Concurrent Mutations in STK11 and KEAP1 Promote Ferroptosis Protection and SCD1 Dependence in Lung Cancer
Corrin A Wohlhieter 1, Allison L Richards 2, Fathema Uddin 3, Christopher H Hulton 4, Àlvaro Quintanal-Villalonga 3, Axel Martin 5, Elisa de Stanchina 6, Umeshkumar Bhanot 7, Marina Asher 7, Nisargbhai S Shah 3, Omar Hayatt 6, Darren J Buonocore 8, Natasha Rekhtman 8, Ronglai Shen 5, Kathryn C Arbour 3, Mark Donoghue 2, John T Poirier 9, Triparna Sen 10, Charles M Rudin 11
Highlights
•STK11/KEAP1 co-mutation promotes cell proliferation, independent of KRAS status
•NRF2 activity is enhanced in STK11/KEAP1 co-mutation beyond KEAP1 loss alone
•STK11 and KEAP1 mutations each independently promote ferroptosis protection
•SCD1 protects STK11/KEAP1 co-mutant LUAD from ferroptosis and is essential for survival
Summary
Concurrent loss-of-function mutations in STK11 and KEAP1 in lung adenocarcinoma (LUAD) are associated with aggressive tumor growth, resistance to available therapies, and early death. We investigated the effects of coordinate STK11 and KEAP1 loss by comparing co-mutant with single mutant and wild-type isogenic counterparts in multiple LUAD models. STK11/KEAP1 co-mutation results in significantly elevated expression of ferroptosis-protective genes, including SCD and AKR1C1/2/3, and resistance to pharmacologically induced ferroptosis. CRISPR screening further nominates SCD (SCD1) as selectively essential in STK11/KEAP1 co-mutant LUAD. Genetic and pharmacological inhibition of SCD1 confirms the essentiality of this gene and augments the effects of ferroptosis induction by erastin and RSL3. Together these data identify SCD1 as a selective vulnerability and a promising candidate for targeted drug development in STK11/KEAP1 co-mutant LUAD.
Introduction
Lung cancer is the leading cause of cancer mortality, accounting for about 12% of newly diagnosed cases and about 18% of total cancer deaths worldwide (Bray et al., 2018). The most common histologic subtype of lung cancer is lung adenocarcinoma (LUAD), making up approximately 50% of cases (Herbst et al., 2008). In recent years, dramatic progress has been made in tailoring therapies for subgroups of patients with LUAD harboring known oncogenic mutations or translocations in genes, including epidermal growth factor receptor (EGFR, 28%), anaplastic lymphoma kinase (ALK, 3.8%), ROS proto-oncogene 1 (ROS1, 2.6%), and others (Jordan et al., 2017). Genetic profiling and targeting of these oncogenic drivers have markedly improved clinical outcomes of patients with these subtypes of LUAD. Unfortunately, only a few patients benefit from those approaches.
At the other end of the spectrum of oncogenic mutations in LUAD are drivers that have not been successfully profiled or targeted to date and that are associated with particularly poor prognosis. Using mutational profiling data from a cohort of more than 1,000 patients with metastatic LUAD, our group recently identified concurrent loss-of-function mutations in two genes as a defining factor of strikingly poor prognosis in a subset of patients with LUAD; serine/threonine kinase 11 (STK11), encoding the protein LKB1, and Kelch-like ECH-associated protein 1 (KEAP1). This co-mutation occurs in 10.2% of metastatic LUAD and defines a patient cohort with a median overall survival of only 7.3 months (Shen et al., 2019). Inactivating mutations in either STK11 or KEAP1 have been previously analyzed in the context of oncogenic
KRAS mutations in LUAD, but their co-association with poor prognosis appears to be independent of KRAS status. Biological mechanisms favoring the coordinated loss of these two genes and clinically tractable therapeutic vulnerabilities of this subset of LUAD have not been defined. Identifying therapeutic strategies for those exceptionally poor prognosis LUAD cases is a critical need.
STK11 is the third most commonly mutated gene in LUAD, after TP53 and KRAS, and has been identified in up to 33% of primary LUADs (Gleeson et al., 2015). STK11 encodes a serine/threonine kinase, LKB1, which activates a family of 12 downstream kinases, including AMP-activated protein kinase (AMPK), and has a role in essential biological functions, including cellular energy regulation. We have previously reported STK11 mutations in the context of KRAS-mutant LUAD to be strongly associated with resistance to immunotherapy (Skoulidis et al., 2018).
Approximately 17% of patients with LUAD have loss-of-function mutations in KEAP1 (Collisson et al., 2014). KEAP1 encodes a key factor controlling the antioxidant response pathway, functioning as a negative regulator of the transcription factor nuclear factor erythroid-1 like 2 (NFE2L2/NRF2) (Rojo de la Vega et al., 2018). Loss of KEAP1 increases both protein stability and nuclear translocation of NRF2, which, in turn, alters the transcription of genes involved in cellular antioxidant, detoxification, and metabolic pathways. We have previously reported that KRAS-mutant LUADs with concomitant KEAP1 loss have an increased dependence on glutaminolysis (Romero et al., 2017) and shorter survival when treated with either chemotherapy or immunotherapy (Arbour et al., 2018).
