM vs 10.84 uM for control shlacz infected cells) (Fig. 4C), supporting NRF2 promoting chemoresistance in these cells.
We next used the NRF2 shRNA2 construct to determine if the elevated NRF2 expresssion in the PIDD-overexpressing H1299 cells (H1299-LV-PIDD) contributes to their increased chemoresistance. As observed in MGH7 cells, NRF2 shRNA2 reduced NRF2 and NQO1 protein levels in H1299-LV-PIDD cells, and importantly, without affecting PIDD expression (Fig. 4D,E). The reduction in NRF2 expression also increased sensitivity to cisplatin (Fig. 4F), supporting the PIDD-mediated increase in NRF2 levels driving increased chemoresistance.
Clinical relevance of PIDD and NRF2 expression in NSCLC. We next investigated PIDD and NRF2 expression in NSCLC patient tissue to obtain potential corroborating evidence for the clinical relevance of our findings. First, we compared PIDD and NRF2 expression in 4 pairs of matched NSCLC and adjacent normal The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. (B) Quantification of NRF2 and NQO1 protein levels in lentivirus infected MGH7 cells. Data are presented as the densitometric means of three immunoblotting replicates and are normalized to expression of GAPDH. Error bars represent standard error of the mean (SEM) (Two tailed Student's t-test, *P = 0.0302; **P = 0.0055; ***P < 0.0001). (C) Viability of MGH7 cells infected with lentiviruses expressing either NC or shNRF2 (NRF2-shRNA2) after treatment with different doses of cisplatin for 72 hours. Viability was quantified by the MTT assay. (D) Western blotting was performed to determine the expression levels of NRF2, NQO1 and PIDD in both NC and NRF2-shRNA infected cells. The upper and lower panels were from the same gel. The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. (E) Quantification of NRF2, NQO1 and PIDD protein levels in lentivirus infected H1299-LV-PIDD cells. Data are presented as the densitometric means of three immunoblotting replicates and are normalized to expression of ß-actin. Error bars represent standard error of the mean (SEM) (Two tailed Student's t-test, **P = 0.0019 for NRF2 in shRNA2; **P = 0.0037 for NQO1 in shRNA2). (F) NRF2 shRNA2-infected H1299-LV-PIDD cells showed decreased viability as compared to the negative control cells after treatment with different doses of cisplatin for 24 hours. Viability was quantified by the MTT assay. Data are mean ± SEM for three experiments (Two way ANOVA; *P < 0.05; **P < 0.01; ***P < 0.001; ns, P > 0.05). (2019) 9:12437 | https://doi.org/10.1038/s41598-019-48763-4 www.nature.com/scientificreports www.nature.com/scientificreports/ tissues using Western blot analysis. NRF2 and PIDD levels were elevated in the tumor samples of 3 of the 4 cases, with 2 cases showing elevation in both PIDD and NRF2 in the tumor (Fig. 5A-C).
We then semi-quantitatively examined expression of PIDD, NRF2, and the NRF2 target gene, MRP1 31 , in 120 cases of treatment-naïve lung adenocarcinoma by immunohistochemistry (Supplemental Table S2, Fig. 5D). Using these data, we then evaluated the degree to which NRF2 might be active in the samples of this cohort by comparing NRF2 and MRP1 expression. The expression of these two proteins was significantly correlated, supporting NRF2 being active in many samples (Fig. 6A). We also found that PIDD expression was correlated with both NRF2 and MRP1, consistent with PIDD levels promoting NRF2 expression and activity (Fig. 6B,C).
