Activation of the NRF2 antioxidant program generates an imbalance in central carbon metabolism in cancer

1Cancer cells rewire their metabolism to sustain uncontrolled cellular proliferation. Alterations to cellular metabolism are required to meet the increased energetic and biosynthetic demands of cancer cells. Warburg first observed that cancer cells have an increased demand for glucose and increased glycolysis compared to non-cancer cells. It is now increasingly clear that during transformation and tumorigenesis cancer cells require additional metabolic alterations, which result in dependencies on exogenous metabolites such as the non-essential amino acid glutamine. Although cells can readily synthesize glutamine, several studies have demonstrated that some cancer cells rely on exogenous glutamine and require supplementation for survival. Glutamine is the most abundant amino acid found in blood and can be used for a variety of purposes including energy production and generation of biosynthetic intermediates by fueling the TCA cycle. Glutaminase catalyzes the conversion of glutamine to glutamate and is the rate-limiting enzyme for glutaminolysis. Glutamine-derived glutamate is also critical for the generation of the most abundant cellular metabolite, the antioxidant glutathione (GSH), making glutamine metabolism important for maintaining redox homeostasis. Numerous oncogenes drive increased glutamine metabolism, hence targeting glutaminolysis in particular genetic subtypes of cancer has emerged as an attractive therapeutic option. Despite extensive studies on the role of glutamine in KRAS-driven tumors, inhibiting this pathway has led to limited responses in both pre-clinical models and ongoing clinical trials; highlighting the importance of identifying genetic subtypes of cancer that may have a greater dependency on glutamine.

2During tumor initiation and progression, increased metabolic output as well as changing environmental conditions result in high oxidative stress in the form of reactive oxygen species (ROS). The role of ROS in the development and progression of cancer has been the subject of considerable study and debate. Emerging evidence supports the idea that cancers increase antioxidant capacity as a stress response mechanism, suggesting that high ROS levels may constitute a barrier to tumor progression. To maintain oxidative homeostasis, cancer cells increase their antioxidant capacity. Under normal physiologic conditions, the KEAP1/NRF2 signaling axis is responsible for coordinating the cell’s antioxidant defenses in response to ROS. KEAP1 binds to NRF2 and enables the CUL3-mediated ubiquitination and proteasomal degradation of NRF2. When ROS increase, KEAP1 oxidation leads to a conformational change enabling the dissociation and stabilization of NRF2, which translocates to the nucleus to induce the transcription of a battery of antioxidant genes, including enzymes regulating the synthesis of the major cellular antioxidant GSH, and enzymes involved in the reduction of oxidized GSH and thioredoxin (TXN). The activation of the antioxidant response requires the flux of glucose and glutamine-derived carbons and cofactors (e.g. NAD/NADH, NADP/NADPH) to be diverted to antioxidant production. Increased flux of nutrients to antioxidant production may limit nutrient availability for other biosynthetic pathways.

3However, in the context of chronic ROS exposure, post-transcriptional or post-translational signaling can stabilize NRF2, or epigenetic and genetic events can lead to NRF2 stabilization, resulting in the chronic activation of NRF2 and its downstream antioxidant program. Interestingly, KEAP1 is the third most frequently mutated gene in lung adenocarcinoma (LUAD) and often co-occurs with oncogenic mutations in KRAS. Activation of the NRF2-driven antioxidant axis may create a unique set of metabolic requirements necessary to sustain this increased antioxidant capacity, creating potential for novel therapeutic vulnerabilities in aggressive lung cancers.

4Here we demonstrate that KEAP1 loss of function (LOF) mutations drive increased dependency on glutamine in both mouse and human KRAS-driven LUAD cell lines. We show that KEAP1 mutant cells have decreased intracellular glutamate pools through increased glutamate consumption for GSH synthesis and by exporting glutamate through the antiporter xCT in exchange for cystine. The low intracellular pools of glutamate lead to increased sensitivity to glutamine deprivation and glutaminase inhibition in an xCT dependent fashion. Using a small molecule activator of NRF2, we provide evidence that acute NRF2 activation is sufficient to rewire cellular metabolism, similar to that of KEAP1 mutant cells, and leads to glutamine dependency due to a basal deficiency in anaplerosis. Finally, we show that this is a phenomenon that occurs across multiple types of cancers with KEAP1 mutations, and demonstrate the importance of sub-stratifying patients based on genotype to maximize therapeutic efficacy of glutaminase inhibition in clinical trials and pre-clinical models where responses have been previously limited.

5To study the role of Keap1 mutations in rewiring lung cancer metabolism, we generated isogenic Kras-driven, Trp53 null (KrasG12D/+; p53-/-; hereafter KP) cells with wild-type (KP) or LOF mutations in Keap1 (KPK) using CRISPR/Cas9-editing. We observed that KPK cells had increased production of GSH compared to KP cells (Figure 1A), as a result of increased levels of glutamate-cysteine ligase catalytic subunit, Gclc, which along with glutamate-cysteine ligase modifier subunit, Gclm synthesize GSH (Figure 1—figure supplement 1A and B). As expected, elevated levels of GSH in KPK cells corresponded with decreased levels of cellular ROS (Figure 1—figure supplement 1C). Both KP and KPK cells had similar growth rates under basal conditions (Figure 1—figure supplement 1D), however when challenged with oxidative stress, KPK cells were more resistant compared to KP cells (Figure 1B and C, Figure 1—figure supplement 1E).