To better define interventional targets for these therapeutically refractory cancers, this study investigated the global changes in gene expression and oncogenic signaling pathways driven by concomitant loss of STK11 and KEAP1 versus loss of either or neither of those genes. We characterized that co-mutation across multiple models, including isogenic human LUAD cell lines generated by selective gene knockout, and cell line xenografts from cancers harboring those mutations de novo. We used the isogenic models in a targeted CRISPR/Cas9 screen to define candidate therapeutic vulnerabilities specifically associated with concomitant STK11/KEAP1 loss. Our data demonstrate that concomitant loss of STK11 and KEAP1 drives ferroptosis protection and identifies a key negative regulator of this cell death pathway, stearoyl-CoA desaturase 1 (SCD1), as a critical and selective dependency in STK11/KEAP1 co-mutant LUAD.
Results
STK11/KEAP1 Co-mutation Predicts Short Overall Survival in Patients with LUAD, Independent of KRAS Status
MSK-IMPACT (Memorial Sloan Kettering-Integrated Mutation Profiling of Actional Cancer Targets) is a clinically deployed next-generation sequencing panel that detects mutations, select translocations, and copy number alterations in more than 340 cancer-associated genes (Cheng et al., 2015). We queried a cohort of 1,235 sequentially profiled metastatic LUAD patients for tumor-specific somatic mutations in STK11 only (n = 43), KEAP1 only (n = 53), or both (n = 57) (Figure 1A). We included a third mutation in our analysis, KRAS (n = 358), to specifically assess the role of KRAS mutation in dictating survival for STK11/KEAP1 co-mutant patients. KRAS mutations often co-occur with STK11 (n = 41), KEAP1 (n = 31), and STK11/KEAP1 (n = 66); however, recent findings suggest that STK11/KEAP1 co-mutation independently predicts a high-risk patient cohort (Shen et al., 2019).
We observed a marked decrease in median overall survival from 26.4 months in patients with wild-type (WT) STK11/KEAP1 alleles to 11.5 months in patients harboring the STK11/KEAP1 co-mutation (KRAS WT) and to 6.5 months in patients harboring KRAS/STK11/KEAP1 triple-mutation, with single mutants having intermediate survival (Figure 1B). In multivariate analysis, STK11/KEAP1 co-mutant status significantly (p < 0.001) predicted poor survival, independent of KRAS status (Figure 1C). To date, the biological link between loss of STK11 and KEAP1 has only been studied in the context of a KRAS-driver mutation. Based on our data, these studies exclude nearly 50% of patients with STK11/KEAP1 co-mutations who do not harbor a KRAS mutation but who are still at exceptionally high risk for early death. Taken together, these findings support the need for a better understanding of the biology driving STK11/KEAP1 co-mutant LUAD, with or without a KRAS mutation, to identify therapeutic vulnerabilities for the entire high-risk patient cohort. Figure 1 Patients with Lung Adenocarcinoma and STK11 and KEAP1 Co-mutation Have Lower Overall Survival, Independent of KRAS Status Given the size of the available clinical dataset, we explored whether particular mutations within STK11 or KEAP1 favored acquisition of the co-mutation and whether genomic profiling of single- and double-mutant tumors could inform the relative timing of STK11 and KEAP1 mutation in LUAD formation. For that analysis, we included patients with LUAD at all stages. Mutation location within the STK11 or KEAP1 genes did not differ between single mutants and co-mutants, consistent with a wide array of mutations resulting in loss of function (Figure S1A). In an attempt to gain insight into the ordering of mutations in tumorigenesis, we analyzed the cancer cell fraction (CCF) of STK11 and KEAP1 mutations in all patients with LUAD harboring the STK11/KEAP1 co-mutation in the MSK-IMPACT-sequencing cohort who had the necessary allele-specific copy number data (n = 292) (Shen and Seshan, 2016). We found that mutations in STK11 and KEAP1 are both clonal in 84% of samples containing both mutations (Figure S1B). Nearly 90% of STK11/KEAP1 co-mutant tumors demonstrate a loss-of-heterozygosity (LOH) event on chromosome 19, where both STK11 and KEAP1 reside. This rate of LOH is significantly higher (p < 0.001) than the occurrence of LOH in either single mutant and is nearly 60% higher than the occurrence in tumors with no loss of STK11 or KEAP1 (Figure S1C). Given the chromosomal proximity of STK11 and KEAP1, LOH as a mechanism of selecting against WT alleles by coordinately deleting a copy of both genes provides insight into the acquisition of the co-mutation in oncogenesis but, coupled with the observed clonality, prevents analysis into whether there is selective pressure in the temporal order of gene disruption. STK11/KEAP1 Co-mutation Promotes Tumor Cell Proliferation In Vitro and Tumor Growth In Vivo Mutations in STK11 and KEAP1 have each independently been reported to promote cell proliferation in the context of KRAS mutations (Murray et al., 2019; Romero et al., 2017). Given our clinical data supporting STK11/KEAP1 cooperativity, regardless of KRAS status, we investigated whether loss-of-function mutations in both STK11 and KEAP1 promote cell proliferation more than either single gene loss in a KRAS-independent manner. We used CRISPR/Cas9 gene editing to create stable knockouts of STK11, KEAP1, or both genes in two LUAD lines, H358 and H292. H358 harbors the oncogenic KRAS mutation G12C, whereas H292 is WT for KRAS. We created singe-cell clones