The LKB1/STK11 tumor suppressor encodes a serine/threonine kinase, which coordinates cell growth, polarity, motility, and metabolism. In non–small cell lung carcinoma, LKB1 is somatically inactivated in 25% to 30% of cases, often concurrently with activating KRAS mutations. Here, we used an integrative approach to define novel therapeutic targets in KRAS-driven LKB1-mutant lung cancers. High-throughput RNA interference screens in lung cancer cell lines from genetically engineered mouse models driven by activated KRAS with or without coincident Lkb1 deletion led to the identification of Dtymk, encoding deoxythymidylate kinase (DTYMK), which catalyzes dTTP biosynthesis, as synthetically lethal with Lkb1 deficiency in mouse and human lung cancer lines. Global metabolite profiling showed that Lkb1-null cells had a striking decrease in multiple nucleotide metabolites as compared with the Lkb1–wild-type cells. Thus, LKB1-mutant lung cancers have deficits in nucleotide metabolism that confer hypersensitivity to DTYMK inhibition, suggesting that DTYMK is a potential therapeutic target in this aggressive subset of tumors.
Significance: Using cell lines derived from the lung cancers occurring in genetically engineered mice, we conducted an integrative genome-wide short hairpin RNA and metabolite screen to identify DTYMK as a potential therapeutic target in Kras/Lkb1–mutant lung cancer. We believe that DTYMK is tractable for the development of novel therapeutics, and show an integrative approach to target identification that reduces false-positive candidates and should have broad applicability for the development of targeted therapeutics. Cancer Discov; 3(8); 870–9. ©2013 AACR.
See related commentary by Marcus and Khuri, p. 843
This article is highlighted in the In This Issue feature, p. 826
LKB1/STK11 functions as a master regulator of cell metabolism and energy stress responses (1, 2). Its best-characterized target is AMP-activated protein kinase (AMPK), which is directly phosphorylated and activated by LKB1 in the context of low cellular ATP levels (2). AMPK in turn modulates nutrient use to restore energy homeostasis through phosphorylation of multiple substrates controlling nutrient uptake and metabolism (1, 2). LKB1 also activates other members of the AMPK-related family of kinases, which regulate diverse aspects of cell metabolism, growth, and polarity (1, 2). LKB1/STK11 deficiency results in broad defects in metabolic control, as evidenced by primary cells and cancer cell lines lacking LKB1 being sensitized to nutrient deprivation and other types of metabolic stress (3–5). LKB1 is also a major tumor suppressor gene that is somatically inactivated in many common types of cancer (3, 4). Human tumor data and genetic studies in mice suggest that LKB1-mutant cancers are biologically distinct from those with LKB1 intact (6). Notably, LKB1 inactivation is the single most prominent biomarker for poor outcome in cervical cancer, predicting survival of 1 year, as compared with 10-year survival for LKB1 wild-type (WT) tumors (7). In mouse models of lung cancer and melanoma, Lkb1 loss synergizes with active KRAS to drive a highly metastatic phenotype not seen in the context of other combinations of mutations (6, 8). Unfortunately, there are currently few drugs available for clinical use that target LKB1 loss specifically, and recent human cancer cell line screens using more than 130 drugs under clinical and preclinical investigation failed to identify known anticancer agents with strong selective activity in this subset of tumors (data not shown; ref. 9).
Here, we sought to use an integrative program to systematically identify novel drug targets in LKB1-mutant lung cancer using a synthetic lethal RNA interference (RNAi) screen and comprehensive metabolomics analysis. For these studies, we took advantage of a series of low-passage lung cancer cell lines derived from genetically engineered mouse models (GEMM) programmed with common mutations in Kras and Trp53, alone or in combination with Lkb1. Although the heterogeneity of human cancer cell lines can obscure synthetic lethal associations, we predicted that this murine cell line panel, developed in the context of a well-defined model system, would effectively enable the discovery of genotype-driven sensitivities.
