Enhanced Vulnerability of LKB1-Deficient NSCLC to Disruption of ATP Pools and Redox Homeostasis by 8-Cl-Ado

Non–small lung cancer (NSCLC) is the leading cause of cancer-related mortality worldwide (1). Despite advances in targeted therapy, patients with NSCLC have a 5-year survival rate of less than 20% (2). After TP53, STK11, which encodes serine/threonine kinase 11/liver kinase B1 (LKB1), is the most commonly mutated gene in NSCLC. STK11/LKB1 mutations occur in approximately 17% of lung adenocarcinomas (3) and poorly differentiated tumors have a higher frequency of LKB1 deficiency (4). In patients with stage III or IV NSCLC, LKB1 deficiency is associated with shorter cause-specific survival (5). STK11/LKB1 mutations or genomic loss frequently co-occur with KRAS alterations (6), resulting in a highly aggressive phenotype and reduced survival rates in both preclinical models (7) and patients with NSCLC (6). In fact, many clinical laboratory improvement amendments–based analyses of NSCLC tumor specimens, besides documenting KRAS alterations, also report back on tumors with and without STK11/LKB1 genetic abnormalities setting the stage of “precision medicine” selection of patients for STK11/LKB1 targeted therapy. Unfortunately, while numerous studies have focused on identifying vulnerabilities unique to LKB1-deficient tumors (8–11), to date there are no effective therapeutic regimens specific for this subset of patients. The urgency to discovery such regimens is additionally highlighted by the resistance of nearly all STK11/LKB1 altered NSCLC to immune checkpoint blockade (12). Thus, a crucial need for beneficial clinical translation is to identify a small molecule or targeted therapy that takes advantage of a NSCLC acquired “vulnerability” generated by STK11/LKB1 abnormalities.

LKB1, a master kinase that initiates the activation of numerous other kinases, plays a major role in the modulation of cell metabolism, polarity, differentiation, and proliferation (13). The effects of LKB1 are largely mediated through its phosphorylation and activation of the energy sensor AMP-activated protein kinase (AMPK; ref. 14). Under low cellular energy levels, AMPK restores intracellular homeostasis by balancing ATP and NADPH concentrations (15). Moreover, AMPK activation results in the inhibition of mTOR signaling, which is a key protein in controlling cellular growth and proliferation (16). Likewise, AMPK activation has been shown to attenuate KRAS-mediated tumor cell proliferation (17). Therefore, LKB1-deficient tumors are unable to modulate cell growth and proliferation to maintain cellular ATP homeostasis in response to high-energy stress. Consequently, LKB1-deficient tumors have an aberrant response to energy reduction (8, 18) and a heightened sensitivity to energy-depleting agents (10, 11, 19), a vulnerability that can be exploited clinically.

8-chloro-adenosine (8-Cl-Ado), a nucleoside analogue that induces depletion of the endogenous ATP pool (20–24), has completed a phase I clinical trial in patients with chronic lymphocytic leukemia (CLL; NCT00714103; ref. 25) and is currently in a phase I/II clinical trial in patients with acute myeloid leukemia (NCT02509546). This analog is rapidly metabolized to 8-Cl-ATP through the adenylate pathway. 8-Cl-ATP metabolite, 8-Cl-ADP, acts as a substrate for ATP synthase for conversion to 8-Cl-ATP, thereby blocking ATP synthase activity and preventing the catalysis of its natural substrate ADP to synthesize ATP (26). In addition to depleting ATP pools, 8-Cl-ATP incorporates into mRNA and prematurely terminates transcription (24), leading to a decline in transcription rates (20, 22, 23). This results in a loss of expression of genes with short lived transcripts which is detrimental for tumor cells which are dependent on these genes (27).

In the current study, we performed an unbiased drug screen on a panel of NSCLC cell lines and identified 8-Cl-Ado as being highly effective in LKB1-deficient NSCLC cells, particularly in cell lines bearing KRAS co-mutations. Moreover, the heightened sensitivity of LKB1-deficient cells to 8-Cl-Ado was associated with an attenuated signaling response to ATP depletion due to inactivation of AMPK/mTOR signaling, which also lead to increased reactive oxygen species (ROS) production. Proteomic analyses revealed that MAPK/MEK/ERK and PI3K/AKT/mTOR pathways are upregulated as a compensatory mechanism in response to 8-Cl-Ado treatment. Hence, 8-Cl-Ado combinations with AMPK, MEK, or PI3K inhibitors resulted in a synergistic antitumor effects in LKB1-deficient tumor cells. Collectively, our data indicate that LKB1-deficient tumors are highly sensitive to ATP depletion, suggesting that treatment with 8-Cl-Ado, alone or in combination with AMPK, MEK, or PI3K inhibitors, is a promising treatment strategy for patients with NSCLC harboring this specific genetic background.

8-Cl-Ado was obtained from V. Rao (Drug Development Branch, NCI, Bethesda, MD). Metformin and 2-deoxyglucose (2-DG) were obtained from Sigma-Aldrich. Trametinib, A66, PI-103 were purchased from Selleck Chemical.

