STK11/LKB1 Mutations in NSCLC Are Associated with KEAP1/NRF2-Dependent Radiotherapy Resistance Targetable by Glutaminase Inhibition

Lung cancer is the leading cause of cancer-related mortality worldwide, estimated to be responsible for nearly 1.76 million deaths annually (1). Among all lung cancer cases, greater than 80% are classified as non–small cell lung cancer (NSCLC). For patients with locally advanced NSCLC, radiotherapy alone or in combination with chemotherapy and adjuvant durvalumab are mainstays of treatment (2, 3). However, locoregional recurrence (LRR) after radiotherapy remains a major clinical challenge, and biomarkers, other than KEAP1 (4), TP53, or KRAS (5–8), predicting treatment response have yet to be identified. Understanding the mechanisms underlying radioresistance in NSCLC is critical for identifying novel effective therapeutic strategies and developing predictive biomarkers to select patients likely to benefit from these therapies.

The Liver kinase B1 (LKB1) protein, encoded by the STK11 gene, is the second most commonly altered tumor suppressor in NSCLC, with mutations or genomic loss occurring in 17%–23% of NSCLC cases (9–11). STK11/LKB1 mutations are typically loss of function and frequently cooccur with KRAS-activating mutations (12). KRAS-mutant tumors with cooccurring mutations in STK11/LKB1 (KL) are associated with an aggressive clinical course, shorter survival rates, and an immunosuppressed phenotype (13–15), and to date there are no validated therapeutic strategies for this tumor subtype. We and others have demonstrated that loss of LKB1 impacts tumor progression through alterations in energy metabolism, cell motility, cytokine suppression, and tumor immune suppression (16–20). STK11/LKB1-mutant NSCLC tumors are also enriched for KEAP1 mutations or biallelic loss. KEAP1 is a negative regulator of NRF2, a transcription factor that binds to antioxidant response elements on the DNA and initiates the transcription of a number of genes involved in regulation of redox balance and cellular detoxification (21). The KEAP1/NRF2 pathway is an established mechanism of radioresistance in many cancers, including lung cancer due to enhanced expression of reactive oxygen species (ROS) scavengers and detoxification pathways in tumor cells with increased NRF2 (4, 22–25). In NSCLC, LKB1-deficient tumors with KEAP1 mutations express higher levels of NRF2-regulated genes (14, 26), and KEAP1/NRF2 mutations were predictive of local recurrence after radiotherapy in patients with NSCLC (4).

Recent findings indicate that loss of LKB1 can induce metabolic changes leading to increased ROS production, while loss of KEAP1 expression can protect against ROS-mediated damage (18, 27, 28). Radiation induces ROS generation within cells, and inadequate removal of ROS results in oxidative stress and DNA damage resulting in cell death, whereas increased expression of antioxidant enzymes or of the presence of free radical scavenger results in lower ROS levels and radioresistance (29, 30). Preclinical reports have suggested that loss of LKB1 in KRAS-driven lung adenocarcinoma may render tumor cells less responsive to radiation (31), but the impact of LKB1 loss in patients with NSCLC receiving radiotherapy is unknown.

Glutamate is a precursor for glutathione (GSH) synthesis, which is the major cellular antioxidant for redox homeostasis. Enhanced NRF2 expression protects tumor cell DNA from ROS-mediated damage, in part through NRF2-mediated upregulation of glutamate-cysteine ligase (GCLC), the enzyme that catalyzes the production of GSH from glutamate (32). Because of this “glutamine addiction,” glutaminase (GLS), the rate-limiting enzyme that converts glutamine to glutamate, has emerged as a potential therapeutic target (33). We recently reported that alterations in LKB1 and KEAP1 cooperatively drive metabolic adaptations in tumor cells that result in glutamine dependence and vulnerability to glutaminase inhibition in KRAS-mutant NSCLC (18). Although the inhibition of glutamine metabolism has been reported as an effective radiosensitizing strategy in lung tumor cells (34), identification of useful biomarkers for predicting therapeutic response to this treatment remain a clinical need.

Here, we report that mutations in STK11/LKB1 are a predictive marker for both local and distant recurrences in patients with NSCLC treated with radiotherapy. Furthermore, we show that LKB1 loss promotes resistance to ionizing radiation and radiation-induced intracellular ROS generation at least in part due to a KEAP1/NRF2 pathway-dependent mechanism that can be targeted via GLS inhibition. Thus, targeting the KEAP1/NRF2 pathway, or glutamine metabolism through GLS inhibitors, are two potential approaches to radiosensitize LKB1-deficient NSCLC tumors.

We queried the MD Anderson Lung Cancer Moon Shot GEMINI Project, a prospective database for patients with lung cancer incorporating collection of molecular profiling and clinical information, for patients with NSCLC undergoing definitive radiation treatment. Patients treated with neoadjuvant chemoradiation who underwent surgery were not included in the analysis. This project is approved by the MD Anderson Cancer Center Institutional Review Board, and all patients signed informed consent form. This project was conducted in accordance with ethical guidelines including Declaration of Helsinki and U.S. Common Rule. Mutation status was determined using one of the next generation sequencing panels of 50, 134, or 409 genes used for routine clinical care at MD Anderson Cancer Center (Houston, TX).

