Mutations in KEAP1 and NFE2L2 (encoding the protein Nrf2) are prevalent in both adeno and squamous subtypes of non–small cell lung cancer, as well as additional tumor indications. The consequence of these mutations is stabilized Nrf2 and chronic induction of a battery of Nrf2 target genes. We show that knockdown of Nrf2 caused modest growth inhibition of cells growing in two-dimension, which was more pronounced in cell lines expressing mutant KEAP1. In contrast, Nrf2 knockdown caused almost complete regression of established KEAP1-mutant tumors in mice, with little effect on wild-type (WT) KEAP1 tumors. The strong dependency on Nrf2 could be recapitulated in certain anchorage-independent growth environments and was not prevented by excess extracellular glutathione. A CRISPR screen was used to investigate the mechanism(s) underlying this dependence. We identified alternative pathways critical for Nrf2-dependent growth in KEAP1-mutant cell lines, including the redox proteins thioredoxin and peroxiredoxin, as well as the growth factor receptors IGF1R and ERBB3. IGF1R inhibition was effective in KEAP1-mutant cells compared with WT, especially under conditions of anchorage-independent growth. These results point to addiction of KEAP1-mutant tumor cells to Nrf2 and suggest that inhibition of Nrf2 or discrete druggable Nrf2 target genes such as IGF1R could be an effective therapeutic strategy for disabling these tumors.

Significance:

This study identifies pathways activated by Nrf2 that are important for the proliferation and tumorigenicity of KEAP1-mutant non–small cell lung cancer.

Lung cancer is the leading cause of cancer death in men and women, so new insights into driver genes for this indication are especially needed. Exome sequencing of 230 tumor/normal pairs from lung adenocarcinoma by The Cancer Genome Atlas (TCGA) consortium showed that KEAP1 was the third most mutated gene, present in 17% cases, with only TP53 and KRAS displaying higher mutation frequencies (1). KEAP1 is a substrate targeting protein for the Cul3 E3 ubiquitin ligase that ubiquitinates the Nrf2 transcription factor, resulting in its proteasomal degradation. NFE2L2, the gene encoding Nrf2, is also found frequently mutated across multiple human tumors, especially in squamous lung (15%; ref. 2). Mutations in Nrf2 are localized around 2 regions that interact with KEAP1, and mutations impair association with KEAP1 (3). In contrast, mutations in KEAP1 are spread throughout the gene, and may play a more complex role in affecting the interaction and ubiquitination of Nrf2 (4).

The KEAP1/Nrf2 pathway plays an important role in the cellular response to reactive oxygen species (ROS). Under nonstressed conditions, KEAP1 dimers maintain low levels of Nrf2 through binding 2 regions of Nrf2 (DLG and ETGE motifs), resulting in its Cullin 3-dependent ubiquitination and degradation. Upon increases in oxidative stress, key cysteine residues in KEAP1 become oxidized, changing the conformation of the KEAP1/Nrf2 complex such that Nrf2 no longer becomes ubiquitinated, leading to stabilization and accumulation in the cytosol and nucleus (5). Many transcriptional targets of Nrf2 have been identified, some of which counteract the cellular increases in ROS. For example, multiple components of the glutathione biosynthesis pathway are direct Nrf2 target genes, as well as enzymes in the pentose phosphate pathway (6), which is one of the major mechanisms generating NADPH, an important source of reducing power. ChIP-seq experiments in human and mouse cells have identified additional Nrf2 target genes under basal and stimulated conditions (7–9), thereby expanding the range of biological processes known to be under Nrf2 transcriptional control. However, the specific transcriptional targets of Nrf2 that generate a selective advantage for cells with mutations in this pathway are not clearly defined. Some studies have shown that supplementation with n-acetyl cysteine, a cell-permeable precursor of cysteine, can rescue viability defects induced by Nrf2 knockdown or knockout, under either basal or challenged conditions (10, 11), but whether this pathway represents the main survival benefit in KEAP1/Nrf2 mutant cells is not clear. In this study, we explore the consequences of KEAP1 mutations on the requirement for Nrf2 activity under different growth environments and show that Nrf2 activity is essential for growth in anchorage independent conditions. Surprisingly Nrf2 dependence is uncoupled from the glutathione synthesis pathway. Rather, through a CRISPR screen, we show that the thioredoxin/peroxiredoxin/thioredoxin reductase pathway is important for Nrf2-driven growth and viability. In addition, we find that growth factors signaling through IGF1R and ERBB3 are critical mediators of the growth of KEAP1-mutant cells, thus identifying multiple avenues for potential therapeutic intervention or biomarker discovery within this mutant context.

Cell culture and creation of doxycycline-inducible shRNA cell lines

All cell lines were obtained from Genentech's cell line core facility gCell. STR profiles are determined for each line using the Promega PowerPlex 16 System. This is performed once and compared with external STR profiles of cell lines (when available) to determine cell line ancestry. Cells are Mycoplasma tested before distribution. Cells were maintained in either RPMI or DMEM in the presence of 10% FBS (v/v) and 2 mmol/L l-glutamine, apart from BEAS2B cells that were maintained in BEGM growth media (Lonza, CC-3170). Doxycycline-inducible cell lines were maintained as above but with 10% Tet-system approved FBS (Clontech, 631101). All cell lines were cultured in a humidified incubator at 37°C / 5% CO2 unless otherwise noted, and discarded or replaced following culture for 3 months.

Nrf2- and nontargeting shRNA pINDUCER 10 lentiviral expression vectors were cotransfected with pCMV-VSVG and pCMV-dR8.9 plasmids into HEK293T cells using Lipofectamine 2000 reagent (Life Technologies, #11668). Viral supernatants were collected after 48 hours and were used with 8 μg/mL of polybrene reagent (Millipore, #TR-1003-G) to infect cell lines. Stably transformed cell line pools were generated by placing the transduced cells under 2 μg/mL puromycin selection for a minimum of 2 weeks.

For all additional targets, A549 or H460 cells were seeded in 6-well plates and grown to ∼50% confluence in media containing Tet-free FBS. The cells were then cotransfected with 250 ng of PiggyBac transposase expression plasmid (pBO, Transposagen) and 750 ng of a PiggyBac transposon plasmid containing a dox-inducible shRNA cassette (pBInducerOD_tRFP_miRE, Genentech), using Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific). Sequences of all shRNA reagents are described in Supplementary Table S1. Three days after transfection, cells were split into selection media containing 2 μg/mL puromycin, and selected for 3 days. Cells were then assayed for knockdown by Western blot analysis after treatment with 500 ng/mL doxycycline for 5 days.

siRNA-mediated depletion of Nrf2 and PRDX1

Sequences of all siRNA reagents are described in Supplementary Table S1. siRNA knockdown of Nrf2 and PRDX1 was conducted in 384 well plate format. Six siRNAs were tested per gene and each was tested in quadruplicate. Briefly, siRNA (20 nmol/L) was printed to plate wells and Lipofectamine RNAiMax was added in 20 μL of serum-free media. Complexes were incubated for 30 minutes at ambient temperature prior to adding cells in 20 μL of media containing 2× serum. Cells were incubated at 37° C for 96 hours prior to assaying with CellTiter Glo (Promega). Data were normalized on a scale set between negative (Ambion Silencer Select Negative Control #2, n = 8) and positive (Qiagen All Stars Cell Death Control, n = 8) siRNA controls. The median activity of all 6 siRNAs was taken as a gene-level score. Data are shown in Supplementary Table S2.