Generation of Lung Cancer Cell Lines from GEMMs
To generate isogenic lung cancer cell lines, somatic KRAS activation and Trp53 loss with or without Lkb1 inactivation were induced in the lungs of genetically engineered mice (Kras+/LSL-G12DTrp53L/L or Kras+/LSL-G12DTrp53L/LLkb1L/L) by intranasal administration of Adenovirus-Cre as previously described (6). Inactivation of Trp53 was included in these models, as inactivation of Tp53 is common in human non–small cell lung carcinoma (NSCLC; >50%; ref. 10). Tumor nodules from mice of defined genotypes were dissected, minced, and cultured, resulting in the derivation of the 634, 855, and 857 lines from Kras+/LSL-G12DTrp53L/L mice (Lkb1-WT) and the t2, t4, and t5 lines from Kras+/LSL-G12DTrp53L/LLkb1L/L mice (Lkb1-null; Supplementary Fig. S1A). Genotype, LKB1 expression, and epithelial origin of the lines were confirmed by PCR, Western blot analysis, and pan-cytokeratin immunostaining (Supplementary Fig. S1B–S1D). These six lines showed similar growth rates (Supplementary Fig. S1E), and the Lkb1-null lines exhibited lower cellular ATP levels compared with Lkb1-WT cells (Supplementary Fig. S1F).
Identification of Selective Essential Genes in Kras/Trp53/Lkb1 GEMM-Derived Cell Lines
To identify genes that induce cell death selectively in Lkb1-null lung cancers, a synthetic lethal screen was conducted using a pooled 40K murine short hairpin RNA (shRNA) lentiviral library with each of the Lkb1-WT and Lkb1-null cell lines. The relative abundance of shRNAs in each cell line sample was determined by deep-sequencing, and for every shRNA, a log2 fold change (log2FC) value was calculated from the difference in relative abundance at a late time point after infection versus the initial shRNA-infected sample. Unsupervised hierarchical clustering analysis of the ranked hairpins from the triplicate pooled shRNA library screens revealed clear clustering of the Lkb1-WT and Lkb1-null cells into distinct groups, and the blue color in the top right corner represents genes for which the abundance of shRNAs is significantly reduced in all Lkb1-null cultures, suggesting a specific effect in the inhibition of Lkb1-null cell growth (Fig. 1A). We collapsed the ranked hairpins using two methods, a RIGER analysis (Kolmogorov–Smirnov t test–based statistics) and a weighted second-best analysis to rank genes that selectively impaired proliferation/viability in Lkb1-null cells. We nominated a union of 344 genes, identified by the top 100 individual hairpins for 88 genes (Supplementary Table S1.1) and the top 200 genes from both the Kolmogorov–Smirnov (Supplementary Table S1.2) and weighted second-best analysis (Supplementary Table S1.3), as our initial prioritized list (Fig. 1B). Of note, 340 shRNAs, targeting 70 candidate genes from this prioritized list, were chosen for validation (Supplementary Table S1.4). These 70 genes consisted of the top 10 candidates from the Kolmogorov–Smirnov analysis, as well as 60 others involved in a range of biologic processes in an attempt to represent all biologic categories in the validation process. Validation was conducted in an array format and identified 13 genes that displayed two or more hairpins with a significant growth disadvantage in the Lkb1-null cells (Supplementary Table S1.5). deoxythymidylate kinase (Dtymk), Chek1, Pdhb, and Cmpk1 were the top four candidates, each with two or more hairpins that scored in the validation assay (Fig. 1C and Supplementary Table S1.5).
Metabolomics Analysis Implicates Dtymk as a Critical Gene in Lkb1-Null Cells
LKB1 is reported to be involved in metabolic reprogramming (4, 11); therefore, we assessed the metabolic profile of Lkb1-WT and Lkb1-null cells and discovered a set of 58 metabolites, including the nucleotide metabolites IMP, AMP, ADP, GMP, dGMP, UMP, UDP, CDP, dCDP, and dTDP, which were present at consistently lower levels in Lkb1-null cells (Fig. 1D). Pathway enrichment analysis showed that metabolites in both purine and pyrimidine metabolism were significantly reduced in Lkb1-null compared with Lkb1-WT cells (Fig. 1D; P = 3.5 × 10−7 and 3.4 × 10−5, respectively), including multiple metabolites involved in dTTP synthesis, such as dTDP (the product of DTYMK) and UDP/CDP/dCDP (products of UMP-CMP kinase; Fig. 1E; refs. 12, 13). Collectively, these two independent sets of data suggest that Lkb1-mutant lung cancer cells exhibit alterations in dTTP metabolism and are particularly sensitized to disruption of intracellular dTTP synthesis, and therefore have potential as important targets in Lkb1-null lung cancer.