The four NSCLC human cell lines (A549, H460, H23, and Calu6) were purchased from the ATCC and maintained in RPMI1640 with 10% FBS, 1% glutamine and 1% penicillin-streptomycin. Isogenic pairs were established by stable expression of wild-type LKB1 cDNA in the naturally LKB1-deficient A549 and H460 cell lines and stable short hairpin (shRNA)-mediated LKB1 knockdown in Calu-6 were achieved by lentiviral transduction as reported previously (18).

Fingerprinting and Mycoplasma test using MycoAlert mycoplasma detection kit (Lonza) were performed periodically to authentication. For in vivo injection, cells were used within 10 passages from thawing. Cell lines were disregarded after 20 passages from thawing.

Western blot analysis of whole-cell extracts was performed as reported previously (18). The following primary antibodies included LKB1 (D60C5; 1:1,000, #3047, Cell Signaling Technology), KEAP1, (1:1,000, #10503-2-AP, Proteintech), NQO1 (A180; 1:1,000, #3187, Cell Signaling Technology), NRF2 (EP1808Y; 1:1,000, ab62352, Abcam), GCLC (1:1,000, ab41463, Abcam), and β-actin (1:5,000, Abcam) were used.

The radiation response and the effect of CB-839 as radiosensitization was assessed in vitro using clonogenic survival assays. Cells were seeded in duplicates at three dilutions into 6-well plates and incubated for 12–16 hours prior to treatment. The cells were treated with either CB-839 (10 nmol/L) or DMSO for 4 hours prior to irradiation. The cells were then irradiated at doses from 0 to 6 Gy using a JL Shepherd Mark I-68A with ¹³⁷Cs source at a dose rate of 3.254 Gy/minute. The medium was then replaced by fresh medium and cells maintained in normal cultural conditions for 10–14 days. The colonies were stained by crystal violet (0.5% w/v). Colonies containing at least 50 cells were counted. The surviving fraction was calculated as previously reported (18, 35) using SigmaPlot 10.0. The assays were repeated three times for each cell line.

Cells were treated with either DMSO or CB-839 (1 μmol/L) for 4 hours followed by irradiation (8 Gy). Intracellular ROS levels were measured 72 hours after irradiation using 2′, 7′-dichlorodihydrofluorescein diacetate (DCFDA; Sigma). The cells were incubated with DCFDA (20 μmol/L) for 30 minutes at 37°C. Fluorescence of oxidized DCF was measured at an excitation wavelength of 488 nmol/L and an emission wavelength of 535 nmol/L for the DCFDA using a FAC Scan flow cytometer (BD Biosciences).

Cells were treated with either DMSO or CB-839 (1 μmol/L) for 4 hours with or without irradiation (8 Gy) as described above. Cells were then collected and stained with phycoerythrin (PE)-conjugated Annexin-V and 7-aminoactinomycin D (7-AAD) according to the manufacturer's instruction. The flow cytometry was analyzed using a FAC Scan flow cytometry (BD Biosciences). The assays were done in triplicate.

CB-839 was provided by Calithera Biosciences (EMD Millipore). CB-839 were resuspended in DMSO to a final concentration of 10 mmol/L. Drug aliquots were stored at −80°C and each aliquot was used only once.

Athymic female nude mice 6–8 weeks old from (Invigo) were used for tumor studies. Animals were maintained in an Association for the Assessment and Accreditation of Laboratory Animal Care approved facility in accordance with current regulations of the U.S. Department of Agriculture and Department of Health and Human Services. Experimental methods were approved by and in accordance with institutional guidelines established by the Institutional Animal Care and Use Committee (approved protocol numbers: 00001213-RN02 and 00001217-RN02). For A549 xenografts, mice were subcutaneously inoculated with a total of 7 × 10⁶ A549-CTR, A549-LKB1, and A549-KEAP1 cells into their right hind legs. When the tumor reached 250–300 mm³, mice were randomized into different treatment groups (7–8 animals per group). Mice were irradiated with 5 Gy of gamma irradiation by using a Mark 1-68A cell irradiator with a Cesium-137 source at a dose rate of 3.254 Gy per minute (J L Shepherd & Associates) to deliver local irradiation to the flank or thigh of lead-shielded mice. Two axes of the tumor (L, longest axis; W, shortest axis) were measured three times per week after ionizing irradiation using Vernier calipers. Tumor volume was calculated using the formula: (L × W²)/2 (mm³). The tumor growth curve was calculated using GraphPad Prism 6.