NSCLC cell lines were established by J.D. Minna and A. Gazdar at The University of Texas Southwestern Medical Center, Dallas, TX, and the NCI or obtained from the ATCC and maintained in RPMI1640 media (Sigma or Mediatech) supplemented with 10% FBS and 1% penicillin-streptomycin (Life Technologies) in the presence of 5% CO₂ at 37°C. Cells were routinely tested for Mycoplasma infection and authenticated by short tandem repeat analysis by MD Anderson's Characterized Cell Line Core Facility. LKB1 proficiency status was based on LKB1 genetic status, RNA, and protein levels (Supplementary Table S1). LKB1 protein deficiency was assigned on the basis of LKB1 levels in cell lines with known STK11 mutations (Supplementary Fig. S1).

LKB1 isogenic A549, H460, H2030, and H23 cell lines (all NSCLC lines who developed in patients by virtue of endogenous STK11/LKB1 abnormalities) were generated with a pLenti-GIII-CMV-GFP-2A-Puro vector with or without a wild-type LKB1–encoding sequence (Applied Biological Materials, Inc.). LKB1-knockdown Calu6 cell line (initially STK11/LKB1 wild type) were generated with a scramble control vector or two independent short hairpin RNAs (shRNA; V3LHS_348649 and V3LHS_639000, GE Healthcare Dharmacon, Inc).

An optimized number of viable cells for each cell line were seeded in 96-well plates (3,000 cells per well) and allowed to attach overnight. Cells were treated with the indicated doses (serial 3-fold dilutions) of 8-Cl-Ado for 72 hours. CellTiter-Glo (Promega) was added to each well as per the manufacturer's protocol, and contents were briefly mixed and incubated for 15 minutes to determine cell viability. Bioluminescence was measured using a FLUOstar OPTIMA Multimode Microplate Reader (BMG Labtech). Average readings from triplicate wells were then expressed as a percentage of average bioluminescence measured from control DMSO wells treated with vehicle (DMSO), representing the highest DMSO concentration in drug-treated cells. A dose–response model was used to estimate IC₅₀ values from cell viability data. Multiple models from the DoseFinding and drc packages were fitted and the best model was selected on the basis of residual SE using the R software.

For combination studies, LKB1-deficient A549, EKVX, H1355, H1944, H2030, H2126, H460, H1573, H1993, H2122, and HCC15 cell lines, and LKB1-proficient Calu1, H1155, H1299, H1819, H1838, H2052, H508, H661, HCC4011, Hop62, Hop92, DFCI032, H2228 cell lines were treated with seven different concentrations (serial 3-fold dilutions) for the following agents either individually or in combination with 8-Cl-Ado for 96 hours. 8-Cl-Ado was combined at indicated fixed ratio with the following drugs: compound C at 100:30 ratio, PI-103 at 100:3 ratio, trametinib at 100:1 ratio, or A66 at 1:1 ratio. The concentration range for each drug was determined on the basis of specific IC₅₀ set as median dose. CalcuSyn software program (Biosoft) was used to calculate combination indexes (Is) to assess the efficacy of drug combinations.

For cell counting proliferation assay, 1 × 10⁵ cells were seeded in 6-well plates. Twenty-four hours later, 8-Cl-Ado or DMSO were added at the indicated concentrations. After 72 hours of incubation, cells were harvested and stained with Trypan Blue 0.4% and counted using Count II Automatic Cell Counter (Invitrogen).

A total of 400 cells were plated in triplicate into 24-well plates containing 0.5 mL of medium. Cells were incubated for 24 hours in a humidified CO₂ incubator at 37°C, and subsequently the medium was then replaced with medium containing indicated drug concentrations, allowing cells to continuously grow for colony formation for 10 to 14 days. To assess clonogenic survival, we stained cells with 1% crystal violet, wash with distillated water and dry out. Quantification of cell density was calculated by covered area using Image J software.

Perchloric acid was used to extract nucleosides and nucleotides from LKB1 isogenic A549 and H460 cell lines treated with or without 8-Cl-Ado, and the neutralized extracts were analyzed by high-pressure liquid chromatography (HPLC) using a Waters 600E System Controller (Waters Corp.) as described previously (21). Three independent experiments were performed for each dose assessed.

Macromolecule synthesis was measured by the addition of either 1 μCi/ml [³H]-thymidine, [³H]-uridine, or [³H]-leucine during the final 30 minutes of a drug treatment as described previously (22). Radioactivity incorporation into macromolecules was assessed in acid insoluble cellular materials retained on multiscreen-GV filters (Millipore Corp.).

We assessed the viability of cells treated for 4 days with the indicated doses of 8-Cl-Ado. Cells were stained with Annexin V-Cy5 (BD Biosciences) and propidium iodide (PI) according to the manufacturer's instructions and analyzed with a FACSCalibur flow cytometer (BD Biosciences) using CellQuest software (BD Biosciences). Cells staining positive for either Annexin V or PI were considered nonviable. A minimum of three independent experiments were performed for each dose assessed.