Dtymk Is a Synthetic Lethal Gene Selectively Required for Lkb1-Null Cell Proliferation
To further examine the role of DTYMK in lung tumorigenesis, we screened five shDtymks and identified shDtymk-1 and shDtymk-3, which knocked down DTYMK to nearly undetectable levels (Supplementary Fig. S2A and Supplementary Table S2). Compared with shGFP, both shDtymk-1 and shDtymk-3 strongly inhibited the growth of the Lkb1-null cells (t2, t4, and t5), while producing a weaker effect in the Lkb1-WT (634, 855, and 857) cell lines (Fig. 2A and Supplementary Fig. S2B). To see if overexpression of shRNA-resistant Dtymk can rescue the shDtymk effect, Dtymk-R1 and Dtymk-R3 were cloned into the pLenti6 vector and then transduced into Lkb1-null t4 cells. Blasticidin-resistant cells were pooled and further transduced with shGFP, shDtymk-1, or shDtymk-3, respectively. Consistently, shDtymk-1 and shDtymk-3 killed Lkb1-null t4 cells within 3 days, whereas Dtymk-R1 and Dtymk-R3 expression largely restored the growth of shDtymk-1– and shDtymk-3–transduced t4 cells (Fig. 2B). Western blot analysis revealed lower DTYMK signals in t4-Dtymk-R1/shDtymk-1 and t4-Dtymk-R3/shDtymk-3 cells, suggesting that some of the blasticidin-resistant cells were not DTYMK-R–positive and thus were killed by shDtymk (Fig. 2B), which likely accounted for the significant but incomplete rescue by Dtymk-R1 or Dtymk-R3. To extend these findings to tumorigenesis in vivo, Lkb1-WT (634 and 857) and Lkb1-null (t2 and t4) cells were transduced with doxycycline-inducible (TetOn) shGFP or shDtymk-3 and then implanted into athymic nude mice. Consistent with the in vitro proliferation assay, doxycycline-induced expression of shDtymk-3 for 3 weeks resulted in a marked impairment in the growth of Lkb1-null tumors while producing more modest effects in the Lkb1-WT tumors (Fig. 2C).
Dtymk Knockdown Alters Pyrimidine Metabolism
DTYMK catalyzes the phosphorylation of dTMP to form dTDP, and it is the first merged step of both the de novo and salvage pathways in the production of dTTP (Fig. 1E). We expected that knockdown of Dtymk would inhibit this pathway and lead to accumulation of the substrate dTMP and a decrease in the product dTDP. Corresponding metabolite analysis of Lkb1-WT 634 and Lkb1-null t4 cells transduced with shDtymk-1 revealed the expected significant increase in dTMP and moderate decrease in dTDP levels in both cell lines (Fig. 2D), indicating that DTYMK is a major source of dTDP in the cells and underscores the importance of this gene in cancer cell proliferation, as dTDP is required for production of dTTP for DNA synthesis.
dTTP Rescues the shDtymk Growth Phenotype
To investigate whether adding dTTP to the medium can rescue shDtymk-induced cell death, Lkb1-WT 634 and Lkb1-null t4 cell lines were transduced with shGFP, shDtymk-1, or shDtymk-3 and cultured in the presence or absence of 100 μmol/L dTTP for 4 days (14). Consistently, shDtymk-1 and shDtymk-3 killed more Lkb1-null t4 cells than Lkb1-WT 634 cells, but not the cells cultured in medium containing exogenous dTTP (Fig. 2E; confirmation of Dtymk knockdown and incorporation of the exogenous dTTP into DNA are shown in Supplementary Fig. S3A and S3B). Collectively, our data indicate that dNTP metabolism is impaired in Lkb1-deficient lung cancer cells, and that targeting Dtymk is synthetically lethal in this setting.