Continuous and ordinal variables were summarized as median and range, categorical variables were reported as counts and percentages. The relationship between variables was assessed with χ² or Fisher exact test, as appropriate. Outcomes were analyzed in terms of LRR, distant metastases (DM), disease-free survival (DFS), and overall survival (OS). LRR was defined as described previously (36). DM was defined as disease recurrence in a different lobe or failure outside the chest. Time-to-failure endpoints were calculated from the date of diagnosis to the date of the first disease recurrence. DFS was calculated from the date of diagnosis to the date of the first disease recurrence, development of second lung primary malignancy, or death from any cause. Second primary lung malignancy was defined as development of a lung nodule that did not meet radiological criteria for metastasis, or that was biopsy proven to have distinct histologic subtype [adenocarcinoma vs. squamous cell carcinoma (SCC)]. OS was calculated from the date of diagnosis to the date of death from any cause or last follow-up. Patients who did not develop an event were censored at the date of last follow-up. The survival probability was computed using the Kaplan–Meier method and heterogeneity in survival rates among strata was assessed using the log-rank test. In the presence of competing risks (death) when performing survival analyses for any disease recurrence, an alternative cumulative incidence competing risk method was used to overcome the overestimated probabilities of disease recurrence. Time to event endpoints were calculated with the Kaplan–Meier method and the log-rank test. HRs and corresponding 95% confidence intervals (95% CI) were calculated according to the Wald method. P ≤ 0.05 was considered as statistically significant.

Results obtained from in vitro and in vivo analysis were shown as means ± SD or SEM of three or more independent experiments. Data were analyzed using unpaired Student t test, Mann–Whitney U test, or one-way ANOVA. All P values are two tailed and for all analyses, P ≤ 0.05 is considered statistically significant, unless otherwise specified. Wilcoxon analysis was performed to determine association of mutation status and radiation sensitivity. All statistical analyses were conducted using GraphPad Prism version 6.00 for Windows (GraphPad Software), SAS 9.4 (SAS Institute Inc.), and R version 3.3.3.

To investigate the association between STK11 mutations and outcome after radiotherapy in NSCLC, we examined 194 consecutive patients consented for the GEMINI database at MD Anderson Center (Houston, TX) who had undergone molecular profiling and received definitive radiation treatment from March/2007 until April/2017 (Supplementary Table S1). The median age of the study population at diagnosis was 64 years, with the majority of patients being female (53.6%). Most patients had a good performance status (PS 0–1: 94.5%) and were current (15%) or former (67.4%) smokers. The most common histology was adenocarcinoma (71.5%), 9.3% presented squamous histology and the vast majority had stage III disease (84.5%). We focused the analysis on stage III patients treated prior to the standard use of durvalumab as maintenance therapy so that LRR and DFS would not be confounded by maintenance durvalumab treatment, given the association between STK11 and response to PD-1/PD-L1 inhibition (37).

We assessed the cohort with stage III disease treated with definitive radiation (N = 164), with or without concurrent chemotherapy (Table 1). After a median follow-up of 4.0 years (range: 0.3–13.1 years), the presence of STK11/LKB1 mutation was associated with significantly higher cumulative LRR rates (P = 0.0108; Table 2; Fig. 1A), shorter DFS (HR 2.53; 95% CI, 1.375–4.657; P = 0.0029, Cox regression model; Table 3; Fig. 1B) and OS (HR 2.198; 95% CI, 1.097–4.405; P = 0.0263, Cox regression model; Table 4; Fig. 1C) than patients with STK11/LKB1 wild-type tumors. DM failure rate was numerically higher in the STK11/LKB1-mutant group compared with the STK11/LKB1 wild-type group but this was not statistically significant (1-year DM rate: 66.7% vs. 39.3%, P = 0.1129; Supplementary Table S2; Fig. 1D). Estimated cumulative 1-year LRR rate using competing risk analysis was significantly higher in the STK11/LKB1 group [25.0% (95% CI: 4.8–53.1) versus 10.8% (95% CI: 6.5–16.5), P = 0.0108; Table 2]. In addition, 5-year DFS and OS rates were significantly lower in the STK11/LKB1 group [5-year DFS: 0% (95% CI: NA) vs. 14.6% (95% CI: 8.9–21.6), P = 0.0020; 5-year: OS: 15.6% (95% CI: 1.1–46.4) vs. 42.0% (95% CI: 32.7–51.0), P = 0.0228; Supplementary Table S3]. Of the 7 patients whose tumor harbored STK11/LKB1 mutations and were censored for 1-year LRR competing risk analysis, 6 patients (85.7%) had subsequent LRR during their disease course. STK11/LKB1 mutation status remained significantly associated with worse OS in a multivariate analysis including STK11/LKB1 status and age (the two characteristics associated with OS in univariate analysis; Table 5), and was the single variable associated with poor DFS in univariate analysis (Table 3). In a subgroup analysis of patients with stage III disease that received concurrent chemoradiation and had PS 0–1, 1-year LRR rate was also significantly higher for the STK11/LKB1 group [27.3% (95% CI, 5.2–56.6) vs. 9.0% (95% CI, 4.9–14.6), P = 0.0023; Supplementary Fig. S1].