harboring either a control non-targeting guide (NTC), STK11 mutation (STK11KO), KEAP1 mutation (KEAP1KO), or both (double-knockout [DKO]) (Figure 2A). To assess cell proliferation, we tracked the growth of H358 and H292 isogenic clones over time. Supporting the tumor-suppressive nature of STK11 and KEAP1, we observed an approximate doubling in growth rates in DKO derivatives of both H358 and H292 relative to WT controls (p = 0.0002 and p < 0.0001, respectively), with single-gene knockouts demonstrating intermediate phenotypes (Figures 2B and 2C). These data support our hypothesis that STK11 and KEAP1 co-mutations cooperate to promote cell growth, independent of KRAS status. Figure 2 STK11 and KEAP1 Co-mutation Promotes Cell Growth, Independent of KRAS Mutation Status STK11 regulates a family of 12 AMPK-related kinases known to have major roles in diverse cellular functions, including growth, survival, and metabolism (Jeon et al., 2012; Lamming and Sabatini, 2013; Shaw et al., 2004). Although ultimately providing a growth advantage, loss of this master regulator would be predicted to alter cellular homeostasis and require cells to adapt to those new conditions. We hypothesized that concomitant loss of KEAP1 and resultant activation of NRF2-dependent antioxidant signaling could help cancer cells survive and proliferate in the context of STK11 loss. We investigated whether loss of function of both STK11 and KEAP1 augments cancer cell proliferation in vitro and tumor progression in vivo. We transduced A549 and H460 LUAD lines, both harboring de novo STK11 and KEAP1 co-mutations, with doxycycline-inducible lentiviral vectors expressing either WT STK11 (LKB1) or KEAP1 (Figure 3A). Both lines demonstrate tightly regulated, doxycycline-inducible expression of LKB1 or KEAP1 proteins and, consistent with KEAP1 protein function, we observed a decrease in NRF2 upon re-expression of KEAP1 (Figure 3B). We observed a significant decrease (p = 0.012) in A549 cell growth in vitro when KEAP1 protein was re-expressed (A549-KEAP1) (Figure 3C). In vivo, we observed a significant survival advantage (p < 0.0001) in doxycycline-treated mice bearing A549-KEAP1 tumors compared with all other groups: mice with tumors re-expressing KEAP1 survived 40 days longer than the next-longest-surviving group (KEAP1-dox) and approximately 90 days longer than A549-GFP and A549-STK11 mice, with or without doxycycline treatment (Figure 3D). In the H460 xenograft model, we observed significantly smaller tumor volume (p = 0.0028) and weight (p = 0.0019) in doxycycline-treated H460-KEAP1 compared with the untreated control group (Figure 3E). Surprisingly, in these two xenograft models, re-expression of LKB1 did not alter proliferation rate in vitro or tumor growth rate in vivo. These results indicate a requirement for KEAP1 loss of function in STK11/KEAP1 tumors to maintain the cell proliferation rate. This is consistent with recent findings suggesting an “NRF2-addicted” phenotype in KEAP1-mutant LUAD (Kitamura and Motohashi, 2018). STK11 loss of function was not required to maintain proliferative potential of the cells, suggesting that loss of STK11 has a role in tumorigenesis that is distinct from simply maximizing proliferative potential. Figure 3 KEAP1 Re-expression Disrupts Growth of STK11/KEAP1 Co-mutant Lung Adenocarcinoma Cells The primary known function of KEAP1 is negative regulation of the transcription factor NRF2. To further assess whether the growth advantage conferred by KEAP1 loss was attributable to NRF2 upregulation, we treated H358-STK11KO cells with Ki-696, an NRF2 activator that increases NRF2 expression by disrupting the protein-protein interaction with KEAP1. The expected increase in NRF2 expression in the treated population was confirmed by western blot (Figure S2A). As predicted, H358-STK11KO cells treated with the NRF2 activator (1 μM, 3 days) grew significantly faster (p = 0.025) than vehicle-treated cells, phenocopying the growth potential of the H358-DKO cells (Figure S2B). Conversely, H358-DKO cells with Cas-9-mediated knockout of NRF2 experienced a decrease in growth rate, as measured by dropout of blue fluorescent protein (BFP), the marker for the single-guide RNA (sgRNA) targeting NRF2, over a period of 20 days (Figures S2C and S2D). Together, these findings confirm that NRF2 activation is primarily responsible for the growth advantage imparted by loss-of-function mutations in KEAP1. STK11/KEAP1 Co-mutant Cells Have a Distinct Transcriptional Profile We next investigated how selective mutation in STK11 or KEAP1 alters the transcriptional profile of cancer cells, and how the transcriptional profile of co-mutant cells might uniquely affect cancer-related pathways, including control of proliferative potential, metabolic homeostasis, and cell death. We performed bulk RNA sequencing (RNA-seq) on NTC, STK11KO, KEAP1KO, and DKO single-cell clones (n = 3 per group) from both H358 and H292 cell lines. Principal component analysis (PCA) of our clones ordered each clone first by cell line, and later PCs separated clones by mutation status as anticipated (Figure S3A). RNA-seq results identified 1,084 differentially expressed genes in DKO clones compared with all other groups (q value < 0.05) (Table S1). Six pathways were identified as significantly over-represented among upregulated genes (adjusted p value < 0.05), whereas 10 pathways were significantly over-represented among downregulated genes (adjusted p value < 0.05) (Figures 4A and S3B). Three of six pathways