Finally, we separated PIDD, NRF2, and MRP1 expression into low and high groups (see Methods) and re-evaluated their expression relative to each other, as well as to clinicopathologic variables, using the Pearson χ 2 test (Table 1). Consistent with the Pearson correlation test analyses, significant positive correlations were found between the expression levels of PIDD and that of NRF2 and MRP1 (p = 0.037 and 0.038). There were also significant positive correlations between PIDD expression and that of tumor size (p = 0.007), lymph node status (p = 0.009), and clinical stage (p = 0.035). Overall, these data indicate that as compared with low PIDD expression, high PIDD levels are associated with larger tumor size, lymph node metastasis, and higher stage. Although not quite significant, Kaplan-Meier analysis indicated that there was a trend for high PIDD expression to also be associated with worse overall survival in this cohort that was treated with surgery alone (p = 0.075, Fig. 6D). These findings support PIDD regulating NRF2 activity in NSCLC and suggest that in addition to promoting chemoresistance, PIDD might also contribute to development of malignant properties in NSCLC.

Discussion
A wide body of work indicates that prolonged activation of the NRF2 transcription factor can promote progression of various cancers, including lung, breast, head and neck, ovarian, and endometrial carcinomas 32 . The poor prognosis of patients with tumors expressing high levels of NRF2 appears to relate to NRF2's ability to both enhance cancer cell proliferation and promote chemo and radio-resistance 26,33,34 . In addition, NRF2 expression is induced during therapy, which further promotes development of drug resistance 35,36 . As such, NRF2 contributes to both intrinsic and acquired chemo-resistance 34,37 . Although NRF2 stabilization through genomic alterations in NRF2, KEAP1, and CUL3 is well documented 8 , additional mechanisms that increase NRF2 activity have remained enigmatic and may play an equal or even greater role in cancer progression. Indeed, we found that although in the TCGA lung adenocarcinoma data, the cumulative alteration frequency of NRF2, KEAP1, and CUL3 is 23% ( Supplementary Fig. S1), 58% of patients within our lung adenocarcinoma cohort had high levels of NRF2 protein (Table 1).
One way through which NRF2 levels may be increased in the absence of genomic alterations in the three core genes is through an indirect mechanism involving competitive sequestration of KEAP1. p62, WTX, PALB2 and DPP3 have been reported to activate NRF2 by disrupting NRF2-KEAP1 binding 20,21,38 . However, the expression of these proteins in patient tumors and their potential contributions to tumor growth or chemoresistance are unknown. Here, we not only identified PIDD as a novel KEAP1-interactor, but also demonstrated its ability to promote NRF2 stabilization, which is accompanied by an increase in chemoresistance in NSCLC cells in vitro and in vivo.
PIDD was originally described as a primary p53 target gene that is induced upon DNA damage 39 . Despite initially being characterized for its role in pro-apoptotic pathways, PIDD was later shown to play a role in DNA-damage induced NF-kB activation, which putatively promotes cell survival and chemoresistance 40,41 . Indeed, later work, including in a KRAS-driven mouse lung cancer model and KRAS mutant human lung cancer cell lines, showed that PIDD overexpression can promote resistance to cisplatin 23,42 . Consistent with these findings, our functional studies using H1299 lung cancer cells also support PIDD expression promoting chemoresistance, and in cells that do not harbor KRAS mutations. However, whereas previous work in murine KRAS-driven tumors found some evidence that cisplatin induced changes in glutathione metabolism, a target pathway of NRF2 43 , those changes were not linked specifically to PIDD, and more evidence was presented that chemoresistance was primarily due to an increase in DNA repair mechanisms 23 . Our findings that high levels of PIDD prevent NRF2 ubiquitination and promote its stabilization, and are associated with increased NRF2 and target gene expression in NSCLC patient samples, suggest that PIDD-mediated NRF2 activation may have also been a major factor driving chemoresistance in the mouse lung tumors.
Our data thus, suggest a new way through which PIDD may affect tumor growth in response to DNA damaging chemotherapeutics. Previous work has described both pro and anti-apoptotic functions of PIDD that are mediated through distinct interactions involving proteins other than KEAP1 44 . The switch in the role of PIDD has also been linked to differential association with either PIDDosome complex containing Caspase-2 and RAIDD for cell death or NEMO (NF-kB)/RIP1 for cell survival 45 . In addition, PIDD also participates directly in DNA damage repair. PIDD binds to PCNA upon UV irradiation through its ZU-5 domains and stimulates PCNA monoubiquitination during translesion DNA synthesis 46 . Therefore, it is likely that multiple interactions between PIDD and other proteins may mediate its many roles in cellular responses to genotoxic stress.