We further analyzed clinical outcomes among patients with stage III NSCLC treated with definitive radiation according to KRAS subgroups. A total of 51 of 164 patients (31.09%) had NSCLCs harboring a KRAS mutation. Of the 51 patients with NSCLC harboring a KRAS mutation, 7 patients (13.72%) had STK11/LKB1 comutation (KL group), 20 patients (39.21%) had TP53 comutation (KP group), and the remaining 24 patients (47.05%) had no TP53 and no STK11/LKB1 comutations (K group; Fig. 2A). Given the size of the cohorts, this analysis of KRAS subgroups had a limited statistical power. Nevertheless, it showed that patients with KL and KP tumors displayed shorter DFS and higher LRR rate compared with K tumors, but this was not statistically significant [5-year DFS: KL 0% vs. KP 10% vs. K 27.3%, Kaplan–Meier P = 0.0701; 1-year LRR rate: K 8.3% (95% CI, 1.4–23.7) vs. KL 14.3% (95% CI, 0.3–53.1) vs. KP 10.5% (95% CI, 1.9–33.7), P = 0.2983]. Pairwise comparison of patients with KL versus KP tumors revealed statistically significant shorter DFS in KL patients (P = 0.0211; Fig. 2B and C; Tables 2 and 3; Supplementary Table S3). No differences were found for DM rate or OS for KL vs. KP tumors (Supplementary Fig. S2A and S2B). Because KRAS mutations are rare in squamous NSCLC, we then excluded patients with SCC from our analysis. A total of 3 of the 51 (5.88%) KRAS-mutant NSCLCs revealed a squamous histology. Of the 48 patients with nonsquamous NSCLC harboring KRAS mutation, 7 patients (14.58%) harbored KL tumors, 17 patients (35.41%) KP, and the remaining 24 patients (50.00%) K tumors (Supplementary Fig. S3A; Supplementary Table S4). Likewise, patients with KL tumors and KP tumors showed numerically shorter DFS and higher LRR rate compared with K tumors, but this was not statistically significant [5-year DFS: KL 0% vs. KP 5.9% vs. K 27.3%, Kaplan–Meier P = 0.0843; 1-year LRR rate: K 8.3% (95% CI, 1.4–23.7) vs. KL 14.3% (95% CI, 0.3–53.1) vs. KP 12.5% (95% CI, 1.9–33.7), P = 0.3573]. DFS for patients harboring KL tumors was significantly shorter when compared with KP patients, P = 0.0462 (Supplementary Fig. S3B and S3C; Supplementary Tables S5 and S6).

To better dissect the impact of KRAS and TP53 mutations on clinical outcome from radiotherapy, we also performed subgroup analysis of tumors harboring cooccurring KRAS and TP53 mutations (KP tumors) versus tumors harboring KRAS or TP53 wild-type (non-KP tumors), and showed no difference in 1-year LRR rate [10.5% (95% CI, 1.6–29.1) vs. 12.1% (95% CI, 7.3–18.1), P = 0.9026; Supplementary Fig. S4A; Table 2]. Comparison of clinical outcomes for patients whose tumors harbored a KRAS mutation (KRAS Mut) vs. patients whose tumors were KRAS wild-type (KRAS WT) showed no difference in 1-year LRR rate [10.0% (95% CI, 3.6–20.3) vs. 12.8% (95% CI, 7.3–19.8), P = 0.4408; Supplementary Fig. S4B]. Finally, we compared 1-year LRR rates between three groups: patients whose tumors harbored a STK11/LKB1 mutation (STK11 Mut), patients whose tumors harbored cooccurring KRAS and TP53 mutations (KP), and patients whose harbor KRAS or TP53 or STK11 wild-type tumors (other; Supplementary Fig. S4C). In this comparison, we observed that the STK11-mutant subgroup had significantly higher 1-year LRR rate compared with KP and all other genotype subgroups [STK11 Mut 25% (95% CI, 4.8–53.1) vs. KP 10.5% (95% CI, 1.6–29.1) vs. other 10.9% (95% CI, 6.2–17.0), P = 0.0379]. STK11 Mut versus other pairwise comparison showed significantly higher 1-year LRR rate (P = 0.0143), but not KP versus other or STK11 Mut vs. KP revealed significantly differences (Supplementary Fig. S4C).

