A Meta-Analysis of Lung Cancer Gene Expression Identifies PTK7 As a Survival Gene in Lung Adenocarcinoma

Published OnlineFirst March 20, 2014; DOI: 10.1158/0008-5472.CAN-13-2775

Cancer Tumor and Stem Cell Biology Research

A Meta-analysis of Lung Cancer Gene Expression Identiﬁes PTK7 as a Survival Gene in Lung Adenocarcinoma

Ron Chen1, Purvesh Khatri2,3, Pawel K. Mazur1, Melanie Polin1, Yanyan Zheng1, Dedeepya Vaka1, Chuong D. Hoang4, Joseph Shrager4, Yue Xu4, Silvestre Vicent1, Atul J. Butte5, and E. Alejandro Sweet-Cordero1

Abstract Lung cancer remains the most common cause of cancer-related death worldwide and it continues to lack effective treatment. The increasingly large and diverse public databases of lung cancer gene expression constitute a rich source of candidate oncogenic drivers and therapeutic targets. To define novel targets for lung adenocarcinoma, we conducted a large-scale meta-analysis of genes specifically overexpressed in adenocarcinoma. We identified an 11-gene signature that was overexpressed consistently in adenocarcinoma specimens relative to normal lung tissue. Six genes in this signature were specifically overexpressed in adenocarcinoma relative to other subtypes of non–small cell lung cancer (NSCLC). Among these genes was the little studied protein tyrosine kinase PTK7. Immunohistochemical analysis confirmed that PTK7 is highly expressed in primary adenocarcinoma patient samples. RNA interference–mediated attenuation of PTK7 decreased cell viability and increased apoptosis in a subset of adenocarcinoma cell lines. Further, loss of PTK7 activated the MKK7–JNK stress response pathway and impaired tumor growth in xenotransplantation assays. Our work defines PTK7 as a highly and specifically expressed gene in adenocarcinoma and a potential therapeutic target in this subset of NSCLC. Cancer Res; 74(10); 2892–902. 2014 AACR.

Introduction Gene expression analysis has been used to classify cancers, Lung cancer is the leading cause of cancer death in the United predict clinical outcomes, and discover disease-associated bio- States and worldwide (1). Despite intensive basic and clinical markers. However, gene expression experiments are usually research, the overall 5-year survival rate of the major histologic analyzed in isolation and are limited to a small number of subtype, non–small cell lung cancer (NSCLC) has only improved samples. Meta-analysis approaches make it possible to combine from 14% to 18% since 1975 (1). Recently, targeted treatment multiple gene expression datasets and increase the statistical based on patient-specific molecular aberrations has led to power for gene discovery. Such meta-analysis approaches have significant response rates in subsets of patients with NSCLC been successfully used for cancers of the breast (5, 6), prostate – (2, 3). However, about half of all patients do not harbor known (7), liver (8), and lung (9), as well as broadly across cancers (10 "driver" mutations and cannot be treated with targeted agents 12). Statistically, individual studies of gene expression in cancer (4). Thus, new approaches for identification of novel regulators are limited by both biologic (e.g., sampling of a particular patient and potential targets for treatment of lung cancer are needed. population) and technical (e.g., only using one expression analysis platform) biases that hinder the broader application of their findings and ultimate translation into clinical practice. Authors' Affiliations: 1Cancer Biology Program, Division of Hematology/ Oncology, Department of Pediatrics; 2Center for Biomedical Informatics Meta-analysis can control for such confounding factors by Research, Department of Medicine; 3Institute for Immunity, Transplant and increasing the statistical power to detect consistent changes Infection; 4Department of Cardiothoracic Surgery; and 5Division of Sys- across multiple datasets. Several groups have made available tems Medicine, Department of Pediatrics, Stanford University School of Medicine, Stanford, California large NSCLC gene expression datasets that consist of tumor-to- normal comparisons (9, 13–21). These datasets represent a large Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). and as yet not fully tapped resource for discovering novel genes relevant to the pathogenesis of lung cancer. We reasoned that a R. Chen and P. Khatri contributed equally to this work. careful meta-analysis that combined multiple patient popula- Current address for S. Vicent: Division of Oncology, Center for Applied tions across many institutions, platforms, and data procure- Medical Research (CIMA), Pamplona, Spain. ment methods would uncover genes with functional relevance Corresponding Authors: E. Alejandro Sweet-Cordero, Stanford Univer- to lung cancer that may have been otherwise overlooked by the sity, 265 Campus Drive, LLSCRB 2078B, Stanford, CA 94305. Phone: 650- isolated analysis of individual gene expression studies. 725-5901; Fax: 650-736-0195; E-mail: [email protected]; and Atul J. Butte, Stanford University School of Medicine, 1265 Welch Road, Room X- We applied a recently proposed meta-analysis approach (22) 163 MS-5415, Stanford CA 94305-5415. Phone: 650-723-3465; Fax: 650- to 13 gene expression datasets consisting of 2,026 lung samples, 723-7070; Email: [email protected] which enabled the discovery and validation of commonly over- doi: 10.1158/0008-5472.CAN-13-2775 expressed genes in lung adenocarcinoma, the predominant 2014 American Association for Cancer Research. subtype of NSCLC. Among the most consistently overexpressed