Western blotting, immunoprecipitation

The following antibodies were used for Western blotting: Nrf2 (EP1808Y) diluted 1:1,000, SLC7A11 (Cell Signaling Technology, #12691) diluted 1:1,000, actin (Cell Signaling Technology, 5125S) diluted 1:5,000, IGF1R (Cell Signaling Technology, 9750S), ERBB3 (Cell Signaling Technology, 12708S), PRDX1 (Cell Signaling Technology, 5499S), TXN (Cell Signaling Technology, 2429S), TXNRD (Cell Signaling Technology, 15140S) all diluted 1:1,000, donkey anti-rabbit (NA9340, GE-Healthcare) diluted 1:5,000, goat antimouse diluted 1:5,000. Monoclonal Anti-HA-agarose antibody (A2095, Sigma) was used for immunoprecipitating HA-tagged ubiquitin. Cells were grown in on tissue culture plates, lysed on ice using RIPA buffer, spun at 14,000 rpm for 20 minutes, and the supernatant diluted in SDS-containing sample buffer. Twenty micrograms of total lysate was run on either 4% to 12% or 4% to 20% Tris-glycine gels, and transferred to polyvinylidene difluoride membranes using the iBlot transfer apparatus (Life Technologies). Primary and secondary antibodies were diluted in 3% dried milk or 5% BSA, and chemiluminescence detected using a fluorchem HD2 system (ProteinSimple). Precision Plus molecular weight markers were from Bio-Rad.

Two-dimensional and three-dimensional viability and confluency assays

For two-dimensional (2D) assays, cells were seeded at 2,000 cells/well in 96W plates, treated with 500 ng/mL doxycycline or the appropriate compound, and viability was assessed using CellTiter-Glo reagent (Promega, #G7572), which measures cellular ATP levels as a surrogate for cell number and growth, although changes in cell metabolism could also contribute to this assay. Cell confluency was examined using the IncuCyte Live Cell Imaging system (Essen Bioscience). For soft agar assays, a base layer of 0.8% agar (in RPMI) was dispensed into each well of a 12W plate (1 mL base layer per well), and allowed to solidify at 4°C. Cells were then seeded in 1 mL 0.4% agar (in RPMI) on top of the base layer and allowed to solidify. Finally, 1 mL of medium was added on top of the cell layer and plates were carried in a 37°C incubator/5% CO2. Compounds and/or doxycycline were added to the plates in the final 1 mL of medium, at 3× concentration to account for the total volume of agar and medium. Fresh medium and compound were added as needed. Once colonies reached appropriate size for quantitation, photos were taken and the colonies were stained by adding 100 μL 12 mmol/L MTT per well, and allowed to stain at 37°C. Colonies were then imaged/quantified via Gelcount.

Lung tumor xenograft models

Eleven- to 12-week-old female C.B-17 SCID.beige mice (Charles River Laboratories) were subcutaneously inoculated in the right lateral flank with 10 × 106 A549 cells stably transduced with respective shRNA vectors in 100 μL Hank's Balanced Salt Solution/Matrigel (BD Biosciences) or with 10 × 106 shRNA-expressing H441 cells in 100 μL Hank's Balanced Salt Solution per mouse. All experiments using mice were approved by the Genentech Institutional Animal Care and Use Committee (IACUC). When tumor volume reached approximately 150 to 250 mm3, mice were randomized to receive drinking water containing 1 mg/mL doxycycline (in 5% sucrose) or no doxycycline (5% sucrose alone) ad libitum. The doxycycline was replaced 3 times a week and the sucrose replaced once a week. Tumor volumes were determined using digital calipers (Fred V. Fowler Company, Inc.) using the formula (L × W × W)/2 and plotted as mean tumor volume (mm3) ± SEM. Tumor growth inhibition (%TGI) was calculated as the percentage of the area under the fitted curve (AUC) for the respective dose group per day in relation to the vehicle, such that %TGI = 100 × 1 − (AUC treatment/day)/(AUC vehicle/day). In a separate study, mice with 150 to 250 mm3 tumors were dosed with 1 mg/mL doxycycline for 5 days before the tumors were excised and analyzed by Western blotting for Nrf2 levels. All individuals participating in animal care and use are required to undergo training by the institution's veterinary staff. Any procedures, including handling, dosing, and sample collection mandates training and validation of proficiency under the direction of the veterinary staff prior to performing procedures in experimental in vivo studies. All animals were dosed and monitored according to guidelines from the IACUC on study protocols approved by Genentech's Laboratory Animal Resource Committee at Genentech, Inc.

RNA-seq

A549 and H441 cells were plated overnight and then treated with a final concentration of 0.5 μg/mL doxycycline for 24 or 48 hours. Cells were harvested and total RNA isolated as previously described (12). Samples were prepared and sequenced using MiSeq on an Illumina instrument.

RNA-seq data were aligned to the human reference genome (GRCh37/hg19) using GSNAP version 2013-10-10 (13) (parameters: -M 2 -n 10 -B 2 -i 1 -N 1 -w 200000 -E 1 –pairmax-rna = 200000). Gene expression levels for RefSeq genes were quantified based on counts and RPKM (reads per kilobase of target and million reads sequenced) values.

Cystine uptake assays

Cystine L-[3,3′-14C] (Perkin-Elmer, #NEC845050UC) was used in uptake experiments in A549 cells as previously described (14). A total of 1 × 10e6 cells/well were plated in 6-well plates. Following treatments with either doxycycline (0.5 μg/mL) for 48 hours and/or 50 μmol/L erastin (Sigma-Aldrich, #E7781) for 12 hours.

Measurement of glutathione levels

Cells were seeded onto 6-well dishes overnight at 37°C, 5% CO2using complete growth medium (RPMI1640 + 10% FBS + 1% L-glu) to nearly 100% confluent.

After cells in 6-well dish were washed twice with cold 1× PBS, 0.5 mL of 0.5N perchloric acid was added to each well. The cells were scraped off and the contents were transferred to 1.5 mL eppendorf tubes. Additional 0.5 mL perchloric acid was added to each well to recover any remaining contents and combined. The samples were sonicated for 30 seconds in water sonication bath, then centrifuged at 10,000 rpm at 4°C for 5 minutes, and the supernatants were stored at −80°C before analysis.

Cells were also harvested by trypsin and then counted on Vicell to give estimated cell counts and average diameter cell size for subsequent calculations.

Samples were analyzed by LC/MS-MS. The chromatography was performed on Shimadzu Nexera on a 100 × 4.6 mm, 3 μ, Hypercard column. The sample injection volume was 1 to 3 μL. The compounds were eluted by a linear gradient of 0.1% formic acid to acetonitrile containing 0.1% formic acid in 5 minutes. Mass spectrometry analysis used AB Sciex 5500 in a positive mode. The MRM transitions were 308/179 for glutathione and 311/182 for isotopically labeled glutathione as international standard.