We previously reported that LKB1-deficient tumors are characterized by metabolic adaptations that result in glutamine dependence and upregulation of NRF2 and its downstream target genes as a compensatory mechanism to maintain redox homeostasis (18). To investigate the impact of STK11/LKB1 mutations over radiotherapy resistance in vitro, we first screened a panel of 33 NSCLC cell lines for radiation sensitivity (Supplementary Fig. S5A). Consistent with clinical data, STK11 and KEAP1 but not KRAS or TP53 mutations, correlated with radiation resistance (Supplementary Fig. S5B), although differences did not reach statistical significance due to a limited number of samples. To further explored radioresistance associated to LKB1 loss, we then established three pairs of isogenic KRAS-mutant LKB1-deficient/proficient NSCLC cell lines by either reexpressing wild-type LKB1 in LKB1-deficient cells (A549 and H460) or stably expressing shRNA-targeting LKB1 in LKB1-proficient NSCLC cells (Calu-6). We previously reported that reexpression of LKB1 results in decreased expression of NRF2 and NRF2-regulated genes, including NAD(P)H Quinone Dehydrogenase 1 (NQO1) in A549 and H460 cells, and knockdown of LKB1 in Calu-6 cells enhanced NRF2 and NQO1 expression (18). To investigate the underlying mechanism by which LKB1 loss contributes to radiation resistance, A549 cells with or without LKB1 expression were exposed to ionizing radiation and effects on cell proliferation were determined by clonogenic survival assay (Fig. 3A). LKB1 reexpression showed an increased radiosensitivity with a DER (dose enhancement ratio) of 1.4 at a surviving fraction of 0.5 compared with LKB1-deficient cells [P = 0.0057 at 2 Gy and P < 0.0001 at 4 and 6 Gy, and SD for control or LKB1 0 Gy; 0.16 or 0.21, 2 Gy; 0.15 or 0.17, 4 Gy; 0.07 or 0.05 and 6 Gy; 0.04 or 0.03, respectively (Fig. 3B)]. Consistently, LKB1 overexpression in H460 cell lines resulted in enhanced radiosensitivity compared with vector control cells, with a mean of the surviving fraction for LKB1-overexpressing H460 cells of 32%, compared with 55% for the H460 vector control cells at the 2 Gy ionizing radiation dose (P = 0.007; Fig. 3C). Similarly, knockdown of LKB1 increased the radioresistance of the LKB1-proficient Calu-6 cell line. The surviving fraction was significantly higher in LKB1 knockdown Calu-6 cells (23.7%) compared with control cells (13.4%; P = 0.023) when cells were treated with 4 Gy dose of ionizing radiation (Fig. 3D).

To determine whether the KEAP1/NRF2 pathway contributes to ionizing radiation resistance induced by LKB1 loss, we stably expressed wild-type KEAP1 in A549 cells (LKB1-mutant/KEAP1-deficient). Reexpression of KEAP1 resulted in decreased expression of NRF2, NQO1, and glutamate-cysteine ligase catalytic subunit (GCLC) as determined by Western blotting (Supplementary Fig. S6A) and significantly enhanced radiosensitivity compared with A549 vector control cells (P < 0.0001 at 2 and 4 Gy, P = 0.01221 at 6 Gy, SD for control or LKB1; 0 Gy; 0.21 or 0.12, 2 Gy; 0.13 or 0.15, 4 Gy; 0.08 or 0.08 and 6 Gy; 0.05 or 0.03, respectively; DER at a surviving fraction of 0.5 = 1.4; Fig. 3E and F). Similarly, depletion of KEAP1 increased the radioresistance of H23 cells (Supplementary Fig. S6B), a KEAP1-proficient NSCLC cell lines. H23 control cells displayed greater sensitivity (1.6% of surviving fraction) to ionizing radiation treatment (2 Gy) than H23 cells with knockdown of KEAP1 (7.4%), although this was not statistically significant (Supplementary Fig. S6B). Likewise, knockout of Lkb1 (KL) and/or Keap1 (KK or KLK) in a Kras-mutant murine cell line (LKR13) significantly enhanced radioresistance compared with Kras-mutant control cells (K; Supplementary Fig. S6C). Collectively, these data support that LKB1 loss promotes radioresistance via a mechanism dependent, at least in part, on the KEAP1/NRF2 pathway in NSCLC tumor cells.

Given recent reports that LKB1 loss results in increased NRF2 expression which, in turn, reduces ROS levels (18), and because radiation-derived cytotoxic effects are mediated, at least in part, by induction of intracellular ROS (29, 30), we next sought to determine whether loss of LKB1 or KEAP1/NRF2 activation protects tumor cells from ROS-mediated damage and contributes to radioresistance. LKB1 and KEAP1 A549 isogenic pairs were irradiated at a dose of 8 Gy and intracellular ROS levels were measured after 72 hours (Fig. 3G). As expected, reexpression of LKB1 reduced intracellular ROS levels, while KEAP1 knockdown significantly increased redox stress in nontreated cells. In all three cell lines, exposure to ionizing radiation resulted in a rise in ROS levels. However, there was a significantly greater increase in ROS production following ionizing radiation in LKB1- and KEAP1-expressing cells compared with A549 control cells lacking LKB1 and KEAP1 (Fig. 3G). These findings suggested that the generation of ROS, as a result of ionizing radiation exposure, is impaired in LKB1- and KEAP1-mutant cells and that this may contribute to the enhanced radiation resistance observed in LKB1- and KEAP1-deficient NSCLC cells. We next evaluated ionizing radiation–induced cell apoptosis by Annexin-V/7-AAD staining. In NSCLC A549 cells, ionizing radiation exposure resulted in a reduced percentage of live cells from baseline in LKB1-proficient cells than in LKB1-deficient parental cells (P = 0.0603). A similar experiment was performed using stably expressed wild-type KEAP1 in A549 cells. NRF2 pathway suppression (via KEAP1 reexpression) significantly enhanced cytotoxic effect induced by ionizing radiation (P = 0.0164; Fig. 3H). Taken together, our findings indicate that LKB1 and KEAP1 loss may protect tumor cells from the cytotoxic effects of ionizing radiation due to a greater compensation of radio-derived intracellular ROS generation by activation of the KEAP1/NRF2 axis.