A549, H460, H2030, H23, and Calu-6 LKB1 isogenic pairs were treated for 72 hours with 5 μmol/L of 8-Cl-Ado. After treatment, cells were incubated with CellRox DeepRed (Life Technologies) at a final concentration of 5 μmol/L for 30 minutes at 37°C. After incubation, cells were trypsinized, washed, and resuspended in FACS buffer for fluorescence emission quantification. ROS levels were assessed by flow cytometry using FACSCanto and FACSDiva software (BD Biosciences).

Reverse-phase protein array (RPPA) analysis was performed and analyzed as described previously (28). Protein lysates were prepared from untreated and treated cells at the indicated times, and serial dilutions of the lysates were printed on nitrocellulose-coated glass FAST Slides (Schleicher & Schuell BioScience, Inc.). Each array was incubated with primary antibodies (Supplementary Table S2) extensively validated by immunoblot analysis.

Protein lysates were collected using RIPA lysis buffer (1% Triton X-100, 50 mmol/L HEPES, 150 mmol/L NaCl, 1.5 mmol/L MgCl₂, 1 mmol/L EGTA, 100 mmol/L NaF, 10 mmol/L Na pyrophosphate, 1 mmol/L Na₃VO₄, 10% glycerol) or 1× Cell Signaling Lysis Buffer with 1 mmol/L PMSF, supplemented with protease (mini-Complete) and phosphatase (phosStop) inhibitors (Roche Diagnostics Corporation). Primary antibodies were rabbit polyclonal antibodies against phospho-p70S6K (Thr389), LKB1 clone 27D10, phospho-ACC Ser79 clone D7D11, phospho-AMPKα (Thr172) clone D79.5E, phospho-ribS6 clone D57.2.2E, p70 S6 clone 49D7, phospho-Erk1/2 clone 197G2, Erk1/2 #9102, Akt #9272, phosphor-Akt (Ser473), phosphor-GSK-3α/β (Ser21/9) #9331 (Cell Signaling Technology); and mouse mAbs against, GSK-3α/β (Santa Cruz Biotechnology, Inc.), GAPDH clone 6C6 (Novus Biologicals, Inc.), vinculin clone H-10 (Santa Cruz Biotechnology, Inc.), and ribS6 clone 54D2 (Cell Signaling Technology).

All graphing, statistical, and regression analyses were performed using either Prism software (GraphPad Software) or the R system for statistical computing. The paired t test, non-parametric Mann–Whitney t test, and Holm-Sidak unpaired t test were used as indicated. Box-and-whisker plots indicate medians with interquartile ranges. Synergy was assessed using CompuSyn software.

To identify therapeutic agents with activity against LKB1-deficient NSCLC, we screened 27 commonly used chemotherapeutic agents against a panel of 57 NSCLC cell lines. LKB1 status was defined on the basis of STK11 genetic status, STK11 mRNA, and LKB1 protein levels (Supplementary Table S1), and deficiency was assigned on the basis of LKB1 levels in cell lines with known STK11 mutations (Supplementary Fig. S1). NSCLC cell lines harboring STK11/LKB1 mutations were significantly more sensitive to 8-amino-adenosine (8-NH2-Ado) and 8-Cl-adenosine (8-Cl-Ado), two 8-substituted-adenosine analogues, when compared with STK11/LKB1 wild type (wt) NSCLC cell lines. Likewise, 8-Cl-Ado IC₅₀ values were significantly lower for STK11/LKB1-mutant cells than STK11 wt NSCLC cells (P = 0.003; Fig. 1A). On the basis of these initial findings, we next expanded our drug screen to a panel of 104 NSCLC cell lines for sensitivity to 8-Cl-Ado. Consistent with our initial observations, LKB1-deficient cells were significantly more sensitive to 8-Cl-Ado than LKB1-proficient cells (P = 0.0080; Fig. 1B). Because STK11/LKB1 is frequently co-mutated with KRAS, we also assessed sensitivity to 8-Cl-Ado in regard to KRAS and STK11/LKB1 mutation status. There was not a significant difference in 8-Cl-Ado IC₅₀ values (Fig. 1C) between KRAS wt and mutant cells. Overall, these results identify STK11/LKB1 alterations as a potential molecular subgroup with enhanced 8-Cl-Ado sensitivity.

We next performed an extensive protein expression analysis to validate the correlation of LKB1 protein expression and sensitivity to 8-Cl-Ado. Sixty-six LKB1-proficient and LKB1-deficient NSCLC cell lines were proteomic profiled by RPPA and correlations between IC₅₀ and levels of protein expression was assessed. Consistently, LKB1 and p-AMPK protein levels strongly correlated with the IC₅₀ values for 8-Cl-Ado (Fig. 1D–F). A similar correlation was observed for p-AMPK^T172 and AMPKα levels when only LKB1-deficient cell lines were analyzed (Supplementary Fig. S2A and S2B), suggesting that inactivation of AMPK axis may play a role in the sensitivity to 8-Cl-Ado. Likewise, 8-Cl-Ado IC₅₀ values negatively correlated with levels of proteins that are downregulated by the LKB1/AMPK signaling pathway, such as MYC, S6, FOXO3, and CD98 (Fig. 1G–J; refs. 13, 29–32).