ROS measurements

A549 cells were reverse transfected with Nrf2 or Control nontargeting siGenome pools (Dharmacon) using Dharmafect 3 reagent (Dharmacon). Two days posttransfection, cells were treated with 50 μmol/L erastin (Sigma-Aldrich, #E7781), 1 mmol/L glutathione, 50 μmol/L tert-butyl hydrogen peroxide or DMSO control for 12 hours, then labeled with 4 μmol/L DCF (InVitrogen, #C6827 or Abcam, ab113851) for 30 minutes as per manufacturer protocol. Then, cells were removed from plates and analyzed with a FACS Caliber 4 instrument or a fluorescent plate reader.

CRISPR-modified xenografts

A549.pLENTI6.3-Cas9 cells were transduced with individual sgRNA expressing lentiviral particles, selected for stable integration using 2 μg/μL puromycin for 5 days, expanded, and injected subcutaneously into the left side flank of Nu/Nu mice at 5 × 106 cells per mouse. Primary xenografts were measured via caliper. Fifteen and 30 days after transplantation, 7 mice respectively were sacrificed and primary tumors were manually dissected. Primary tumors were flash frozen in liquid nitrogen. All animal work was performed with approval by the IACUC at Genentech Inc.

KEAP1-mutant lung cancer cells are selectively addicted to Nrf2

We previously used a panel of wild-type (WT) and mutant KEAP1 lung cancer cell lines to define a Nrf2 gene signature (15). Here we wished to explore the consequences of KEAP1 mutations on Nrf2 levels and activity using a panel of lung cancer cell lines. Blotting for Nrf2 levels in KEAP1 and Nrf2 mutant cell lines showed a strong increase in both the cytosol and the nucleus compared with WT lines (Supplementary Fig. S1A), which was independent of changes in Nrf2 mRNA (Supplementary Fig. S1B). These suggested posttranslational regulation of Nrf2 by KEAP1. Indeed, phosphorylated Nrf2 was detected (e.g., Supplementary Fig. S1C), and ubiquitinated Nrf2 could be detected in some WT KEAP1 cell lines following proteasomal inhibition, but not mutant cell lines (Supplementary Fig. S1D and S1E).

We next wanted to examine the consequences of loss of Nrf2 across WT and mutant KEAP1 and Nrf2 cell lines. We therefore established stable cell lines expressing 3 independent Nrf2 shRNAs under the control of doxycycline, as well as 3 independent nontargeting controls (NTCs and full blots shown in Supplementary Figs. S2 and S3). Fig. 1A shows that these Nrf2 shRNAs were effective at reducing Nrf2 protein levels in 5 mutant KEAP1, 2 mutant Nrf2, and 5 WT lung cancer cell lines, as well as in immortalized but nontransformed lung epithelial BEAS2B cells. Upon doxycycline addition, ATP levels (a surrogate for viability) of most cell lines were decreased to varying extents, with the mutant KEAP1 cell lines generally resulting in larger decreases (Fig. 1B), a consequence that was statistically significant (Fig. 1C). Knockdown of Nrf2 by siRNA in a larger panel of lung cancer cell lines confirmed a genotype-dependent effect (Fig. 1D; Supplementary Table S2).

Figure 1.

Nrf2 knockdown decreases ATP levels of mutant KEAP1 cell lines. A, The indicated cell lines were infected with lentiviruses expressing three independent Nrf2 shRNA sequences (sh1, sh3, sh10) and, following puromycin selection, were incubated for 48 hours ±500 ng/mL doxycycline. Nrf2 levels were detected by Western blotting. B, The cell lines shown in A were incubated ±500 ng/mL doxycycline for 7 days and then cell ATP levels assessed using CellTiter-Glo. Each circle represents a unique gRNA and is the average of 6 technical replicates, and values were normalized to the average percentage viability (as assessed using CellTiter-Glo assays to measure cellular ATP content) of three independent NTCs + dox. The average value for the three independent target shRNAs is shown by the boxplot. C, The averages from the three target shRNAs in B for each WT and mutant KEAP1 cell line are plotted, and significance was calculated using a Student t test. D, The ATP levels of 39 lung cancer cell lines following treatment with Nrf2 siRNA relative to NTC treatment. Cells are grouped by KEAP1 genotype. Significance was calculated using a Student t test. E, Mice were implanted with A549 or H441 cells expressing Nrf2 sh10. When tumors reached ∼200 mm3, 1 mg/mL doxycycline (D) or 5% sucrose (S) was added to the drinking water. Five days later, tumor extracts were Western blotted for Nrf2. The asterisk indicates a putative nonspecific band that is seen using the Nrf2 antibody in some cell lines. F, Mice were implanted with A549 or H441 cell lines expressing Nrf2sh10. When tumors reached ∼200 mm3, mice were randomized into groups of 10, and either 1 mg/mL doxycycline or 5% sucrose added to the drinking water. Tumors were measured over a 28-day period. Error bars, SEM (n = 10). Differences between doxycycline and sucrose treated animals were assessed using Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 1.

Nrf2 knockdown decreases ATP levels of mutant KEAP1 cell lines. A, The indicated cell lines were infected with lentiviruses expressing three independent Nrf2 shRNA sequences (sh1, sh3, sh10) and, following puromycin selection, were incubated for 48 hours ±500 ng/mL doxycycline. Nrf2 levels were detected by Western blotting. B, The cell lines shown in A were incubated ±500 ng/mL doxycycline for 7 days and then cell ATP levels assessed using CellTiter-Glo. Each circle represents a unique gRNA and is the average of 6 technical replicates, and values were normalized to the average percentage viability (as assessed using CellTiter-Glo assays to measure cellular ATP content) of three independent NTCs + dox. The average value for the three independent target shRNAs is shown by the boxplot. C, The averages from the three target shRNAs in B for each WT and mutant KEAP1 cell line are plotted, and significance was calculated using a Student t test. D, The ATP levels of 39 lung cancer cell lines following treatment with Nrf2 siRNA relative to NTC treatment. Cells are grouped by KEAP1 genotype. Significance was calculated using a Student t test. E, Mice were implanted with A549 or H441 cells expressing Nrf2 sh10. When tumors reached ∼200 mm3, 1 mg/mL doxycycline (D) or 5% sucrose (S) was added to the drinking water. Five days later, tumor extracts were Western blotted for Nrf2. The asterisk indicates a putative nonspecific band that is seen using the Nrf2 antibody in some cell lines. F, Mice were implanted with A549 or H441 cell lines expressing Nrf2sh10. When tumors reached ∼200 mm3, mice were randomized into groups of 10, and either 1 mg/mL doxycycline or 5% sucrose added to the drinking water. Tumors were measured over a 28-day period. Error bars, SEM (n = 10). Differences between doxycycline and sucrose treated animals were assessed using Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

We next determined the consequence of Nrf2 knockdown in tumor xenografts. The KEAP1-mutant A549 cell line and the KEAP1 WT H441 cell lines expressing dox-inducible Nrf2 shRNAs were implanted into the flanks of female SCID mice. Fig. 1E shows the effective knockdown of Nrf2 in doxycycline treated mice in both tumors (full blots shown in Supplementary Fig. S4). Fig. 1F shows that Nrf2 knockdown in the KEAP1-mutant A549 line shows a dramatic effect on tumor growth, resulting in complete tumor regression in 5/10 tumors. In contrast, the effect on H441 growth was more modest, showing a 37% reduction in tumor growth, with all animals displaying maintained tumor burden.