Several inhibitors of NRF2 activity have been described. One group consists of non-specific natural products such as brusatol. Although brusatol sensitizes a broad spectrum of cancer cells to antitumor drugs and reduces cell growth 13,47 , its potent cytotoxic effects appear to act independently of NRF2 inhibition and are more consistent with the profile of a protein translation inhibitor 48 . Other non-specific inhibitory natural products include luteolin and alkaloid trigonelline 49,50 . By contrast a more specific NRF2 inhibitor, ML385, was recently identified through screening a drug library for compounds that inhibit NRF2-dependent transcriptional activity 51 . ML385 binds directly to NRF2 and enhances the cytotoxicity of platinum-based drugs on NSCLC cell lines. However, despite the promise of these anti-NRF2 therapies, the development of clinically useful drugs is still in its infancy. www.nature.com/scientificreports www.nature.com/scientificreports/ Representative western blot analysis. GAPDH was used as a loading control. The upper and lower panels were from the same gel. The gel was transferred to the same membrane, which was cut to probe with different antibodies. The entire image of each exposed membrane is shown. (B,C) Quantification of NRF2 and PIDD expression in matched normal and tumor NSCLC tissue by western blotting. Three independent western blotting experiments were performed from the same patient tissue lysates. The data are the mean ± SEM (Two tailed Student's t-test; ***P < 0.001; ns, P > 0.05). (D) Immunohistochemical staining of PIDD (a1-a4), NRF2 (b1-b4), and the NRF2 target, MRP1 (c1-c4) in NSCLC specimens. Paraffin-embedded tissue sections were stained with antibodies against PIDD, NRF2, and MRP1, and counterstained with hematoxylin. Representative images from adjacent tissue normal (a1-c1,), and well differentiated (a2-c2), moderately differentiated (a3-c3), and poorly differentiated tumor tissue (a4-c4) are shown.
Our finding that PIDD and NRF2 are widely co-expressed in clinical NSCLC support the PIDD-KEAP1 interaction as being prominent both in chemo-naïve and chemo-treated tumors. Thus, drugs that disrupt the PIDD-KEAP1 interaction may be especially effective in overcoming chemoresistance. However, given that PIDD can interact with different proteins to perform distinct functions, drugs should ideally be designed to interfere with the KEAP1 interaction, but potentially allow interaction with pro-apoptotic RAIDD. Indeed, some drugs are currently in clinical development that target specific protein-protein interactions 52 . For example, drugs are being developed to specifically interfere with human murine double minute 2 (MDM2) interaction withTP53. When MDM2 interacts with TP53, it promotes TP53 degradation, thereby antagonizing its tumor suppressor function 53 . Drugs such as NVP-CGM097 and NVP-HDM201 block the MDM2-TP53 interaction and are currently being tested in phase 1 clinical trials as potential anti-cancer therapies for tumors with wild-type TP53 54 .
In conclusion, our work supports continued study of non-genomic mechanisms of NRF2 activation and further characterization of PIDD and KEAP1 interactions with their distinct binding partners.