with upregulated genes were related to regulation of cellular metabolism. As an important internal validation of our data, the most significantly enriched of those pathways was glutathione metabolism (adjusted p value = 2.4E−05), consistent with previously published observations made in A549 KRAS/STK11/KEAP1-mutant cells (Galan-Cobo et al., 2019). Genes in that list include GCLM and GSR, two major components of the glutathione metabolism pathway that aid in reduction of reactive oxygen species (ROS). Interestingly, the second most enriched gene set among upregulated genes in DKO was ferroptosis (adjusted p value = 1.3E−03). Ferroptosis is a non-apoptotic form of cell death, characterized by a failure in glutathione-dependent antioxidant defenses, resulting in unchecked lipid peroxidation and subsequent cell death (Dixon et al., 2012). Ferroptosis can be induced by either impaired elimination or over-production of lipid peroxides leading to accumulation to lethal levels. Our analysis identified a transcriptional profile for STK11/KEAP1 co-mutant cells, independent of KRAS status, in which glutathione metabolism is upregulated to maintain metabolic homeostasis, and a ferroptosis-protective gene signature is enriched, which may promote cell survival. Figure 4 Ferroptosis-Protective Genes Are Upregulated in STK11/KEAP1 Co-mutants and STK11 and KEAP1 Mutations Independently Protect Cells from Ferroptosis STK11/KEAP1 Co-mutant Cells Have Higher Expression of Genes Involved in Ferroptosis Protection and Are Resistant to Ferroptosis-Inducing Agents The coordinated induction of an anti-ferroptosis program suggested a possible therapeutic vulnerability in STK11/KEAP1 co-mutant LUAD. We hypothesized that the metabolic shift associated with the loss of STK11 and KEAP1 might lower a cellular threshold to ferroptosis, making those cells dependent on the observed upregulation of anti-ferroptotic factors. To explore that possibility, we investigated how ferroptosis regulators might promote survival of STK11/KEAP1 co-mutant LUAD when challenged with ferroptosis induction. Notably, NRF2 regulates expression of many genes in the ferroptosis pathway; however, our system differentiates gene-expression changes between KEAP1KO clones and DKO clones. That methodology provides an opportunity for comparison of single-mutant and co-mutant effects and identifies enhanced NRF2 activity in DKO mutants beyond that attributed to KEAP1 mutation alone. Comparing our DKO clones relative to all other groups, we identified significant upregulation of genes encoding negative regulators of ferroptosis. Those genes include components of the Xc- antiporter SLC7A11 (q value = 0.0058, β = 1.1) and SLC3A2 (q value = 0.011, β = 0.61) and regulators of GPX4-mediated reduction of lipid peroxidation, including members of the glutathione pathway GCLC (q value = 0.0090, β = 0.92), GCLM (q value = 0.031, β = 0.99), and G6PD (q value = 0.0042, β = 1.07) (Figures 4B and S4A). Based on the increased expression of ferroptosis-protective genes in DKO clones, we postulated that those cells might be less responsive to known inducers of ferroptosis, including erastin, an inhibitor of the Xc- antiporter, and RSL3, an inhibitor of GPX4. We treated H358 isogenic clones with erastin at a dose of 2 μM for 72 h and observed a decrease in cell viability by 70% in NTC compared with 30% in H358-STK11KO, 40% in H358-KEAP1KO, and 35% in H358-DKO (Figure S4B) (p < 0.001 for comparison of DKO versus NTC). The difference in cell viability among genotypes was not as striking as we anticipated. The target of erastin, the Xc- antiporter, is upstream of the actual mechanism of ferroptosis induction, peroxidation of lipids. When blocking the Xc- antiporter, and therefore, the transport of cystine into the cell, glutathione/glutathione disulfide (GSH/GSSG) redox is inhibited from neutralizing ROS within the cell. High-cellular ROS can activate both ferroptosis and apoptosis cell-death mechanisms. To address whether apoptosis is a contributing cause of cell death via erastin, independent of genotype, we performed a flow cytometry assay for Annexin V/DAPI after treatment with erastin. We observed that H358 cells treated with erastin for up to 72 h do, in fact, induce apoptosis (Figure S4C). In contrast to the upstream mechanism of erastin, RSL3 directly inhibits GPX4, the protein responsible for neutralizing lipid peroxides and, therefore, directly inhibiting ferroptosis. We treated our H358 isogenic clones with RSL3 for 72 h at 370 nM and observed resistance in H358-STK11KO (p < 0.0001) and H358-DKO (p = 0.0002) compared with NTC (Figure 4C). Although it has been reported that KEAP1 loss induces NRF2-regulated expression of genes involved in ferroptosis protection, these data indicate that STK11 loss may have an important role in cell survival by ferroptosis evasion. To further dissect the mechanism of ferroptosis-protection we identified in STK11/KEAP1 co-mutant LUAD, we compared the production of lipid peroxides in H358 cells (STK11 and KEAP1 WT) to A549 cells (STK11 and KEAP1 mutant) treated with 500 nM RSL3. H358 cells demonstrated lipid peroxidation after 1 h of RSL3 treatment, and complete cell death by 3 h. In contrast, A549 cells were completely protected from lipid peroxidation for the entire 3-h treatment (Figure 4D). To determine whether STK11 and KEAP1 both have an independent role in ferroptosis protection, we used our A549 model for dox-inducible re-expression of LKB1 or KEAP1. Upon treatment with RSL3 at 500 