Nrf2 requirement depends on growth conditions and is independent of glutathione metabolism

To understand the dramatic contrast, we observed between the effects of Nrf2 knockdown on tumor propagation in xenografts versus growth on plastic we tested several additional cell culture environments that might underlie differences between these conditions. Nrf2 knockdown in cells grown in low adherence plates, low oxygen (0.5%), or a combination of both, showed similar consequences to cells grown on plastic (Supplementary Fig. S5A). In contrast, the growth of KEAP1-mutant cell lines upon Nrf2 knockdown was severely compromised when cultured in soft agar (Fig. 2A and B, NTC treatments shown in Supplementary Fig. S5B), on micropatterned plastic films (“Scivax,” Supplementary Fig. S6A and S6B), or in methyl cellulose (Supplementary Fig. S6C). We therefore used growth in soft agar to characterize the consequences of Nrf2 knockdown in more detail. Although knockdown of Nrf2 completely abolished colony formation in 3 KEAP1-mutant cell lines, it had almost no effect in H1048 and H441, 2 WT KEAP1 lung cancer cell lines (Fig. 2A and B). We also addressed the role of the glutathione pathway in the response to Nrf2 knockdown, as this pathway has previously been shown to mediate survival properties facilitated by high Nrf2 activity (10). Knockdown of Nrf2 decreased reduced glutathione levels (Fig. 2C). Although addition of reduced glutathione generally increased the ability of all tested cell lines to form colonies in soft agar, it was unable to rescue the consequences of Nrf2 knockdown. Similar negative results were seen with the more cell permeable glutathione precursor N-acetyl cysteine (Supplementary Fig. S6D). Exogenously added reduced glutathione was able to reduce reactive oxygen levels, as measured by dichlorofluorescein staining (Fig. 2D).

Figure 2.

Requirement for Nrf2 under anchorage independent growth is independent of exogenously added reduced glutathione. A, The indicated cell lines stably expressing Nrf2sh1 were plated in soft agar and treated with either vehicle, 500 ng/mL doxycycline (dox), or 1 mmol/L reduced glutathione (GSH) as indicated for the duration of the experiment. Representative areas of the plate were photographed. B, Quantification of the colonies shown in A from biological triplicate wells. Error bars represent SD from triplicate wells, and P values are from a Student t test. C, GSH levels were measured in A549 and H1437 cells in the absence or presence of 500 ng/mL doxycycline. Error bars, SD (n = 4). D, ROS levels in A549 cells under the conditions shown as measured using 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA). Error bars, SD from triplicate wells.

Figure 2.

Requirement for Nrf2 under anchorage independent growth is independent of exogenously added reduced glutathione. A, The indicated cell lines stably expressing Nrf2sh1 were plated in soft agar and treated with either vehicle, 500 ng/mL doxycycline (dox), or 1 mmol/L reduced glutathione (GSH) as indicated for the duration of the experiment. Representative areas of the plate were photographed. B, Quantification of the colonies shown in A from biological triplicate wells. Error bars represent SD from triplicate wells, and P values are from a Student t test. C, GSH levels were measured in A549 and H1437 cells in the absence or presence of 500 ng/mL doxycycline. Error bars, SD (n = 4). D, ROS levels in A549 cells under the conditions shown as measured using 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA). Error bars, SD from triplicate wells.

Close modal

Induction of ROS is not sufficient to decrease viability of KEAP1-mutant lung cancer cells

To explore the effects of the glutathione pathway in Nrf2 responses in more detail, we monitored the expression and activity of the xCT glutamate/cysteine antiporter, 1 of the rate-limiting steps in glutathione synthesis. As expected from a direct Nrf2 target gene (15–17), SLC7A11 expression was reduced following Nrf2 knockdown (Fig. 3A), causing a decrease in cystine uptake (Fig. 3B). Concurrent with this, Nrf2 knockdown also caused a large increase in ROS levels (Fig. 3C). To determine whether inhibition of SLC7A11 expression and cystine uptake contributed to decreased viability following Nrf2 knockdown, we inhibited xCT function using erastin (18). This compound inhibited cystine uptake (Fig. 3B), and induced large increases in oxidative stress (Fig. 3C). However, this was not sufficient to decrease the viability of the KEAP1-mutant cell line A549 (Fig. 3D), or indeed most KEAP1-mutant cell lines (Supplementary Fig. S7A). The combination of erastin and Nrf2 knockdown did result in a dramatic decrease in cell viability however (Fig. 3D). Similarly, the glutathione synthase inhibitor BSO or the glutaminase inhibitor BPTES also did not display preferential toxicity for KEAP1-mutant cell lines (Supplementary Fig. S7B and S7C). Therefore, supplementation with glutathione is not sufficient to rescue viability effects induced by Nrf2 knockdown, nor is depletion of glutathione sufficient to decrease viability of KEAP1-mutant cell lines.

Figure 3.

Effects of Nrf2 knockdown on SLC7A11 expression and cystine uptake. A, A549 cells expressing Nrf2 sh10 were treated with vehicle or 500 ng/mL doxycycline (dox) for the indicated time points and were Western blotted using SLC7A11 and actin antibodies. B, A549 cells expressing NTC1 or Nrf2 sh10 were incubated with vehicle or doxycycline for 48 hours, then incubated with 0.5 μCi 14C-Cystine for 20 minutes. Four hundred μmol/L cold cystine and 400 μmol/L erastin were added at uptake (cold cystine) or 20 hours before uptake (erastin) as controls. Cells were lysed and intracellular cystine was measured by liquid scintillation counting. C, ROS levels were measured using H2DCFDA in A549 cells subjected to Nrf2 siRNA, erastin (50 μmol/L), or both for 24 hours. D, A549 cells expressing NTC1 or Nrf2 sh10 were treated with the indicated concentrations of erastin for 72 hours, in the absence or presence of the indicated concentration of doxycycline, and cell ATP levels were measured using CellTiter-Glo.

Figure 3.

Effects of Nrf2 knockdown on SLC7A11 expression and cystine uptake. A, A549 cells expressing Nrf2 sh10 were treated with vehicle or 500 ng/mL doxycycline (dox) for the indicated time points and were Western blotted using SLC7A11 and actin antibodies. B, A549 cells expressing NTC1 or Nrf2 sh10 were incubated with vehicle or doxycycline for 48 hours, then incubated with 0.5 μCi 14C-Cystine for 20 minutes. Four hundred μmol/L cold cystine and 400 μmol/L erastin were added at uptake (cold cystine) or 20 hours before uptake (erastin) as controls. Cells were lysed and intracellular cystine was measured by liquid scintillation counting. C, ROS levels were measured using H2DCFDA in A549 cells subjected to Nrf2 siRNA, erastin (50 μmol/L), or both for 24 hours. D, A549 cells expressing NTC1 or Nrf2 sh10 were treated with the indicated concentrations of erastin for 72 hours, in the absence or presence of the indicated concentration of doxycycline, and cell ATP levels were measured using CellTiter-Glo.