Lkb1-Null Cells Are More Prone to DNA Damage than Lkb1-WT Cells
Knockdown of Dtymk will consequently reduce dTTP but also increase dUTP levels. Such changes have been previously linked to dUTP misincorporation and DNA damage, when high expression levels of ribonucleotide reductase R2 subunit activate nucleotide excision repair (15–18). We noted that, although Lkb1-null and Lkb1-WT cells have similar R2 expression, Lkb1-null cells have much lower DTYMK expression (Fig. 3A), potentially creating a cellular state favorable for dUTP misincorporation. Supportively, Lkb1-null cells have a large 4N peak in DNA content (Fig. 3B), are more sensitive to CHEK1 inhibition (Fig. 3C and Supplementary Fig. S4A), and have slightly increased basal phospho-CHEK1 (Fig. 3A), consistent with the activation of a G2 DNA damage checkpoint during replication in Lkb1-null cells (19–22). This pathway seems relevant in vivo. Lkb1-null tumors exhibited increased γH2AX and phospho-CHEK1 signals as compared with Lkb1-WT tumors (Supplementary Fig. S5A–S5C), suggesting there is more DNA damage in vivo than under in vitro culture conditions. In line with this evidence for more DNA damage in Lkb1-null cells, it is notable that Chek1 ranked second in our screen (Fig. 1C and Supplementary Table S1.5), suggesting a dependence of Lkb1-null cell survival on CHEK1 function. Knockdown of Dtymk over a short period (i.e., 2.5 days after shDtymk-transduction) resulted in comparable increases in the phosphorylation of CHEK1 and H2AX in both cell types, whereas the phosphorylation of replication protein A 32 kDa subunit (RPA32) was much more pronounced in Lkb1-null cells (Fig. 3A), suggesting more DNA damage and elevation in nucleotide excision repair in Lkb1-null cells (23). Interestingly, the expression of total RPA32 was increased in Lkb1-WT cells (Fig. 3A), suggesting that LKB1 may positively regulate RPA32 expression following Dtymk knockdown and more DNA damage. Collectively, these data suggest that Lkb1 loss sensitizes cells to DTYMK-depletion–induced DNA damage and replication stress, as equivalent knockdown of Dtymk in Lkb1-null and Lkb1-WT cells leads to more robust DNA damage in the Lkb1-null cell lines.
DNA Replication Is More Sensitive to Dtymk Knockdown in Lkb1-Null Cells than in Lkb1-WT Cells
To examine how the knockdown of Dtymk affects DNA synthesis, Lkb1-WT and Lkb1-null cells were pulse-labeled with 5-iododeoxyuridine (IdU) for 20 minutes at 0, 2.5, and 3.5 days posttransduction with shDtymk-1. As shown in Fig. 3D, the proportion of IdU-labeled cells decreased upon Dtymk knockdown regardless of Lkb1 status, although the decrease was much greater in the Lkb1-null cells (dropping from 43.1% to 5.8% in 3.5 days, a decrease of 86.5%) as compared with those with Lkb1-WT (decreasing from 57.7% to 22.3%, a decrease of 61.2%). The lower degree of labeling of Lkb1-null cells compared with Lkb1-WT cells observed under basal conditions (43.1% vs. 57.7%) may be related to the broad reductions in dNTP metabolism in Lkb1-null cells. After shDtymk transduction, Lkb1-null cells appeared normal for 3 days, but by day 4 there was massive cell death leaving virtually no surviving cells, although there was no evidence of apoptosis (data not shown). After 3.5-day knockdown of Dtymk, the remaining Lkb1-null t4 cells showed deformed and fragmented nuclei, indicative of thymineless death (Fig. 3E; refs. 24–27).
LKB1-Mutant Human NSCLC Cell Lines Are Hypersensitive to DTYMK Knockdown
We further sought to determine whether our observations in mouse lung cancer cells could be recapitulated in human LKB1-deficient NSCLC cell lines. We first screened five shRNAs targeting DTYMK and identified two, shDTYMK-D8 and shDTYMK-D10, that gave efficient knockdown (Supplementary Fig. S6A and Supplementary Table S2). Next, we screened LKB1-WT and LKB1-deficient NSCLC cell lines (Fig. 4A) for proliferation in response to DTYMK knockdown and found that LKB1-deficient H2122 and A549 cell lines had heightened sensitivity as compared with LKB1-WT H358 and Calu-1 cell lines (Fig. 4B). Knockdown of DTYMK was confirmed by Western blot analysis (Fig. 4C). ShDTYMK-D10 transduction showed an increased lethality in LKB1-WT H358 and Calu-1 cells, trending toward that of the LKB1-deficient cell lines (Fig. 4B). One possible explanation could involve a differential threshold of DTYMK knockdown, as the remaining DTYMK protein levels after shDTYMK-D10 transduction were lower than those of shDTYMK-D8 (Fig. 4C). This suggests that there may be a differential sensitivity to absolute DTYMK reduction between LKB1-deficient and LKB1-WT cells. DTYMK was reported to be synthetic lethal with doxorubicin in colon cancer cells independent of p53 (28). Consistently, the synthetic lethal interaction of LKB1-deficient and DTYMK knockdown may be independent of p53 status, as LKB1-deficient cell lines with (A549) or without (H2122) functional p53 behave similarly (Fig. 4B).