2892 Cancer Res; 74(10) May 15, 2014

PTK7 Lung Adenocarcinoma Meta-analysis

genes was PTK7, a member of the receptor tyrosine kinase pooled effect size and its standard error using the random family conserved across Hydra, Drosophila, Japanese puffer effects inverse-variance technique. The z-statistic was com- fish, chicken, and human (23). PTK7 contains seven immuno- puted as a ratio of the pooled effect size to its standard error for globulin (Ig) domains, a transmembrane domain, and a cata- each gene, and the result was compared with a standard lytically inactive kinase domain within the cytoplasmic tail normal distribution to obtain a nominal P value. P values were (24). It was first discovered in melanocytes (24) and subse- corrected for multiple hypotheses testing using the Benjamini– quently found to be overexpressed in colon carcinoma (25). In Hochberg correction (34). Xenopus, PTK7 acts at the level of Frizzled and Dishevelled to A second nonparametric meta-analysis that combines regulate the Wnt/planar cell polarity (PCP) pathway (26). Ptk7 P values from individual experiments to identify genes with knockout mice die perinatally and display developmental a large effect size in all datasets was also used. A t-statistic was defects of the inner ear and of neural tube closure, consistent calculated for each gene in each study. After computing one- with a role in regulating PCP (27). Overexpression of PTK7 has tail P values for each gene, these were corrected for multiple been described in several cancers, including colon (25), gastric hypotheses using the Benjamini–Hochberg correction. Next, (28), esophageal (29), and acute myelogenous leukemia (AML; Fisher sum of logs method (35) was applied. Briefly, this ref. 30). However, the precise role of PTK7 in regulating method sums the logarithm of corrected P values across all oncogenesis remains unclear. In colon cancer, PTK7 may play datasets for each gene, and compares the sum against a c2 a role in the Wnt/b-catenin pathway (31), and knockdown distribution with 2k degrees of freedom, in which k is the leads to caspase-10–mediated apoptosis (32). number of datasets used in the analysis. Consistent with the gene expression analysis, we found consistently elevated expression of PTK7 protein in primary Leave-one-out validation and classification adenocarcinoma patient samples. Knockdown of PTK7 dem- To control for the influence of single large experiments on onstrated that it is essential for the viability of a subset of the meta-analysis results, leave-one-out meta-analysis was NSCLC cell lines, and PTK7 disruption increased Mapk kinase performed. One dataset at a time was excluded and both – kinase-7 (MKK7) JNK (c-jun-NH2-kinase) pathway activity. meta-analysis methods were applied to the remaining datasets. Xenotransplantation studies revealed the requirement of PTK7 We hypothesized that the minimal set of genes that are in tumor growth. These results demonstrate the power of using significantly overexpressed, irrespective of the set of datasets publicly available patient data to uncover oncogenic drivers analyzed, would constitute a robust gene expression signature and suggest that PTK7 may