We next sought to determine whether LKB1 expression has a causal role in mediating sensitivity to 8-Cl-Ado. We generated human NSCLC LKB1 isogenic pairs using A549, H460 (LKB1-deficient), and Calu-6 (LKB1-proficient) cell lines (Fig. 2A). Consistent with our previous finding, reexpression of LKB1 significantly reduced sensitivity to 8-Cl-Ado in A549 and H460 LKB1-deficient cell lines assessed by IC₅₀ estimation (Fig. 2B). Similar data were obtained using two additional LKB1-deficient cell lines, H23 and H2030 (Supplementary Fig. S3A and S3B). Likewise, cell proliferation assessed by clonogenic and cell counting assays revealed a significantly reduced sensitivity to 8-Cl-Ado in A549, H460 and H2030 LKB1-proficient compared with LKB1-deficient cells (Supplementary Fig. S3C and S3D). These methods displayed differences for a more limited range of concentrations when compared with cell viability by Cell-Titer Glo, which may be due to differences in technical precision as well as end time points and experimental design used for each method. Conversely, shRNA-mediated knockdown of LKB1 in Calu-6 cells demonstrated a trend toward increased sensitivity (Fig. 2B). To determine whether 8-Cl-Ado treatment causes cytostatic or cytotoxic effects, we assessed the apoptotic rate based on annexin-V/PI staining in A549, H460, and Calu-6 isogenic pairs treated with 0, 2, and 20 μmol/L of 8-Cl-Ado for 4 days (Fig. 2C–F). The fraction of late apoptotic (Annexin V/PI positive) and necrotic (PI positive) cells was significantly higher in A549 LKB1-deficient cells following treatment with 8-Cl-Ado. The early apoptotic population (Annexin V positive) was also significantly increased in LKB1-deficient cells compared with LKB1-proficient cells when treated with 20 μmol/L of 8-Cl-Ado (Fig. 2C). Likewise, we quantified the percentage of Annexin V/PI positive cells after treating A549, H460, and Calu-6 LKB1 isogenic pairs with 2–50 μmol/L 8-Cl-Ado. Although the dose needed to induce substantial levels of cell death exceed IC₅₀, the fraction of apoptotic cells was significantly increased after 8-Cl-Ado treatment in LKB1-deficient cells (Fig. 2D–F). These data suggest that the greater sensitivity in LKB1-deficient cells to lower doses may be associated with induction of cytostatic rather than cytotoxic effects. Taken together, these results indicate that LKB1 loss is associated with increased sensitivity to 8-Cl-Ado.

We previously demonstrated that 8-Cl-Ado inhibits RNA and DNA synthesis in hematologic malignancies which at least in part mediate its antitumor activity (33). To evaluate whether the differential sensitivity of LKB1-deficient and LKB1-proficient cell lines is associated with decreased DNA, RNA, or protein synthesis impairment, we quantified incorporation of ³[H] thymidine, ³[H] uridine and ³[H] leucine respectively after treatments with 0–20 μmol/L of 8-Cl-Ado (Fig. 3). Consistent with the earlier finding in other tumor types, 8-Cl-Ado treatments resulted in an overall reduction of RNA and proteins synthesis in both LKB1-deficient and LKB1-proficient cell lines. These effects, however, were overall not greater in LKB1-deficient as compared with LKB1-intact cells. We observed a modest but significant difference in the effects of 8-Cl-Ado on RNA synthesis in H460-deficient compared with proficient cells after treatment with 2 and 20 μmol/L of 8-Cl-Ado, similar differences were not observed in the A549 model. Likewise, there was no significant difference in the impact of 8-Cl-Ado on proteins synthesis between LKB1-proficent and LKB1-deficient models (Fig. 3). It is interesting to note that the 8-Cl-Ado–induced inhibition of DNA synthesis was greater in LKB1-proficient A549 cells at 2 and 20 μmol/L than in LKB1-deficient cells. Although we did not specifically address this issue experimentally, this phenomenon may be due, at least in part, to the lack of mTOR pathway suppression observed under conditions of ATP depletion in LKB1-deficient cells, discussed in more detail below. These results indicate that while 8-Cl-Ado significantly reduced DNA, RNA, or protein synthesis, this reduction is not the primary mechanism of the selective vulnerability of LKB1-deficient cells to 8-Cl-Ado.

To further investigate potential mechanisms underlying the enhanced sensitivity of LKB1-deficient cells to 8-Cl-Ado, we assessed the effect of 8-Cl-Ado on ribonucleoside triphosphates (ATP, GTP, CTP, and UTP) levels. Baseline ATP levels were increased by 20% when LKB1 was expressed in H460 and A549 cells (Fig. 4A). 8-Cl-Ado treatments reduced ATP levels to a greater extent in LKB1-deficient cells than in LKB1-proficient cells, showing an ATP depletion of 60% in H460-deficient cells and only 25% in LKB1-proficient cells compared with control LKB1-deficient nontreated cells (Fig. 4B). This greater ATP depletion found in LKB1-deficient cells was not due to a greater accumulation of the drug, because accumulation of active triphosphate form of 8-Cl-Ado (8-Cl-ATP) was comparable or lower in LKB1-deficient cells compared with their LKB1-proficient control (Fig. 4C). Evaluation of all four ribonucleoside triphosphates revealed a significant depletion of purine ribonucleotides, with similar levels of ATP and GTP depletion. In contrast, the effects of 8-Cl-Ado on pyrimidine ribonucleoside triphosphates (CTP and UTP) levels did not show an overall significant reduction either in LKB1-deficient or in LKB1-proficient cells (Fig. 4D; Supplementary Fig. S4). For comparison, H460 and A549 LKB1-proficient and LKB1-deficient isogenic pairs were next treated with two other ATP-depleting agents which have undergone clinical testing, metformin and 2-DG, at concentrations comparable with or greater than those typically achieved in the clinic (10, 11, 19). Consistent with the earlier findings, we observed that LKB1-deficient cells were more sensitive to ATP depletion than LKB1-proficient cells for all three drugs (Fig. 4E). The levels of ATP depletion by 8-Cl-Ado were comparable with, or greater than, those of metformin and 2-DG.