We next sought to determine whether loss of LKB1 and/or KEAP1 impacts radiation sensitivity in vivo. Supporting in vitro data, radiation did not impair the growth of A549 (A549 control) tumors, which harbor comutations in KRAS, STK11/LKB1, and KEAP1 genes (Fig. 4A), there were not significant differences in progression-free survival (PFS) rate (Fig. 4B) between irradiated and control animals. In contrast, radiotherapy was modestly effective in A549 tumors with LKB1 reexpression (A549 LKB1; Fig. 4C), and no differences were observed in the survival rate between ionizing radiation and control arms (Fig. 4D). Radiotherapy significantly impaired tumor growth in A549 KEAP1–expressing cells (A549 KEAP1; Fig. 4E) and prolonged the median PFS of A549 KEAP1 tumor-bearing mice (P ≤ 0.04; Fig. 4F). Collectively, these data provide additional evidence that loss of STK11/LKB1 and more significantly, KEAP1 are potential mediators of radioresistance in vivo.

As previously described, enhanced NRF2 can protect LKB1-deficient cells from ROS-mediated damage in part by the NRF2-mediated upregulation of GCLC enzyme that catalyzes the production of GSH from glutamate. Our previous reports indicate that conversion of glutamine to glutamate by GLS is a key step in the maintenance of redox homeostasis in LKB1-deficient tumors, particularly those with upregulation of the KEAP1/NRF2 pathway, and these tumors therefore display heightened sensitivity to GLS inhibition (18). Therefore, we next analyzed whether GLS inhibition could sensitize LKB1-deficient tumors to radiotherapy. We treated A549 cells with GLS inhibitors (also called CB-839) and then measured intracellular ROS levels 72 hours after ionizing radiation treatment GLS inhibition (1 μmol/L CB-839) markedly increased ionizing radiation–induced ROS in A549 control cells (Fig. 5A), while ROS induction was reduced when cells reexpressed LKB1 (Fig. 5B) or KEAP1 (Fig. 5C). Combination of 1 μmol/L CB-839 with ionizing radiation significantly induced intracellular ROS production in A549 control cells when compared with ionizing radiation alone (Fig. 5A). Although ionizing radiation treatment alone significantly increased ROS levels in A549-LKB1 cells, there were no greater induction of ROS levels when combined with GLSi (Fig. 5B). In A549-KEAP1 cells, GLSi treatment plus ionizing radiation resulted in a modest increase in ROS compared with ionizing radiation alone, although this increase was far smaller than the increase induced by GLSi in the A549 control cells (Fig. 5C). These data indicate that the GLSi-induced increase in ROS is at least in part dependent on the LKB1 and KEAP1 pathways.

We then explored whether inhibition of GLS enhanced ionizing radiation cytotoxic effects in radio-resistant LKB1-deficient tumor cells. Consistent with previous redox induction data, treatment with 1 μmol/L CB-839 enhanced ionizing radiation–induced apoptosis compared with control, CB-839 or ionizing radiation alone in A549 parental cells, although differences did not reach statistically significance (P = 0.119; Fig. 5D). Conversely, combined treatment of CB-839 and ionizing radiation did not significantly increase apoptosis induction compared with ionizing radiation-only–treated cells in A549-LKB1 (Fig. 5E) nor A549-KEAP1 (Fig. 5F) cells. We next treated A549 control, A549/LKB1, and A549/KEAP1 cells with CB-839 and evaluated sensitivity to ionizing radiation by clonogenic survival assay (Fig. 5G and H). Quantification of surviving fractions (Fig. 5I) revealed that GLS inhibition strongly radiosensitized A549 control cells (LKB1/KEAP1-deficient) with a DER at a surviving fraction of 0.5 of 1.8. Reexpression of LKB1 abrogated the CB-839-induced radiosensitization with a DER at a surviving fraction of 0.5 were 0.8. Similar results were found in KEAP1 reexpressing cells after GLS inhibition (Fig. 5I). As expected, GLS blockade significantly radiosensitized KK and KLK cells with a DER at surviving fraction of 0.7 of 2.19 and 1.63, respectively. Conversely, CB-839 treatment did not enhance radiotherapy effects in K or KL cells (Supplementary Fig. S7). These data reveal that in LKB1 and/or KEAP1-deficient tumors, inhibition of glutamine metabolism using GLS inhibition impairs redox homeostasis and tolerance to oxidative stress, resulting in radiosensitization of LKB1-deficient tumor cells.

