Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHODS FOR CLASSIFYING AND TREATING ADENOCARCINOMAS
Document Type and Number:
WIPO Patent Application WO/2014/152377
Kind Code:
A1
Abstract:
The present disclosure provides methods for classifying and treating cancers, particularly adenocarcinomas, including lung cancers, breast cancers and cervical cancers. The methods utilize a gene signature that identifies cancers as being LKB 1 deficient and, further, being responsive to MEK inhibition therapy.

Inventors:
KAUFMAN JACOB (US)
CARBONE DAVID (US)
Application Number:
PCT/US2014/027272
Publication Date:
September 25, 2014
Filing Date:
March 14, 2014
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
UNIV VANDERBILT (US)
International Classes:
C12Q1/68; G01N33/574
Domestic Patent References:
WO2013019927A12013-02-07
Foreign References:
US20110236903A12011-09-29
US20100130527A12010-05-27
US20110119776A12011-05-19
Other References:
GRIGORIEVA ET AL.: "D-Glucuronyl C5-epimerase suppresses small- cell lung cancer cell proliferation in vitro and tumour growth in vivo", BRITISH JOURNAL OF CANCER, vol. 105, 7 June 2011 (2011-06-07), pages 74 - 82
ZHANG ET AL.: "Unraveling the mystery of cancer metabolism in the genesis of tumor-initiating cells and development of cancer", BIOCHIMICA ET BIOPHYSICA ACTA, vol. 1836, 21 March 2013 (2013-03-21), pages 49 - 59
Attorney, Agent or Firm:
HIGHLANDER, Steven, L. (1120 S. Capital of Texas HighwayBuilding One, Suite 20, Austin TX, US)
Download PDF:
Claims:
WHAT IS CLAIMED:

1. A method of identifying a patient with an LKB 1 -deficient cancer comprising: a) obtaining a tumor sample from said patient; and b) determining a gene expression profile from cells of said tumor sample for AVPI1, BAG1, CPS1, DUSP4, FGA, GLCE, HAL, IRS2, MUC5AC, PDE4D, PTP4A1, RFK, SNF 1LK, TACC2, TFF1 and TSC, wherein an increased level of expression for a majority of BAG1, HAL, IRS2, PTP4A1, RFK, SNF1LK, TACC2, TFF1 and TSC, and a decreased level of expression for a majority of AVPI1, CPS1, DUSP4, FGA, GLCE, MUC5AC and PDE4D, as compared to that observed in a cell expressing functional LKB1, indicates that said tumor is formed from an LKB 1 -deficient cancer.

2. The method of claim 1, wherein 10, 11, 12, 13, 14, 15, or all 16 of the genes are found to have altered expression.

3. The method of claim 1, further comprising administering to said patient a MEK inhibitor when said patient is determined to have an LKB 1 -deficient cancer.

4. The method of claim 3, further comrprising administering to said patient a Raf inhibitor.

5. The method of claim 3, wherein said MEK inhibitor is selumetinib or trametinib.

6. The method of claim 1, wherein said cancer is a breast cancer, lung cancer or cervical cancer.

7. The method of claim 1, wherein said cancer is an adenocarcinoma.

8. The method of claim 1, wherein said patient is a human.

9. The method of claim 1, wherein said patient is a smoker.

10. The method of claim 1, wherein said tumor is recurrent, metastatic and/or multi-drug resistant.

11. The method of claim 1, wherein said tumor sample is a biopsy.

12. The method of claim 1, wherein said tumor sample is a resected tumor tissue.

13. The method of claim 1, further comprising conducting histopathology on said tumor sample.

14. The method of claim 1, further comprising conducting immunohistochemistry on said tumor sample.

15. The method of claim 1, further comprising assessing Lkbl expression and or mutational status on said tumor sample.

16. The method of claim 1, wherein step (b) comprises RT-PCR.

17. The method of claim 1, wherein step (b) further comprises preparing cRNA.

18. The method of claim 1, wherein step (b) further comprises preparing cDNA.

19. The method of claim 1, wherein step (b) comprises microarray hybridization.

20. A kit comprising oligo- or polynucleotides permitting detection and quantification of at least majority of mRNAs products selected from the group consisting of AVPI1, BAGl, CPSl, DUSP4, FGA, GLCE, HAL, IRS2, MUC5AC, PDE4D, PTP4A1, RFK, SNF1LK, TACC2, TFF1 and TSC.

21. The kit of claim 20, further comprising reagents for assessing Lkbl expression and'or mutational status.

Description:
DESCRIPTION

METHODS FOR CLASSIFYING AND TREATING ADENOCARCINOMAS

This application claims benefit of priority to U.S. Provisional Application Serial No. 61/784,757, filed March 14, 2013, and U.S. Provisional Application Serial No. 61/830,087, filed June 1, 2013, the entire contents of both applications being hereby incorporated by reference.

BACKGROUND OF THE INVENTION This invention was made with government support under grant no. NCI

U01CA1 14771 awarded by the National Cancer Institute. The government has certain rights in the invention.

I. FIELD OF THE INVENTION

The invention relates generally to the fields of oncology and molecular biology. More specifically, it relates to the classification and treatment of LKB1 -negative cancers.

II. RELATED ART

Non-small cell lung cancer (NSCLC) is the most common and lethal cancer worldwide. At least three major histologies of NSCLC are described: squamous carcinoma (48%), large cell carcinoma (12%) and adenocarcinoma (40%) The standard treatment for these patients is systemic chemotherapy. However, systemic chemotherapy has modest efficacy and has not greatly prolonged the median survival (8-12 months) or 5-year survival rates (2%) in these patients. Although these sub-types differ markedly in histologic appearance and gene expression, each is highly lethal, and until recently, little clinical distinction has been made among these entities.

It is anticipated that a better understanding of the molecular mechanisms involved in the initiation and progression lung tumorigenesis - as well as the impact of environmental exposures on lung carcinogenesis - would help guide the development of better and more targeted lung cancer therapeutics as well as potential prevention strategies. The recent identification of the oncogenic kinase domain mutations in the epidermal growth factor receptor (EGFR) in human lung adenocarcinomas and their association with sensitivity to small molecule EGFR kinase inhibitors such as gefitinib and erlotinib further support the that the molecular understanding of the mechanisms involved in lung tumorigenesis will lead to advances in patient screening, development of better targeted therapeutics, and identification of patients who are best suited for each type of targeted treatment. Activating K-RAS, EGFR, and BRAF mutations comprise the most common oncogenic mutations in human NSCLC. However, their interaction with other concurrent tumor suppressor loss is not well understood.

Serine-threonine kinase 11 (STK11 or LKBl) is a tumor suppressor and one of the most commonly mutated genes in non-small cell lung cancers 6"8 . It has been shown to play important roles in embryologic development, cellular polarity, motility, and transcriptional regulation and represents a 'metabolic checkpoint' via its interactions with AMPK and the mTOR pathway 9 ' 10 . Although the mTOR pathway has been reported to be active in tumors lacking LKBl, interactions between mTOR and other dysregulated pathways and the resulting tumor phenotypes remain poorly characterized. Better understanding of the biology of these tumors may identify crucial pathways that can be targeted to improve patient outcome with individualized therapy. To further this understanding, murine models examining LKBl loss in cancer have been used to probe phenotypic and gene expression changes associated with alterations in this pathway in both lung cancer and melanoma. Intriguing results from these studies implicate the TGF-beta and SRC pathways in the metastasis of LKB l -deficient tumors, and have been used to probe the efficacy of MEK inhibition as a potential treatment strategy 2,3 . However, the validity of these models in predicting human disease phenotypes has not been evaluated.

SUMMARY OF THE INVENTION

Thus, in accordance with the present disclosure, there is provided a method of identifying a patient with an LKBl -deficient cancer comprising a) obtaining a tumor sample from said patient; and b) determining a gene expression profile from cells of said tumor sample for AVPI1, BAG1, CPS1, DUSP4, FGA, GLCE, HAL, IRS2, MUC5AC, PDE4D, PTP4A1, RFK, SNF1LK, TACC2, TFF1 and TSC, wherein an increased level of expression for a majority of BAG1, HAL, IRS2, PTP4A1, RFK, SNF 1LK, TACC2, TFF1 and TSC, and a decreased level of expression for a majority of AVPI1, CPS 1, DUSP4, FGA, GLCE, MUC5AC and PDE4D, as compared to that observed in a cell expressing functional LKBl, indicates that said tumor is formed from an LKBl -deficient cancer. The method may result in 10, 11, 12, 13, 14, 15, or all 16 of the genes are found to have altered expression. LKB 1- deficiency may result from the lack of expression of any gene product, from expression of a non- functional protein or a altered function protein.

The method may further comprise administering to said patient a MEK inhibitor when said patient is determined to have an LKBl -deficient cancer. The method may also further comprising administering to said patient a af inhibitor. The MEK inhibitor may be selumetinib or trametinib. The cancer may be is a lung cancer, and/or an adenocarcinoma. The patient may be a human, and/or may be a smoker. The tumor may be recurrent, metastatic and/or multi-drug resistant. The tumor sample may be a biopsy or a resected tumor tissue. The method may further comprise conducting histopathology and/or immunohistochemistry on said tumor sample. The method may further comprise assessing Lkbl expression and/or mutational status on said tumor sample. Step (b) may comprise RT- PCR, including further further preparing cR A or preparing cDNA, or and/or microarray hybridization.

It is contemplated that any method or composition described herein can be implemented with respect to any other method or composition described herein.

The use of the word "a" or "an" when used in conjunction with the term "comprising" in the claims and/or the specification may mean "one," but it is also consistent with the meaning of "one or more," "at least one," and "one or more than one."

These, and other, embodiments of the invention will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following description, while indicating various embodiments of the invention and numerous specific details thereof, is given by way of illustration and not of limitation. Many substitutions, modifications, additions and/or rearrangements may be made within the scope of the invention without departing from the spirit thereof, and the invention includes all such substitutions, modifications, additions and/or rearrangements.

BRIEF DESCRIPTION OF THE DRAWINGS

The following drawings form part of the present specification and are included to further demonstrate certain aspects of the present invention. The invention may be better understood by reference to one or more of these drawings in combination with the detailed description of specific embodiments presented herein. FIGS 1A-E. LKBl loss produces a characteristic pattern of gene expression. (FIG. 1A) The significance of gene overlap is shown for pairwise comparisons of the top 200 genes over-expressed in tumors with LKB l loss in 14 studies of lung adenocarcinomas. Asterisks indicate comparisons between cell lines expressing vector control and those expressing wild- type LKBl . P-values from a hypergeometric test are color coded according to the legend (see also FIG. 5). (FIG. IB) Unsupervised hierarchical clustering of 178 resected lung adenocarcinomas using a 129 gene signature of LKB l loss. Tumors are shown on the horizontal axis, with loss of LKB l highlighted in red; genes are shown on the vertical axis, with four clusters of gene expression highlighted in red. (FIG. 1C) Sensitivity and specificity of the LKBl classifier for prediction of LKBl mutations across independent testing sets; p- value represents the result of the Fisher's exact test (FIGS. 1D-E) Expression of LKBl mRNA is shown as standard deviations from the mean among resected lung adenocarcinomas classified as LKBl loss or LKBl wild-type among tumors in which LKBl has been sequenced (FIG. ID) or with unknown LKBl mutation status (FIG. IE). Each dot represents one tumor and p-values are derived from the student's t-test comparing indicated groups.

FIG. 2. Pharmacologic and genetic perturbations affect the expression of transcriptional nodes. The results of gene set enrichment analysis are shown for selected perturbations that affect the expression of the three transcriptional clusters upregulated in LKBl -deficient tumors. P-values represent the significance of a hypergeometric test for the indicated comparisons.

FIGS. 3A-D. Restoring wild-type LKBl in cell lines harboring mutations slows growth and attenuates the expression of the LKBl-deficient gene signature. (FIG. 3A) Immunoblots of whole-cell lysates from A549, H2122, and H460 stably expressing emtpy pBABE vector, LKBl or K78I LKBl after puromycin selection. (FIGS. 3B-C) Changes in gene expression of A549 (FIG. 3B) or H2122 (FIG. 3C) cell lines after re-expressing LKBl or K78I LKBl were compared to the gene lists for each of the four LKBl -associated clusters using a hypergeometric test. (FIG. 3DA) Activity of CRE-luciferase is shown for A549, H2122, and H460 cell line after stable expression of LKBl or K78I LKBl. Reporter activations were determined relative to a control luciferase with mutated CRE sites, and are shown relative to the pBABE control. P-values show the significance of unpaired student's t- tests. FIGS 4A-G. LKB1 loss confers sensitivity to selumetinib. (FIGS. 4A-B) Maximum inhibitory effect of selumetinib is shown for cell lines with high expression of the L B1 classifier compared to those with low expression in both a training (FIG. 4A) and testing (FIG. 4B) cohort from the CCLE dataset. (FIGS. 4C-F)Association between selumetinib and LKB l-loss signature is shown for cell lines with mutations in LKB1 (FIG. 4C), BRAF (FIG. 4D), KRAS (FIG. 4E) and wild-type for BRAF, KRAS, NRAS, and HRAS (FIG. 4F). For (FIGS. 4A-F) cell lines classified as LKBl-loss are marked as 'classifier positive' while those with a wild-type signature are marked as 'classifier negative'. Distributions and medians are plotted and P-values represent the significance of Welch's t-test. (FIG. 4G) Cell viability was determined by colorimetric Alamar Blue assay, and is shown for cell lines stably transduced with either pBABE, LKB1, or LKB 1 K78I and treated with 1.0 μΜ selumetinib for 72 hours. Mean viability is shown relative to untreated controls with error bars representing standard deviation. P-values represent the significance of unpaired student's t-tests.