LKB1 loss affects not only ATP levels, but also redox homeostasis (8). Therefore, we next investigated whether the cytotoxic effect of 8-Cl-Ado in LKB1-deficient cells is mediated by an increase in intracellular ROS. We measured intracellular ROS levels in A549, H460, H23, H2030, and Calu-6 LKB1 isogenic pairs with and without 5 μmol/L of 8-Cl-Ado treatment for 72 hours. As expected, basal ROS levels were significantly higher in LKB1-deficient than LKB1-proficient cells (Supplementary Fig. S5A). Furthermore, in LKB1-deficient cells 8-Cl-Ado significantly increased ROS levels, which is indicative of higher redox stress (Fig. 4F; Supplementary Fig. S5B). In contrast, 8-Cl-Ado did not significantly increase ROS levels in the LKB1-proficient cells. Thus, ROS levels were significantly lower in 8-Cl-Ado–treated LKB1-proficient cells when compared with 8-Cl-Ado–treated LKB1-deficient cells, suggesting that LKB1 plays an important role in protecting cells from 8-Cl-Ado–mediated ROS accumulation.

Altogether, these data suggest that 8-Cl-Ado cytotoxicity in LKB1-defient tumor cells may be mediated by ATP depletions as well as an accumulation of intracellular ROS.

Alterations in ATP or ROS levels can lead to activation of the cellular energy sensor AMPK through both LKB1-dependent and -independent mechanisms (14, 34). Previously, we showed that 8-Cl-Ado induces AMPK activation in breast cancer and CLL cells (25, 35). Therefore, we next evaluated the effects of 8-Cl-Ado on AMPK activation in LKB1-deficient NSCLC cells. H460, A549, and Calu-6 isogenic pairs were treated with 5 or 10 μmol/L of 8-Cl-Ado and p-AMPK ^T172 levels were analyzed by Western blotting. 8-Cl-Ado increased p-AMPK^T172 in LKB1-proficient but not LKB1-deficient cells (Fig. 5A). Likewise, 8-Cl-Ado increased the phosphorylation of the AMPK downstream target protein acetyl CoA carboxylase (ACC) in LKB1-proficient but not LKB1-deficient cells, demonstrating that activation of AMPK after 8-Cl-Ado treatment is LKB1 dependent (Fig. 5B and C).

AMPK regulates cellular energy in part by inhibiting the mTOR pathway to restore energy balance through the reduction of cell proliferation and therefore ATP-consuming processes. 8-Cl-Ado has been shown to inhibit mTOR signaling in breast cancer cells (35). In LKB1-proficient cells, but not LKB1-deficient cells, 8-Cl-Ado treatment resulted in inhibition of mTOR activity, as evidenced by the phosphorylation of its downstream targets p70S6K and S6 (Fig. 5B and C). Consistent with these findings, 8-Cl-Ado suppressed DNA synthesis to a greater extent in LKB1-proficient compared with LKB1-deficient cells (Fig. 3). These data suggest that LKB1 loss impairs the ability of the cell to suppress DNA synthesis through mTOR signaling under conditions of ATP depletion induced by 8-Cl-Ado, which may lead to an inability to maintain ATP and redox homeostasis, and contribute to the enhanced sensitivity to 8-Cl-Ado in LKB1-deficient tumor cells.

To determine the role of AMPK in 8-Cl-Ado sensitivity, we then evaluated the synergistic effects of compound C, an AMPK inhibitor (36–38) with 8-Cl-Ado in a panel of 14 LKB1-proficient and 10 LKB1-deficient NSCLC cell lines (Fig. 5D and E). IC₅₀ values of compound C were significantly reduced in LKB1-deficient and proficient cells when combined with 8-Cl-Ado (Fig. 5D). To further determine synergistic effect between these two drugs, we calculated the CI for each cell line at 25%, 50%, and 75% the effective dose (ED₂₅, ED₅₀, and ED₇₅, respectively); CIs <1.0 were considered to indicate synergy (39). CI values indicated synergistic effects of compound C when combined with 8-Cl-Ado in both, LKB1-proficient and LKB1-deficient cells (Fig. 5E), although the synergistic effect was more pronounced in LKB1-deficient cell lines. These results indicate that LKB1-deficient cells are more sensitive to energy-depleting agents and that dysregulation of AMPK/mTOR axis may be partially responsible of the higher sensitivity associated with LKB1 loss.