In this study, we investigated the association between alterations in STK11/LKB1, the second most common tumor suppressor in NSCLC, and response to radiotherapy as well as potential therapeutic approaches to improve outcomes. We show that in patients with stage III, locally advanced NSCLC treated with radiotherapy, STK11/LKB1 mutations were associated with significantly higher cumulative rates of locoregional failure and shorter DFS and OS than tumors with wild-type STK11/LKB1. We next investigated the impact of LKB1 loss in preclinical models. In vitro, LKB1-deficient tumors displayed a radio-resistant phenotype, and modulation of LKB1 expression modified sensitivity to ionizing radiation, supporting our clinical observations and providing evidence for a causative role of LKB1 in modulating response to radiotherapy.

We also investigate the mechanism(s) underlying the radioresistance induced by LKB1 loss. Prior studies have demonstrated that STK11/LKB1 mutant tumors display metabolic alterations that render them sensitive to energetic and oxidative stress, and that upregulation of the KEAP1/NRF2 pathway (via KEAP1 loss or other mechanisms) in these tumors is critical to maintain redox homeostasis and cell survival (14, 18). More broadly, overexpression of NRF2 has been found to be one of the mechanisms for chemotherapy and/or radiation resistance in lung cancer (21). Inhibition of NRF2 induced ROS generation and resulted in enhanced sensitivity to chemotherapy and radiation-induced cytotoxicity (22, 24, 32, 38). Our data indicate that LKB1-deficient tumors are particularly dependent on the KEAP1/NRF2 pathway for radioresistance and that GLS inhibition is a potential therapeutic approach to overcome this resistance. However, we were unable to assess the impact of KEAP1 alterations in this clinical cohort as KEAP1 genetic alterations were not covered in our genomic profiling panel. Enhanced NRF2 activation protects tumor cells from ROS-mediated damage, in part by the NRF2-mediated upregulation of GCLC enzyme that catalyzes the production of GSH, one of the major antioxidant molecule, from glutamate. Interestingly, GSH plays an important role in cellular defense against radiation (39). Likewise, KEAP1 loss is associated with radiotherapy resistance in different tumor types (22, 24, 32, 38). Consistent with this observation, our data showed that reexpression of KEAP1 sensitized tumor cells to ionizing radiation in vitro and in animal studies. We also showed that modulation of LKB1 expression alone was sufficient to activate NRF2-dependent antioxidant response.

We previously observed that the LKB1 and KEAP1/NRF2 pathway cooperate to promote enhanced dependence on glutamine metabolism and vulnerability to glutaminase inhibitor CB-839 (18). Here, we observe that LKB1-deficient tumors are vulnerable to therapies that target NRF2-mediated detoxification of ROS. With these protections removed, LKB1-deficient redox stressed tumors are also potentially vulnerable to ROS-mediated DNA damage. A growing body of evidence supports the idea that inhibition of GLS with small-molecule inhibitors or by genetic knockdown leads to glutathione depletion, which leads to marked changes in response to radiation treatment (34, 40, 41). Here, we demonstrate that the GLS inhibition could sensitize LKB1-deficient tumor cells to ROS-mediated DNA damage induced by radiation, overcoming radioresistance associated with STK11/LKB1 and/or KEAP1 mutations. Therefore, inhibition of critical NRF2 targets, such as GLS, may allow targeted radiosensitization in patients with STK11/LKB1 mutations.

It is noteworthy that the retrospective clinical analysis presented here was from patients treated with radiotherapy prior to the routine use of the PD-L1 inhibitor durvalumab as maintenance therapy after chemoradiotherapy, based on the phase III PACIFIC study (42, 43). Therefore, one limitation of the analysis is that the association between LKB1 and clinical outcome may be impacted by the inclusion of durvalumab. Previously, our group and others have reported that LKB1-deficient NSCLC are characterized by a markedly suppressed immune microenvironment within the tumor (14, 44), increased expression of T-cell exhaustion markers and tumor-promoting cytokines (19) and lack of response to PD-1/PD-L1 inhibitors (14, 37, 45). In light of this association between LKB1 loss and an immunosuppressed tumor microenvironment, it is noteworthy that the immune system plays an important role in prolonged response to radiotherapy (46, 47). Ionizing radiation promotes an inflamed tumor microenvironment attracting the immune cells to the tumor site, which triggers innate and adaptive immune responses leading to tumor regression (48). While the preclinical studies here focused on cell-autonomous effects of LKB1 loss on radiotherapy response, it is possible that LKB1 loss may exert additional effects on radiotherapy response via tumor immunosuppression. Further studies using immunocompetent in vivo models are needed to further elucidate the contributions of cell autonomous versus immune-mediated mechanisms of radiotherapy resistance in LKB1-deficient tumors.