6GSH is a tri-peptide composed of glutamate, cysteine, and glycine. Therefore, we conjectured that high levels of GSH in KPK cells might result in increased demand for glutamate and consequently KPK cells would be more dependent on glutamine-derived glutamate compared to KP cells. Indeed, decreasing glutamine concentrations in the media substantially decreased proliferation of KPK cells with minimal effects on KP cells (Figure 1—figure supplement 1F). Glutaminase catalyzes the conversion of glutamine to glutamate and is the rate-limiting enzyme for glutaminolysis. Here we use a small molecule inhibitor of glutaminase, CB-839, which is currently in clinical phase I trials for multiple cancer types, including LUAD. CB-839 blocks glutamine catabolism and decreases the availability of glutamate in the cell for GSH antioxidant production and the generation of α-ketoglutarate (αKG) as a carbon source for the TCA cycle and other biosynthetic intermediates. Consistent with glutamine deprivation, KPK cells had increased sensitivity to CB-839, (Figure 1—figure supplement 1G). In line with CB-839 decreasing glutamate availability, cellular glutathione levels were reduced after CB-839 treatment (Figure 1D). Increased sensitivity to both glutamine deprivation as well as glutaminase inhibition suggests that Keap1 mutant lung cancers depend on glutamine-derived glutamate to support GSH synthesis.

7To elucidate the mechanism of glutamine dependency in Keap1 mutant cells resulting from increased cellular glutamate demand, we measured both intra- and extra-cellular glutamate levels. Surprisingly, KPK cells had lower intracellular but higher extracellular levels of glutamate when compared to KP cells (Figure 1E and F). Glutamate is required for the synthesis of GSH, but is also necessary for the import of cystine via the xc- antiporter system (xCT), which exchanges glutamate for cystine across the plasma membrane in order to support intracellular cysteine pools for antioxidant production (Figure 2A). Previously, expression of xCT has been linked with glutamine sensitivity and to antagonize glutamine anaplerosis. xCT is a heterodimer of Slc7a11, a bona-fide Nrf2 target gene, and Slc3a2. Consistent with prior findings, gene expression analysis revealed dramatically higher levels of Slc7a11 in KPK cells compared to KP controls (Figure 2B). We hypothesized that the Nrf2-driven increase in expression of Slc7a11 and Gclc, and higher levels of GSH in KPK cells may result in export of glutamate in exchange for import of cystine (Figure 2A). To determine if xCT/Slc7a11 is responsible for the KPK cell dependency on exogenous glutamine, we treated cells with a small molecule inhibitor of xCT, Erastin. Indeed, Erastin treatment was sufficient to rescue the growth suppressive effects of both glutamine deprivation and CB-839 treatment in KPK cells with little effect on KP cells (Figure 2C and Figure 2—figure supplement 1A). Furthermore, through the examination of a panel of human LUAD cell lines, we demonstrate that KEAP1 mutants are robustly sensitive to glutaminase inhibition (Figure 2—figure supplement 1B). Importantly, pre-treatment with Erastin overcomes sensitivity to glutaminase inhibition in KEAP1 mutant lines (Figure 2—figure supplement 1B). xCT/Slc7a11 functions based on the intra- and extra- cellular gradients of cystine and glutamate in a concentration dependent manner. Therefore, we asked whether we could rescue glutamine dependency of KPK cells by forcing the export of cystine and the import of glutamate by modulating the concentration of these two amino acids in the culture media. By either raising the levels of glutamate or lowering the levels of cystine we were able to rescue proliferation in KPK cells after CB-839 treatment (Figure 2D and E). To further validate that CB-839 sensitivity in KPK cells is dependent on xCT/Slc7a11, we tested multiple doxycycline inducible mirE based shRNAs targeting Slc7a11 (Figure 2—figure supplement 2A and B). In line with expectations, knockdown of Slc7a11 by two independent shRNAs rescued KPK cells from both CB-839 treatment as well as glutamine deprivation while shCTRL had no effect (Figure 2F and Figure 2—figure supplement 2C). In addition, modulating the levels of glutamate or cystine in the media did not provide additional benefits to depletion of Slc7a11 in regards to sensitivity to CB-839 or glutamine deprivation (Figure 2F and Figure 2—figure supplement 2C). Taken together, these results indicate that increased expression of xCT/Slc7a11 through Nrf2 hyperactivation in Keap1 mutant cells is necessary for the increased glutamine dependency and sensitivity to glutaminase inhibitors.

8Keap1 mutations that activate Nrf2 result in elevated GSH production thereby leaving Keap1 mutant cells highly dependent on glutaminolysis and susceptible to glutaminase inhibition in vitro. However, the extracellular environment can greatly affect cellular metabolism, and culturing cells in vitro versus in vivo can significantly alter their metabolic requirements. To assess the importance of xCT/Slc7a11 in driving sensitivity to glutaminase inhibition in KPK cells in vivo, we implanted KP and KPK cells that expressed inducible shRNA constructs against Slc7a11 or CTRL subcutaneously in mice. After tumors were established, mice were randomized to groups to receive oral CB-839 or vehicle twice daily and to receive a doxycycline or control diet to induce shRNA expression. CB-839 treatment suppressed tumor growth only in KPK cells, but had no effect on KP tumors (Figure 2G and H and Figure 2—figure supplement 2D and E). Knockdown of Slc7a11 completely abrogated the effect of CB-839 on KPK tumors (Figure 2H and Figure 2—figure supplement 2E). These data strongly support the notion that Keap1 mutations result in a metabolic liability to glutaminase inhibition in vivo driven by increased expression of xCT/Slc7a11 in KPK cells.