represent a novel therapeutic of adenocarcinoma across multiple independent cohorts. target in adenocarcinoma. A very stringent threshold [false discovery rate (FDR) 1 10 5] for selecting differentially overexpressed genes in ade- Materials and Methods nocarcinoma was used. Furthermore, we analyzed heteroge- Data collection, preprocessing, and normalization neity of the effect sizes across all studies. Genes with significant Gene expression data for 13 human lung cancer studies heterogeneity (P 0.05) were removed from the overexpressed were downloaded from the National Center for Biotechnology genes identified using stringent FDR criteria. The geometric Information Gene Expression Omnibus (GEO; accession num- mean of the remaining significant genes was computed and bers GSE10072, GSE2514, GSE7670, GSE19188, GSE11969, used to create a univariate binomial linear model for classifying GSE21933, GSE42127, GSE41271, GSE37745, GSE28571, and a lung sample as a normal or adenocarcinoma sample, or as GSE20853) and websites http://www.broadinstitute.org/mpr/ adenocarcinoma or SCC sample. lung/ and https://array.nci.nih.gov/caarray/project/details. action?project.id¼182. The histologic phenotypes were Immunohistochemistry defined as in the corresponding original publications. Datasets Immunohistochemistry was performed as previously were curated to include only normal, adenocarcinoma, squa- described (36) with the following antibodies: rabbit antibody mous cell carcinoma (SCC), and large cell carcinoma (LCC) to phospho-histone H3 (pHH3; 1:500; Upstate), rabbit antibody samples. All datasets were normalized individually using Gua- to cleaved caspase-3 (CC3; 1:400; Cell Signaling Technology; nine Cytosine Robust Multi-Array Average (gcRMA; ref. 33). For 9664), rabbit antibody to PTK7 (1:1,000; Sigma; SAB3500340). PTK7 gene expression, RNA was extracted and prepared using PTK7 staining was performed using pepsin antigen retrieval. established protocols and hybridized to Affymetrix Human Gene 1.0 ST Array. Raw data are available in GEO (GSE50138). Human primary adenocarcinoma samples This study complied with federal, state, and local regulations Meta-analysis of gene expression data of the Human Research Protection Program and was approved Two meta-analysis approaches were applied to the normal- by the Stanford Institutional Research Board. Informed con- ized data (22). The first approach combines effect sizes from sent was obtained from all patients included in the study. each dataset into a meta-effect size to estimate the amount of change in expression across all datasets. For each gene in each Establishment of patient-derived xenograft tumors from dataset, an effect size was computed using Hedges adjusted g.If primary human adenocarcinoma multiple probes mapped to a gene, the effect size for each gene Surgically removed human NSCLC tumor tissues were kept was summarized using the fixed effect inverse-variance model. in ice-cold Hank's Balanced Salt Solution (Life Technologies) Next, study-specific effect sizes were combined to obtain the until use. Tumors were cut into 1-mm pieces and implanted in

www.aacrjournals.org Cancer Res; 74(10) May 15, 2014 2893

Chen et al.