Next, we showed that knockdown of DTYMK in A549 cells reduced dTDP levels (Fig. 4D), suggesting that DTYMK is a major source of dTDP in human lung cancer cells. We further showed that LKB1-deficient cells H2122 and A549 were more sensitive than LKB1-WT H358 and Calu-1 cell lines to treatment with the selected CHEK1 inhibitors (Fig. 4E and Supplementary Fig. S4B), suggesting more DNA damage in LKB1-deficient cells than in LKB1-WT cell lines. This pathway seems relevant in vivo as LKB1 loss was associated with elevated CHEK1 expression in KRAS-mutant NSCLCs (Supplementary Fig. S6B).
In the current study, we created cell lines using Lkb1-null lung tumor nodules and conducted multiple screens that identified Dtymk as a putative synthetic lethal candidate with Lkb1 loss. Furthermore, we showed that depletion of DTYMK in mouse and human NSCLC cells diminished the dTDP pool and led to greater growth inhibition in Lkb1/LKB1–deficient cells; and that LKB1 loss in mouse and human cells was linked to more DNA damage. These results suggest that DTYMK is a potential therapeutic target in LKB1-mutant human cancer. In addition, the parallel results observed in both mouse and human cell lines suggest that GEMM-derived tumor cell lines can be used successfully for in vitro synthetic lethal screening.
One possible explanation for the synthetic lethality of Lkb1 loss and Dtymk knockdown is partly because of the lower expression of DTYMK in Lkb1-null cell lines, leading them to be more dependent on the dTTP synthesis pathway. ShDtymk depletes the absolute amount of DTYMK protein below a critical threshold, resulting in thymineless death in Lkb1-null cells but not in Lkb1-WT cells (Supplementary Fig. S7). In addition, unlike Lkb1-WT cell lines, Lkb1-null cell lines lack feedback upregulation of RPA32 expression upon Dtymk knockdown (Supplementary Fig. S7). Because RPA32 is involved in binding and stabilizing ssDNA during repair and replication, the lack of feedback upregulation of RPA32 expression may hinder DNA repair. A preliminary study revealed lower Dtymk and Chek1 transcripts in Lkb1-null cell lines (Supplementary Fig. S8A and S8B), suggesting that transcriptional regulation contributes to the lower DTYMK and CHEK1 protein levels in Lkb1-null cell lines. More work will be needed to decipher the roles of LKB1 in the regulation of DTYMK, CHEK1, and RPA32 expression. Although LKB1-deficient NSCLC cell lines did not apparently show less DTYMK expression, the shDTYMK data still suggested a differential sensitivity to absolute DTYMK reduction between LKB1-deficient and LKB1-WT cells. As an essential gene governing dTTP biosynthesis and DNA replication, DTYMK is necessary to all dividing cells, and overdepletion of DTYMK below the threshold is lethal to all dividing cells, especially to the tumor cells carrying low levels of deoxynucleotide pools yet maintaining a fast growth rate. This may explain the eventual death of Lkb1/LKB1-WT cells after Dtymk/DTYMK knockdown.