9Keap1 mutant cells are dependent on glutamine due to basally decreased intracellular pools of glutamate, partially because of increased secretion via xCT and increased production of GSH. Glutamate can fuel both GSH synthesis, or via transamination be converted to α-KG, a carbon source for the TCA cycle (Figure 3A). To assess whether limited glutamate availability for GSH synthesis leads to proliferation defects, we pretreated KPK cells with the antioxidants trolox (Vitamin E analog) or N-acetyl cysteine (NAC). However, supplementation with these antioxidants did not rescue proliferation after glutamine deprivation or glutaminase inhibition in KPK cells (Figure 3—figure supplement 1A and Figure 3B). We hypothesized that high antioxidant production in KPK cells driven by Nrf2 activation increases cellular demand for glutamate. Glutamate is required in order to synthesize glutathione while limiting availability for other cellular processes such as fueling the TCA cycle. Consistent with this hypothesis, pretreatment with: (1) dimethyl 2-oxoglutarate (DMG), a cell permeable α-KG analogue, or (2) pyruvate, a glucose-derived TCA cycle carbon source, rescued sensitivity to CB-839 and glutamine deprivation in KPK cells (Figure 3B and Figure 3—figure supplement 1A). These results suggest that sensitivity to glutaminase inhibition is due to decreased glutamate availability for anaplerosis or other biosynthetic reactions, not synthesis of the antioxidant GSH. In line with this, reducing GSH levels in KPK cells with L-Buthionine sulfoximine (BSO) did not alter sensitivity to CB-839 (Figure 3—figure supplement 1B and C).

10Our data suggests that hyperactivation of Nrf2 by Keap1 mutation genetically reprograms cells to produce high levels of antioxidants at the expense of contributing carbon to replenish the TCA cycle. Consistent with this data, KPK cells have reduced total levels of TCA cycle intermediates (Figure 3C) and supplementation with either 2 mM pyruvate or DMG increased total levels of TCA cycle intermediates in KPK cells equivalent to or in excess to those in KP cells (Figure 3C). To determine the fate of glutamine, we used a stable-isotope labeled glutamine tracer (U-C13-L-glutamine) (Figure 3D). We observed that KPK cells incorporate less carbon from glutamine into the TCA cycle as compared to KP cells (Figure 3E). Taken together, these data strongly suggest that KPK cells have a basal deficiency in glutamine metabolism that contributes to their increased sensitivity to loss of glutamine-derived glutamate for anaplerosis. A functional output of the TCA cycle is oxidative phosphorylation in the mitochondria, as the TCA cycle generates NADH and FADH2 that will be used by the mitochondrial electron transport system to produce ATP. Measuring mitochondrial respiration revealed that KPK cells have reduced oxygen consumption rates (OCR) compared to KP cells indicating that Keap1 mutants have lower mitochondrial respiration basally (Figure 3F). The overall decrease in respiration in KPK compared to KP cells is in line with decreased TCA cycle output (Figure 3C). However, supplementation with pyruvate or DMG did not significantly increase basal OCR in KPK cells (Figure 3—figure supplement 3A). This is likely a result of KPK cells having reduced reserve respiratory capacity as compared to KP cells (Figure 3—figure supplement 3B) in normal media conditions. As expected, CB-839 treatment reduced respiration in both KP and KPK cells, however, the drop in OCR after CB-839 administration was dramatically more pronounced in KPK cells (−50.8%) compared to KP cells (−21.6%), despite baseline levels being lower in KPK cells (Figure 3F and G). The drop in OCR in response to CB-839 was rescued by co-treatment with DMG or pyruvate (Figure 3G). The ability of both DMG and pyruvate to rescue overall levels of TCA cycle metabolites as well as cellular proliferation and mitochondrial respiration in response to CB-839 treatment in KPK cells suggests that the basal defects in TCA cycle metabolism in KPK cells causes a metabolic bottleneck due to limited glutamate availability which is further exacerbated by glutaminase inhibition via CB-839. As an αKG analogue, DMG can directly fuel the TCA cycle and bypass glutaminase inhibition. However, the ability of pyruvate to rescue TCA cycle defects in KPK cells is less direct. One explanation is that pyruvate offers an alternative anaplerotic substrate via increased flux through pyruvate carboxylase, which catalyzes the conversion of pyruvate to oxaloacetate. To determine if treatment with CB-839 enhanced carbon flux through pyruvate carboxylase (PC), we performed stable isotope tracing using [3 C13]-D-glucose to look at PC flux in basal conditions as well as with [1 C13]-pyruvate to look at flux in rescue conditions (Figure 3H). In line with this hypothesis, we observed increased flux through pyruvate carboxylase in response to CB-839 in normal media conditions as well as when cells are supplemented with pyruvate (Figure 3I and J). This is in agreement with previous results indicating that KP tumors have increased PC flux.