FIG. 5. Association matrix for down-regulated genes associated with LKBl loss. The significance of gene overlap is shown for pairwise comparisons of the 200 genes with the most significantly decreased expression in tumors with LKBl loss across 14 studies of lung adenocarcinomas. Asterisks indicate comparisons between cell lines expressing vector control and those expressing wild-type LKBl. P-values from a hypergeometric test are color coded according to the legend. FIGS. 6A-F. Distribution of LKBl loss scores and receiver operating characteristics in NSCLC cell lines and resected lung adenocarcinomas. Distribution of LKBl loss scores and receiver operating characteristics in NSCLC cell lines and resected lung adenocarcinomas. (FIG. 6A) Distribution of LKBl loss scores among NSCLC cell lines with or without LKBl mutations. (FIG. 6B) Receiver operating characteristics for the prediction of NSCLC cell line mutations using the LKBl loss score. (FIGS. 6C-D) Analysis of pooled resected lung adenocarcinomas showing score distribution (FIG. 6C) and receiver operating characteristics (FIG. 6D). (FIGS. 6E-F) Analysis of resected lung adenocarcinomas studied by the TCGA showing score distribution (FIG. 6E) and receiver operating characteristics (FIG. 6F). In FIGS. 6B, 6D and 6E, red points correspond to the score cutoff of 0.2 used throughout the paper. AUC refers to area under the curve and the p-value results from the Mann-Whitney U test. FIGS. 7A-B. Association between patient outcome and LKBl loss. Resected LUAD tumors from the Director's Challenge Consortium (n=441) were classified as LKB1- loss or LKB l WT using the LKBl -classifier score. Kaplan-Meier curves were used to plot cumulative events for these two groups for overall survival (FIG. 7A) or progression-free survival (FIG. 7B). P-values represent the results of the log-rank test; the number of evaluable tumors remaining are given at yearly intervals below each plot.

FIG. 8. Comparison of LKBl loss scores derived from independent training cohorts. LKB l-loss scores are plotted for two distinct classifiers: Score A is the primary 16- gene LKB l-loss classifier used throughout this study and Score B results from an independent classifier derived by applying the same statistical approach to a different training cohort. The concordance is the percentage of tumors that are given the same LKB l-loss classification by each of the two classifiers.

FIGS. 9A-B. TGF-beta mRNA expression is decreased in tumors with LKBl loss.

(FIG. 9A) The distribution of TGF-beta mRNA expression is shown for resected lung adenocarcinomas in the Director's Challenge Consortium (n=449) that are predicted by the LKBl-loss classifier to exhibit LKBl loss or wild-type LKBl; the P-value represents the result of a student's t-test comparing these groups. (FIG. 9B) Induction of TGF-beta mRNA relative to pBABE vector control is shown for A549 and H2122 cell lines after stable expression of wild-type LKBl . The range is plotted and p-values represent the result of student's t-test of the indicated comparisons.

FIG. 10. TGF-beta and c-src pathways antagonize the expression of genes in the CREB cluster. The significance of gene overlap is shown for comparisons of the CREB signature to the genes perturbed by TGF-beta or dasatinib treatment of the LKBl-mutant cell line A549 at the various time points or concentrations shown. P-values from a hypergeometric test are shown on the ordinate axis with positive values indicating an induction of CREB-associated genes and negative values indicating repression.

FIGS. 11A-B. Stable expression of LKBl in cell lines slows proliferation.

Proliferation was assessed in LKBl-mutant cell lines A549 (FIG. 11A) and H2122 (FIG. 1 IB) after stable transduction of empty pBABE vector, LKB1-WT or LKB l K78I. Alamar blue assay was used to determine relative cell counts, which are plotted as the mean for each time point relative to values at day 1, with error bars representing standard deviations. P-values represent the results of a student's t-test comparing LKB1-WT to LKB 1-K78I. FIGS. 12A-C. Expression of wild-type LKB1 in A549, H2122, or HeLa cell lines decreases the expression of the genes in the CREB transcriptional node. (FIGS. 12A-C) Microarray gene expression was measured in triplicate after stable expression of pBABE vector or LKB1 -wild-type in A549 (FIG. 12A), H2122 (FIG. 12B) or HeLa (FIG. 12C). For each gene comprising the CREB transcriptional node the average change in gene expression (log base 2) is plotted comparing LKB1 wild-type to pBABE control. Error bars represent standard deviations; P-values represent the results of student's t-test comparing these groups. NA indicates minimal basal expression.

FIGS. 13A-B. Cell lines with expression of the LKB1 classifier score show increased susceptibility to MEK inhibition. (FIG. 13 A) The associations between L B1 classifier score and IC50 values for four MEK inhibitors are shown for the GDSC training cohorts, and for two MEK inhibitors in the CCLE training and testing cohorts. (FIG. 13B) Associations between LKBl-loss score and the maximum inhibitory effect of selumetinib or PD-0325901 are shown in the CCLE training and testing cohorts. The values plotted are the linear regression coefficients relating the LKBl-loss score to the indicated inhibitor IC 50 , with bars representing their 95% confidence interval. P-values represent the significance associated with the linear regression coefficient.

FIGS. 14A-C. Decreased LKB1 mRNA in association with LKB1 loss signature in resected breast cancer, lung squamous, and cervical cancer. Expression of LKB 1 mRNA is shown for tumors classified as LKBl loss or LKB 1 WT in a pooled cohort of lung squamous cell carcinomas (GSE4573; TCGA) (FIG. 14A), among the TCGA breast cancer specimens (TCGA) (FIG. 14B), or in a pooled cohort of cervical squamous cell carcinoma (GSE38964; GSE20167; TCGA) (FIG. 14C). Bottom panel - prediction of LKB l mutations in breast, lung squamous, and endometrial tumors analyzed by TCGA. FIGS. 15A-B. Comparison of protein and gene expression differences associated with known LKBl mutations or associated with predicted LKBl loss among LKBl WT tumors. (FIG. 15 A) Differences in protein expression determined by TCGA RPPA are shown for proteins with significant association with LKBl mutations (having p<0.01). (FIG. 15B) Differences in mRNA expression determined by TCGA RNAseq analysis are shown for genes with significant association with LKBl mutations (having ρ<Γ 6 ). Dots represent individual proteins or genes. The x-axis shows the difference in average expression between LKBl mutant tumors and LKBl-WT tumors with WT classification score. The y-axis shows the difference in average expression between LKB1-WT tumors with a LKBl-loss classification score vs LKB1-WT tumors with a WT classification score.

FIGS. 16A-E. Responsiveness to MEK inhibition for three previously published MEK signatures, in contrast to LKBl loss signature. Changes in gene expression corresponding to different signatures of MEK sensitivity are shown after treatment with MEK inhibitor. (FIG. 16A) Analysis of 15 gene 'MEK functional activation' signature reported in Dry, et al. (2010). (FIG. 16B) Analysis of 58 genes reported to be correlated with MEK sensitivity in Garnett et al. (2012). (FIG. 16C) Analysis of 90 gene signature reported in Loboda et al. (2010). (FIG. 16D) Analysis of 16 gene signature of LKBl loss reported here. (FIG. 16E) Analysis of top 200 genes associated with the 16 gene LKBl loss signature. The y-axis represents the log-transformed p-values for hypergeometric test of overlap significance between each signature and the genes perturbed by MEK inhibition in 34 cell lines (22 pancreatic, six skin, three breast, two colon, one lung) from two studies. Pancreatic cell lines were treated with 2 mM CI- 1040 for 24 hours, while other cell lines were treated with 50 nM PD0325901 for eight hours.

DETAILED DESCRIPTION OF THE INVENTION

Inactivation of STK11/LKB1 is one of the most common genetic events in lung cancer, and understanding the cellular phenotypes and molecular pathways altered as a consequence will aid the development of therapeutic strategies targeting LKBl -deficient cancers. A murine model of LKB l loss in lung tumors has given insight into these questions, yielding gene expression signatures that identify dysregulated pathways in primary tumors and metastases, and showing alterations in drug susceptibility in a mouse clinical trial 1"3 . However, the fidelity of this mouse model in representing human disease phenotypes is unknown and no large-scale study of LKBl loss in human tumors has been reported.

Here, the inventors report the comprehensive analysis of gene expression patterns associated with LKBl functional loss in human lung adenocarcinomas, through which they identify up-regulation of the mitochondrial respiratory chain and activation of CREB and NRF2 transcription factors as key features in these tumors. A 16-gene signature is predictive of both mutational and non-mutational LKBl loss across human tumors, including resected lung adenocarcinomas and cell lines of both lung and non-lung histology. Moreover, cell lines expressing this signature show increased sensitivity to MEK inhibition across two large studies 4,5 , independent of the sensitivity conferred by mutations in RAS and RAF family members. The inventors also show that restoration of LKBl in lung cancer cell lines down- regulates the gene expression pattern, attenuates CREB activation, and induces resistance to MEK inhibition. These results are distinct from those reported in the murine model, and further the understanding of the biology of LKBl -deficient tumors. Furthermore, loss of LKB 1 is shown to be a determinant of MEK sensitivity and indicates positive response to MEK inhibitors. I. Non-Small Cell Lung Cancer

Non-small-cell lung carcinoma (NSCLC) is any type of epithelial lung cancer other than small cell lung carcinoma (SCLC). As a class, NSCLCs are relatively insensitive to chemotherapy, compared to small cell carcinoma. When possible, they are primarily treated by surgical resection with curative intent, although chemotherapy is increasingly being used both pre-operatively (neoadjuvant chemotherapy) and post-operatively (adjuvant chemotherapy). The most common types of NSCLC are squamous cell carcinoma, large cell carcinoma, and adenocarcinoma, but there are several other types that occur less frequently, and all types can occur in unusual histologic variants and as mixed cell-type combinations. Sometimes the phrase "non-small-cell lung cancer" ("not otherwise specified," or NOS) is used generically, usually when a more specific diagnosis cannot be made. This is most often the case when a pathologist examines a small amount of malignant cells or tissue in a cytology or biopsy specimen. Lung cancer in never-smokers is almost universally NSCLC, with a sizeable majority being adenocarcinoma. On relatively rare occasions, malignant lung tumors are found to contain components of both SCLC and NSCLC. In these cases, the tumors should be classified as combined small cell lung carcinoma (c-SCLC), and are (usually) treated like "pure" SCLC.

Squamous cell carcinoma (SCC) of the lung is more common in men than in women. It is closely correlated with a history of tobacco smoking, more so than most other types of lung cancer. According to the Nurses' Health Study, the relative risk of SCC is approximately 5.5, both among those with a previous duration of smoking of 1 to 20 years, and those with 20 to 30 years, compared to never-smokers. The relative risk increases to approximately 16 with a previous smoking duration of 30 to 40 years, and approximately 22 with more than 40 years.

It most often arises centrally in larger bronchi, and while it often metastasizes to locoregional lymph nodes (particularly the hilar nodes) early in its course, it generally disseminates outside the thorax somewhat later than other major types of lung cancer. Large tumors may undergo central necrosis, resulting in cavitation. A squamous cell carcinoma is often preceded for years by squamous cell metaplasia or dysplasia in the respiratory epithelium of the bronchi, which later transforms to carcinoma in situ. In carcinoma in situ, atypical cells may be identified by cytologic smear test of sputum, bronchoalveolar lavage or samples from endobronchial brushings. However, squamous-cell carcinoma in situ is asymptomatic and undetectable on X-ray radiographs. Eventually, it becomes symptomatic, usually when the tumor mass begins to obstruct the lumen of a major bronchus, often producing distal atelectasis and infection. Simultaneously, the lesion invades into the surrounding pulmonary substance. On histopathology, these tumors range from well differentiated, showing keratin pearls and cell junctions, to anaplastic, with only minimal residual squamous cell features.

Currently, four variants (papillary, small cell, clear cell, and basaloid) of squamous cell carcinoma of the lung are recognized. Of these variants, there is some evidence that the basaloid and poorly differentiated small-cell variants may have worse prognoses than "conventional" squamous cell carcinomas. The papillary variant occurs more frequently as a primarily superficial, endobronchial lesion, with a modestly better prognosis. Very little data is currently available on the clear cell variant of squamous cell carcinoma, and no consensus has been reached on the prognostic implications of clear cell changes in lung cancer.

Recently, four mRNA expression subtypes (primitive, basal, secretory, and classical) were identified and validated within squamous cell carcinoma. The primitive subtype correlates with worse patient survival. These subtypes, defined by intrinsic expression differences, provide a possible foundation for improved patient prognosis and research into individualized therapies.