Human DTYMK was cloned by functional complementation of a Saccharomyces cerevisiae cell-cycle mutant cdc8, an essential gene for DNA synthesis (29). DTYMK is the first enzymatic step following the convergence of the de novo and salvage pathways in dTTP biosynthesis. In the de novo pathway, the DTYMK substrate dTMP is synthesized from methylation of dUMP by thymidylate synthase. In the salvage pathway, dTMP is produced from phosphorylation of thymidine by thymidine kinase. The next step in both pathways is the DTYMK-mediated phosphorylation of dTMP to form dTDP (30, 31). The production of dTDP is in contrast to that of the other deoxyribonucleotides used in DNA synthesis—dADP, dGDP, dCDP, and dUDP, which are synthesized from ADP, GDP, CDP, and UDP by ribonucleotide reductase (12, 32). Therefore, the unique dTTP biosynthesis pathway is a good target for drugs. There are multiple precedents of inhibition of the key enzymes in the de novo dTTP synthesis pathway, including thymidylate synthase by 5-fluorouracil or pemetrexed (15) and ribonucleotide reductase by hydroxyurea (33). We have targeted thymidylate synthase and ribonucleotide reductase in both the mouse and human NSCLC Lkb1/LKB1-mutant cell lines using the same drugs and have not seen any selective effect on Lkb1/LKB1–deficient cell growth, likely because of an escape mechanism from the salvage pathway (Supplementary Fig. S9). In summary, the lack of redundant pathways for dTTP biosynthesis and the vital role of DTYMK in this process together make DTYMK a new anticancer target. In this regard, expression of DTYMK is increased in the majority of lung adenocarcinomas in comparison with normal lung (Supplementary Fig. S10A), and elevated DTYMK levels are correlated with poor survival (Supplementary Fig. S10B). Unfortunately, LKB1 mutation status was not determined in these datasets.
In addition to identifying DTYMK as a potential therapeutic target for LKB1-deficient lung cancer, we have also provided proof-of-principle that GEMM-derived cancer cell lines can provide a genetically homogeneous and therefore tractable substrate for high-throughput screens to identify novel therapeutic targets. To date, the weakness of genome-wide RNAi screening has been high false discovery rates and cell line–specific off-target effects, and consequently, despite the implementation of multiple very large-scale cell line screening programs, few actionable genotype-associated sensitivities have been uncovered to date. Our study suggests that integrating RNAi screening with metabolite profiling is an effective strategy to leverage the strengths and mitigate the weaknesses of each approach. We propose that this approach will have broad applicability, and enable more rapid development of additional targeted therapeutics for an array of genetic abnormalities occurring in cancer.
Detailed protocols for all sections are described in the Supplementary Methods.
RNAi Screening and Metabolite Profiling
Large-scale pooled screening and data analysis were conducted at the Broad Institute's RNAi Platform as recommended previously (34), and metabolite extraction and targeted mass spectrometry analysis were conducted as reported previously (35, 36).
Cell Lines and Cell Culture
Fresh murine lung tumor nodules were minced and cultured in 100-mm dishes with RPMI-1640/10% FBS/1% penicillin–streptomycin. Calu-1, H358, H2122, and A549 (obtained from American Type Culture Collection) were cultured in RPMI-1640/10% FBS/1% penicillin–streptomycin; and 293ft (Invitrogen) was cultured in Dulbecco's Modified Eagle Medium (DMEM)/10% FBS/1% penicillin–streptomycin. All cells were cultured at 37°C in a humidified incubator with 5% CO2.
Plasmid Constructs and Mutagenesis
pLKO.1–shRNAs were purchased from the Broad Institute (Cambridge, MA). DTYMK (BC030178) cDNAs were purchased from Thermo Scientific. shRNA-resistant DTYMKs were made by mutagenesis PCR and subcloned into the BamHI and XhoI sites of pLenti6 vector (Invitrogen). All mutagenized cDNAs were confirmed by sequencing.
Multiple Routine In Vitro Studies
Lentiviral production and target cell transduction, proliferation assay, quantitative real-time PCR, Western blot analysis, flow cytometry, and immunofluorescence microscopy were conducted as described in the Supplementary Methods.
In Vivo Study
Lkb1-WT and Lkb1-null cells were transduced with pTetOn-shGFP (puromycin) or pTetOn-shDtymk-3 (puromycin), and then 1 million puromycin-resistant cells per transduction were implanted into athymic nude mice. When tumors grew to a diameter of 3 mm, the mice were maintained on doxycycline diet for 3 weeks to allow 634/shGFP tumors reach about 1,000 mm3.