11Taken together these data show that Keap1 mutations result in basal defects in central carbon metabolism, with KPK cells having overall decreased levels of TCA cycle intermediates and reduced contribution of carbon from glutamine to these metabolites. This metabolic imbalance in KPK cells can be therapeutically exploited by the use of glutaminase inhibitors to further limit glutamate availability.

12Having established that Keap1 mutations increase sensitivity to glutaminase inhibition, we sought to determine if activation of Nrf2 in wild type cells would be sufficient to sensitize cells to CB-839. Using a small-molecule activator that disrupts the binding of Keap1 to Nrf2, KI696, we first validated that treatment with KI696 could increase Nrf2 protein levels as well as increase expression of downstream target genes in KP cells (Figure 4—figure supplement 1A–C and Figure 4 1A). Pretreatment with KI696 and subsequent Nrf2 activation was sufficient to induce strong growth suppression in response to CB-839 treatment in KP cells (Figure 4B), similar to what we observed in KPK cells (Figure 2C). Additionally, effects of CB-839 were rescued both by pretreatment of anaplerotic precursors, pyruvate and DMG, and by modulating extracellular glutamate and cystine concentrations, but not by the antioxidants NAC and trolox (Figure 4B).

13Additionally, we took a genetic approach to elevate GSH levels in KP cells in order to reproduce the Keap1 mutant phenotype in wild type cells and sensitize them to CB-839. Using the CRISPR synergistic activation mediate (CRISPRa SAM) system (Figure 4C), which takes advantage of the DNA targeting capabilities of a catalytically dead Cas9 protein fused to VP64 and a modified sgRNA that recruits and targets transcriptional activators to promoter regions upstream of transcriptional start sites of genes of interest to enhance endogenous transcription. Using sgRNAs that target the promoters of Gclc, Gclm, and Slc7a11, we were able to increase mRNA expression in cells that expressed single sgRNAs to each gene as well as in cells that expressed two sgRNAs to increase expression of Slc7a11 and Gclc or Gclm in combination (Figure 4C). Increased expression of these Nrf2 target genes resulted in significant increases in total GSH in all cell lines (Figure 4—figure supplement 1D). However, we only observed increased sensitivity to CB-839 in KP cells that had increased expression of Slc7a11 in combination with either Gclc or Gclm (Figure 4D).

14Similar to mouse KP cells, human KEAP1 wild type lung cancer cells, H2009, responded to KI696 with nuclear accumulation of NRF2 and subsequent up regulation of target gene expression and total GSH levels (Figure 4E and Figure 4—figure supplement 2A–C). In line with KP cells, human H2009 cells responded strongly to co-treatment with CB-839 and KI696 and this synergistic effect was rescued by pretreatment of anaplerotic precursors, modulating extracellular glutamate and cystine concentrations but not by antioxidants (Figure 4F). We next asked if KI696 treatment would replicate the metabolic phenotypes observed in KPK cells. KP cells treated with KI696 had reduced total levels of TCA cycle intermediates (Figure 4G), equivalent to that observed in KPK cells (Figure 4G and Figure 3C). Additionally, overall mitochondrial respiration was reduced upon treatment of wild type KP cells with KI696, and further reduced by the addition of CB-839 (Figure 4H). These data demonstrate that Nrf2 activation is sufficient to rewire cellular metabolism, resulting in increased glutamine dependency and sensitivity to CB-839 treatment in both mouse and human LUAD lines.

15In addition to LUAD, KEAP1/NRF2 mutations are found in other cancers. To assess whether NRF2 promotes glutamine dependency in other KEAP1 mutant cancers, we tested CB-839 on a panel of human cancer cell lines with WT or mutant KEAP1, including melanoma, bone, colon, renal, squamous and urinary tract cancers. For each cancer type, mutations in KEAP1 predicted sensitivity to CB-839 (Figure 5A). Importantly, co-treatment with KI696 was able to sensitize KEAP1 wild type cancer cells that were relatively insensitive to CB-839 (Figure 5B and Figure 5—figure supplement 1). Taken together, our data suggests that glutaminase is an attractive target in both lung adenocarcinoma and in other cancers in which the KEAP1/NRF2 axis is mutated and the glutaminase inhibitor CB-839 could be an effective therapeutic agent in patients with these genetic subtypes of cancer. Additionally, glutaminase inhibition could be more broadly applicable in a variety of cancer types when combined with a small molecule activator of NRF2, which renders KEAP1 wild type cells highly sensitive to glutaminase inhibition.

16Tumorigenesis requires cancer cells to increase their metabolic output to support tumor growth. Higher metabolic activity increases oxidative stress in the form of ROS, which can lead to oxidative damage of macromolecules. Under physiologic conditions of increased ROS, the NRF2 stress response promotes an antioxidant network that will scavenge and clear ROS. After ROS clearance NRF2 will once again be degraded through its interaction with KEAP1 and CUL3-mediated ubiquitination. However, in the context of chronic ROS exposure, post-transcriptional or post-translational signaling which stabilize NRF2, or epigenetic and genetic events that lead to NRF2 stabilization, NRF2 and its downstream antioxidant program will be chronically activated (Figure 5C). NRF2 activation subsequently leads to changes in cellular metabolism to support increased synthesis of antioxidants, which have recently been shown to promote tumorigenesis. In this study, we show that NRF2 dependent metabolic alterations result in a unique set of metabolic requirements in KEAP1 mutant cells that can consequently be theraputically exploited.