Materials and Methods
Mass spectrometry. One round of discovery immumoprecipitation-mass spectrometry (IP-MS) analysis was performed using H1299 NSCLC cells. Cells were lysed in immunoprecipitation buffer (25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, and 1% NP-40, pH 7.8) and precleared with protein G sepharose (Sigma) for 2 h. Then, 100 mg of protein lysate was immunoprecipitated overnight at 4 °C with 5 μg of anti-KEAP1 antibody (H-190; SC 33569, Santa Cruz Briotechnology, Inc.), or rabbit normal immunoglobulin G (IgG) (Bioworld Technology, MN). Immune complexes were recovered with protein G sepharose for 2 h. After three washes in lysis buffer, the samples were run on a 10% SDS-PAGE gel and stained with Coomassie brilliant blue. The gel lanes were cut out and proteins extracted and subjected to mass spectrometry analysis as previously described 55 . www.nature.com/scientificreports www.nature.com/scientificreports/ Tissue specimens. Permission for the use of the human fresh tissue samples, tissue sections, and associated patient clinical data for this project was granted by the Ethics Committee of the Affiliated Hospital of Nantong University, Jiangsu Province, China. Written informed consent was obtained from all the patients, and patient specimens were collected in accordance with the Declaration of Helsinki 2013. For Western blot analyses, a total of four paired fresh NSCLC tumorous and adjacent non-tumorous lung tissues were used. These samples were harvested immediately after surgical resection and stored at −80 °C. For immunohistochemistry analyses, 120 lung adenocarcinoma specimens collected by informed consent from 2008 to 2013, were obtained from the Department of Pathology, Affiliated Hospital of Nantong University. None of the patients had received pre-surgical therapies such as immunotherapy, radiation, or chemotherapy. The paraffin sections were stained with hematoxylin and eosin (H&E). Demographic information and clinical and pathological data of these 120 patients are shown in Supplementary Table S3. Immunohistochemical staining. For immunohistochemical staining, deparaffinized sections were processed for antigen retrieval by heating in sodium citrate buffer (pH 6.0). The following panel of primary antibodies was used: (1) NRF2 (1:100, ab76026, Abcam, Inc.), (2) MRP1 (1:50, ab24102, Abcam, Inc.), and (3) PIDD (1:100, sc-32161, Santa Cruz Briotechnology, Inc.). The sections were incubated with primary antibody for 1 h at 37 °C and 4 °C overnight. After washing with phosphate-buffered saline (PBS), tissues were incubated with a horseradish peroxidase-conjugated anti-rabbit or anti-mouse Ig polymer as a second antibody (Dako, Hamburg, Germany) for 20 min at room temperature, according to the manufacturer's instructions. Finally, the peroxidase reaction was visualized by incubating the sections with DAB (0.1% phosphate buffer solution, 0.02% diaminobenzidine tetrahydrochloride, and 3% H 2 O 2 ). After rinses in water, the sections were counterstained with hematoxylin and then dehydrated and mounted in resin mount. Negative control slides were processed in parallel using a non-specific immunoglobulin IgG (Sigma Chemical Co, St. Louis, MO, USA) at the same concentration as the primary antibody.
Immunohistochemical analysis. All samples were evaluated and independently scored by 2 pathologists using a prespecified protocol. Five high-power fields in each section were selected randomly, and at least 300 cells were counted per field. Each slide was evaluated using a semiquantitative scoring system for both the intensity of the stain and the percentage of positive malignant cells. The intensity of staining was coded as follows: 1 (negatively or poorly stained), 2 (moderately stained), and 3 (strongly stained). The percentage of staining was scored Western blot analysis and immunoprecipitation. The tissues and harvested cells were promptly homogenized in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 5 mM EDTA, and a protease inhibitor cock-tail (Roche Diagnostics, Basel, Switzerland)), and then centrifuged to collect the supernatant. Protein concentrations were measured with a Bio-Rad protein assay (Bio-Rad, Hercules, CA, USA). Next, the same total protein was separated by SDS-PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane (Immobilon; Millipore). The membranes were first blocked with 5% non-fat milk in Tris-buffered saline with Tween 20 (TBST) (150 mM NaCl, 20 mM Tris, 0.05% Tween 20), washed with TBST for three times after 2 h blocking at room temperature. The membrane was then cut according to the position of corresponding protein, and probed overnight with each primary antibody. After three washes, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (Pierce Biotechnology) for 2 h at room temperature, and proteins detected by ECL (enhanced chemiluminescence) (Pierce, Rockford, IL). Alternatively, secondary antibody incubation was performed using fluorescent secondary antibodies (LI-COR Biosciences, Lincoln, NE), and protein bands were visualized using an Odyssey system (LI-COR Biosciences). For immunoprecipitation assays, H1299 cells were grown to confluence in 10-cm dishes and washed with cold phosphate-buffered saline (PBS). For HEK293T cells, they were transiently transfected with 10 µg of the expression plasmids by using Lipofectamine (Invitrogen) according to the manufacturer's instructions. At 48 hours after transfection, the growth medium was aspirated and cells were washed three times with cold PBS. The supernatants of these cell lysates were immune-precipitated with the primary antibodies against PIDD (sc-32161, Santa Cruz Briotechnology, Inc.), KEAP1 (H-190; SC 33569, Santa Cruz Briotechnology, Inc.), HA (ab9110, Abcam), FLAG (ab1240, Abcam), or control IgG (Santa Cruz Biotechnology) conjugated to protein G-Sepharose beads. Ten micrograms of antibody were added to 500 micrograms of cell lysate, followed by a 2 h incubation with gentle rotation at 4 °C. The precipitates were washed three times with PBS and collected for Western blot analysis.