Large cell lung carcinoma (LCLC) is a heterogeneous group of undifferentiated malignant neoplasms originating from transformed epithelial cells in the lung. LCLC's have typically comprised between around 10% of all NSCLC in the past, although newer diagnostic techniques seem to be reducing the incidence of diagnosis of "classic" LCLC in favor of more poorly differentiated squamous cell carcinomas and adenocarcinomas. LCLC is, in effect, a "diagnosis of exclusion," in that the tumor cells lack light microscopic characteristics that would classify the neoplasm as a small-cell carcinoma, squamous-cell carcinoma, adenocarcinoma, or other more specific histologic type of lung cancer. LCLC is differentiated from small cell lung carcinoma (SCLC) primarily by the larger size of the anaplastic cells, a higher cytoplasmic -to-nuclear size ratio, and a lack of "salt-and-pepper" chromatin.

Adenocarcinoma of the lung is currently the most common type of lung cancer in "never smokers" (lifelong non-smokers). Adenocarcinomas account for approximately 40% of lung cancers. Historically, adenocarcinoma was more often seen peripherally in the lungs than small cell lung cancer and squamous cell lung cancer, both of which tended to be more often centrally located. Interestingly, however, recent studies suggest that the "ratio of centrally-to-peripherally occurring" lesions may be converging toward unity for both adenocarcinoma and squamous cell carcinoma.

II. LKB1

Serine/threonine kinase 11 (STK1 1) also known as liver kinase Bl (L B1) or renal carcinoma antigen NY-REN- 19 is a protein kinase that in humans is encoded by the STK11 gene. Testosterone and DHT treatment of murine 3T3-L1 or human SGBS adipocytes for 24 h significantly decreased the mRNA expression of LKB 1 via the androgen receptor and consequently reduced the activation of AMPK by phosphorylation. In contrast, Πβ-estradiol treatment increased L B 1 mRNA, an effect mediated by estrogen receptor a. The STK11/LKB1 gene, which encodes a member of the serine/threonine kinase, regulates cell polarity and functions as a tumor suppressor.

However in ER-positive breast cancer cell line MCF-7, estradiol caused a dose- dependent decrease in LKBl transcript and protein expression leading to a significant decrease in the phosphorylation of the LKBl target AMPK. ERa binds to the STK1 1 promoter in a ligand-independent manner and this interaction is decreased in the presence of estradiol. Moreover, STKl 1 promoter activity is significantly decreased in the presence of estradiol. LKBl is a primary upstream kinase of adenine monophosphate-activated protein kinase (AMPK), a necessary element in cell metabolism that is required for maintaining energy homeostasis. It is now clear that LKB l exerts its growth suppressing effects by activating a group of other -14 kinases, comprising AMPK and AMPK-related kinases. Activation of AMPK by LKB l suppresses growth and proliferation when energy and nutrient levels are scarce. Activation of AMPK-related kinases by LKBl plays vital roles maintaining cell polarity thereby inhibiting inappropriate expansion of tumor cells. A picture from current research is emerging that loss of LKBl leads to disorganization of cell polarity and facilitates tumor growth under energetically unfavorable conditions.

Germline mutations in this gene have been associated with Peutz-Jeghers syndrome, an autosomal dominant disorder characterized by the growth of polyps in the gastrointestinal tract, pigmented macules on the skin and mouth, and other neoplasms. Recent studies have uncovered a large number of somatic mutations of the LKBl gene that are present in lung, cervical, breast, intestinal, testicular, pancreatic and skin cancer.

LKBl is activated allosterically by binding to the pseudokinase STRAD and the adaptor protein M025. The LKB 1-STRAD-M025 heterotrimeric complex represents the biologically active unit, that is capable of phosphorylating and activating AMPK and at least 12 other kinases that belong to the AMPK-related kinase family. The crystal structure of the LKB 1-STRAD-M025 complex was elucidated using X-ray crystallography, and revealed the mechanism by which LKBl is allosterically activated. LKB l has a structure typical of other protein kinases, with two (small and large) lobes on either side of the ligand ATP-binding pocket. STRAD and M025 together cooperate to promote LKBl active conformation. The LKB l activation loop, a critical element in the process of kinase activation, is held in place by M025, thus explaining the huge increase in LKBl activity in the presence of STRAD and M025. III. MEK Inhibitors

MEK inhibitors are defined herein as a chemical or drug that inhibits MEKl and/or MEK2, mitogen-activated protein kinase kinase enzymes. They can be used to affect the MAPK/ERK pathway which is often overactive in some cancers. Hence, MEK inhibitors have potential for treatment of some cancers, especially BRAF-mutated melanoma,http://en.wikipedia.org/wiki/MEK Inhibitor - cite note-ASCO2012-2 and KRAS/BRAF mutated colorectal cancer. In the context of the present disclosure, they will find particular use in treating LKB 1 -deficient cancers. Some specific MEK inhibitors include trametinib (GSK1 120212), selumetinib, MEK162, PD-325901, XL518, CI- 1040, and PD035901.

Trametinib (GSK1120212) is experimental cancer drug. It is a MEK inhibitor drug with anti-cancer activity. It inhibits both MEKl and MEK2. Trametinib had good results for V600E mutated metastatic melanoma in a hase III clinical trial. Its structure is shown below:

Selumetinib (AZD6244) is a drug being investigated for the treatment of various types of cancer, for example non-small cell lung cancer (NSCLC). The gene BRAF is part of the MAPK/ERK pathway, a chain of proteins in cells that communicates input from growth factors. Activating mutations in the BRAF gene, primarily V600E, are associated with lower survival rates in patients with papillary thyroid cancer. Another type of mutation that leads to undue activation of this pathway occurs in the gene KRAS and is found in NSCLC. A possibility of reducing the activity of the MAPK/ERK pathway is to block the enzyme MAPK kinase (MEK), immediately downstream of BRAF, with the drug selumetinib. More specifically, selumetinib blocks the subtypes MEKl and MEK2 of this enzyme. In addition to thyroid cancer, BRAF-&c\vtdX g mutations are prevalent in melanoma (up to 59%), colorectal cancer (5-22%), serous ovarian cancer (up to 30%), and several other tumor types. KRAS mutations appear in 20 to 30% of NSCLC cases and about 40% of colorectal cancer. Its structure is shown below:

The foregoing compounds are exemplary in nature and do not limit the scope of the invention. Indeed, any MEK inhibitor may be used in accordance with the present invention.

Where clinical applications are contemplated, it will be necessary to prepare pharmaceutical compositions containing MEK inhibitors. Generally, this will entail preparing compositions that are essentially free of pyrogens, as well as other impurities that could be harmful to humans or animals.

One will generally desire to employ appropriate salts and buffers to render materials stable and allow for uptake by target cells. Aqueous compositions of the present invention comprise an effective amount of the vector to cells, dissolved or dispersed in a pharmaceutically acceptable carrier or aqueous medium. Such compositions also are referred to as inocula. The phrase "pharmaceutically or pharmacologically acceptable" refers to molecular entities and compositions that do not produce adverse, allergic, or other untoward reactions when administered to an animal or a human. As used herein, "pharmaceutically acceptable carrier" includes any and all solvents, dispersion media, coatings, antibacterial and antifungal agents, isotonic and absorption delaying agents and the like. The use of such media and agents for pharmaceutically active substances is well know in the art. Except insofar as any conventional media or agent is incompatible with the vectors or cells of the present invention, its use in therapeutic compositions is contemplated. Supplementary active ingredients also can be incorporated into the compositions.

The active compositions of the present invention may include classic pharmaceutical preparations. Administration of these compositions according to the present invention will be via any common route so long as the target tissue is available via that route. Such routes include oral, nasal, buccal, rectal, vaginal or topical route. Alternatively, administration may be by orthotopic, intradermal, subcutaneous, intramuscular, intraperitoneal, or intravenous injection. Such compositions would normally be administered as pharmaceutically acceptable compositions, described supra. Of particular interest is direct intratumoral administration, perfusion of a tumor, or admininstration local or regional to a tumor, for example, in the local or regional vasculature or lymphatic system, or in a resected tumor bed. The active compounds may also be administered parenterally or intraperitoneally.

Solutions of the active compounds as free base or pharmacologically acceptable salts can be prepared in water suitably mixed with a surfactant, such as hydroxypropylcellulose. Dispersions can also be prepared in glycerol, liquid polyethylene glycols, and mixtures thereof and in oils. Under ordinary conditions of storage and use, these preparations contain a preservative to prevent the growth of microorganisms.

The pharmaceutical forms suitable for injectable use include sterile aqueous solutions or dispersions and sterile powders for the extemporaneous preparation of sterile injectable solutions or dispersions. In all cases the form must be sterile and must be fluid to the extent that easy syringability exists. It must be stable under the conditions of manufacture and storage and must be preserved against the contaminating action of microorganisms, such as bacteria and fungi. The carrier can be a solvent or dispersion medium containing, for example, water, ethanol, polyol (for example, glycerol, propylene glycol, and liquid polyethylene glycol, and the like), suitable mixtures thereof, and vegetable oils. The proper fluidity can be maintained, for example, by the use of a coating, such as lecithin, by the maintenance of the required particle size in the case of dispersion and by the use of surfactants. The prevention of the action of microorganisms can be brought about by various antibacterial and antifungal agents, for example, parabens, chlorobutanol, phenol, sorbic acid, thimerosal, and the like. In many cases, it will be preferable to include isotonic agents, for example, sugars or sodium chloride. Prolonged absorption of the injectable compositions can be brought about by the use in the compositions of agents delaying absorption, for example, aluminum monostearate and gelatin.

Sterile injectable solutions are prepared by incorporating the active compounds in the required amount in the appropriate solvent with various of the other ingredients enumerated above, as required, followed by filtered sterilization. Generally, dispersions are prepared by incorporating the various sterilized active ingredients into a sterile vehicle which contains the basic dispersion medium and the required other ingredients from those enumerated above. In the case of sterile powders for the preparation of sterile injectable solutions, the preferred methods of preparation are vacuum-drying and freeze-drying techniques which yield a powder of the active ingredient plus any additional desired ingredient from a previously sterile-filtered solution thereof.

As used herein, "pharmaceutically acceptable carrier" includes any and all solvents, dispersion media, coatings, antibacterial and antifungal agents, isotonic and absorption delaying agents and the like. The use of such media and agents for pharmaceutical active substances is well known in the art. Except insofar as any conventional media or agent is incompatible with the active ingredient, its use in the therapeutic compositions is contemplated. Supplementary active ingredients can also be incorporated into the compositions.

For oral administration the polypeptides of the present invention may be incorporated with excipients and used in the form of non-ingestible mouthwashes and dentifrices. A mouthwash may be prepared incorporating the active ingredient in the required amount in an appropriate solvent, such as a sodium borate solution (Dobell's Solution). Alternatively, the active ingredient may be incorporated into an antiseptic wash containing sodium borate, glycerin and potassium bicarbonate. The active ingredient may also be dispersed in dentifrices, including: gels, pastes, powders and slurries. The active ingredient may be added in a therapeutically effective amount to a paste dentifrice that may include water, binders, abrasives, flavoring agents, foaming agents, and humectants. The compositions of the present invention may be formulated in a neutral or salt form.

Pharmaceutically-acceptable salts include the acid addition salts (formed with the free amino groups of the protein) and which are formed with inorganic acids such as, for example, hydrochloric or phosphoric acids, or such organic acids as acetic, oxalic, tartaric, mandelic, and the like. Salts formed with the free carboxyl groups can also be derived from inorganic bases such as, for example, sodium, potassium, ammonium, calcium, or ferric hydroxides, and such organic bases as isopropylamine, trimethylamine, histidine, procaine and the like.

Upon formulation, solutions will be administered in a manner compatible with the dosage formulation and in such amount as is therapeutically effective. The formulations are easily administered in a variety of dosage forms such as injectable solutions, drug release capsules and the like. For parenteral administration in an aqueous solution, for example, the solution should be suitably buffered if necessary and the liquid diluent first rendered isotonic with sufficient saline or glucose. These particular aqueous solutions are especially suitable for intravenous, intramuscular, subcutaneous and intraperitoneal administration. In this connection, sterile aqueous media which can be employed will be known to those of skill in the art in light of the present disclosure. For example, one dosage could be dissolved in 1 ml of isotonic NaCl solution and either added to 1000 ml of hypodermoclysis fluid or injected at the proposed site of infusion, (see for example, "Remington's Pharmaceutical Sciences," 15th Edition, pages 1035-1038 and 1570-1580). Some variation in dosage will necessarily occur depending on the condition of the subject being treated. The person responsible for administration will, in any event, determine the appropriate dose for the individual subject. Moreover, for human administration, preparations should meet sterility, pyrogenicity, general safety and purity standards as required by FDA Office of Biologies standards.

IV. Gene Signature and Detection Methods

Accordingly, the invention provides methods of detecting and evaluating the presence or absence of LKB1 function in a subject having cancer, including identifying those cells/cancers in which L B1 is entirely absent or expressed in a non-functional form. In addition, the differentially expressed genes identified herein are used for predicting the success of certain therapeutic approaches to treating adenocarcinomas. The subject is generally a mammal, and is typically human female or human male. The subject typically will have been diagnosed as having cancer, and possibly has already undergone treatment for the cancer. Diagnosis of cancer is typically made through the identification of a mass on an examination, though it may also be through other means such as a radiological diagnosis, or ultrasound.