Close modal

A Nrf2 target gene CRISPR screen reveals pathways important for viability

To understand the potential importance of pathways activated due to Nrf2 activation/KEAP1 loss, we performed a pooled, lentiviral CRISPR screen using a library of genes either decreased in expression upon Nrf2 knockdown in A549 cells, and/or elevated in a panel of KEAP1-mutant lung cancer cell lines relative to Keap1 WT (Supplementary Table S3). As distinct consequences were observed following Nrf2 knockdown in 2D, three-dimensional (3D), and xenograft growth conditions, we performed the screen in KEAP1-mutant A549 cells under all 3 environments to determine whether discrete dependencies could be found. The full data for the screen is in Supplementary Table S4. Fig. 4A–C show the 15-day time point for all 3 conditions, illustrating that all 3 screens performed similarly, with gRNAs representing only a small number of genes showing significant drop-out (Table 1). NFE2L2 and its binding partner MAFG were among the most significantly depleted genes, showing that the screen performed as expected. The pentose phosphate pathway genes PGD, G6PD, and TKT, known Nrf2 target genes (19), also showed strong drop-out. However, these genes show depletion in many cell lines in CRISPR screens (20), so they are unlikely to mediate the strong genotype-dependency of Nrf2 knockdown. The other strong hits in the screen were 2 growth factor receptors, IGF1R and ERBB3, and 3 components of a redox signaling relay, PRDX1, TXN, and TXNRD1 (21). These latter 3 genes were selected for inclusion in this CRISPR library due to their elevated expression in KEAP1-mutant lung cancer cell lines, as well as their decreased expression following Nrf2 knockdown in KEAP1-mutant A549 cells (Supplementary Table S3; Supplementary Fig. S8A–S8F). ERBB3 was selected due to its decreased expression following Nrf2 knockdown (although its expression was also elevated in KEAP1-mutant lung cancer cell lines (Supplementary Fig. S8G, S8H) and IGF1R was selected due to its elevated expression in KEAP1-mutant cell lines (although its expression was also decreased following Nrf2 knockdown (Supplementary Fig. S8I, S8J).

Figure 4.

TXN, PRDX1, and TXNRD1 are required for KEAP1-mutant lung cancer cells under competitive growth conditions. A, A549 cells were infected with lentivirus (0.3 MOI at ×1,000 coverage) expressing a gRNA library comprising 481 Nrf2/KEAP1 target genes and 37 control genes. Puromycin-resistant cells were then plated into 2D plastic tissue culture flasks, grown in methyl cellulose, or implanted into nude mice. After various time points, cells were collected, genomic DNA was isolated and PCR amplified, and gRNAs identified by Next Gen sequencing. Average gRNA expression per gene (log2-fold change relative to 2D, day 0) at the 15-day time point for 2D and 3D growth is shown. B and C, Same as A, but plots for 2D vs. xenograft and 3D vs. xenograft are displayed. D, Parental A549 cells were plated 1:1 with A549 cells expressing either NTC or Nrf2sh10 shRNAs and treated with 500 ng/mL doxycycline. Cells were imaged for total and red cell confluence 5 days later. E, Parental A549 cells were mixed 1:1 with A549 cells expressing the indicated shRNAs and treated with 500 ng/mL doxycycline. Cells were split after 5 days and red (left) and total (right) cell confluence were quantified over the following 5 days.

Figure 4.

TXN, PRDX1, and TXNRD1 are required for KEAP1-mutant lung cancer cells under competitive growth conditions. A, A549 cells were infected with lentivirus (0.3 MOI at ×1,000 coverage) expressing a gRNA library comprising 481 Nrf2/KEAP1 target genes and 37 control genes. Puromycin-resistant cells were then plated into 2D plastic tissue culture flasks, grown in methyl cellulose, or implanted into nude mice. After various time points, cells were collected, genomic DNA was isolated and PCR amplified, and gRNAs identified by Next Gen sequencing. Average gRNA expression per gene (log2-fold change relative to 2D, day 0) at the 15-day time point for 2D and 3D growth is shown. B and C, Same as A, but plots for 2D vs. xenograft and 3D vs. xenograft are displayed. D, Parental A549 cells were plated 1:1 with A549 cells expressing either NTC or Nrf2sh10 shRNAs and treated with 500 ng/mL doxycycline. Cells were imaged for total and red cell confluence 5 days later. E, Parental A549 cells were mixed 1:1 with A549 cells expressing the indicated shRNAs and treated with 500 ng/mL doxycycline. Cells were split after 5 days and red (left) and total (right) cell confluence were quantified over the following 5 days.

Close modal
Table 1.

All genes whose aggregate gRNAs showed a significant (P < 0.05) reduction in abundance at the 15-day time-point in any 1 of the 3 growth conditions

All genes whose aggregate gRNAs showed a significant (P < 0.05) reduction in abundance at the 15-day time-point in any 1 of the 3 growth conditions
All genes whose aggregate gRNAs showed a significant (P < 0.05) reduction in abundance at the 15-day time-point in any 1 of the 3 growth conditions

We created stable A549 cell lines expressing 7 independent dox-regulated hairpins against PRDX1, TXN, and TXNRD1, and chose the most effective 2 for further study (Supplementary Fig. S9A). Despite effective knockdown, reducing the expression of these proteins had minimal consequences on cell proliferation, either in 5-day CellTiter-Glo assays, 12-day clonogenic growth assays, or 14-day soft agar assays (Supplementary Figs. S9B–S9E). siRNA knockdown of PRDX1 also showed minimal effect on growth in a 5-day CellTiter-Glo assay in a panel of 39 lung cancer cell lines (Supplementary Fig. S9F; Supplementary Table S2). One difference between these validation assays and the CRISPR screen is that the screen is performed under competitive growth conditions where the majority of the neighboring cells do not lack these proteins. Therefore, we tested whether knockdown of these genes could reduce growth of cells when cultured together with parental A549 cells. Induction of shRNA expression is accompanied by expression of RFP in these vectors (22). The area of red expressing cells can therefore be used as an indicator for shRNA expressing cells, and measured independently from total cell area. As a positive control, Fig. 4D shows that induction of Nrf2 shRNA reduces the red area compared with the negative control NTC shRNA. A modest reduction in competitive proliferation could be detected for PRDX1, TXN, and TXNRD1 shRNAs 5 days following induction (Supplementary Fig. S10A), but this was more dramatic when these cells were split and then monitored for a further 5 days (Fig. 4E). A similar effect was also seen in H460 cells; little effect of PRDX1, TXN, or TXNRD1 knockdown on cellular ATP levels, clonogenic, or anchorage-independent growth (Supplementary Figs. S10B–S10E), but decreased representation of knockdown cells in competitive growth assays (Supplementary Figs. S11A). The disconnect between competitive and noncompetitive growth was also confirmed using a completely orthogonal method to induce gene ablation, using gRNAs against the target genes of interest (Supplementary Fig. S11B–S11D).

KEAP1-mutant lung cancer cells are selectively dependent on IGF1R signaling for anchorage-independent growth

Another class of hits were the growth factor receptors ERBB3 and IGF1R. ERBB3 knockdown had minimal effects on the ATP levels of A549 cells grown in 2D, or the colony number seen in 3D environments (Supplementary Fig. S12A–S12C). Consistent with this, the use of 2 independent ERBB3 inhibitory antibodies (23) also had no effect on A549 ATP levels in 2D or colony growth in 3D conditions (Supplementary Fig. S13A and S13B). However, we could demonstrate the requirement for ERBB3 in A549 tumor propagation in a xenograft model (Supplementary Fig. S13C), confirming previous observations (24). Administration of the YW57.88.5 antibody showed a 65% reduction in tumor growth, with no effect on body weight (Supplementary Fig. S13C).