nM for 4 h, we saw that re-expression of both LKB1 and KEAP1 independently sensitized A549 cells to lipid peroxidation (Figure 4E). Taken together, our findings indicate that STK11/KEAP1 co-mutant tumors develop multi-modal ferroptosis-protective mechanisms, regulated by the loss of both STK11 and KEAP1. AKR1C Family Genes Are Significantly Upregulated and Aid Ferroptosis Evasion in STK11/KEAP1 Co-mutant Cells Among the NRF2-regulated ferroptosis-protective genes, members of the aldo-keto-reductase-1C (AKR1C) family, AKR1C1, AKR1C2, and AKR1C3, were among the top genes upregulated in DKO cells as compared with all other groups (Wald test β = 3.83, 3.52, and 2.54, respectively) (Figures 5A and 5B). These increases in RNA transcript expression correlate with protein expression levels across cell lines (Figure 5C). Notably, protein expression was also partially induced in KEAP1KO lines. The substantial increase in AKR1C expression in DKO versus KEAP1KO supports our finding that NRF2 activity is enhanced in STK11/KEAP1 co-mutant LUAD beyond that induced by KEAP1 mutation alone. AKR1C members have been shown to promote cell proliferation, metastasis, and chemo-resistance in multiple cancer models (Chang et al., 2019; Tian et al., 2016; Zhu et al., 2018) and have been shown to protect melanoma cells from ferroptosis (Gagliardi et al., 2019). Figure 5 AKR1C Family Is Highly Upregulated in STK11/KEAP1 Co-mutants To assess whether the upregulation of AKR1C1 in our KEAP1KO and DKO cells was reflective of STK11/KEAP1 mutational status in patients with LUAD, we stained tissue microarrays containing 119 resected LUADs with an antibody to AKR1C1. We observed the highest expression of AKR1C1 in co-mutant samples with 79% (49/62) of samples with a co-mutation scoring as strong expressors, compared to 65% (26/40) of KEAP1 mutants, 14% (12/86) of STK11 mutants, and 9% (4/44) of double WT samples (Figure 5D). Taken together, RNA and protein expression analyses of our isogenic clones across cell lines, as well as patient data categorized by mutation subtype, consistently supports the finding that STK11/KEAP1 co-mutant tumors markedly upregulate AKR1C family members. A challenge in defining a selective therapeutic strategy for ferroptosis induction in STK11/KEAP1 co-mutant tumors is to identify a protein target that can be inhibited without substantial toxicity in healthy tissues. Given that the upregulation of AKR family members is specific to STK11/KEAP1 mutants and that these proteins have a known role in ferroptosis protection, we hypothesized that they could represent effective therapeutic targets for this subtype of LUAD. We investigated whether genetic inhibition of AKR1C1 would be selectively effective against co-mutant LUAD. We transduced H358-DKO clones with a lentivirus containing a BFP-marked guide against AKR1C1 and observed a steady decrease in the percentage of BFP+ cells over 20 days, indicating a decrease in the cell growth rate (Figure S5A). Given that a viable population of BFP+ cells was maintained, we concluded that knockout of AKR1C1 alone was not sufficient to decrease the viability of H358-DKO cells. The other AKR1C family members (AKR1C2/3) have substantial redundancy and may compensate for the loss AKR1C1. Therefore, we treated H358 isogenic clones with a titration of medroxyprogesterone 17-acetate (MPA), a weak pan-AKR1C inhibitor, for 3 days. We observed a 35% decrease in cell viability in H358-DKO clones at a high dose of 10 μM MPA, a significantly greater response compared with NTC (p = 0.0056) (Figure S5B). Although MPA selective efficacy is supportive of AKR1C having a role in DKO survival, inhibition of AKR1C members by MPA as a single agent is not a clinically viable strategy because of our inability to reach the half-maximal inhibitory concentration (IC50) at a dose as high as 10 μM. Interestingly, the combination of MPA (10 μM) with the ferroptosis-inducer erastin (5 μM) resulted in a 4-fold decrease in cell viability (from ∼80% in erastin treated cells to ∼20% in cells treated with the combination of erastin and MPA) in H358-DKO cells (Figure S5C). These results support a role for AKR1C family members in protecting STK11/KEAP1 co-mutant cells from ferroptosis but do not support the pan-AKR1C inhibitor MPA as a therapeutic strategy. CRISPR/Cas9 Screen Identifies Ferroptosis Regulator SCD as an Essential Gene Required for Proliferation and Survival of STK11/KEAP1 Co-mutant Cells To identify genetic vulnerabilities selective for STK11/KEAP1 co-mutant tumors, we performed a CRISPR/Cas9 screen in our isogenic in vitro models. That screen was performed with a curated “druggable genome” sgRNA library that targets 1,463 genes encoding proteins that are direct targets of currently available drugs or are immediately downstream of a directly targetable protein (Table S2). That approach was taken so that any hits identified in the screen as preferentially maladaptive when disrupted in co-mutant clones could be readily targeted with a pharmacologic approach. That strategy increases the translational potential of our findings by providing immediate therapeutic options for this particularly aggressive subset of LUAD. We performed that screen in our two cell lines, H358 (KRAS mutant) and H292 (KRAS WT) in all four mutant groups (NTC, STK11KO, KEAP1KO, and DKO) and across three isogenic clones per group. Clones were passaged separately throughout the screen and tracked for the number of doublings to control for variation in proliferation rate. Each