17We show that constitutive (KEAP1 mutation) or pharmacological activation of NRF2 leads to alterations in glutamine metabolism and defective glutamate-dependent fueling of central carbon metabolism (Figure 5D). Glutamine is a critical nutrient for proliferating cells as it contributes to supporting cellular bioenergetics by fueling ATP production as well as supplying the cell with necessary intermediates for biosynthesis. Both the carbon and nitrogen derived from glutamine are used for the synthesis of nucleotides, amino acids, and other macromolecules. Nitrogen from glutamine contributes to de novo nucleotide synthesis making an adequate supply of glutamine to support purine and pyrimidine synthesis necessary to maintain rapid rates of proliferation. Additionally, carbon and nitrogen from both glutamine and glutamate support amino acid synthesis via the activity of a variety of transaminases and aminotransferases. Depleted intracellular glutamate levels in KEAP1 mutant cells likely alters the activity of various transaminases due to limited substrate availability and may also result in the dependency on other exogenously supplied metabolites. The decreased availability of intracellular glutamate we observe in KEAP1 mutant cells results in dependency on exogenous glutamine (Figure 1—figure supplement 1F), which can be therapeutically exploited by glutaminase inhibition (Figures 1–5).

18The dependency on glutamine-derived glutamate that we observe in cells with NRF2 activation is primarily due to defective TCA cycle anaplerosis and not the limited availability of glutamate for antioxidant production, since neither antioxidant supplementation (Figure 3B, Figure 3—figure supplement 1A, and Figure 4A) nor pharmacological decrease of GSH levels (Figure 3—figure supplement 1B and C) rescued glutamine dependency or CB-839 sensitivity. The ability of both pyruvate and DMG to rescue overall levels in TCA cycle metabolites as well as cellular proliferation and reduced mitochondrial respiration in response to CB-839 treatment highlights that glutaminase inhibition further depletes limited intracellular glutamate stores in KEAP1 mutant cells and exacerbates basal defects in the TCA cycle. Reduced TCA cycle output in KEAP1 mutant cells can have wide ranging effects beyond immediate use in biosynthesis pathways, it may also impact the activity of αKG-dependent deoxygenases, a family of enzymes that regulate signaling, metabolism, and epigenetic modifications. αKG, succinate, and fumarate levels can have wide ranging effects beyond immediate use in biosynthesis pathways by altering hypoxia-inducible transcription factor (HIF) signaling and altering gene expression through epigenetic modifications.

19We show that the decrease in intracellular glutamate availability for the TCA cycle, in the context of increased NRF2, occurs through the active secretion of glutamate through xCT/SLC7A11. We hypothesize that xCT/SLC7A11-mediated secretion of glutamate and import of cystine is driven by the increased intracellular demand for cysteine, a key amino acid for many NRF2-induced antioxidants (e.g. GSH, TXN). Alternatively, the levels of glutamate and cystine in culture medium in vitro or in serum in vivo can also modulate the directionality of xCT/SLC7A11, which may also change the availability of intracellular glutamate for central carbon metabolism and the dependency on exogenous glutamine (Muir, A. et al, co-submitted with this manuscript). Similar to our findings, depletion of xCT/SLC7A11 improves cancer cell viability after glucose withdrawal through conservation of intracellular glutamate levels to maintain mitochondrial respiration. In this context, SLC7A11 expression was also dependent on the transcription factor ATF4, which is known to cross talk with NRF2 in cancer cell metabolism.

20NRF2 is a pleiotropic transcription factor that regulates many genes and pathways that could contribute to the depletion of intracellular glutamate stores in Keap1 mutants and facilitate metabolic rewiring. In line with this, using pharmacological and genetic means to modulate GSH levels specifically was not sufficient to alter sensitivity to CB-839 (Figure 3—figure supplement 1D and Figure 4D). Interestingly, increased endogenous expression of Slc7a11 in combination with Gclc or Gclm using SAM CRISPRa, was sufficient to induce sensitivity to CB-839 in KP SAM CRISPRa cells (Figure 4D). However, the sensitivity to CB-839 was not as robust as in KPK cells, suggesting that the moderate increase in target genes and GSH synthesis in double sg KP SAM CRISPRa cells was not enough to fully recapitulate the metabolic phenotype of Keap1 mutant cells.

21Tumor progression occurs in a complex microenvironment with infiltrating immune cells and fibroblasts, which can impact tumor growth. In addition to its cell-autonomous effects, glutamate secretion can signal to cells in the microenvironment through non cell-autonomous mechanisms. Increased glutamate secretion and paracrine signaling from tumors with NRF2 activation may have a profound impact on both the function and the cellular composition of cells in the microenvironment. Altered levels of both glutamine and glutamate in the tumor microenvironment of KEAP1 mutant tumors can effect cell signaling and regulate cell growth and proliferation pathways.