The signature includes the genes AVPI1 (arginine vasopressin-induced 1; GenelD: 60370), BAG1 (BCL2-associated athanogene; GenelD: 573), CPS1 (carbamoyl-phosphate synthase 1, mitochondrial; GenelD: 1373), DUSP4 (dual specificity phosphatase 4; GenelD: 1846), FGA (fibrinogen alpha chain; GenelD: 2243), GLCE (glucuronic acid epimerase; GenelD: 26035), HAL (histidine ammonia-lyase; GenelD: 3034), IRS2 (insulin receptor substrate 2; GenelD: 8660), MUC5AC (mucin 5AC, oligomeric mucus/gel-forming; GenelD: 4586), PDE4D (phosphodiesterase 4D, cAMP-specific; GenelD: 5144), PTP4A1 (protein tyrosine phosphatase type IVA, member 1; GenelD: 7803), RFK (riboflavin kinase; GenelD: 55312), SIK1 (salt-inducible kinase 1; GenelD: 150094), TACC2 (transforming, acidic coiled-coil containing protein 2; GenelD: 10579), TFF1 (trefoil factor 1; GenelD: 7031), TESC (tescalcin; GenelD: 54997). Of these AVPI1, CPS1, DUSP4, FGA, GLCE, MUC5AC and PDE4D were found to be down-regulated in LKB1 -negative cells, whereas BAG1, HAL, IRS2, PTP4A1, RFK, SNF1LK, TACC2, TFF1 and TSC were found to be up-regulated in LKB1 -negative cells.

Table A - Gene Signature Genes

Gene Symbol 1 Gene Name ! E TREZ GE RefSeq Trans 6B_ACC

2 AVPI1 1 arginine yasopressin-induced 1 I 60370 MM.021732 NMJ21732

3 BAQ1 ! BCL2-associated athanogene ! 573 NM.00 2 NMJ04323

4 CPS1 jcar&amoyl-phosphate synthase 1, mitochondrial ! 1373 NMjOO1 22 AF154830

5 DUSP4 dual specificity phosphatase 4 I 1846 NM_001394 MJ01394

5 FGA fibrinogen alpha chain I 2243 _000508 NMJ00508

7 GLCE i glucuronic acid epimerase I 26035 N J 5554 W87398

8 HAL fiistitiine ammonia-iyase I 3034 N _002 08 NMJ02108

9 IRS2 I insulin receptor substrate 2 I 8660 N _003749 AF073310

10 MUC5AC I mucin 5AC, oiigomeric mticus gel-forming I 4586 XM..003119 AVV192795

11 PDF.4D Iphosphodissterase 4D, cA P-specific I 5144 MM .001 04 R40917

1?. PTP4A1 protein tyrosine phosphatase type IVA, member 1 i 7803 NMJTO3463 BF576710

13 RFK riboflavin kinase I 55312 NM J013339 BF340123

14 SIK1 salt-inducible kinase 1 1 150094 N _i 73354 MJ30751

15 TACC2 !transtorrrting, acidic coiled-coll containing protein 2 i 10579 NM_006997 NMJ06997

16 TPr jireioii iaetor 1 I 7031 ΝΜ_00322δ N!W_003225

17 TESC isscaicii " : 1 54997 N JJ01 68 NMJ1 899

Expression may be determined at the protein level, i.e., by measuring the levels of proteins encoded by the gene products described herein, or activities thereof. Such methods are well known in the art and include, e.g., immunoassays based on antibodies to proteins encoded by the genes. Immunoassays carried out in accordance with the present invention may be homogeneous assays or heterogeneous assays. In a homogeneous assay the immunological reaction usually involves the specific antibody, a labeled analyte, and the sample of interest. The signal arising from the label is modified, directly or indirectly, upon the binding of the antibody to the labeled analyte. Both the immunological reaction and detection of the extent thereof are carried out in a homogeneous solution. Immunochemical labels which may be employed include free radicals, radioisotopes, fluorescent dyes, enzymes, bacteriophages, or coenzymes. More typically, expression levels of the markers described here will be determined at the nucleic acid level using any method known in the art. For example, Northern hybridization analysis using probes which specifically recognize one or more of these sequences can be used to determine gene expression. Alternatively, expression is measured using reverse-transcription-based PCR assays, e.g., using primers specific for the differentially expressed sequence of genes. RNA is isolated from cancer cells according to standard methodologies (Sambrook et al, 1989). The RNA may be converted to a complementary DNA. In one embodiment, the RNA is whole cell RNA; in another, it is poly- A RNA. Normally, the nucleic acid is amplified. The following provides a general discussion of nucleic acid assays for expression.

A. Primers and Probes

The term primer, as defined herein, is meant to encompass any nucleic acid that is capable of priming the synthesis of a nascent nucleic acid in a template-dependent process. Typically, primers are oligonucleotides from ten to twenty base pairs in length, but longer sequences can be employed. Primers may be provided in double-stranded or single-stranded form, although the single-stranded form is preferred. Probes are defined differently, although they may act as primers. Probes, while perhaps capable of priming, are designed to binding to the target DNA or RNA and need not be used in an amplification process. In particular

32 14 35 3 embodiments, the probes or primers are labeled with radioactive species ( P, C, S, H, or other label), with a fluorophore (rhodamine, fluorescein) or a chemiluminescent (luciferase).

B. Template Dependent Amplification Methods

A number of template dependent processes are available to amplify the marker sequences present in a given template sample. One of the best known amplification methods is the polymerase chain reaction (referred to as PCR) which is described in detail in U.S. Patent Nos. 4,683, 195, 4,683,202 and 4,800,159, , each of which is incorporated herein by reference in its entirety.

Briefly, in PCR, two primer sequences are prepared that are complementary to regions on opposite complementary strands of the marker sequence. An excess of deoxynucleoside triphosphates are added to a reaction mixture along with a DNA polymerase, e.g., Taq polymerase. If the marker sequence is present in a sample, the primers will bind to the marker and the polymerase will cause the primers to be extended along the marker sequence by adding on nucleotides. By raising and lowering the temperature of the reaction mixture, the extended primers will dissociate from the marker to form reaction products, excess primers will bind to the marker and to the reaction products and the process is repeated.

A reverse transcriptase PCR amplification procedure may be performed in order to quantify the amount of mRNA amplified. Methods of reverse transcribing RNA into cDNA are well known and described in Sambrook et al. (1989). Alternative methods for reverse transcription utilize thermostable, RNA-dependent DNA polymerases. These methods are described in WO 90/07641 filed Dec. 21, 1990. Polymerase chain reaction methodologies are well known in the art.

Another method for amplification is the ligase chain reaction ("LCR"), disclosed in EPO No. 320 308, incorporated herein by reference in its entirety. In LCR, two complementary probe pairs are prepared, and in the presence of the target sequence, each pair will bind to opposite complementary strands of the target such that they abut. In the presence of a ligase, the two probe pairs will link to form a single unit. By temperature cycling, as in PCR, bound ligated units dissociate from the target and then serve as "target sequences" for ligation of excess probe pairs. U.S. Patent No. 4,883,750 describes a method similar to LCR for binding probe pairs to a target sequence.

Qbeta Replicase, described in PCT Application No. PCT/US87/00880, may also be used as still another amplification method in the present invention. In this method, a replicative sequence of RNA that has a region complementary to that of a target is added to a sample in the presence of an RNA polymerase. The polymerase will copy the replicative sequence that can then be detected.

An isothermal amplification method, in which restriction endonucleases and ligases are used to achieve the amplification of target molecules that contain nucleotide 5'-[alpha- thio] -triphosphates in one strand of a restriction site may also be useful in the amplification of nucleic acids in the present invention, Walker et al. (1992).

Strand Displacement Amplification (SDA) is another method of carrying out isothermal amplification of nucleic acids which involves multiple rounds of strand displacement and synthesis, i.e., nick translation. A similar method, called Repair Chain Reaction (RCR), involves annealing several probes throughout a region targeted for amplification, followed by a repair reaction in which only two of the four bases are present. The other two bases can be added as biotinylated derivatives for easy detection. A similar approach is used in SDA. Target specific sequences can also be detected using a cyclic probe reaction (CPR). In CPR, a probe having 3' and 5' sequences of non-specific DNA and a middle sequence of specific RNA is hybridized to DNA that is present in a sample. Upon hybridization, the reaction is treated with RNase H, and the products of the probe identified as distinctive products that are released after digestion. The original template is annealed to another cycling probe and the reaction is repeated.

Still another amplification methods described in GB Application No. 2 202 328, and in PCT Application No. PCT/US89/01025, each of which is incorporated herein by reference in its entirety, may be used in accordance with the present invention. In the former application, "modified" primers are used in a PCR.TM.-like, template- and enzyme- dependent synthesis. The primers may be modified by labeling with a capture moiety (e.g., biotin) and/or a detector moiety (e.g., enzyme). In the latter application, an excess of labeled probes are added to a sample. In the presence of the target sequence, the probe binds and is cleaved catalytically. After cleavage, the target sequence is released intact to be bound by excess probe. Cleavage of the labeled probe signals the presence of the target sequence.

Other nucleic acid amplification procedures include transcription-based amplification systems (TAS), including nucleic acid sequence based amplification (NASBA) and 3SR (Kwoh et al, 1989; PCT Application WO 88/10315, incorporated herein by reference in their entirety). In NASBA, the nucleic acids can be prepared for amplification by standard phenol/chloroform extraction, heat denaturation of a clinical sample, treatment with lysis buffer and minispin columns for isolation of DNA and RNA or guanidinium chloride extraction of RNA. These amplification techniques involve annealing a primer which has target specific sequences. Following polymerization, DNA/RNA hybrids are digested with RNase H while double stranded DNA molecules are heat denatured again. In either case the single stranded DNA is made fully double-stranded by addition of second target specific primer, followed by polymerization. The double-stranded DNA molecules are then multiply transcribed by an RNA polymerase such as T7 or SP6. In an isothermal cyclic reaction, the RNA's are reverse transcribed into single-stranded DNA, which is then converted to double stranded DNA, and then transcribed once again with an RNA polymerase such as T7 or SP6. The resulting products, whether truncated or complete, indicate target specific sequences.

EP 0 329 822 (incorporated herein by reference in its entirety) disclose a nucleic acid amplification process involving cyclically synthesizing single-stranded RNA ("ssRNA"), ssDNA, and double-stranded DNA (dsDNA), which may be used in accordance with the present invention. The ssRNA is a template for a first primer oligonucleotide, which is elongated by reverse transcriptase (RNA-dependent DNA polymerase). The RNA is then removed from the resulting DNA:RNA duplex by the action of ribonuclease H(RNase H, an RNase specific for RNA in duplex with either DNA or RNA). The resultant ssDNA is a template for a second primer, which also includes the sequences of an RNA polymerase promoter (exemplified by T7 RNA polymerase) 5' to its homology to the template. This primer is then extended by DNA polymerase (exemplified by the large "Klenow" fragment of E. coli DNA polymerase I), resulting in a double-stranded DNA ("dsDNA") molecule, having a sequence identical to that of the original RNA between the primers and having additionally, at one end, a promoter sequence. This promoter sequence can be used by the appropriate RNA polymerase to make many RNA copies of the DNA. These copies can then re-enter the cycle leading to very swift amplification. With proper choice of enzymes, this amplification can be done isothermally without addition of enzymes at each cycle. Because of the cyclical nature of this process, the starting sequence can be chosen to be in the form of either DNA or RNA.

PCT Application WO 89/06700 (incorporated herein by reference in its entirety) disclose a nucleic acid sequence amplification scheme based on the hybridization of a promoter/primer sequence to a target single-stranded DNA ("ssDNA") followed by transcription of many RNA copies of the sequence. This scheme is not cyclic, i.e., new templates are not produced from the resultant RNA transcripts. Other amplification methods include "RACE" and "one-sided PCR".

Methods based on ligation of two (or more) oligonucleotides in the presence of nucleic acid having the sequence of the resulting "di-oligonucleotide," thereby amplifying the di-oligonucleotide, may also be used in the amplification step of the present invention. Wu et al. (1 89), incorporated herein by reference in its entirety.

C. Separation Methods

It normally is desirable, at one stage or another, to separate the amplification product from the template and the excess primer for the purpose of determining whether specific amplification has occurred. In one embodiment, amplification products are separated by agarose, agarose-acrylamide or polyacrylamide gel electrophoresis using standard methods. See Sambrook gi a/. (1989).

Alternatively, chromatographic techniques may be employed to effect separation. There are many kinds of chromatography which may be used in the present invention: adsorption, partition, ion-exchange and molecular sieve, and many specialized techniques for using them including column, paper, thin-layer and gas chromatography (Freifelder, 1982).

D. Chip/Array Technologies

Specifically contemplated by the present inventors are chip-based DNA technologies.

A DNA chip is a device that is convenient to compare expression levels of a number of genes at the same time. A DNA chip comprises immobilized high-density probes to detect a number of genes. Thus, expression levels of many genes can be estimated at the same time by a single-round analysis. Namely, the expression profile of a specimen can be determined with a DNA chip. Commercially available reagents and kits are available to perform each of the steps required for analysis.