clone was passaged for 16 doublings (Figure 6A). In agreement with a recent study (Galan-Cobo et al., 2019), we identified glutaminase (GLS) as a top-ranked hit in our H358 screen (KRAS mutant), increasing our confidence in the screen results. We did not, however, identify GLS as a hit in our H292 screen (KRAS WT), supporting the need for investigation of KRAS-independent genetic vulnerabilities in STK11/KEAP1 co-mutant LUAD. We compared the log-fold change (LFC) in hits from the H358 screen to that of the H292 screen to identify shared dependencies (Figure 6B). Notably, although AKR1C1/2/3 guides were in the sgRNA library, those genes were not detected as significant hits in either screen, again consistent with a possible functional redundancy among these family members. Figure 6 SCD1 Activity Is Essential for Survival of STK11/KEAP1 Co-mutant Adenocarcinoma Two of the top 50 hits in both H358-DKO and H292-DKO compared with their NTC controls have known critical functions in ferroptosis protection, NQO1 and SCD1. NADPH quinone dehydrogenase 1 (NQO1) is an NRF2-regulated gene that has a role in quinone and hydroquinone reduction, preventing the production of free radical species. NQO1 has been previously implicated as a potential target for cancer therapeutics (Oh and Park, 2015). SCD/SCD1 is a regulator of lipid composition, which converts saturated fatty acids (SFAs) to monounsaturated fatty acids (MUFAs) (Igal, 2010). MUFAs have been shown to compete with polyunsaturated fatty acids (PUFAs) to integrate into the plasma membrane, thereby decreasing the level of PUFAs available for lipid peroxidation and subsequent ferroptosis (Magtanong et al., 2019). SCD was depleted in H358-DKO with a LFC of −2.6 and in H292-DKO with a LFC of −1.9. To assess whether SCD was transcriptionally altered in STK11/KEAP1 co-mutants, we plotted RNA-seq differential-expression effect size (comparing DKO to other groups within the individual cell line) versus CRISPR LFC for H358 and H292 (Figures S6A and S6B). SCD stood out as a gene that is upregulated at the transcript level in STK11/KEAP1 co-mutants, is depleted in a dropout-dependency screen in both H358-DKO and H292-DKO, and has a role in ferroptosis protection. Unlike many of the ferroptosis-protective genes, SCD1 is not known to be regulated by NRF2. Together, these findings indicate that SCD1 may have an essential role in STK11/KEAP1 co-mutant LUAD, which is distinct from the many NRF2-regulated antioxidant response proteins. Genetic and Pharmacological Inhibition of SCD1 Prevents the Growth of STK11/KEAP1 Co-mutant Cells and Sensitizes Those Cells to Ferroptosis Induction SCD1 has gained recognition in the past decade as a central regulator of cancer metabolism (Igal, 2016) and, more recently, as a protein responsible for protection against ferroptosis (Tesfay et al., 2019). In our RNA-seq dataset, we identified SCD as a top differentially expressed gene in our DKO subgroup compared with all other groups (q value = 3.6E−4, β = 0.84) (Figure 6C). At the protein level, both full-length SCD1 and cleaved SCD1 (Luyimbazi et al., 2010) were upregulated in DKO mutants compared with NTC and KEAP1KO mutants (Figure 6D). Cleaved SCD1 has been noted in the literature, but distinct functions for these two forms have not been reported. Interestingly, SCD1 was upregulated at the protein level in H358-STK11KO mutants compared with NTC as well, suggesting a role for STK11 in regulating SCD1 levels. To genetically validate the essentiality of SCD1, we transduced our Cas9-containing H358 clones with a BFP-expression vector containing an sgRNA against SCD or a safe-targeting sgRNA as a control. This approach allowed us to track percentage of BFP over time as a marker for SCD1 knockout. Effective knockout of SCD1 was validated by western blot (Figure S6C). BFP expression was tracked over a period of 2 weeks to determine whether cells with knockout of SCD1 could survive in culture. Proliferation and viability of H358-NTC clones were unchanged with SCD1 knockout, confirming that SCD is not an essential gene in all cancer cells. Strikingly, we saw a dropout of ∼90% of BFP+ cells in the DKO population over a period of 12 days after transduction (Figure 6E), suggesting that STK11/KEAP1 co-mutant cells cannot survive if SCD1 expression is lost. We saw ∼30%–40% dropout of BFP+ cells in STK11KO and KEAP1KO single mutants suggesting that SCD1 plays an important role in each single mutation but is less essential to survival (Figure 6E). These results confirm that SCD is an essential gene for the survival of STK11/KEAP1 co-mutant cells. To ensure that this genetic dependency on SCD1 is not specific to our particular DKO cells, we performed a clonal competition experiment, as described in our previous work (Hulton et al., 2020), in A549-pSpectre cells expressing doxycycline-inducible Cas9. Briefly, these cells were transduced with a lentivirus containing either mCherry-sgNTC (non-targeting control) or BFP-sgSCD1. The positively selected populations were mixed 50/50 and treated with or without dox for 17 days. Dox-induced Cas9 was detected by western blot, and SCD1 knockout was confirmed at 10 days after dox treatment (Figure S6D). At 17 days, or 1 week after SCD1 knockout, we observed no change in the mCherry/BFP ratio in the −dox population. In the +dox population, we observed a depletion of BFP+ cells from 50% to 11%, confirming the lethality of SCD1 knockout in STK11/KEAP1 co-mutant A549 cells (Figure S6E). The process of lipid