22Glutamine metabolism is positively regulated by a variety of oncogenes, rendering many cancer types addicted to glutamine and making glutamine metabolism an attractive therapeutic target. Hence, there are several small molecule glutaminase inhibitors, one of which, CB-839, is currently in phase I clinical trials. However, our data as well as previous studies highlight the potential importance of stratifying patients based on genotype for a successful response to glutaminase inhibitors. In a previous study, we have reported that KEAP1 mutant tumors respond to CB-839 treatment, while growth of wild type tumors is unaffected in vivo. Taken together with the current study, which provides mechanistic insight into how KEAP1 mutations rewire tumor metabolism and why glutamine deprivation results in growth suppression, these data provide critical pre-clinical evidence supporting the treatment of patients with KEAP1 mutations with glutaminase inhibitors. We further provide evidence that mutations in KEAP1 are sufficient to predict sensitivity to glutaminase inhibitors regardless of tissue of origin. CB-839 treatment significantly suppressed cell growth in a panel of tumor cells including melanoma, bone, colon, renal, squamous, and urinary tract cancers with KEAP1 mutations. Additionally, use of KI696, a small molecule activator of NRF2, which blocks the interaction between NRF2 and KEAP1, was able to sensitize both mouse and human KEAP1 WT cell lines to glutaminase inhibition. These data argue that the use of glutaminase inhibitors along with small molecule activators of NRF2 could be an effective therapeutic strategy with broad application across various KEAP1 mutant and wild-type cancers.

23KP and KPK isogenic clonal cell lines were previously established. Human cell lines were acquired from ATCC. All lines were tested negative for mycoplasma. All cells were cultured in a humidified incubator at 37°C and 5% CO2. Cells were maintained in either DMEM or RPMI-1640 (Cellgro, Corning; Manassas, Virginia) supplemented with 10% fetal bovine serum (Sigma Aldrich, St. Louis, Missouri) and gentamicin (Invitrogen, ThermoFisher; Waltham, Massachusetts).

24For cell proliferation assays conducted under different drug or media conditions as indicated, cells growing in DMEM were trypsinized, counted and plated into 12 well plate dishes (BD/Falcon, Corning; Manassas, Virginia) in 1 mL of RPMI media. For glutamine deprivation experiments, cells were seeded in RPMI containing either 2 mM or 0.5 mM glutamine at the time of plating. After cell attachment, cells were treated with the indicated drugs: vehicle (DMSO), 500 nM erastin (Sigma Aldrich) 6 mM glutamate (Sigma Aldrich), 50 uM Trolox (Acros Organics, Fisher Scientific; Waltham, Massachusetts), 0.5 mM N-acetyl-L-cysteine (NAC, Sigma Aldrich), 2 mM pyruvate (Gibco, ThermoFisher Scientific; Waltham, Massachusetts), 2 mM dimethyl-2-oxoglutarate (DMG, Sigma Aldrich), L-Buthionine sulfoximine (Sigma Aldrich), hydrogen peroxide (LabChem, Zelienople, Pennsylvania), menadione (Sigma Aldrich) or 1 μM KI696 (provided by Craig Thomas). Cells exposed to galactose oxidase (Sigma Aldrich) were cultured in media supplemented with 5 mM galactose (Sigma Aldrich). After 12 hr, cells were treated with 250 nM CB839. For media in which cystine concentrations were lowered to 20 μM, RPMI was prepared from a powder mix without amino acids (US Biological, Salem, Massachusetts) according to manufacturers instructions. All amino acids besides cystine were added to the RPMI mixture in the same concentrations as present in RPMI-1640 formulation. Proliferation experiments were carried out for 5 days post drug treatment and collected either by staining or cell counting. Cell counts were collected by trypan blue exclusion on a Countess II automated cell counter (Life Technologies, ThermoFisher Scientific; Waltham, Massachusetts). Cells were stained with a 0.5% crystal violet solution in 25% methanol. Plates were then washed, dried, and crystal violet was eluted in 400 uL of 10% acetic acid.

25For cell viability assays cells were plated in a white, opaque 96-well plate with clear bottom at a density of 1000 cells/well in RPMI. Drugs were added at the same concentration and manner as described above for the indicated amount of time. After 4 days, cell viability in the presence of all compounds was assessed by cell titer glo (Promega #G7570, Madison, Wisconsin).

26mRNA was collected from cells with RNeasymini kit (Qiagen, Germany) according to manufacturers protocol. Complementary DNA (cDNA) was synthesized from extracted mRNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems #4368814, ThermoFisher Scientific; Waltham, Massachusetts) according to manufacturers protocol. Gene expression was analyzed by quantitative reverse transcription polymerase chain reaction on a CFX96 Real-Time System (BioRad Technologies, Hercules, California) with the following primers, Slc7a11: gattcatgtccacaagcacac and gagcatcaccatcgtcagag, SLC7A11: ccatgaacggtggtgtgtt, Gclc: agatgatagaacacgggaggag and tgatcctaaagcgattgttcttc, GCLC: atgccatgggatttggaat and gatcataaaggtatctggcctca, Nqo1: acgctgccatgtatgacaaa and ggatcccttgcagagagtaca, NQO1: acgctgccatgtatgacaaa and ggatcccttgcagagagtaca, Gclm: tgactcacaatgacccgaaa and tcaatgtcagggatgctttct.