We next sought to identify other proteins or signaling pathways associated with sensitivity to energy depletion. Proteomic profiling revealed that basal levels of ERK1/2 and MEK1/2 proteins correlated with 8-Cl-Ado sensitivity (Fig. 6A–E). Among LKB1-deficient cell lines, ERK2 as well as MEK levels, were specifically associated with 8-Cl-Ado sensitivity (Supplementary Fig. S6A and S6B). Likewise, AKT significantly correlated with 8-Cl-Ado sensitivity (Fig. 6F) and we observed same trend in LKB1-deficient cells analysis (Supplementary Fig. S6C). These data suggest that the MEK/ERK and PI3K/AKT pathways may play a role in the response to 8-Cl-Ado.

To further investigate the role of MEK/ERK and PI3K/AKT pathways on 8-Cl-Ado sensitivity, we assessed 8-Cl-Ado–mediated changes in the proteome of LKB1-deficient A549, H1355, H1944, H2030, and H2126 cell lines and LKB1-proficient H1155, HOP-62, H1299, H1819, H1838, HCC15, and H2052 cell lines. Cells were treated with 10 μmol/L 8-Cl-Ado for 16, 24, or 48 hours and then evaluated by RPPA. p-ERK1/2^T202Y204 (p-MAPK^T202Y204), p-AKT^S473, and the AKT target p-GSK3β^S21/9 showed the greatest significant increase after 8-Cl-Ado treatment in LKB1-deficient and LKB1-proficient cells (Fig. 7A; Supplementary Fig. S7A and S7B). Furthermore, 8-Cl-Ado treatment resulted in activation of these signaling molecules regardless of the LKB1 status (Supplementary Fig. S7C and S7D). To confirm whether 8-Cl-Ado induces phosphorylation of these proteins and evaluate the impact of LKB1 expression on this modulation, we then analyzed total and phosphorylated levels of ERK 1/2, AKT and GSK3β by Western blotting in three LKB1 isogenic pairs after treatment with 10 μmol/L 8-Cl-Ado for 24 hours (Fig. 7B). 8-Cl-Ado induced ERK1/2^(T202/204), AKT^S473, and GSK3β^S9 phosphorylation in both LKB1-deficient and LKB1-proficient cells.

Altogether, these finding suggest that the MAPK/ERK and PI3K/AKT pathways play a compensatory role in response to the energetic stress induced by 8-Cl-Ado. Therefore, higher expression of these proteins following 8-Cl-Ado treatment as seen in LKB1-proficient cells may mediate reduced sensitivity to 8-Cl-Ado.

Because cells with higher levels of ERK1/2 and MEK1/2 were less sensitive to 8-Cl-Ado, we sought to determine the efficacy of combined MEK inhibitor and 8-Cl-Ado treatment. We treated 10 LKB1-deficient and 14 LKB1-proficient cell lines with 8-Cl-Ado and the MEK inhibitor trametinib, alone or in combination. In LKB1-deficient cells, the IC₅₀ values for 8-Cl-Ado in combination with trametinib were significantly lower than those for 8-Cl-Ado or trametinib alone, suggesting that the cells had enhanced sensitivity to 8-Cl-Ado when MEK was inhibited (Fig. 7C). Similarly, in LKB1-proficient cells, the IC₅₀ values for 8-Cl-Ado in combination with trametinib were significantly lower than single-agent treatments IC₅₀, although to a lesser extent compared with LKB1-deficient cells (Fig. 7C). To discern whether 8-Cl-Ado synergizes with trametinib, we calculated the CI value for each cell line. Strong synergy was noted in the LKB1-deficient cell lines, where all but one cell line had a CI ≤0.5. Moreover, CI values of most LKB1-proficient cell lines indicated synergistic effects; however, two cell lines indicated additive effects at ED₇₅ and three cell lines even displayed antagonistic effects (Fig. 7D). These results indicate that inhibition of the MEK pathway could synergize with 8-Cl-Ado in LKB1-deficient and some LKB1-proficient cells.

Because we found that 8-Cl-Ado also mediated induction of PI3K/AKT signaling, we next tested the effect of 8-Cl-Ado in combination with two PI3K inhibitors, PI-103 and A66, in the same set of cell lines. Combination of 8-Cl-Ado with the PI3K inhibitor A66 significantly reduced IC₅₀ in LKB1-deficient and LKB1-proficient cell lines, although increment in sensitivity was more pronounced in LKB1-deficient compared with LKB1-proficient cell lines (Fig. 7E). Strong synergy was found in all cell lines regardless LKB1 expression. CI values for these combinations yielded synergistic effect in a portion of cell lines and additive effects in others, including in LKB1-deficient cell lines (Fig. 7F). Similarly, the addition of PI3K inhibitor, PI-103, significantly enhanced sensitivity to 8-Cl-Ado in all cell lines (Supplementary Fig. S8A), although synergy appeared more robust with the A66 compound. These findings provide evidence that inhibition of the PI3K pathway may enhance the cytotoxic effects and may compensate for ATP-depletion induced by 8-Cl-Ado in LKB1-deficient and, to a lesser extent, in LKB1-intact cells.