synthesis, storage, and degradation are finely regulated to maintain cellular homeostasis and protect cells from ferroptotic death. SCD1 has been shown to promote ferroptosis protection by promoting MUFA production, thereby regulating the levels of membrane PUFAs available for lipid peroxidation. To determine whether SCD1 expression alters pharmacologically induced lipid peroxidation, we knocked out SCD1 in A549 cells (STK11/KEAP1 co-mutant) and overexpressed SCD1 in H358 cells (STK11 and KEAP1 WT) and treated these cells with 500 nM RSL3 for 2 h. We observed that SCD1 knockout sensitizes A549 cells to RSL3, whereas SCD1 overexpression protects H358 cells from RSL3-induced lipid peroxidation (Figures 7A and 7B). These results validate the ferroptosis-protective mechanism of SCD1 and the specificity of this effect to STK11/KEAP1 co-mutant LUAD. Figure 7 Pharmacologic Inhibition of SCD1 Is Effective in STK11/KEAP1 Co-mutants In Vivo and In Vitro Alone or in Combination with a Ferroptosis Inducer To test the effects of SCD1 pharmacological inhibition, we treated H358 isogenic clones with an SCD1 inhibitor (CVT-11127) at 1 μM for 4 days and observed a significant decrease in cell viability in the DKO group compared with NTC (p = 0.0004) (Figure 7C). We note that inhibition of SCD1 by CVT-11127 decreased expression of the cleaved form of SCD1 (Figure S7A). To ensure that the response to SCD1 inhibition is not cell line specific, we tested a panel of LUAD cell lines with or without STK11/KEAP1 co-mutation with CVT-11127 at 1 μM for 4 days. Viability of STK11/KEAP1 cell lines was significantly decreased (p < 0.0001) compared with control lines lacking those mutations (Figure S7B). That the pharmacological responses observed with CVT-11127 are not as pronounced as the more dramatic genetic evidence of SCD1 dependence points to the need for more potent and selective SCD1 inhibitors. However, these results agree with and further support out genetic validation of SCD1 as a selective dependency in the context of STK11/KEAP1 co-mutant LUAD. We next explored whether pharmacologic SCD1 inhibition could prime STK11/KEAP1 co-mutant LUAD for response to other ferroptosis-targeting agents. The combination of CVT-11127 (1μM) with the ferroptosis inducer erastin (2 μM) completely reversed the erastin resistance observed in DKO cells relative to single-mutant or NTC cells, reducing viability in DKO cell from 75% with single-dose erastin to 23% with the combination after 4 days treatment (p < 0.005) (Figure 7D). Interestingly, STK11KO clones did not respond to the combination, suggesting a model of complementary roles: STK11 loss upregulating SCD1 expression, and KEAP1 loss promoting dependence on SCD1 activity, together making co-mutants specifically SCD1 dependent. Considering that both AKR1C1/2/3 and SCD1 have been implicated in ferroptosis protection, we hypothesized that the combination of MPA and CVT-11127, the small molecule inhibitors of pan-AKR1C and SCD1, respectively, would have a more potent killing effect on STK11/KEAP1 co-mutant cells compared with either single agent. Supporting this hypothesis, we saw a dose-dependent decrease in the IC50 of CVT-11127 in H358-DKO cells when combined with MPA. The IC50 of CVT-11127 was decreased ∼1,000-fold in the presence of 0.1 μM MPA and ∼5,000-fold in the presence of 10 μM MPA (Figure S7C). This combination was selectively effective against H358-DKO cells compared with H358-STK11KO, H358-KEAP1KO and H358-NTC cells (Figure S7D). These data suggest that AKR1C family members and SCD1 have complementary roles in maintaining cell survival of STK11/KEAP1 co-mutant LUAD. Optimization of selective combinatorial strategies targeting ferroptosis in STK11/KEAP1 co-mutant LUAD deserves further investigation. In Vivo Inhibition of SCD1 Defines a Selective Therapeutic Target in STK11/KEAP1 Co-mutant Adenocarcinoma To further explore the translational implications of our in vitro findings, we assessed the efficacy of an SCD1 inhibitor in vivo. Given that the SCD1 inhibitor CVT-11127 is not suitable for in vivo experimentation, we chose an alternative SCD1 inhibitor, A939572, which has been formulated for in vivo use. A939572 has demonstrated preclinical efficacy in vivo in various cancer types, including clear cell renal cell carcinoma and EGFR-mutant lung cancer in combination with tyrosine kinase inhibitors (TKIs) (von Roemeling et al., 2013; She et al., 2019). We treated mice harboring H358-NTC tumors or H358-DKO tumors with vehicle or A939572 and assessed the tumor growth and overall survival over time (n = 5 per group). Our findings show that the single agent SCD1 inhibitor provides significant selective growth inhibition in H358-DKO tumors (p = 0.008) (Figure 7E), resulting in a survival benefit in this cohort relative to control (Figure 7F). In the H358-NTC model, however, mice treated with A939572 grew similarly to those treated with vehicle control (Figures S7E and S7F). Taken together, our genetic and pharmacologic data across isogenic cell lines nominate SCD1 as a therapeutic target in STK11/KEAP1 co-mutant LUAD irrespective of KRAS status. Our findings support the development of potent and specific SCD1 inhibitors as a therapeutic strategy that may be selectively effective in this exceptionally high-risk patient cohort. Discussion Targeted therapies for LUAD have been successful in a subset of patients with tumors harboring some single driver mutations, most notably mutations leading to activation of oncogenic