In one embodiment, the DNA chip-based method of the present invention comprises the steps of:

(1) synthesizing cRNAs or cDNAs corresponding to the marker genes;

(2) hybridizing the cRNAs or cDNAs with probes for marker genes; and

(3) detecting the cRNA or cDNA hybridizing with the probes and quantifying the amount of mRNA thereof.

The cRNA refers to RNA transcribed from a template cDNA with RNA polymerase. A cRNA transcription kit for DNA chip-based expression profiling is commercially available. With such a kit, cRNA can be synthesized from T7 promoter-attached cDNA as a template by using T7 RNA polymerase. On the other hand, by PCR using random primer, cDNA can be amplified using as a template a cDNA synthesized from mRNA.

On the other hand, the DNA chip comprises probes, which have been spotted thereon, to detect the marker genes of the present invention. There is no limitation on the number of marker genes spotted on the DNA chip, and for the purposes of the present invention, it is only required that the chip contain nucleic acids for the 16 gene signature, and may optionally include control nucleic acids, as well as any other genes of interest. For a control, a probe for a gene whose expression level is rarely altered may be spotted on the DNA chip. Such a gene can be used to normalize assay results when assay results are intended to be compared between multiple chips or between different assays.

The probes are designed for each marker gene selected. Such a probe may be, for example, an oligonucleotide comprising 5-50 nucleotide residues. A method for synthesizing such oligonucleotides on a DNA chip is known to those skilled in the art. Longer DNAs can be synthesized by PCR or chemically. A method for spotting long DNA, which may be synthesized by PCR or the like, onto a glass slide is also known to those skilled in the art. The prepared DNA chip is contacted with cRNA, followed by the detection of hybridization between the probe and cRNA. The cRNA can be previously labeled with a fluorescent dye or any other label. A fluorescent dye such as Cy3(red) and Cy5 (blue) can be used to label a cRNA. cRNAs from a subject and a control are labeled with different fluorescent dyes, respectively. The difference in the expression level between the two can be estimated based on a difference in the signal intensity. The signal of fluorescent dye on the DNA chip can be detected by a scanner and analyzed by using a special program. For example, the Suite from Affymetrix is a software package for DNA chip analysis. E. Non-Enzymatic Target Assessment

U.S. Patent Publication 2013/0017971 (Nanostring Technologies, Seattle WA) describes methods for detecting the relative expressions of a plurality of target nucleic acid molecules in one assay. The methods use a plurality of probe molecules which specifically bind to one target nucleic acid molecule of a plurality of target nucleic acids in a sample, and a plurality of reference molecules that represent each of the plurality of target nucleic acid molecules, where the probe molecules specifically bind to the plurality of reference molecules, and each of the plurality of reference molecules is present in known amounts in the composition. The methods permit multiplexed detection of a plurality of target nucleic acid molecules from a biological sample including a plurality of probe molecules. Significantly, the probe molecules are capable of enzymatic or non-enzymatic direct detection of the target nucleic acid molecules, and preferably the detection of the target nucleic acid molecules occurs without target nucleic acid amplification. Furthermore, the methods can be adapted to quantifying the expression of the plurality of target nucleic acid molecules. See also Geiss et al. (2008).

The plurality of probe molecules can include about 8 to about 50 probe molecules, about 15 to about 50 probe molecules, about 25 to about 50 probe molecules, about 50 to about 100 probe molecules or more than 100 probe molecules. The probe molecules can be nucleic acid probes. Each nucleic acid probe can include: (i) a target-specific region that specifically binds to a target nucleic acid molecule; and (ii) a region including a plurality of label-attachment regions linked together, wherein each label attachment region is attached to a plurality of label monomers that create a unique code for each target-specific probe, the code having a detectable signal that distinguishes one nucleic acid probe which binds to a first target nucleic acid from another nucleic acid probe that binds to a different second target nucleic acid molecule. The plurality of label-attachment regions can include at least four, at least five, at least six, at least seven label attachment regions. The plurality of label monomers includes at least four, at least five, at least six, at least seven label monomers. The number of label monomers used can vary depending on the complexity of the plurality of target nucleic acid molecules. Each of the label monomers can be selected from the group consisting of a fluorochrome moiety, a fluorescent moiety, a dye moiety and a chemiluminescent moiety. The nucleic acid probe can further include an affinity tag.

The aforementioned application also provides kits including a composition for the multiplexed detection of a plurality of target nucleic acid molecules from a biological sample including a plurality of probe molecules, where each probe molecule in the plurality specifically binds to one target nucleic acid molecule in the sample, and instructions for the multiplexed detection of a plurality of target nucleic acid molecules. The composition included within the kit can further include a plurality of reference molecules that represent each of the plurality of target nucleic acid molecules, wherein the probe molecules specifically bind to the plurality of reference molecules, and wherein each of the plurality of reference molecules is present in known amounts. The probe molecules are capable of enzymatic or non-enzymatic direct detection of the target nucleic acid molecules. Preferably, the probe molecules are capable of non-enzymatic direct detection of the target nucleic acid molecules. The kit can further include an apparatus which includes a surface suitable for binding, and optionally detecting, the probe molecules included with the kit. Preferably, the probe molecules are hybridized to the target nucleic acids or the reference molecules when bound to the surface. The probe molecules may be bound to the surface by any means known in the art. The kit can further include a composition for the extraction of the target nucleic acids from a biological sample. The kit can further include a reagent selected from the group consisting of a hybridization reagent, a purification reagent, an immobilization reagent and an imaging reagent.

In a particular aspect, the present invention may utilized the NanoString nCounter® Analysis System to determine the expression levels of any or all of the genes described herein. The NanoString nCounter® Analysis System (also referred to as a nanoreporter code system) delivers direct, multiplexed measurements of gene expression through digital readouts of the relative abundance of hundreds of mRNA transcripts. The nCounter® Analysis System uses gene-specific probe pairs that hybridize directly to the mRNA sample in solution, eliminating any enzymatic reactions that might introduce bias in the results. After hybridization, all of the sample processing steps are automated on the nCounter® Prep Station. First, excess capture and Reporter Probes are removed, followed by binding of the probe-target complexes to random locations on the surface of the nCounter® cartridge via a streptavidin-biotin linkage. Finally, probe/target complexes are aligned and immobilized in the nCounter® sample cartridge. The Reporter Probe carries the fluorescent signal; the Capture Probe allows the complex to be immobilized for data collection. Up to 800 pairs of probes, each specific to a particular gene, can be combined with a series of internal controls to form a CodeSet. After sample processing has completed, sample cartridges are placed in the nCounter®. Digital Analyzer for data collection. Each target molecule of interest is identified by the "color code" generated by six ordered fluorescent spots present on the Reporter Probe. The Reporter Probes on the surface of the cartridge are then counted and tabulated for each target molecule.

The nCounter® Analysis System is comprised of two instruments, the nCounter® Prep Station used for post-hybridization processing, and the Digital Analyzer used for data collection and analysis. The assay also requires a heat block and microcentrifuge for R A extraction and a low- volume spectrophotometer for measuring the concentration and purity of the RNA output. A heat block with a heated lid is required to run the hybridization at a constant elevated temperature, and a swinging bucket centrifuge is required for spinning the Prep Plates prior to insertion into the Prep Station.

The nCounter® Prep Station is an automated fluid handling robot that processes samples post-hybridization to prepare them for data collection on the nCounter® Digital Analyzer. Prior to processing on the Prep Station, total RNA or alternatively other RNA molecules extracted from FFPE (Formalin-Fixed, Paraffin-Embedded) tissue samples, or other sample types, are hybridized with the Reporter Probes and Capture Probes according to the nCounter® protocol. Hybridization to the target RNA is driven by excess probes. To accurately analyze these hybridized molecules they are first purified from the remaining excess probes in the hybridization reaction. The Prep Station isolates the hybridized mRNA molecules from the excess Reporter and Capture Probes using two sequential magnetic bead purification steps. These affinity purifications utilize custom oligonucleotide-modified magnetic beads that retain only the tripartite complexes of mRNA molecules that are bound to both a Capture Probe and a Reporter Probe. Next, this solution of tripartite complexes is washed through a flow cell in the NanoString sample cartridge. One surface of this flow cell is coated with a polyethylene glycol (PEG) hydrogel that is densely impregnated with covalently bound streptavidin. As the solution passes through the flow cell, the tripartite complexes are bound to the streptavidin in the hydrogel through biotin molecules that are incorporated into each Capture Probe. The PEG hydrogel acts not only to provide a streptavidin-dense surface onto which the tripartite complexes can be specifically bound, but also inhibits the non-specific binding of any remaining excess Reporter Probes.

After the complexes are bound to the flow cell surface, an electric field is applied along the length of each sample cartridge flow cell to facilitate the optical identification and order of the fluorescent spots that make up each Reporter Probe. Because the Reporter Probes are charged nucleic acids, the applied voltage imparts a force on them that uniformly stretches and orients them along the electric field. While the voltage is applied, the Prep Station adds an immobilization reagent that locks the reporters in the elongated configuration after the field is removed. Once the reporters are immobilized the cartridge can be transferred to the nCounter® Digital Analyzer for data collection. All consumable components and reagents required for sample processing on the Prep Station are provided in the nCounter® Master Kit. These reagents are ready to load on the deck of the nCounter® Prep Station which can process a sample cartridge containing 12 flow cells per run in approximately 2 hours. The 12 flow cells can comprise a mixture of test samples and reference samples as required for the particular test.

The nCounter® Digital Analyzer collects data by taking images of the immobilized fluorescent reporters in the sample cartridge with a CCD camera through a microscope objective lens. Because the fluorescent Reporter Probes are small, single molecule barcodes with features smaller than the wavelength of visible light, the Digital Analyzer uses high magnification, diffraction-limited imaging to resolve the sequence of the spots in the fluorescent barcodes. The Digital Analyzer captures hundreds of consecutive fields of view (FOV) that can each contain hundreds or thousands of discrete Reporter Probes. Each FOV is a combination of four monochrome images captured at different wavelengths. The resulting overlay can be thought of as a four-color image in blue, green, yellow, and red. Each 4-color FOV is captured in just a few seconds and processed in real time to provide a "count" for each fluorescent barcode in the sample. Because each barcode specifically identifies a single mRNA molecule or other nucleic acid molecule tested, the resultant data from the Digital Analyzer is an accurate inventory of the abundance of each mRNA or nucleic acid of interest in a biological sample.

The resulting test sample data from the Digital Analyzer are normalized to the reference sample data to generate a test result. Other transformations may be included as part of the algorithm in order to generate a test result, but in the described method, at least one of the steps includes a normalization of the test sample data to the reference sample. F. Kit Components

All the essential materials and reagents required for detecting expression of target genes may be assembled together in a kit. This generally will comprise preselected primers and/or probes, primer and/or probe sets, or arrays such as those on a DNA chip. Also included may be enzymes suitable for amplifying nucleic acids including various polymerases (RT, Taq, Sequenase®, etc.), deoxynucleotides and buffers to provide the necessary reaction mixture for amplification. Kits may also contain substances serving as control reagents for detection and/or quantification steps. Such kits also generally will comprise, in suitable means, distinct containers for each individual reagent.

V. Examples

The following examples are included to demonstrate particular embodiments of the invention. It should be appreciated by those of skill in the art that the techniques disclosed in the examples which follow represent techniques discovered by the inventor to function well in the practice of the invention, and thus can be considered to constitute particular modes for its practice. However, those of skill in the art should, in light of the present disclosure, appreciate that many changes can be made in the specific embodiments which are disclosed and still obtain a like or similar result without departing from the spirit and scope of the invention.

Example 1- Materials and Methods

Gene expression. Gene expression datasets for resected lung adenocarcinomas, cell lines, and murine tumors were obtained from publicly available sources or derived as part of this work and made publicly available. Sources and statistical comparisons performed are given in Table 1.

LKBl classifier. Genes associated with LKB1 loss with p-value less than 0.01 in each of two training cohorts comprised a 129-gene LKBl-loss signature. A smaller 16-gene set having the strongest association with LKB l loss in the same training cohort and was used as the LKB 1 classifier throughout the study. Clinical and mutational data were taken from the various publicly available datasets and differences between tumors classified as LKBl-loss and LKBl -wild-type were compared using the Fisher's exact test.

Drug sensitivity. Publicly available datasets of in vitro drug sensitivity in cell lines were obtained from the GDSC 5 and CCLE 4 studies, which were used as training and testing sets. General linear models were used to determine drug sensitivity significantly associated with LKBl-loss. Multivariate linear regression was used to test the association between the LKB l classifier score and MEK sensitivity with incorporation of known mutations in the RAS/RAF pathway and previously published signatures of MEK sensitivity 5 ' 15,16 . Cell line experiments. Stable expression of LKB l and LKB l K78I was achieved by retroviral transduction with puromycin selection for at least two weeks. Protein ly sates and mRNA were collected from cells in logarithmic growth phase. Drug sensitivity was determined in 96-well plates with inhibitors added 24 hours after seeding cells and relative cells numbers quantitated after 72 hours of drug exposure using the Alamar Blue colorimetric assay. Activation of the CREB transcription factor was determined using a luciferase reporter driven by the CRE consensus sequence, the activity of which was measured as the ratio to an identical reporter with mutated CRE sequence.