27Cells were lysed in 150 uL of ice-cold RIPA buffer supplemented with 1X Complete Mini inhibitor mixture (Roche, #11 836 153 001, Switzerland) and mixed on a rotator at 4°C. Protein concentration of lysates was determined using the Bio-Rad DC Protein Assay (#500–0114). 20–40 ug of total protein was separated on a 4–12% Bis-Tris gradient gel (Bio-Rad) by SDS-PAGE and then transferred to nitrocellulose membrane (Bio-Rad). The following antibodies were used for immunoblotting: anti-Nrf2 (Cell Signaling Technologies, Danvers, Massachusetts, #12721, 1:1000), anti-HSP90 (BD, Sparks, Maryland, #610418, 1:4000), anti-H3 (AbCam, Cambridge, Massachusetts, #1991, 1:4000), and anti-Slc7a11 (Novus Biologicals, Littleton, Colorado, #NB300-318).

28Total glutathione was measured after treatment with 250 nM CB-839 or treatment with BSO for 24 hr where indicated and collected according to manufacturer’s protocol (Promega, V6611). Data was normalized to additional wells that received identical conditions using Cell Titer Glo (Promega, G7570).

29ROS in cultured cells were measured by incubating cells with 5 μM CM-H2DCFDA (C6827, Life Technologies) for 30 min at 37°C. DCF fluorescence was acquired on the Attune NxT (ThermoFisher) flow cytometer and analyzed using FlowJo software (Tree Star, Ashland, Oregon).

30Oxygen consumption rate (OCR) experiments were performed using the XFe-96 apparatus from Seahorse Bioscience (Agilent, Santa Clara, California). Cells were seeded to ~80% confluence in at least triplicate for each condition in RPMI. The following day, wells were treated with 250 nM CB-839 for 4 hr. Media was completely replaced with reconstituted RPMI with 12.5 mM glucose and 2 mM glutamine (no sodium bicarbonate) adjusted to pH 7.4 and incubated for 45 min at 37°C in a CO2-free incubator prior to measurements. Maximal respiration measurements were obtained by performing a mito-stress test. Briefly, cells were prepared as described above. After reaching steady-state respiratory flux (basal respiration), 100 μM oligomycin was injected to inhibit ATP synthase. Next, respiration was uncoupled from oxidative phosphorylation by addition of carbonylcyanide p-trifluoromethoxyphenylhydrazone (FCCP) (Sigma)until maximal respiratory capacity was reached (up to 2 μM in 0.5 μM increments). Finally, respiration was inhibited by the addition of 50 μM rotenone and 50 μM antimycin A (Sigma).

31All animal studies described were approved by the NYU Langone Medical Center Institutional Animal Care and Use Committee. 1 × 106 cells were implanted subcutaneously into NSG mice. After tumor establishment phase (13 days post implantation), animals were randomized and assigned to the following groups: vehicle treated and doxycycline diet, vehicle treated and normal diet, CB-839 treated and doxycycline diet, or CB-839 treated and normal diet. Animals were treated with 200 mg/kg CB-839 or vehicle (Calithera, San Francisco, California) twice daily administered through oral gavage as before. The vehicle contained 25% (w/v) hydroxypropyl-β-cyclodextrin in 10 mmol/L citrate (pH 2.0), and CB839 was formulated at 20 mg/mL for a final dosing volume of 10 mL/kg.

32Extracellular glutamate was measured by collecting fresh and spent medium after 24 hr of growth. Cells were assumed to grow exponentially over the culture period. Media was analyzed using a YSI biochemistry analyzer (Yellow Springs Instruments, Yellow Springs, Ohio).

33Doxycyline-induced knockdown of xCT was achieved by cloning miR-E shRNAs targeting Slc7a11 into the LT3GEPIR vector as previously described in detail. Briefly, LT3GEPIR was digested with XhoI and EcoRI, and purified with a gel extraction kit (Qiagen). Single stranded ultramers were amplified with forward primer miRE-XhoI (5’‐ TGAACTCGAGAAGGTATATTGCTGTTGACAGTGAGCG‐3’) and reverse primer miRE‐EcoRI (5’‐TCTCGAATTCTAGCCCCTTGAAGTCCGAGGCAGTAGGC‐3’). Amplicons were gel purified, digested with XhoI and EcoRI, cleaned up with PCR purification kit (Qiagen) and ligated into the cut LT3GEPIR vector with T4 DNA Ligase at a 3:1 insert:vector molar ratio. Vectors were transduced into cells and selected with 3 and 6 ug/ml puromycin for three plus three days. Knockdown of xCT/Slc7a11 was verified by western blot and qPCR analysis following 72 hr of treatment with 1 ug/ml doxycycline. Cell lines transduced with the first two efficient shRNAs (#1 and #3), along with a non-efficient shRNA as control (#5), were used in the study. Sequences of ultramers obtained from Integrated DNA Technologies (Coralville, Iowa):