Construction of PIDD overexpressing H1299 cells by lentiviral infection. The PIDD lentiviral
overexpression plasmid was constructed and purchased from Genechem (Genechem Co., Ltd, Shanghai, China). For amplification of PIDD by polymerase chain reaction (PCR), a pair of primers were designed and synthesized as follows: Forward primer: 5′-GAGGATCCCCGGGTACCGGTCGCCACCAT GGCTGCAACGGTGGAGGGGCCAGAG-3′; Reverse primer: 5′-TCCTTGTAGTCCATACCGGCCTGGGCA GGCTCTGGGGGCTGTG-3′. The resulting PIDD cDNA was inserted into the lentiviral vector GV358 (sequence elements: Ubi-MCS-3XFLAG-SV40-GFP-IRES-puromycin; Shanghai Genechem Company, China) to create the lentiviral vector, LV-PIDD. Sequencing was performed to verify the correctness of the sequence. For lentivirus packaging, 293 T cells were cotransfected with the recombinant PIDD expression vector (LV-PIDD), as well as the packaging plasmids, Helper 1.0 and Helper 2.0 (Genechem Company, Shanghai, China) to generate the target lentivirus with an infectious viral titer of 2 × 10 8 TU/ml, which was measured using a fluorescence assay method. The harvested lentiviruses were concentrated, purified and stored at −80 °C.
For PIDD overexpression studies, H1299 cells were exposed to either LV-PIDD lentiviral particles in the presence of polybrene (8 μg/ml, Sigma-Aldrich) or mock conditions (control), in which cells were exposed to media containing polybrene, but no virus. After 8 h, the medium was replaced. Stable PIDD overexpressing cells were selected by treatment with puromycin (Sigma-Aldrich), while mock-infected cells were grown in the absence of puromycin.
NRF2 knockdown in MGH7 and H1299-LV-PIDD cells. NRF2 shRNA lentiviral constructs in the pLKO.1 backbone were purchased from Sigma-Aldrich, St. Louis, USA. The NRF2 targeting sequences included CCGGGCACCTTATATCTCGAAGTTTCTCGAGAAACTTCGAGATATAAGGTGCTTTTT for shRNA1 and CCGGGCTCCTACTGTGATGTGAAATCTCGAGATTTCACATCACAGTAGGAGCTTTTT for shRNA2. The NRF2-shRNA plasmids or non-targetting lacZ shRNA negative control plasmid, along with standard helper packaging plasmids, were transfected into 293T cells using FuGENE 6 (Promega, Madison, WI), following the manufacturer's instructions, to make virus. The viral supernatant was then applied to MGH7 and H1299-LV-PIDD cells, and after 48 hours, the cells were subjected to puromycin selection (Sigma-Aldrich, St. Louis, USA, 3 ug/ml) for 72 hours.
Tumor formation in nude mice. BALB/c athymic nude male mice (4 to 6 weeks old) were purchased from Shanghai SLAC Animal Center and raised in a specific pathogen-free environment in Laboratory Animal