Effective inducible shRNAs against IGF1R were designed and established in A549 cells (Fig. 5A). Knockdown of IGF1R did not show strong effects in 5-day CellTiter-Glo assays on plastic (Supplementary Fig. S14A). However, knockdown of IGF1R showed substantial inhibition of colony growth in soft agar (Fig. 5B; Supplementary Fig. S14B). We next tested the potent and selective IGF1R small molecule inhibitor OSI-906/linsitinib (25). Similar to IGF1R knockdown, this compound showed little effect on ATP levels when tested in 3 WT and 3 mutant KEAP1 lung cancer cell lines grown in 2D (Supplementary Fig. S14C). However, this compound was very potent at inhibiting colony growth of A549 cells in soft agar (IC50 ∼20 nmol/L). Moreover, when tested against a large panel of lung cancer cell lines, there was a selective effect of this compound inhibiting colony growth of KEAP1-mutant cell lines (Fig. 5C and D). This is true for cells cultured both in the presence (Fig. 5C and D) and absence (Supplementary Fig. S14D and S14E) of reduced glutathione. A similar selective effect on KEAP1-mutant cell lines when grown under anchorage independent conditions was also seen with an independent IGF1R inhibitor NVP-AEW541 (Supplementary Fig. S14F). Interrogation of the Broad Institute's Project Achilles genome-wide shRNA knockdown and CRISPR datasets support the relationship between KEAP1-mutant lung cancer and sensitivity to IGF1R (Fig. 5E). Moreover, the cell lines that were most sensitive to linsitinib treatment were also the most sensitive to IGF1R depletion (Supplementary Fig. S14G). In support of this, some KEAP1-mutant cell lines [A549 (26, 27), NCI-H838, NCI-H322 (28)] have been shown to be sensitive to IGF1R inhibition in vivo. Therefore, dependence on IGF1R signaling may represent a feature of KEAP1-mutant lung cancers.

Figure 5.

IGF1R signaling is selectively required for proliferation of KEAP1-mutant cells in anchorage independent growth. A, Seven independent doxycycline (Dox)-inducible shRNA's targeting IGF1R were transfected into A549 cells using the PiggyBac system and selected with 2 μg/mL puromycin. Western blots were performed 72 hours following 5 days of treatment with 500 ng/mL doxycycline. B, A549 cells expressing NTC or IGF1R shRNAs were plated into soft agar, then treated ±500 ng/mL doxycycline for 24 hours. Colonies were stained with MTT and counted after 14 days. Error bars, SD from triplicate wells. P values were calculated using Student t test. C, The indicated cell lines were grown in soft agar in the presence of 1 mmol/L reduced glutathione and treated with either vehicle (0.1% DMSO) or increasing concentrations of the IGF1R inhibitor linsitinib. Colonies were stained with MTT and counted when colonies reached sufficient sizes for imaging (10–15 days). KEAP1 WT and mutant cell lines are colored black and red respectively. Error bars, SD of triplicate wells. D, AUCs of data from C were calculated and P values calculated using a Student t test. E, Dependencies of lung cancer cell lines by KEAP1 mutation/deletion following IGF1R deletion (left), or knockdown (right). Data were downloaded from Depmap.org (18Q3 release), and P values are from a Student t test. NS, nonsignificant.

Figure 5.

IGF1R signaling is selectively required for proliferation of KEAP1-mutant cells in anchorage independent growth. A, Seven independent doxycycline (Dox)-inducible shRNA's targeting IGF1R were transfected into A549 cells using the PiggyBac system and selected with 2 μg/mL puromycin. Western blots were performed 72 hours following 5 days of treatment with 500 ng/mL doxycycline. B, A549 cells expressing NTC or IGF1R shRNAs were plated into soft agar, then treated ±500 ng/mL doxycycline for 24 hours. Colonies were stained with MTT and counted after 14 days. Error bars, SD from triplicate wells. P values were calculated using Student t test. C, The indicated cell lines were grown in soft agar in the presence of 1 mmol/L reduced glutathione and treated with either vehicle (0.1% DMSO) or increasing concentrations of the IGF1R inhibitor linsitinib. Colonies were stained with MTT and counted when colonies reached sufficient sizes for imaging (10–15 days). KEAP1 WT and mutant cell lines are colored black and red respectively. Error bars, SD of triplicate wells. D, AUCs of data from C were calculated and P values calculated using a Student t test. E, Dependencies of lung cancer cell lines by KEAP1 mutation/deletion following IGF1R deletion (left), or knockdown (right). Data were downloaded from Depmap.org (18Q3 release), and P values are from a Student t test. NS, nonsignificant.

Close modal

It is important to identify the spectrum of genes regulated by transcription factors that act as oncogenes. Several studies have performed this analysis for Nrf2, predominantly using ChIP-seq and RNAseq-based approaches (7–9). Although modulation of glutathione synthesis is clearly a well-documented and important consequence of Nrf2 activity, we were surprised to find that it does not seem critical for the oncogenic functions of Nrf2 in the context of KEAP1 mutations described in this study. Neither excess reduced glutathione nor its precursor N-acetyl cysteine could rescue the consequences of Nrf2 knockdown on 2D or 3D growth. Similarly, glutathione synthesis or cystine uptake inhibitors did not show a preferential effect on KEAP1-mutant cell lines, despite causing large increases in cellular ROS levels. In addition, several glutathione pathway genes present in the Nrf2 target gene library, did not show any drop-out in any of the conditions used in the CRISPR screen. The glutaminase inhibitor BPTES did show a trend for more activity in KEAP1-mutant lung cancer cell lines (Supplementary Fig. S7B), consistent with greater activity of glutaminase inhibitors reported in KEAP1-deleted mouse lung tumor cells (29, 30). Interestingly, genetic screens have generated conflicting data on the interaction between GLS and KEAP1 mutations in lung cancer cell lines, with CRISPR screens showing no interaction, whereas shRNA screens showing increased dependency (Supplementary Fig. S15A, data from DepMap.org). Whether this represents a genuine discrepancy between knockout versus knockdown, or a reflection of increased number of cell lines in the knockdown study providing a higher powered analysis, remains to be seen. This increased requirement for glutaminase activity may be to elevate glutamate levels to support increased activity of the SLC7A11 glutamate/cysteine antiporter, which is highly expressed in KEAP1-deleted lung tumor cells (15).

Proteins involved in a protein oxidation relay emerged as strong hits in the CRISPR screen of Nrf2/KEAP1 target genes. Peroxiredoxin and thioredoxin are thought to modulate electron flow from cellular oxidants such as hydrogen peroxide to downstream signaling proteins, thereby regulating their activities (21). Exactly which target proteins showing differential oxidation by peroxiredoxin/thioredoxin are important for KEAP1/Nrf2-dependent tumor growth is an important avenue for future investigation. Proteins involved in translation have been shown to be differentially oxidized in a Nrf2-dependent manner (10).