STK11 (LKB1) is the second most altered tumor suppressor gene in NSCLC, and loss of LKB1 expression leads an aggressive tumor phenotype with worse prognosis and reduced responsiveness to PD1/PD-L1 inhibitors (6, 12, 40). There are currently no approved treatment strategies for this subset of patients. Here, we identified 8-Cl-Ado, an adenosine analog which induces an imbalance in intracellular ATP/AMP levels, as active agent for LKB1-deficient NSCLC cells. In addition, we show that 8-Cl-Ado treatment results in the activation of the MAPK/MEK/ERK and PI3K/AKT/mTOR pathways and targeting of these pathways synergistically enhances the antitumor activity of 8-Cl-Ado in LKB1-deficient tumor cells, identifying these regimens as a potential therapeutic strategy for patients with NSCLC harboring LKB1-deficient tumors.

LKB1 is a master regulator of cellular growth and metabolism. Under cellular stress condition, LKB1/AMPK axis senses low levels of ATP and NADPH to restore intracellular energy homeostasis (15, 41) at least in part by modulating cell proliferation through activation of the mTOR pathway (16). LKB1-deficient tumors display a disrupted balance between energy consumption and production, bypassing normal growth inhibitory signals, resulting in unrestricted cell growth which reduces cellular ATP levels, therefore increasing energetic and oxidative stress (8, 42). Thus, LKB1-deficient tumors display increased vulnerability to energetic stress induced by ATP depletion.

The C8-modified adenosine analogues 8-Cl-Ado and 8-NH2-Ado are cytotoxic to a number of liquid and solid tumor cells (43, 44) and currently, 8-Cl-Ado is under a phase I/II clinical trial for hematologic malignancies (NCT02509546; ref. 21). Compared with nucleoside analogues established for cancer therapeutics, these analogs are unique as they possess a ribose sugar. We previously identified that 8-Cl-Ado is metabolized to 8-Cl-AMP, 8-Cl-ADP, and 8-Cl-ATP (21). 8-Cl-ADP and 8-Cl-ATP act as mimetics to disrupt the metabolism of the endogenous adenylate pools (26), which results in diminished cellular ATP levels (20, 22, 45). Increased 8-Cl-ATP also inhibits mRNA transcript synthesis by incorporating into the RNA as a chain terminator or by inhibiting polyadenylation (24). Previous studies have shown that 8-Cl-Ado and 8-Cl-cAMP, a prodrug form of the analog, inhibit the growth and/or survival of murine and human lung cancer cells. Although these studies did not assess the context in which LKB1 affects sensitivity and energy depletion, they did show that the mechanism of action of 8-Cl-Ado in NSCLC cells was associated with G₂–M arrest and mitotic catastrophe (43, 44). In agreement with these previous findings, our data indicate that basal ATP depletion in LKB1-deficient tumor cells selectively increases vulnerability to 8-Cl-Ado. Treatment with this compound significantly impairs cell proliferation by reducing intracellular ATP levels to greater extent in LKB1-deficient tumor cells. While other ATP-depleting agents (metformin and 2-DG) also show preferential ATP depletion in LKB1-deficient cells, the concentrations of these agents needed to achieve similar levels of ATP depletion have not been shown to be clinically achievable (10, 11, 19). Moreover, our data showed that 8-Cl-Ado activates AMPK signaling in a LKB1-dependent manner, indicating that LKB1-AMPK axis activation may play a key role in compensating for the ATP reduction induced by 8-Cl-Ado in LKB1-proficient tumors.

ROS are a by-product of mitochondrial respiration. Low to moderate ROS levels stimulate cell proliferation and differentiation (46), by contrast excess ROS levels cause oxidative stress and subsequent cell death by damaging cellular components, such as DNA, proteins, and lipids (47). As a consequence of increased energetic and metabolic stress, LKB1-deficient cells generate elevated levels of ROS (8, 42), and the induction of ROS production has also been shown to promote AMPK phosphorylation and activation but may occur in an LKB1-independent manner (34). Glutamine-derived glutamate is the main precursor of glutathione (GSH), the major cellular antioxidant for redox homeostasis. We recently reported that LKB1 alterations drive metabolic adaptations resulting in a glutamine “addiction” which leads to reduced intracellular GSH stores and, consequently vulnerability to oxidative stress (8). Indeed, LKB1-deficient tumor cells used in the current work showed increased level of intracellular ROS. Furthermore, 8-Cl-Ado treatment raised ROS levels primarily in LKB1-deficient cells, but not in LKB1-proficient cells, suggesting that induction of cell death after 8-Cl-Ado treatment may be also driven by higher ROS stress.