kinases. Unfortunately, that approach does not provide a therapeutic strategy for patients whose tumors harbor more-complicated tumor mutation profiles, including concomitant loss-of-function mutations in STK11 and KEAP1. In this study, we used multiple single- and double-gene knockout clones across two independent cell lines to specifically interrogate the differences in gene expression and gene dependency of STK11/KEAP1 co-mutant LUAD relative to isogenic single-mutant and WT counterparts. We identified a therapeutic vulnerability for this particularly aggressive subtype of lung cancer. Our approach may pave the way for future studies to identify therapeutic strategies for other inadequately treated and genetically defined malignancies. RNA sequencing identified a transcription profile specific to STK11/KEAP1 co-mutation, independent of KRAS status, in which co-mutants express higher levels of NRF2-regulated genes and upregulate ferroptosis-protective mechanisms. We identified ferroptosis evasion as a particularly enriched pathway in STK11/KEAP1 co-mutant LUAD. These cells upregulate several ferroptosis-protective genes, resulting in resistance to ferroptosis inducers. Multiple genes regulated by NRF2 encode proteins implicated in ferroptosis control, including regulators of glutathione metabolism SLC7A11, GCLC, GCLM, and GSS (Dodson et al., 2019). More notably, we were able to identify an NRF2-driven gene signature in STK11/KEAP1 co-mutant cells that substantially enhances the component of NRF2-regulated transcription beyond that of KEAP1 mutants and, therefore, increases the genetic signatures of glutathione metabolism and ferroptosis protective mechanisms. These results indicate a cooperativity between STK11 and KEAP1 loss-of-function, independent of KRAS mutation status, in which cells augment and become more dependent on certain aspects of NRF2-dependent ferroptosis-protection. Within this ferroptosis-protective gene list, we identified members of the aldo-keto reducatase-1C (AKR1C) family as having striking increases in transcript and protein expression in vitro, and corresponding increased protein expression in patient tumors harboring STK11/KEAP1 co-mutation. AKRs have a regulatory role in ferroptosis through promoting detoxification of reactive intermediates of aldehyde and ketone agents (Jung et al., 2013). Gene expression of AKR1C members was greater in co-mutant cells compared with KEAP1 mutants alone further supporting the enhanced NRF2-depedence we identified in STK11/KEAP1 co-mutants. Despite the impressive increase in protein expression across this family, our efforts to pharmacologically target AKR1C as a single-agent therapeutic strategy were limited by the potency and selectivity of the available drugs. We performed CRISPR/Cas9 screens to identify genes that are essential to maintain growth and/or survival of STK11/KEAP1 co-mutant cells. We rationalized that a dropout screen using a curated druggable genome library could uncover selective genetic vulnerabilities in this aggressive form of LUAD that we could then validate with available drugs. Through this approach, we identified SCD1, a master regulator of lipid metabolism and, interestingly, a known regulator of ferroptosis. The transcriptional regulation of SCD1 is not fully understood; however, the promoter region contains transcription factor binding sites for several well-known transcription factors including NF-1, AP-2, SREBP, and PPAR. A number of mitogens have been shown to stimulate SCD1 expression, including epidermal growth factor, retinoic acid, and members of the fibroblast growth factor family (Igal, 2010). In this study, we identified a link between STK11 loss of function and expression of SCD1; STK11 single-mutant cells upregulated SCD1 compared with control cells, and STK11/KEAP1 co-mutant cells upregulated SCD1 to a greater extent. Through both genetic and pharmacologic manipulation, we were able to validate SCD as an essential gene specific to STK11/KEAP1 co-mutant tumors. In addition, we found a strongly synergistic effect when inhibiting both AKR1C and SCD1, two ferroptosis-protective genes primarily affected by KEAP1 loss and STK11 loss, respectively. Taken together, our findings show that in STK11/KEAP1 co-mutant LUAD, loss of each of these two genes has a distinct and potentially complementary role in promoting ferroptosis evasion allowing cells to survive and persist. Upregulation of this pathway appears to be necessary for maintenance of STK11/KEAP1 co-mutation in this context, as demonstrated by the strong selective lethality of genetically targeting a key regulator node, SCD1. During the past decade, numerous studies have defined ways in which SCD1 contributes to the progression of cancer through effects on lipid metabolism, cell proliferation, migration, invasion, and metastasis (Tracz-Gaszewska and Dobrzyn, 2019). Despite robust pre-clinical findings supporting SCD1 as a therapeutic target for cancer, development of highly potent and specific SCD1 inhibitors has not been a primary therapeutic focus, and clinical deployment of existing SCD1 inhibitors has been limited to the treatment of type 2 diabetes (Zhang et al., 2014). Our study identifies SCD as an essential gene in STK11/KEAP1 RSL3 co-mutant LUAD. Further studies to design and test targeted SCD1 inhibitors, either alone or in conjunction with agents targeting ferroptosis, represents a promising strategy to improve outcomes in this cohort of patients with limited therapeutic options and poor prognosis.