Although several prior studies have suggested a role of LKB1 expression in modulating response to radiotherapy based on preclinical experiments (31, 49, 50), none to our knowledge have examined the association of STK11/LKB1 mutations with radioresistance in clinical data. We find that these alterations are associated with both LRR and shorter DFS/OS in patients treated with definitive radiotherapy. This analysis does have several limitations, including its retrospective nature, the limited sample size of the LKB1 cohort, which limits the statistical power of the analysis, and the fact that the analysis was conducted before the current standard of care changed to maintenance durvalumab as noted above. Also, it is noteworthy that TP53 mutations have been previously reported as a marker for predicting radiotherapy resistance in the clinic (5–8). In our cohort, we observed a nonsignificant trend for shorter LRR and DFS for patients harboring KP mutations compared with K (KRAS mutant, wt TP53, wt STK11). In the KL subgroup, however, we observed a stronger association with radioresistance in our preclinical and clinical analysis. Nevertheless, they do provide a foundation for further studies potentially targeting vulnerabilities associated with LKB1-deficient tumors. We find that radiotherapy resistance in LKB1-deficient tumors is dependent on the KEAP1/NRF2 pathway in preclinical models, and we identify inhibition of glutamine metabolism with GLS inhibitors as a potential therapeutic strategy that merits further investigation for this highly refractory subgroup of patients with NSCLC.

A. Galan-Cobo reports grants from Lung SPORE P50CA07907, The LKB1 R01 CA205150, CPRIT CP160652, CCSG CA016672, and SU2C/AACR; other from The University of Texas MD Anderson Cancer Center, The Lung Cancer Moon Shot, including donations from Kyte Family, Jeff Hepper, and Normal Godinho, Rexanna's Foundation for Fighting Lung Cancer, Weaver Foundation, and Jane Ford Petrin donation during the conduct of the study. M.V. Negrao reports other from Mirati Therapeuics, Novartis, Checkmate Pharmaceuticals, Ziopharm Oncology, AstraZeneca, and Pfizer Inc. outside the submitted work. J. Zhang reports grants from Merck; grants and personal fees from Johnson & Johnson; personal fees from Bristol Myers Squibb, AstraZeneca, GenePlus, Innovent, OrigMed, and Roche outside the submitted work. J.A. Roth reports other from Genprex, Inc. and grants from Varian Medical System during the conduct of the study; other from Genprex, Inc. and grants from Varian Medical System outside the submitted work; in addition, J.A. Roth has a patent for Genprex, Inc. pending and issued and reports consultancy with and stocks from Genprex, Inc. J.V. Heymach reports grants from Lung SPORE P50CA07907, R01 CA205150, CPRIT CP160652, The Lung Cancer Moon Shot, CCSG CA016672, SU2C/AACR; other from Jane Ford Petrin donation, Kyte Family donation, Jeff Hepper donation, Normal Godinho donation, Rexanna's Foundation for Fighting Lung Cancer donation, Weaver Foundation donation; and nonfinancial support from Calithera during the conduct of the study; grants and personal fees from GlaxoSmithKline, AstraZeneca; grants, personal fees, and other from Spectrum; and personal fees from Boehringer Ingelheim, Bristol-Myers Squibb, Catalyst, EMD Serono, Foundation Medicine, Hengrui Therapeutics, Genentech, GSK, Guardant Health, Eli Lilly, Merck, Novartis, Pfizer, Seattle Genetics, Sanofi, and Takeda outside the submitted work. No disclosures were reported by the other authors.

P. Sitthideatphaiboon: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing-original draft, project administration, writing-review and editing. A. Galan-Cobo: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, writing-original draft, writing-review and editing. M.V. Negrao: Data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. X. Qu: Data curation, formal analysis, methodology, writing-review and editing. A. Poteete: Data curation, methodology. F. Zhang: Data curation, methodology. D.D Liu: Data curation, formal analysis, methodology, writing-review and editing. W.E. Lewis: Data curation, formal analysis, writing-review and editing. H.N. Kemp: Data curation. J. Lewis: Data curation. W. Rinsurongkawong: Data curation. U. Giri: Data curation, formal analysis, investigation, methodology. J.J. Lee: Resources, data curation, formal analysis, investigation, methodology, writing-review and editing. J. Zhang: Resources, data curation, formal analysis, writing-review and editing. J.A. Roth: Data curation, formal analysis, writing-review and editing. S. Swisher: Data curation, formal analysis, writing-review and editing. J.V. Heymach: Conceptualization, data curation, formal analysis, supervision, funding acquisition, project administration, writing-review and editing.

The authors thank Monique Nilsson for critical review of the article and The Faulhaber Foundation. They also thank Calithera Biosciences, San Francisco, CA, for providing CB-839.

This work was supported by The University of Texas Southwestern Medical Center and The University of Texas MD Anderson Cancer Center Lung UT; Lung SPORE P50CA07907; The LKB1 R01 CA205150; CPRIT CP160652; The Lung Cancer Moon Shot, including donations from Kyte Family, Jeff Hepper, and Normal Godinho; Rexanna's Foundation for Fighting Lung Cancer; Weaver Foundation; CCSG CA016672; a Stand Up to Cancer-American Cancer Society Lung Cancer Dream Team Translational Research Grant (SU2C-AACR-DT17-15), and a Jane Ford Petrin donation. Stand Up to Cancer is a division of the Entertainment Industry Foundation. The indicated SU2C grant is administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.