Interestingly, although peroxiredoxin, thioredoxin, and thioredoxin reductase were strong hits from the CRISPR screen, their effects on the growth properties of several cell lines following knockdown or knockout were inconsequential or very modest (Supplementary Figs. S9 and S10). In contrast, when cells depleted of these proteins were grown in the presence of parental cells (a condition that more closely mimics that of the screens), the effects on cell proliferation were more dramatic (Fig. 4; Supplementary Figs. S10 and S11). It is not clear what underlies this discrepancy; it could be that competitive growth is a more sensitive assay that can visualize modest effects on cell proliferation, or there might be some cellular communication between knockdown and parental cells that remains to be understood. Therefore, although PRDX1, TXN, and TXNRD1 validate as hits in the CRISPR screen using competitive growth assays, it is unclear whether inhibition of these targets would be effective in tumors where mutations in KEAP1 are generally clonal.

The requirement of KEAP1-mutant cells for growth factor signals such as ERBB3 and IGF1R was somewhat unexpected. However, it is reminiscent of a recent study showing that increased EGF signaling is also a consequence of Nrf2 activity (10). Interestingly, interrogation of a large number of human lung adenocarcinomas showed that EGFR mutations were the most significantly mutually exclusive genetic alteration in KEAP1 mutated tumors (Supplementary Fig. S15B, S15C). This is consistent with the notion that KEAP1-mutant tumors can supply their own growth factors, obviating the need for activating EGFR mutations. In the case of EGF, this appeared to be through Nrf2-mediated posttranslational cleavage of EGF from its precursor through the activity of ADAM10 (10). For IGF1R, the dependency may arise at least in part from IGFR1 transcription itself, as IGF1R mRNA is elevated in KEAP1-mutant cell lines (Supplementary Fig. S8J) as well as primary tumors (Supplementary Fig. S15D). Validation studies showed that there was a much stronger requirement for IGF1R in 3D versus 2D growth conditions. Similar observations in response to growth factor signaling pathway inhibition have been made recently (31), and more dramatic consequences of IGF1R expression have also been observed in 3D versus 2D environments in breast cancer cell lines (32). Nevertheless, IGF1R gRNAs were effectively depleted in the 2D arm of the CRISPR screen, pointing to sensitivity of this type of screening approach. Although there was no expectation that hits from this screen performed only in a KEAP1-mutant background would show selectivity for KEAP1-mutant cells, IGF1R does appear to show this property. Although IGF1R inhibitors have shown promise in clinical trials, progress has been hampered due to lack of robust biomarkers (33). It is possible that incorporating analysis of KEAP1 and Nrf2 mutation status, in addition to measuring IGF1R levels may help in identifying patients who might benefit most from this type of therapy.

C. Klijn has ownership interest (including patents) in Roche. G. Schaefer has ownership interest (including patents) in Roche. Ryan J. Hartmaier is a senior scientist at Foundation Medicine, is an associate director at AstraZeneca; and has ownership interest (including patents) in Foundation Medicine and AstraZeneca. S.E. Martin reports receiving an other commercial research support from Genentech/Roche and has ownership interest (including patents) in Genentech/Roche. M. Merchant is an associate director at Roche/Genentech and has ownership interest (including patents) in Roche/Genentech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Lee, M. Bagniewska, D. Stokoe

Development of methodology: S. Vartanian, J. Lee, M. Bagniewska, D. Zhang, T. Cuellar, T. Lau, M.R. Costa, B. Haley, D. Stokoe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Vartanian, J. Lee, D. Zhang, J. Tan, S.A. Watson, Y. Liang, D. Kan, R.J. Hartmaier, T. Lau, M.R. Costa, M. Merchant

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Vartanian, J. Lee, C. Klijn, F. Gnad, L. Liu, C. Watanabe, R.J. Hartmaier, T. Lau, S.E. Martin, M. Merchant, D. Stokoe

Writing, review, and/or revision of the manuscript: S. Vartanian, J. Lee, M. Bagniewska, G. Schaefer, L. Liu, D. Kan, M.R. Costa, D. Stokoe

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Lee, M. Bagniewska, G. Schaefer, H. Chen, B. Haley, D. Stokoe