Analysis of publicly available gene expression data. Publicly available datasets were downloaded from GEO and ArrayExpress or from individual websites, as listed in Table 1. Processed data uploaded to these sites by their original contributors were downloaded as 'series matrix.txt' files. In cases where data were presented as linear expression values, log2 transformed values were used. For analyses in which gene expression data from several studies were pooled, probeset expression values were standardized within each dataset by subtracting the mean value and dividing by the standard deviation. To collapse gene lists such that each gene was represented only once in the analyses, standardized scores from multiple probesets representing the same gene were averaged to give a single value.

Analysis of gene expression after expressing LKBl in H2122 and A549 cells. For the inventors' LKBl perturbation analyses, mRNA was isolated from three biological replicates of A549 and H2122 after stable expression of pBABE, LKBl or LKB l K78I using a Qiagen mRNA isolation kit, with trizol extraction reagent. RNA concentrations were measured, and the RNA integrity number and 28s: 18s ratio were calculated for quality control purposes. Amplification of 130 ng of total RNA was performed using Ambion WT Expression kit, and in vitro transcription was carried out overnight. cRNA was subsequently cleaned using Ambion-WT bead cleanup kit. 10.5 μg of cRNA was used for second cycle cDNA synthesis and resulting cDNA was cleaned using Ambion-WT bead cleanup kit. 5.5 μg of purified cDNA products were used in fragmentation and labeling reactions. Samples were hybridized overnight to a HT Human Gene 1.1 ST PM16 array plate utilizing a GeneTitan instrument. They were then scanned on the Affymetrix Gene Titan AGCC v. 3.2.3 and then analyzed on Affymetrix Expression Console v. 1.1 using a RMA normalization algorithm producing log base 2 results.

Analysis of gene expression associated with LKBl mutations. For clinical and cell line datasets in which LKBl status was known, a student's t-test was performed to determine statistical significance of differences in gene expression between LKB l these comparisons. For cell line data, LKB l mutation status was annotated using the Catalog of Somatic Mutations in Cancer (COSMIC) database, the Cancer Cell Line Encyclopedia (CCLE) resource, and individual publications. For data from the Directors Challenge Lung consortium, LKBl mutation status was unknown, and associations with LKBl expression were determined using p-values derived from linear regression modeling, fitting the expression of each probeset to the expression of each of the two probesets corresponding to LKBl : 204292_x_at and 41657_at. These analyses were performed using the R Bioconductor software platform, with the lm() function in the limma package. Finally, for gene expression data from LKBl -mutant cell lines stably transduced to over-express LKBl, average differences were determined for each probeset, comparing expression with wild-type LKBl to expression with vector only or LKBl K78I control.

Overlap significance analysis for univariate LKBl associations. Associations with LKB l were made for comparisons of sample groups listed in Table 1. Genes represented by probesets on the Affymetrix U133A chips were used for further analysis, with each gene represented by a single probeset, resulting in 13,211 genes. The top 200 genes over-expressed and under-expressed in LKBl -deficient samples were taken as 'up' and 'down' gene lists for each of these comparisons. Numerical overlap was determined between each pair of gene lists, and the statistical significance of this overlap was calculated using a hypergeometric distribution with the phyper() function in R bioconductor' s limma package. Overlap significance for all pairwise comparisons of 'up' lists was color coded and represented graphically using Cluster 3.0 and Java TreeView software, and is shown in FIG. 1A. Similarly, overlap significance for all pair-wise comparisons of 'down' lists is shown in FIG. 5.

Development of LKBl-deficient gene signature. The inventors used a training and testing approach to develop and test a gene signature capable of classifying LKBl -deficient tumors. They generated three gene lists using statistical comparisons from two training sets: the Wash U set with comparisons to documented LKB l mutations and the Michigan samples from the Director's Challenge Consortium with comparisons to LKB l expression. The LKBl classifier was taken as the intersection of these three lists:

ListA in Wash U:

All probesets ' ' such that raw p-value < 0.01 for student's t-test comparing LKBl mut (n=7) vs LKB l WT (n=34), resulting in 601 selected probesets. ListB in Mich:

All probesets ' ' such that raw p-value < 0.01 for linear regression model of 178 tumors, resulting in 3679 probesets:

expr (204292 _x_at) ~ a * expr(x) + b in Mich:

All probesets 'χ' such that raw p-value < 0.01 for linear regression model of 178 tumors, resulting in 3467 probesets:

expr (41657 _at) ~ a * expr(x) + b

Classifier = (ListA) (Ί (ListB) Π (ListC)

Lists B and C show a high degree of overlap, sharing more than half their genes, as they are derived from the same source and represent association with the two disticnt LKBl probesets. The intersection of the three lists results in a classifier of 167 probesets, a significantly larger intersection than expected by chance (p-value = 6.8e-38 by hypergeometric test). Standardized values were then averaged across probesets representing the same gene to give a final set of 129 standardized gene expression values. Using different p-value cutoffs gave similar classification results of unknown lung cancer samples (classification concordance greater than 90%).

Characterization of four transcriptional nodes comprising LKBl signature. Expression data for each gene was mean centered and normalized, and unsupervised hierarchical clustering was performed with Gene Cluster 3.0 utilizing uncentered Pearson's correlations and the centroid linkage method. Similar results were obtained for clustering of genes and tumors when using Spearman's rank correlation as the similarity metric (data not shown). Resulting heat maps were visualized using Java TreeView application. Unsupervised clustering revealed correlation patterns within the genes of the LKBl -deficient signature that were reproducible across multiple resected LUAD datasets. Four distinct sets of co-regulated genes were identified from resulting dendrograms as gene clusters with internal centroid correlation values greater than 0.5, including three transcriptional nodes showing increased expression among LKBl -deficient tumors and one with decreased expression. The same four nodes could be observed in classifiers independently derived from either the Michigan or non-Michigan patients of the Director's Challenge consortium, demonstrating the reproducibility of these clustering patterns across multiple datasets. A numeric score for each of these transcriptional nodes was calculated by taking the average of the standardized expression values for the genes comprising the node.

The inventors hypothesized that the expression of these transcriptional nodes were driven by different underlying phenotypes. For subsequent analyses to characterize the biological pathways reflected by these nodes they required larger gene lists corresponding to each node. Thus, the inventors further characterized gene expression correlations with these four gene clusters using a generalized linear model applied to gene expression data from the Director's Challenge Consortium dataset. The lm() function in the limma package of R bioconductor platform was used to determine the best fitting parameters to relate the expression of each probeset to the scores of the four LKBl-asociated gene clusters; interaction terms were not included in the model: expr(x) ~ a * expr(CREB) + b * expr(Mito)+

c * expr(NRF2)+ d * expr(Down)

Gene lists for each of the four nodes were then constructed taking the top 200 most significantly up-regulated genes for that node as determined by the corresponding p-value from this model.

Linear regression analysis to classify clinical samples. To provide a scheme of patient classification that was not dependent on hierarchical clustering analysis, a linear regression model was used to determine the association between L B 1 loss and each of the four transcriptional nodes observed. Using the expression data from the Michigan training cohort, the cluster scores identified in the previous section were used as four variables in a linear regression model to determine the best fit for LKB 1 mRNA expression, as measured by the 41657_at probeset: expr(41657_at) ~ a * expr(CREB) + b * expr(Mito)+

c * expr(NRF2)+ d * expr(Down) The CREB score was found to have the strongest association with LKB1 loss and inclusion of additional variables in the classification model did not substantially affect its accuracy, with concordance in sample classification greater than 90% and equivalent performance in detecting LKB 1 mutations (22 of 26 using CREB alone versus 23 of 26 for a combined model). Thus, the 16 gene CREB score was used to classify the LKB1 loss status of samples in the remainder of this study. A cutoff of 0.2 was used to delineate LKBl loss from LKBl wild-type, resulting in the classification of approximately 30-35% of lung adenocarcinomas as having loss of LKB l, similar to the fraction observed by hierarchical clustering.

To demonstrate that these results were not dependent on idiosyncracies of the training set, the inventors replicated their approach using an independent training cohort, with MSKCC2 for mutation associations and the non-Michigan samples from the Director's Challenge Consortium dataset for LKBl expression associations. The same data analysis methods described previously were applied to this second training cohort, resulting in an independent classifier score that was highly correlated to the primary classification score used throughout the text (FIG. 8) and yielding concordant classifications in 96% of tumors. Furthermore, the second classification score performed equally well at detecting LKBl mutations in resected LUAD and NSCLC cell lines, and showed the same molecular and clinical associations as those observed with the primary training cohort (Table 3).

Determination of clinical and molecular associations. LKBl classification score was used to predict LKB l -loss status for unknown samples from eight collections of resected lung adenocarcinomas (total n=851). The accuracy of predicting LKB l mutations was assessed in resected LUAD using the pooled MSKCC2, UNC, and USC datasets, while predictions of LKBl mutations in cell lines were assessed in the pooled Sanger and CCLE datasets (FIG. 1C; FIGS. 6A-F; Tables 2-3). Association with LKB l mRNA expression was assessed in tumors with unknown LKBl mutation status among the samples from the directors' challenge consortium that had not been used in the training cohort. Additionally, in tumors with known mutation status, three groups of tumors were considered: tumors with identified mutations in LKBl, tumors without observed mutations predicted to have loss of LKB l and tumors without mutations predicted to be LKBl wild-type. Expression of LKBl mRNA was compared between these groups using a student's t-test (FIG. ID).

LKBl mutation data was also available for the MSKCC1 dataset and these samples represented another potential test set. However, for unknown reasons univariate analysis comparing reported LKBl mutant and wild-type tumors in this dataset yielded fewer significant gene associations than would be expected by chance. In this dataset only five probesets out of 22000 passed a p-value cutoff of 0.001 ; in contrast for the Wash U and MSKCC2 cohorts, 118 probesets and 162 probesets passed this cutoff, respectively. Furthermore, the top ranked genes associated with LKB l mutations in this dataset showed no significant overlap with the consistent pattern of gene expression observed in each of the other clinical and cell line datasets (FIG. 1A; FIG. 5). Based on these findings, the inventors considered this dataset an outlier and excluded these data from the validation.

For clinical associations with survival, smoking, and gene mutations, data were pooled from all studies in which it was available. All samples were classified as LKB 1 wild- type or LKB1 -deficient and statistical associations were made using the Fisher's exact test for categorical data. Kaplan-Meier analysis of overall survival and progression free survival was performed between these two groups for patients in the Director's Challenge Consortium, using the R statistical platform with the survival package. Statistical significance was determined using the log rank test.

Analyses to determine biological significance of transcriptional nodes. The inventors used several approaches to identify candidate pathways that could potentially drive the expression of the four transcriptional nodes observed in this analysis. Gene set enrichment analysis was performed using the molecular signatures database (world-wide-web at broadinstitute.org/gsea/msigdb/) to determine enrichment of transcription factor consensus sequences in the promoter regions of these gene lists. This tool was also used to compare the eight gene lists to previously characterized perturbation and cancer-derived signatures. The connectivity map was used to determine significant similarities between the eight gene lists and gene perturbations induced in the cell lines MCF7, HL60, and PC3 by six hours of treatment with 1309 different small molecules. The inventors uploaded the gene lists onto the connectivity map online analysis tool to rank compounds that were significantly associated with the gene expression phenotypes they observed.

Finally, the inventors generated an association matrix using searches of GEO and ArrayExpress to obtain perturbations of interest to this study. Because the connectivity map did not employ a lung cancer derived cell line, they searched for all perturbations made to A549, a commonly studied lung adenocarcinoma cell line with a mutation in LKB1. The inventors next performed targeted queries for perturbations related to the hypotheses suggested by the GSEA and connectivity map analyses; specifically, they searched for perturbations involving pharmacologic or genetic modulations of the CREB pathway, the NRF2 transcription factor, mitochondria, and protein translation. Also, for the connectivity map associations highlighted in FIG. 2, the inventors downloaded Affymetrix .CEL files for the corresponding perturbations and controls and performed the analysis of gene expression changes. The inventors eliminated redundant probesets to reduce the association matrix to a single probeset per gene, and then determined the top 200 over-expressed and underexpressed genes associated with each perturbation (roughly the top and bottom 2% of changes). Numeric overlap was then determined with each of the eight cluster scores and statistical significance calculated using a hypergeometric distribution by the phyper() function in the Bioconductor limma package.

Analysis of drug sensitivity. Data for drug sensitivity across two large collections of cell lines were obtained from the GDSC and CCLE studies. Two large, multi-histology collections of cell line gene expression data were combined to calculate the LKB1 classifier score for each cell line. The cell lines from each collection were merged giving a pooled set of 1244 independent cell lines, and scores for each of the four transcriptional nodes were averaged for cell lines represented in both datasets.

To identify inhibitors that may show differential sensitivity in tumors lacking LKB 1, the inventors performed univariate linear regression analysis to determine the association between the LKB1 classifier score and the IC50 values for 131 different compounds included in the GDSC study. To allow for training and testing analyses the CCLE study was split into two cell line groups. The set of cell lines that overlapped those included in the GDSC study was used as a training set confirmation, while samples that were not included among GDSC cell lines were used as an independent validation set. Linear regression was used to determine associations between cluster scores and the IC50 values and maximum inhibitory effects seen for each of the 24 inhibitors included in the CCLE study. Distributions were also compared for groups of cell lines given a binary classification as high or low LKB 1 loss score and student's t-tests were used to compare drug sensitivity between the two groups.