Response to energy depletion involves shifting of cellular metabolism to restore the energy balance by stimulating catabolism and inhibiting ATP-consuming processes. Several pathways are involved in regulating cell proliferation and metabolism to guarantee the correct maintenance of cellular homeostasis. MAPK and PI3K modulate extracellular signals to control cell growth, differentiation, and migration (48, 49). Consequently, combination of mTOR and PIK3 inhibitors has shown therapeutic effects in LKB1-deficient NSCLC (50). Our proteomic analysis revealed differential baseline levels of MAPK and PI3K pathway signaling molecules associated with response to 8-Cl-Ado. Consequently, 8-Cl-Ado treatment activated these signaling in sensitive and resistant cell lines, and suggested that targeting these pathways may potentiate the effects of 8-Cl-Ado. In agreement with this, the combination of 8-Cl-Ado with the MEK inhibitor trametinib, displayed a strong synergistic effect particularly in LKB1-deficient cell lines. Phosphorylation of AKT^S473 and its target, GSK3^S21/9 was also induced by 8-Cl-Ado, and basal AKT levels correlated with 8-Cl-Ado sensitivity. Moreover, two different PI3K inhibitors (PI-103 and A66) showed increased effects when combined with 8-Cl-Ado in most NSCLC cell lines. These results demonstrate that combinations of 8-Cl-Ado with either MEK inhibitors or PI3K inhibitors merits further investigation as a potential therapeutic regimen for patients harboring LKB1-deficient tumors.

Taken together, our findings indicate that LKB1-deficient NSCLC cells are selectively vulnerable to energy depletion. Treatment with 8-Cl-Ado demonstrated specific antitumor activity, depleted ATP levels and raised redox stress in LKB1-deficient cells. Dysregulation of AMPK/mTOR axis associated with LKB1 loss may be responsible for greater sensitivity to 8-Cl-Ado by reducing ability to compensate energy and redox imbalance induced by this compound. Data provided herein support the future clinical testing of energy-depleting agents such as 8-Cl-Ado combined with signaling pathway inhibitors in patients with LKB1-deficient NSCLC and other tumor types.

A. Galan-Cobo reports grants from NIH/SPORE P50CA07907, NIH/R01 CA205150, CPRIT CP160652, and other support from The Lung Cancer Moon Shot during the conduct of the study. M.L. Ayres reports grants from NIH/NCI P30CA016672 during the conduct of the study. M. Nilsson reports personal fees from Spectrum Pharmaceuticals outside the submitted work. W.G. Wierda reports other support from GSK/Novartis, Abbvie, Genentech, Pharmacyclics LLC, AstraZeneca/Acerta Pharma Inc., Gilead Sciences, Juno Therapeutics, KITE Pharma, Sunesis, Miragen, Oncternal Therapeutics, Inc., Cyclacel, Loxo Oncology, Inc., Janssen, and Xencor outside the submitted work. F. Skoulidis reports personal fees from Amgen, Bristol-Myers Squibb, RV Mais Promocao Eventos LTDS, Intellisphere LLC, Navire Pharma, and Beigene; grants from Amgen, Mirati Therapeutics, Boehringer Ingelheim, Merck and Co, Novartis, and Pfizer; other support from BioNTech SE, Moderna Inc, Tango Therapeutics, AstraZeneca Pharmaceuticals, and Amgen outside the submitted work. J.D. Minna reports grants from NIH during the conduct of the study; personal fees from NIH and University of Texas Southwestern Medical Center outside the submitted work. V. Gandhi reports grants from NIH/NCI P30CA016672 and other support from Helen H. Laughery during the conduct of the study; grants from Abbvie, Acerta, ClearCreek Bio, Gilead, LOXO, Pharmacyclics, and Sunesis; personal fees from AstraZeneca and Dava Oncology outside the submitted work. J.V. Heymach reports grants from Lung SPORE, NIH/NCI, CPRIT, and MD Anderson Lung Cancer Moon Shot; other support from AstraZeneca, Boehringer-Ingelheim, Catalyst, Genentech, GlaxoSmithKline, Guardant Health, Foundation Medicine, Hengrui Therapeutics, Eli Lilly, Novartis, Spectrum, Sanofi, Takeda Pharmaceuticals, Mirati Therapeutics, Bristol-Myers Squibb, BrightPath Biotherapeutics, Janssen Global Services, Nexus Health Systems, EMD Serono, Pneuma Respiratory, Kairos Venture Investments, Leads Biolabs, and RefleXion during the conduct of the study. No disclosures were reported by the other authors.

A. Galan-Cobo: Formal analysis, supervision, investigation, writing–original draft, writing–review and editing. C.M. Stellrecht: Formal analysis, supervision, validation, investigation, writing–original draft, writing–review and editing. E. Yilmaz: Formal analysis, investigation. C. Yang: Formal analysis, investigation. Y. Qian: Data curation. X. Qu: Investigation. I. Akhter: Investigation. M.L. Ayres: Data curation, investigation. Y. Fan: Data curation. P. Tong: Data curation. L. Diao: Software, formal analysis. J. Ding: Data curation. U. Giri: Data curation. J. Gudikote: Data curation. M. Nilsson: Writing–review and editing. W.G. Wierda: Formal analysis, writing–review and editing. J. Wang: Software, formal analysis. F. Skoulidis: Investigation. J.D. Minna: Resources, funding acquisition, writing–review and editing. V. Gandhi: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing. J.V. Heymach: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

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 (grant number: SU2C-AACR-DT17-15), a Jane Ford Petrin donation, and Helen H. Laughery donation (V. Gandhi). The indicated Stand Up To Cancer 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.