Study supervision: D. Stokoe

We would like to thank the Genentech DNA/RNA sequencing, gCell (cell culture), and gCSI (drug screening) core facilities and Yuxin Liang and Honglin Chen for packaging the virus library. We would also like to thank Shiva Malek and Jeff Settleman for helpful advice throughout this study. Funding for this study was supplied by Genentech Inc. and Foundation Medicine Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Cancer Genome Atlas Research N
. 
Comprehensive molecular profiling of lung adenocarcinoma
.
Nature
2014
;
511
:
543
50
.
2.
Cancer Genome Atlas Research N
. 
Comprehensive genomic characterization of squamous cell lung cancers
.
Nature
2012
;
489
:
519
25
.
3.
Fukutomi
T
,
Takagi
K
,
Mizushima
T
,
Ohuchi
N
,
Yamamoto
M
. 
Kinetic, thermodynamic, and structural characterizations of the association between Nrf2-DLGex degron and Keap1
.
Mol Cell Biol
2014
;
34
:
832
46
.
4.
Hast
BE
,
Cloer
EW
,
Goldfarb
D
,
Li
H
,
Siesser
PF
,
Yan
F
, et al
Cancer-derived mutations in KEAP1 impair NRF2 degradation but not ubiquitination
.
Cancer Res
2014
;
74
:
808
17
.
5.
Tong
KI
,
Kobayashi
A
,
Katsuoka
F
,
Yamamoto
M
. 
Two-site substrate recognition model for the Keap1-Nrf2 system: a hinge and latch mechanism
.
Biol Chem
2006
;
387
:
1311
20
.
6.
Leinonen
HM
,
Kansanen
E
,
Polonen
P
,
Heinaniemi
M
,
Levonen
AL
. 
Role of the keap1-nrf2 pathway in cancer
.
Adv Cancer Res
2014
;
122
:
281
320
.
7.
Malhotra
D
,
Portales-Casamar
E
,
Singh
A
,
Srivastava
S
,
Arenillas
D
,
Happel
C
, et al
Global mapping of binding sites for Nrf2 identifies novel targets in cell survival response through ChIP-Seq profiling and network analysis
.
Nucleic Acids Res
2010
;
38
:
5718
34
.
8.
Chorley
BN
,
Campbell
MR
,
Wang
X
,
Karaca
M
,
Sambandan
D
,
Bangura
F
, et al
Identification of novel NRF2-regulated genes by ChIP-Seq: influence on retinoid X receptor alpha
.
Nucleic Acids Res
2012
;
40
:
7416
29
.
9.
Hirotsu
Y
,
Katsuoka
F
,
Funayama
R
,
Nagashima
T
,
Nishida
Y
,
Nakayama
K
, et al
Nrf2-MafG heterodimers contribute globally to antioxidant and metabolic networks
.
Nucleic Acids Res
2012
;
40
:
10228
39
.
10.
Chio
II
,
Jafarnejad
SM
,
Ponz-Sarvise
M
,
Park
Y
,
Rivera
K
,
Palm
W
, et al
NRF2 promotes tumor maintenance by modulating mRNA translation in pancreatic cancer
.
Cell
2016
;
166
:
963
76
.
11.
Singh
A
,
Boldin-Adamsky
S
,
Thimmulappa
RK
,
Rath
SK
,
Ashush
H
,
Coulter
J
, et al
RNAi-mediated silencing of nuclear factor erythroid-2-related factor 2 gene expression in non-small cell lung cancer inhibits tumor growth and increases efficacy of chemotherapy
.
Cancer Res
2008
;
68
:
7975
84
.
12.
Chirgwin
JM
,
Przybyla
AE
,
MacDonald
RJ
,
Rutter
WJ
. 
Isolation of biologically active ribonucleic acid from sources enriched in ribonuclease
.
Biochemistry
1979
;
18
:
5294
9
.
13.
Wu
TD
,
Nacu
S
. 
Fast and SNP-tolerant detection of complex variants and splicing in short reads
.
Bioinformatics
2010
;
26
:
873
81
.
14.
Murphy
TH
,
Miyamoto
M
,
Sastre
A
,
Schnaar
RL
,
Coyle
JT
. 
Glutamate toxicity in a neuronal cell line involves inhibition of cystine transport leading to oxidative stress
.
Neuron
1989
;
2
:
1547
58
.
15.
Goldstein
LD
,
Lee
J
,
Gnad
F
,
Klijn
C
,
Schaub
A
,
Reeder
J
, et al
Recurrent loss of NFE2L2 exon 2 is a mechanism for Nrf2 pathway activation in human cancers
.
Cell reports
2016
;
16
:
2605
17
.
16.
Sasaki
H
,
Sato
H
,
Kuriyama-Matsumura
K
,
Sato
K
,
Maebara
K
,
Wang
H
, et al
Electrophile response element-mediated induction of the cystine/glutamate exchange transporter gene expression
.
J Biol Chem
2002
;
277
:
44765
71
.
17.
DeNicola
GM
,
Chen
PH
,
Mullarky
E
,
Sudderth
JA
,
Hu
Z
,
Wu
D
, et al
NRF2 regulates serine biosynthesis in non-small cell lung cancer
.
Nat Genet
2015
;
47
:
1475
81
.
18.
Dixon
SJ
,
Lemberg
KM
,
Lamprecht
MR
,
Skouta
R
,
Zaitsev
EM
,
Gleason
CE
, et al
Ferroptosis: an iron-dependent form of nonapoptotic cell death
.
Cell
2012
;
149
:
1060
72
.
19.
Mitsuishi
Y
,
Taguchi
K
,
Kawatani
Y
,
Shibata
T
,
Nukiwa
T
,
Aburatani
H
, et al
Nrf2 redirects glucose and glutamine into anabolic pathways in metabolic reprogramming
.
Cancer Cell
2012
;
22
:
66
79
.
20.
Wang
T
,
Birsoy
K
,
Hughes
NW
,
Krupczak
KM
,
Post
Y
,
Wei
JJ
, et al
Identification and characterization of essential genes in the human genome
.
Science
2015
;
350
:
1096
101
.
21.
Netto
LE
,
Antunes
F
. 
The roles of peroxiredoxin and thioredoxin in hydrogen peroxide sensing and in signal transduction
.
Mol Cells
2016
;
39
:
65
71
.
22.
Meerbrey
KL
,
Hu
G
,
Kessler
JD
,
Roarty
K
,
Li
MZ
,
Fang
JE
, et al
The pINDUCER lentiviral toolkit for inducible RNA interference in vitro and in vivo
.
Proc Natl Acad Sci U S A
2011
;
108
:
3665
70
.
23.
Schaefer
G
,
Haber
L
,
Crocker
LM
,
Shia
S
,
Shao
L
,
Dowbenko
D
, et al
A two-in-one antibody against HER3 and EGFR has superior inhibitory activity compared with monospecific antibodies
.
Cancer Cell
2011
;
20
:
472
86
.
24.
Mirschberger
C
,
Schiller
CB
,
Schraml
M
,
Dimoudis
N
,
Friess
T
,
Gerdes
CA
, et al
RG7116, a therapeutic antibody that binds the inactive HER3 receptor and is optimized for immune effector activation
.
Cancer Res
2013
;
73
:
5183
94
.
25.
Mulvihill
MJ
,
Cooke
A
,
Rosenfeld-Franklin
M
,
Buck
E
,
Foreman
K
,
Landfair
D
, et al
Discovery of OSI-906: a selective and orally efficacious dual inhibitor of the IGF-1 receptor and insulin receptor
.
Future Medicinal Chemistry
2009
;
1
:
1153
71
.
26.
Goetsch
L
,
Gonzalez
A
,
Leger
O
,
Beck
A
,
Pauwels
PJ
,
Haeuw
JF
, et al
A recombinant humanized anti-insulin-like growth factor receptor type I antibody (h7C10) enhances the antitumor activity of vinorelbine and anti-epidermal growth factor receptor therapy against human cancer xenografts
.
Int J Cancer
2005
;
113
:
316
28
.
27.
Schanzer
JM
,
Wartha
K
,
Croasdale
R
,
Moser
S
,
Kunkele
KP
,
Ries
C
, et al
A novel glycoengineered bispecific antibody format for targeted inhibition of epidermal growth factor receptor (EGFR) and insulin-like growth factor receptor type I (IGF-1R) demonstrating unique molecular properties
.
J Biol Chem
2014
;
289
:
18693
706
.
28.
Wang
Y
,
Pachter
JA
,
Wang
Y
,
Liu
M
. 
Biomarkers for pre-selection of patients for anti-IGF1R therapy
.
Patent US7811562B2
2010
;
Fig. 2
.
29.
Romero
R
,
Sayin
VI
,
Davidson
SM
,
Bauer
MR
,
Singh
SX
,
LeBoeuf
SE
, et al
Keap1 loss promotes Kras-driven lung cancer and results in dependence on glutaminolysis
.
Nat Med
2017
;
23
:
1362
8
.
30.
Sayin
VI
,
LeBoeuf
SE
,
Singh
SX
,
Davidson
SM
,
Biancur
D
,
Guzelhan
BS
, et al
Activation of the NRF2 antioxidant program generates an imbalance in central carbon metabolism in cancer
.
eLife
2017
;
6
.
pii
:
e28083
.
31.
Riedl
A
,
Schlederer
M
,
Pudelko
K
,
Stadler
M
,
Walter
S
,
Unterleuthner
D
, et al
Comparison of cancer cells in 2D vs 3D culture reveals differences in AKT-mTOR-S6K signaling and drug responses
.
J Cell Sci
2017
;
130
:
203
18
.
32.
Zhang
Y
,
Moerkens
M
,
Ramaiahgari
S
,
de Bont
H
,
Price
L
,
Meerman
J
, et al
Elevated insulin-like growth factor 1 receptor signaling induces antiestrogen resistance through the MAPK/ERK and PI3K/Akt signaling routes
.
Breast Cancer Res
2011
;
13
:
R52
.
33.
Iams
WT
,
Lovly
CM
. 
Molecular pathways: clinical applications and future direction of insulin-like growth factor-1 receptor pathway blockade
.
Clin Cancer Res
2015
;
21
:
4270
7
.