Disclosure of Potential Conflicts of Interest
K. Marks is employed as Associate Director at Agios Pharmaceuticals and has ownership interest (including patents) in the same. E.M. Driggers has commercial research support from Agios Pharmaceuticals. P.A. Janne has received commercial research grants from Pfizer, Boehringer Ingelheim, Sanofi-Aventis, AstraZeneca, Roche, and Genentech. J.A. Engelman has ownership interest (including patents) in Agios Pharmaceuticals and is a consultant/advisory board member of the same. R. Scully has ownership interest (including patents) in Dana-Farber Cancer Institute. A. Kimmelman is a consultant/advisory board member of Forma Therapeutics. L.C. Cantley is on the Board of Directors of Agios Pharmaceuticals, has ownership interest (including patents) in Agios Pharmaceuticals, and is a consultant/advisory board member of the same. K.-K. Wong has commercial research support from Millennium, AstraZeneca, and Infinity, has ownership interest (including patents) in G1 Therapeutics, and is a consultant/advisory board member of MolecularMD. No potential conflicts of interest were disclosed by the other authors.
The Editor-in-Chief of Cancer Discovery (Lewis C. Cantley) is an author of this article. In keeping with the AACR's Editorial Policy, the article was peer reviewed and a member of the AACR's Publications Committee rendered the decision about acceptability.
Conception and design: Y. Liu, K. Marks, G.S. Cowley, P. Gao, Z. Chen, N.E. Sharpless, N. Bardeesy, R. Scully, A.L. Kung, N.S. Gray, D.E. Root, L.C. Cantley, K.-K. Wong
Development of methodology: Y. Liu, G.S. Cowley, T.J.F. Nieland, T.J. Cohoon, E.M. Driggers, J. Zhang, N. Bardeesy, J.M. Asara, L.A. Byers, A.L. Kung, D.E. Root, K.-K. Wong
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Liu, K. Marks, G.S. Cowley, T.J.F. Nieland, C. Xu, T.J. Cohoon, P. Gao, Y. Zhang, Z. Chen, A.B. Altabef, J.H. Tchaicha, X. Wang, E.M. Driggers, J. Zhang, S.T. Bailey, D.N. Hayes, N.M. Patel, P.A. Janne, J.M. Asara, R. Scully, L.A. Byers, D.L. Gibbons, I.I. Wistuba, J.V. Heymach, W.Y. Kim, A.L. Kung, N.S. Gray, D.E. Root, K.-K. Wong
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Liu, K. Marks, G.S. Cowley, J. Carretero, Q. Liu, T.J.F. Nieland, Z. Chen, J.H. Tchaicha, S. Choe, E.M. Driggers, J. Zhang, N.E. Sharpless, D.N. Hayes, N. Bardeesy, J.A. Engelman, B.D. Manning, R.J. Shaw, J.M. Asara, R. Scully, A. Kimmelman, L.A. Byers, D.L. Gibbons, J.V. Heymach, D.J. Kwiatkowski, A.L. Kung, D.E. Root, K.-K. Wong
Writing, review, and/or revision of the manuscript: Y. Liu, K. Marks, Q. Liu, T.J. Cohoon, Z. Chen, E.M. Driggers, N.E. Sharpless, D.N. Hayes, P.A. Janne, N. Bardeesy, J.A. Engelman, R.J. Shaw, R. Scully, A. Kimmelman, L.A. Byers, I.I. Wistuba, J.V. Heymach, D.J. Kwiatkowski, W.Y. Kim, A.L. Kung, N.S. Gray, L.C. Cantley, K.-K. Wong
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Liu, Q. Liu, X. Wang, J. Zhang, D.N. Hayes, K.-K. Wong
Study supervision: Y. Liu, R. Scully, A.L. Kung, K.-K. Wong
This work is supported by the NIH (CA122794, CA140594, CA163896, CA166480, CA154303, and Lung SPORE P50CA090578), United against Lung Cancer Foundation, American Lung Association, and Susan Spooner Research Fund (to K.-K. Wong); and by NIH CA142794 as well as the Damon Runyon Cancer Research Foundation (to W.Y. Kim).
The authors thank Ozan Alkan for helping with the RNAi screen; Hin-Koon Woo, Min Yuan, and Susanne Breitkopf for conducting mass spectrometry; and Jacob B. Reibel for proofreading.
Note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).
- Received January 10, 2013.
- Revision received May 13, 2013.
- Accepted May 13, 2013.
- ©2013 American Association for Cancer Research.