To demonstrate that the association between the CREB signature and MEK sensitivity was a novel observation not accounted for by previous findings, the inventors used a multivariable general linear model relating maximum inhibition to selumetinib to the LKB 1- loss score and each of three previously published MEK sensitivity signatures (Barretina et al, 2012; Dry et al, 2010; Loboda et al, 2012), as well as additional variables representing mutations in KRAS, NRAS, HRAS, BRAF, and LKB l. The published gene signatures were used to calculate sensitivity scores for each cell line by averaging standardized expression for each of the published probesets. The correlations between the genes comprising these predictive gene signatures were examined visually in heat maps to ensure they were strongly correlated to one another, such that each signature could be justifiably represented as a single numeric value. Mutations were determined based on data from COSMIC and the CCLE. Linear regression modeling was performed using the R statistical platform with the Limma package. CRE-luciferase reporter. The inventors designed a dual-luciferase reporter driven by a 3x CRE consensus binding sequence in the promoter region in addition to a TATA box, which was inserted into an FG12 lentiviral construct. Luciferase activity from this reporter was compared to a control reporter that was identical but with mutated CRE sites. Cells were stably transduced to express CRE wild-type or mutant reporters and ratios between the two were compared after subsequent perturbations.

Cell culture and gene transduction. A549, H2122, and H460 cell lines were generously shared with us by John Minna and Luc Girard (University of Texas, Southwestern). They were tested to ensure that they were mycoplasma negative, and were cultured in RPMI1640 containing 5% FBS, without antibiotics. Empty pBABE viral plasmids, pBABE-LKBl and pBABE-LKBl-K78I were obtained from AddGene. Phoenix cells were transfected with viral plasmids and retroviral particles were harvested from media supernatant 48 hours after transfection. Viruses were added to target cells with polybrene, and selection with 1 μg/ml puromycin was begun 48-72 hours after infection. Cells were then selected under puromycin for one to two weeks before performing subsequent experiments, with experiments being completed within two months.

Proliferation and drug sensitivity assays. In vitro proliferation assays were performed in 96-well plates after seeding 1000 cells in each well. Quantitation of relative cell growth was made using the Alamar Blue colorimetric assay. Similarly, for drug sensitivity assays, 1000 cells per well were added to 96-well plates. Inhibitors were added at the specified concentrations 24 hours after seeding, and relative cell viability was quantified 72 hours after adding inhibitors using Alamar Blue. Selumetinib was purchased from Chemitek.

Immunoblots. Cell lysates were harvested while cells were in exponential growth phase in lysis buffer containing phosphatase and protease inhibitors. Phospho-ACC and LKB1 antibodies were obtained from Cell Signaling Technology.

Example 2- Results and Discussion

To address this issue and to improve understanding of the biology of L B1 -deficient lung tumors, the inventors performed a comprehensive analysis of all publicly available gene expression datasets in which L B1 mutation status had been determined, including resected human lung tumors, lung and non-lung cell lines, and murine tumors with LKB1 loss. Data sources are shown in Table 1. Comparison of gene expression associated with LKB1 loss revealed strong similarity among the human datasets (median p-value = 1.0e-19 for 36 pair- wise comparisons), but insignificant overlap between human and mouse profiles (FIG. 1A, FIG. 5).

A 129-gene signature of LKB 1 loss was derived using two studies as a training cohort. Unsupervised clustering of these genes identified a subset of roughly 30% of lung adenocarcinomas that express an LKB 1 -deficient signature (FIG. IB). A numeric LKB 1- classifier score derived from 16 of these genes identifies tumors with LKB1 mutations with high sensitivity and specificity in independent clinical testing datasets as well as in non-small cell lung cancer (NSCLC) cell lines and in non-lung cell lines of various tissue origins (FIG. 1C; FIGS. 6A-F). Moreover, in the clinical testing sets the expression of LKB 1 mRNA is significantly lower among tumors predicted to have L B 1 loss, including the tumors that were sequenced as LKB 1 wild-type (FIGS. 1D-E). These tumors likely represent unrecognized cases of LKB 1 loss that could occur by undetected mutation, intragenic deletion, chromosomal loss, or by an epigenetic mechanism. This finding suggests that the specificity of this classifier exceeds the observed 76%, and that the LKBl-loss classifier is more sensitive than DNA sequencing for the detection of functional L B 1 loss.

Tumors predicted to have LKB 1 loss also had associations with several clinical and molecular covariates previously associated with LKB1 mutations (Table 2). The inventors observed a strong association with tobacco smoking history (P = 6.5 "9 ), significant exclusion of EGFR mutations (P = 1.9 "10 ) and a moderately increased prevalence of KRAS mutations (P = 0.00035). No significant difference in survival was observed between groups expressing LKB 1 -wild-type signature and those expressing the LKB 1 deficient signature (FIGS. 7A-B). These findings are consistent with previously published associations between LKB1 loss and clinical or mutational covariates 6 ' 11 . To ensure that the results were not influenced by the choice of training set, the inventors also confirmed the findings using a second classifier derived from an independent training cohort (FIG. 8; Table 3). The inventors also found the signature to be associated with LKB 1 loss in other cancer types including breast adenocarcinoma and cervical squamous carcinoma (FIGS. 14A-C). Furthermore, other genes and proteins differentially expressed by tumors with known LKBl mutations are concordantly dysregulated among the tumors with predicted loss (FIGS. 15A-B).

In addition to recapitulating previously established associations with LKBl loss in human lung cancer, the inventors' interpretation of the differentially expressed genes allowed us to identify novel biological features of these tumors. Four sets of transcriptionally correlated genes were observed within the 129-gene signature of LKBl loss. A multivariable general linear model was used to derive the two hundred genes most strongly associated with each of the four transcriptional clusters, and these were compared with published data from

12

drug-induced gene expression perturbations , as well as studies of transcription factor binding sites 13 ' 14 and gene sets derived from a variety of individual publications to elucidate the molecular pathways that give rise to these LKBl -associated gene sets. Interpretations were obtained for each of the three up-regulated clusters that were supported by consistent findings in multiple cell lines and among multiple studies (FIG. 2).

Two clusters of genes were closely related to tumor metabolism. One, referred to as the 'Mitochondrial' cluster, had high expression of oxidative phosphorylation and mitochondria-associated genes as well as genes involved in protein translation; a second, referred to as the 'NRF2' cluster, expresses oxidative stress response genes driven by the NRF2 transcription factor; this phenotype was expressed by approximately half of tumors with LKBl loss. The final up-regulated cluster was found to be enriched in genes with CREB consensus sequences in their promoter regions, and was also found to be up-regulated by the cAMP inducer forskolin; hence it was referred to as the 'CREB' cluster. TGF-beta and stroma-related genes comprised a component of the down-regulated genes, but multiple phenotypes likely contributed to this transcriptional node (FIGS. 9A-B). Interestingly, the inventors observed that in the A549 cell line TGF-beta induced a significant subset of these down-regulated genes, while simultaneously attenuating both the NRF2 and CREB transcriptional components; conversely, c-SRC inhibition induced CREB activation, suggesting antagonism between these two pathways and CREB (FIG. 10).

To test the direct effects of LKBl on the regulation of the observed gene expression pattern, the inventors stably expressed LKBl or mutated LKB l K78I in NSCLC cell lines (H2122, A549 and H460) lacking functional tumor suppressor. Wild-type, but not mutant, LKB 1 resulted in phopshorylation of a well recognized downstream target of AMPK and caused significant decrease in cell growth (GIG. 3A; FIGS. 1 1A-B). Microarray analysis of gene expression changes in A549 and H2122 demonstrated restoring LKB l significantly (P < 1.0e-30 by hypergeometric test) down-regulates the CREB-driven gene cluster and increases expression of a subset of the down-regulated genes, while mitochondrial and NRF2 associated clusters are unaffected (FIGS. 3B-C; FIGS. 12A-C). The attenuation of CREB activation was confirmed in three cell lines using a luciferase reporter driven by the CRE- consensus sequence, which showed a reduction in reporter activity of 30-40% (FIG. 3D; p- value less than 0.05 for each cell line).

To identify molecular pathways that represent effective candidates for targeted therapy among patients with LKB l -deficient tumors, the inventors investigated drug sensitivity associations using data from the Genomics of Drug Sensitivity in Cancer (GDSC) study, which measured in vitro susceptibility of several hundred cell lines to 131 diverse pharmacologic inhibitors 5 . Sensitivity to mTOR inhibitors, metformin and AICAR were not associated with expression of the LKBl -loss signature, which was unexpected, given the prominent role that LKBl and AMPK play in regulating mTOR activity and response to metabolic stress. However, cells with high expression of the LKBl -loss signature were significantly more sensitive to three of the four different MEK inhibitors included in the study - PD-0325901, selumetinib (AZD6244), Cl-1040, and RDEA119 (FIGS. 13A-B). This novel association was confirmed using an independent testing set of cell lines from the Cancer Cell Line Encyclopedia (CCLE), a second large-scale analysis of in vitro drug susceptibility that included data on both selumetinib and PD-0325901 4 (FIGS. 4A-B; FIGS. 13A-B).

To determine whether LKBl loss was an independent determinant of MEK sensitivity, the inventors used a multivariable general linear model to account for previously reported associations with MEK sensitivity: mutations in KRAS, NRAS, HRAS, and BRAF, as well as previously reported gene signatures from three studies 4 ' 5 ' 15 ' 16 . This analysis demonstrated that the signature of LKBl loss was still significantly associated with sensitivity after controlling for other factors and thus represents a novel independent predictor of response to this class of drugs (FIGS. 4C-F). The inventors next wanted to determine whether this association represented a causal link between LKBl mutations and MEK sensitivity. They tested the effect of re-expressing LKBl on sensitivity to selumetinib in A549, H2122, and H460 cell lines and found that the relative inhibition by selumetinib was significantly reduced in each of the three lines after LKBl was expressed (FIG. 4G). The overall expression of the 16-gene signature was not affected by MEK inhibition, the only significant interaction observed being the downregulation of DUSP4, a known component of the MAPK pathway (FIG. 16).

This study gives several novel insights into the biology and potential treatment of tumors with loss of LKBl. CREB activation is known to be regulated downstream of LKBl 17"19 , but is further shown here to be a cardinal feature of LKBl loss in human tumors, occurring in 85% of all tumors and cell lines with known LKB l mutations as well as in tumors with non-mutational inactivation of this pathway. Furthermore, the inventors report the novel observation that NRF2 activation occurs preferentially among lung tumors lacking LKBl. NRF2 is a key activator of the oxidative stress response and also plays a role in metabolic reprogramming of cancer cells 20 ' 21 . LKB l -deficient tumors have been shown to be susceptible to oxidative stress, as they are unable to make the appropriate adaptive responses

22

in metabolism and biosynthesis . It has been shown that NRF2 is activated by somatic mutations in NRF2 or KEAP1 in NSCLC 23-27 , and the high frequency of NRF2 activation among LKBl -deficient tumors suggests that selective pressure exists for these mutations as a secondary protective mechanism.

The mouse model of LKBl loss is an important platform for studying the behavior of these tumors in vivo. However, the analysis shows that LKBl loss produces distinctly different phenotypes in humans. Thus, there are key features of the biology of LKB1- deficient patient tumors that may not be recapitulated in the murine model. In contrast to human tumors with LKBl loss, murine tumors did not express genes characteristic of CREB or NRF2 activation, while TGF-beta and related signaling pathways show increased expression in the murine model. By manipulating one or more of these pathways in the murine model it may be possible to engineer tumors that more closely reflect the phenotype of LKBl loss in human lung cancer.

The differences in gene expression are also reflected in clinically relevant phenotypes produced by these models. In contrast to the MEK resistance observed in the murine model, the inventors find that both NSCLC and non-lung cell lines expressing the LKBl -loss signature show higher susceptibility to MEK inhibitors. MEK inhibition has recently shown efficacy in combination with RAF inhibitors in BRAF mutant melanoma 28 and promising results have been presented from a phase II clinical trial in KRAS mutant advanced stage NSCLC 29 . Analysis of LKBl loss in clinical trials of MEK inhibition will be informative as to whether this phenotype is predictive of patient outcome and whether it could be used prospectively to guide treatment decisions.