34#1 -

35TGCTGTTGACAGTGAGCGCCAGGAAGAGACACAAGTCTAATAGTGAAGCCACAGATGTATTAGACTTGTGTCTCTTCCTGATGCCTACTGCCTCGGA

36#2 -

37TGCTGTTGACAGTGAGCGCAAGGAAGATTTTAGTTATTAATAGTGAAGCCACAGATGTATTAATAACTAAAATCTTCCTTTTGCCTACTGCCTCGGA

38#3 –

39TGCTGTTGACAGTGAGCGCCAGGAGCTTTATGTAAATGTATAGTGAAGCCACAGATGTATACATTTACATAAAGCTCCTGATGCCTACTGCCTCGGA

40#5 –

41TGCTGTTGACAGTGAGCGCCAGATTATACTAGAAGTTGTATAGTGAAGCCACAGATGTATACAACTTCTAGTATAATCTGTTGCCTACTGCCTCGGA

42#6 –

43TGCTGTTGACAGTGAGCGCTAAGTTCTATGTGATACAGAATAGTGAAGCCACAGATGTATTCTGTATCACATAGAACTTATTGCCTACTGCCTCGGA

44#7 –

45TGCTGTTGACAGTGAGCGCAGGAACTATGTTTGTACTAAATAGTGAAGCCACAGATGTATTTAGTACAAACATAGTTCCTATGCCTACTGCCTCGGA

46#8 -

47TGCTGTTGACAGTGAGCGCCCAGAAGACTCTAAAGAATTATAGTGAAGCCACAGATGTATAATTCTTTAGAGTCTTCTGGTTGCCTACTGCCTCGGA

48Transcriptional activation of various genes was achieved by cloning sgRNAs targeting the region upstream of the transcriptional start sites for Slc7a11, Gclc, and Gclm into the lenti sgRNA(MS2)_puro and zeo backbones (Addgene plasmid #73797 and 61247, respectively) as previously described. Briefly, both backbones were digested with Bsmb1 (New England Biosciences, Ipswich, Massachusetts) and purified with a gel extraction kit (Qiagen). Single stranded oligo’s were annealed and phosphorylated using T4 PNK (New England Biosciences). Phosphorylated and annealed oligos were then ligated into the purified digested backbones using Quick Ligase (New England Biosciences). Vectors were transduced into KP cells already containing the other two component vectors for the SAM system and selected with 3 and 6 ug/ml puromycin for three plus three days. Transcriptional activation was subsequently verified by qPCR. Sequences of oligos obtained from Integrated DNA Technologies:

49sgSlc7a11:

50sense- CACCGCCTGTCACACCAACTTACT

51antisense- AAACAGTAAGTTGGTGTGACAGGC

52sgGclc:

53sense- CACCGAATAAGGACTGAAAAGTCT

54antisense- AAACAGACTTTTCAGTCCTTATTC

55sgGclm:

56sense: CACCGGTGACTGGCCCTCGGCCTG

57antisense: AAACCAGGCCGAGGGCCAGTCACC

58For glutamine tracing, 2 × 105 cells were seeded in 2 mL of RPMI-1640 in 6 well plates. Where indicated, cells were pretreated for 12 hr with 1 μM of KI696. Media was then replaced with 2 mL of fresh RPMI-1640 containing 2 mM of [U-13C]-L-glutamine (Cambridge Isotope Laboratory, Tewksbury, Massachusetts) and 1 μM of KI696 where indicated. For PC flux analysis, 1 × 105 cells were seeded in 1 mL of RPMI Cells were cultured for 8 hr to reach steady state labeling of TCA cycle intermediates. Cells were washed 1X in ice cold saline and then collected by scraping in 600 uL of 80% (v/v) of ice cold methanol containing 1.4 ug/mL norvaline (Sigma Aldrich). Samples were vortexed for 10 min at 4°C and then centrifuged at max speed for 10 min. Supernatant was transferred to fresh tubes and then dried under nitrogen. Dried and frozen metabolite extracts were then derivatized with 16 uL of MOX reagent (ThermoFisher) for 60 min at 37°C and N-tert-butyldimethylchlorosilane (Sigma Aldrich) for 30 min at 60°C. After derivatization, samples were analyzed by GC-MS using a DB-35MS column (Agilent Technologies) in an Agilent 7890A gas chromatograph coupled to an Agilent 5997B mass spectrometer. Helium was used as the carrier gas at a flow rate of 1.2 mL/minute. One microliter of sample was injected in split mode (split 1:1) at 270°C. After injection, the GC oven was held at 100°C for 1 min and then increased to 300°C at 3.5 °C/minute. The oven was then ramped to 320°C at 20 °C/minute and held for 5 min at 320°C.

59The MS system operated under electron impact ionization at 70 eV and the MS source and quaddrupole were held at 230°C and 150°C respectively. The detector was used in scanning mode, and the scanned ion range was 10–650 m/z. Mass isotopomer distributions were determined by integrating appropriate ion fragments for each metabolite using MatLab (Mathworks, Natick, Massachusetts) and an algorithm adapted from Ferandez and colleagues that corrects for natural abundance.

60Values are presented as mean ± SEM. For statistical analyses, we used Graphpad Prism software v.7.02 (La Jolla, California): one-way ANOVA with Tukey’s post-hoc test for cell growth assays, OCR measurements, Glutathione levels and TCA Metabolites in Figure 4e; Mann-Whitney test for intra and extra cellular Glutamate levels and [U-13C]-L glutamine labeled metabolites in Figure 3e; 2-sided Student’s t-test for gene expression analyses and TCA Metabolites in Figure 3c; and two-way ANOVA for tumor growth. All experiments were repeated at least twice unless otherwise stated. All n indicate biological replicates.