TABLE 1 - Data Sources and Statistical Comparisons

Name Tissue Source Comparison

FIG. 1A

MSKCC q lung adeno Chitale et al, 2009 t-test: LKB1 mut (16) vs LKB1 WT (75)

MSKCC2 lung adeno Chitale et al, 2009 t-test: LKB1 mut (n=12) vs LKB 1 WT (n=90)

Wash U lung adeno GSE12667 t-test: LKB1 mut (n=7) vs LKB1 WT (n=34)

Michigan lung adeno Shedden et al, 2008 Linear regression with LKB1 probeset 41657_at (n=178)

UNC lung adeno GSE26939 t-test: LKB1 mut (n=6) vs LKB1 WT (n=75)

use lung adeno GSE32861 t-test: LKB1 mut (n=8) vs LKB1 WT (n=48)

CCLE NSCLC cell lines broadinstitute.org/ccle/home t-test: LKB1 mut (n=23) vs LKB 1 WT (n=45)

Sanger NSCLC cell lines broadinstitute.org/cgi-bin/cancer/datasets 1 t-test: LKB1 mut (n=15) vs LKB 1 WT (n=30)

A549 NSCLC cell line GSEXXXXX avg diff: LKB 1 WT (n=3) vs pBABE vector (n=3)

H2122 NSCLC cell line GSEXXXXX avg diff: LKB1 WT (n=3) vs pBABE vector (n=3)

Ji (A) Mouse lung adeno GSE6135 t-test: LKB1/KRAS primary adeno (n=5) vs KRAS primary adeno (n=5)

Ji (B) Mouse lung adeno GSE6135 t-test: LKB1/KRAS primary adeno (n=5) vs KRAS/p53 primary adeno (n

Carretero Mouse lung adeno GSE21581 t-test: LKB1/KRAS primary adeno (n=9) vs KRAS primary adeno (n=9)

Carretero Mets Mouse lung adeno GSE21581 t-test: LKB1/KRAS metastases (n=17) vs LKB 1/KRAS primary (n=9)

FIG. 2

LY-294002 H L60 broadinstitute. org/cmap/# avg diff: HL60; 10uM LY-294002 (n=9) vs DMSO

LY-294002 MCF7 broadinstitute. org/cmap/# avg diff: MCF7; 10uM LY-294002 (n=18) vs DMSO

LY-294002 PC3 broadinstitute. org/cmap/# avg diff: PC3; 10uM LY-294002 (n=6) vs DMSO

Sirolimus HL60 broadinstitute. org/cmap/# avg diff: HL60; 100nM sirolimus (n=9) vs DMSO

Sirolimus MCF7 broadinstitute. org/cmap/# avg diff: MCF7; 100nM sirolimus (n=19) vs DMSO

Sirolimus PC3 broadinstitute. org/cmap/# avg diff: PC3; 100nM sirolimus (n=6) vs DMSO

PD0325901 multiple GSE10087 paired t-test: 12 cell lines treated 8hr with 50nM PD-0325901 (n=1 rep each) vs

DMSO

Mitochondria multiple broadinstitute.org/gsea/msigdb Mootha_Mitochondria gene set (n=441 genes) enrichment

PGC1A C2C12 broadinstitute.org/gsea/msigdb Mootha_PGC gene set (n=412 genes)

Colforsin MCF7 broadinstitute. org/cmap/# avg diff: MCF7; 0.5uM (n=1 ) or 50uM (n=1 ) vs DMSO

Colforsin PC3 broadinstitute. org/cmap/# avg diff: PC3; 0.5uM (n=2) vs DMSO

Forskolin H295R GSE5553 avg diff: H295R; 10uM forskolin (n=2) vs DMSO

Forskolin PC12 GSE2071 avg diff: PC12; 10uM forskolin (n=4) vs DMSO

Forskolin MI N6 GSE2060 avg diff: MI N6; 10uM forskolin (n=2) vs DMSO

CRTC1 /MAML2 H292 broadinstitute.org/cgi-bin/cancer/datasets 1 avg diff/standard deviation: H292 cell line vs NSCLC cell lines

DN CREB MIN6 GSE2060 avg diff: MI N6; DN CREB (n=2) vs GFP

15dPGJ2 MCF7 broadinstitute.org/cmap/# avg diff: MCF7; 10uM 15-delta prostaglandin J2 (n=5) vs DMSO

.cgi

{00157807} 41

15dPGJ2 HL60 broadinstitute.Org/cmap/# avg diff: H L60; 10uM 15-delta prostaglandin J2 (n=3) vs DMSO

15dPGJ2 PC3 www.broadinstitute.org/cmap/# avg diff: PC3; 10uM 15-delta prostaglandin J2 (n=2) vs DMSO

keapW- mouse liver GSE1 1287 avg diff: KEAP1-/- liver (n=3) vs control

KEAP1 mut LUSQ tcga-data.nci.nih.gov/tcga/ t-test: KEAP1 mut (n=22) vs KEAP1 /NRF2 WT (n=171 )

NRF2 mut LUSQ tcga-data.nci.nih.gov/tcga/ t-test: N RF2 mut (n=24) vs KEAP1 /NRF2 WT (n=171 )

siNRF2 A549 GSE28230 avg diff: A549; si RF2 (n=3) vs control siR A

FIG. 10

TGF-beta A549 GSE17708 avg diff: A549; 5ng/ml TGF-beta at various times (n=3 reps each) vs control

Dasatinib A549 E-TAMB-585 avg diff: variable concentrations Dasatinib (n=1 rep each) vs DMSO

{00157807} 42

TABLE 2 - Association of Primary LKBl Classifier with Clinicopathologic

Characteristics

LKB1 -loss classifier Fisher test

Number of samples Fraction LKB1 loss Odds Ratio

Resected LUAD

LKB1 mutant 22/26 84.6 16.8 (5.4, 70.0) 2.8e-09 LKB1 wild-type 52/213 24.4

KRAS mutant 50/1 11 45 2.3 (1 .4, 3.7) 0.00035 KRAS wild-type 85/322 26.3

EGFR mutant 4/76 5.3 0.085 (0.02, 0.24)

EGFR wild-type 116/293 39.6

never smoker 12/1 16 1 1.3 0.20 (0.01 , 0.38) 6.5e-09 ever smoker 200/553 36.2

NRF2 active 118/198 59.6 5.1 (3.6, 7.3) <1 e-16 NRF2 low 146/653 22.4

NSCLC Cell Lines

LKB1 mut 31 /33 93.9 97.3 (19.3, 977) 1.7e-15 LKB1 wt 8/63 12.7

Non-lung Cell Lines

LKB1 mut 26/35 74.3 13.3 (5.8, 33.2) 4.0e-12 LKB1 wt 98/551 17.8

TABLE 3 - Association of Alternative LKB1 Classifier with Clinicopathologic

Characteristics

LKB1 -IOSS classifier Fisher test

Number of samples Fraction LKB1 loss Odds Ratio

(LKB1 loss / total) (%) (95% CI.)

Resected LUAD

LKB1 mutant 18/21 85.7 23.8 (6.3, 134) 4.0e-09 LKB1 wild-type 31 /157 19.7

KRAS mutant 46/1 1 1 41.4 2.1 (1 .3, 3.4) 0.0016 KRAS wild-type 81/322 25.2

EGFR mutant 4/76 5.3 0.096 (0.025, 0.27) 7.0e-09 EGFR wild-type 108/293 36.9

never smoker 6/1 16 5.2 0.097 (0.034, 0.22) 4.7e-13 ever smoker 199/553 36.0

NRF2 active 1 19/198 60.1 5.8 (4.0, 8.3) <1 e-16 NRF2 low 135/653 20.7

NSCLC Cell Lines

LKB1 mut 32/33 97.0 175 (23.9, 7506) 3.4e-16 LKB1 wt 9/63 14.3

Non-lung Cell Lines

LKB1 mut 20/35 57.1 14.2 (6.4, 31 .9) 8.0e-12

LKB1 wt 47 / 551 8.5

* * * * * * * * * * * * * * * * * * * * *

All of the compositions and/or methods disclosed and claimed herein can be made and executed without undue experimentation in light of the present disclosure. While the compositions and methods of this invention have been described in terms of preferred embodiments, it will be apparent to those of skill in the art that variations may be applied to the compositions and/or methods and in the steps or in the sequence of steps of the method described herein without departing from the concept, spirit and scope of the invention. More specifically, it will be apparent that certain agents which are both chemically and physiologically related may be substituted for the agents described herein while the same or similar results would be achieved. All such similar substitutes and modifications apparent to those skilled in the art are deemed to be within the spirit, scope and concept of the invention as defined by the appended claims. VI. References

The following references, to the extent that they provide exemplary procedural or other details supplementary to those set forth herein, are specifically incorporated herein by reference:

1. Ji, H. et al. LKBl modulates lung cancer differentiation and metastasis. Nature 448,

807-810 (2007).

2. Carretero, J. et al. Integrative genomic and proteomic analyses identify targets for Lkbl -deficient metastatic lung tumors. Cancer Cell 17, 547-559 (2010).

3. Chen, Z. et al. A murine lung cancer co-clinical trial identifies genetic modifiers of therapeutic response. Nature 483, 613-617 (2012).

4. Barretina, J. et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603-307 (2012).

5. Garnett, M. J. et al. Systematic identification of genomic markers of drug sensitivity in cancer cells. Nature 483, 570-575 (2012).

6. Ding, L. et al Somatic mutations affect key pathways in lung adenocarcinoma. Nature

455, 1069-1075 (2008).

7. Imielinski, M. et al. Mapping the hallmarks of lung adenocarcinoma with massively parallel sequencing. Cell 150, 1 107-1120 (2012).

8. Chitale, D. et al. An integrated genomic analysis of lung cancer reveals loss of DUSP4 in EGFR-mutant tumors. Oncogene 28, 2773-2783 (2009).

9. Shackelford, D. B. & Shaw, R. J. The LKB1-AMPK pathway: metabolism and growth control in tumour suppression. Nature Rev. Cancer 9, 563-575 (2009).

10. Alessi, D. R., Sakamoto, K. & Bayascas, J. R. LKBl -dependent signaling pathways.

Annu. Rev. Biochem. 75, 137-163 (2006).

11. Koivunen, J. P. et al. Mutations in the LKBl tumour suppressor are frequently detected in tumours from Caucasian but not Asian lung cancer patients. Br. J. Cancer 99, 245-252 (2008). Lamb, J. et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 313, 1929-1935 (2006). Malhotra, D. et al. Global mapping of binding sites for Nrf2 identifies novel targets in cell survival response through ChlP-Seq profiling and network analysis. Nucleic Acids Res. 38, 5718-5734 (2010). Zhang, X. et al. Genome-wide analysis of cAMP-response element binding protein occupancy, phosphorylation, and target gene activation in human tissues. Proc. Natl. Acad. Sci. U.S.A. 102, 4459^1464 (2005). Loboda, A. et al. A gene expression signature of RAS pathway dependence predicts response to PI3K and RAS pathway inhibitors and expands the population of RAS pathway activated tumors. BMC Med. Genomics 3, 26 (2010). Dry, J. R. et al. Transcriptional pathway signatures predict MEK addiction and response to selumetinib (AZD6244). Cancer Res JO, 2264-2273 (2010). Feng, Y. et al. The CRTC1-NEDD9 signaling axis mediates lung cancer progression caused by LKB l loss. Cancer Res. 72, 6502-6511 (2012). Gu, Y. et al. Altered LKBl/CREB-regulated transcription co-activator (CRTC) signaling axis promotes esophageal cancer cell migration and invasion. Oncogene 31, 469^79 (2012). Komiya, T. et al. Enhanced activity of the CREB co-activator Crtcl in LKBl null lung cancer. Oncogene 29, 1672-80 (2010). Mitsuishi, Y. et al Nrf2 redirects glucose and glutamine into anabolic pathways in metabolic reprogramming. Cancer Cell 22, 66-79 (2012). DeNicola, G. M. et al. Oncogene-induced Nrf2 transcription promotes ROS detoxification and tumorigenesis. Nature 475, 106-109 (2011). Jeon, S.-M., Chandel, N. S. & Hay, N. AMPK regulates NADPH homeostasis to promote tumour cell survival during energy stress. Nature 485, 661-665 (2012). Shibata, T. et al. Cancer related mutations in NRF2 impair its recognition by Keapl- Cul3 E3 ligase and promote malignancy. Proc. Natl. Acad. Sci. U.S.A. 105, 13568- 13573 (2008).

24. Taguchi, K., Motohashi, H. & Yamamoto, M. Molecular mechanisms of the Keapl- Nrf2 pathway in stress response and cancer evolution. Genes Cells 16, 123-140 (201 1).

25. Singh, A. et al. Dysfunctional KEAP1-NRF2 interaction in non-small-cell lung cancer.

PLoS Med. 3, e420 (2006).

26. Solis, L. M. et al. Nrf2 and Keapl abnormalities in non-small cell lung Ccrcinoma and association with clinicopathologic features. Clin. Cancer Res. 16, 3743-3753 (2010).

27. Cancer Genome Atlas Research Network. Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519-525 (2012).

28. Flaherty, K. T. et al. Combined BRAF and MEK inhibition in melanoma with BRAF V600 mutations. N. Engl. J. Med. 367, 1694-1703 (2012).

29. Janne, P. A., Shaw, A. T., Pereira, J. R. & Jeannin, G. Selumetinib plus docetaxel for KRAS-mutant advanced non-small-cell lung cancer: a randomised, multicentre, placebo-controlled, phase 2 study. Lancet Oncol. 14, 38^7 (2013).

Wu et al, Genomics, 4:560, 1989.

Walker et al., Proc. Nat'l Acad. Sci. USA, 89:392-396 1992.

Kwoh et al., Proc. Nat'l Acad. Sci. USA, 86: 1173, 1989.

Sambrook et al., In: Molecular Cloning: A Laboratory Manual, 1989.

Geiss et al, Nature Biotechnol. 26:317-325, 2008.