the subrenal capsules in NOD-SCID-IL2Rg (NSG) mice (The ing, cell lines were plated at 5 105 cells per 6-cm plates and Jackson Laboratory). stained with Annexin-V (FITC-Annexin-V; 556419; BD Bio- sciences) and propidium iodide (50 pl/mL final concentration) Tissue microarray following the manufacturer's protocol at indicated time points. Lung adenocarcinoma tissue microarray with normal lung A C6 Flow Cytometer (Accuri) was used for fluorescence- tissue, containing 20 cases of lung adenocarcinoma and 10 activated cell sorting (FACS) analysis and the manufacturer's normal lung tissue (BC04119b; US Biomax) was immunohis- software and FlowJo v5 for analysis. tochemically stained for PTK7. Staining was scored as negative (0), weak (1), or strong (2). Immunoblotting Cells were scraped and lysed in radioimmunoprecipita- Cell culture tion assay buffer with protease inhibitor cocktail (Roche), All NSCLC cell lines were maintained in RPMI supplemented 25 mmol/L sodium fluoride, 1 mmol/L phenylmethylsulfo- with 10% FBS and 1% penicillin–streptomycin. Cell lines nylfluoride, and 1 mmol/L sodium orthovanadate (Sigma- included NCI-H1299, NCI-H2009, NCI-H23, A549, NCI-H441, Aldrich). Protein samples were resolved by SDS-PAGE, trans- NCI-H460, NCI-H1792, NCI-H1975, NCI-H2126, NCI-H358, NCI- ferred to Amersham Hybond-P membranes (GE Healthcare), H727, NCI-H1568, NCI-H1650, and NCI-H2087. Immortalized and blocking buffer (5% bovine serum albumin in TBS-T) for normal lung epithelial cell line SALE (37) was cultured in 1 hour before addition of primary antibody, which was serum-free media SAGM (CC-3118; Lonza). appliedovernightat4C. The following antibodies from Cell Signaling Technology were used (1:1,000 dilution unless Short hairpin RNA and virus production otherwise noted): rabbit anti-Cleaved PARP (#5625), mouse Human short hairpin RNA (shRNA) constructs against PTK7 anti-pERK (# 9106), rabbit anti-ERK (#4695), rabbit anti- were purchased from OpenBiosystems. The human PTK7 pAKT (T308; #9275; 1:500), rabbit anti-AKT (#9272), rabbit shRNA set was cat # RHS4533-NM_002821. Of this set, anti-pMKK7 (S271/T275; #4171), rabbit anti-MKK7 (#4172), TRCN0000006434 and TRCN0000006435 were labeled as rabbit anti-pJNK (T183/Y185; #9251), rabbit anti-JNK (# shPTK7–1 and 2, respectively. Control hairpins against GFP 9252), rabbit anti-pJUN (S73; # 9164), rabbit anti-p-P38 and luciferase in the pLKO.1 backbone were used (see Sup- (T180/Y182; #9211), and rabbit anti-P38 (#9212). Mouse plementary Methods for target sequences). Transfection-qual- anti-b-actin (Sigma-Aldrich, clone 1A4; 1:10,000) was used ity DNA was extracted using Qiagen DNA kits. Lentivirus was as a loading control. produced by transfection into 293FT cells as previously described (38), filtered, and applied directly to cells. Puromycin Xenograft selection was started 2 days after lentiviral infection for a Cell lines infected with specific shRNA were resuspended in duration of 2 days at 2 mg/mL. serum-free RPMI and injected subcutaneously at 1 million cells per tumor into the two lower flanks of nude mice (The Jackson Quantitative real-time PCR analysis Laboratory). Between 4 and 8 tumors were injected for each RNA was isolated 5 days after lentiviral infection and puro- condition. One week after injection, tumor dimensions were mycin selection with TRIzol reagent (Invitrogen) following the measured approximately every 3 days and tumor volume was manufacturer's protocol. cDNA was synthesized with a DyNA- calculated using the formula p/6 [(lengthþwidth)/2] (39). mo cDNA Synthesis Kit (F470; New England Biolabs), and All animal experiments were approved by the Stanford Uni- quantitative real-time PCR (qRT-PCR) was carried out in versity School of Medicine Committee on Animal Care. triplicate by SYBR Green (Quanta Biosciences) using a C1000 Thermal Cycler (BioRad). See Supplementary Methods Statistical analyses for primer sequences. Unpaired two-tailed t tests were used for comparisons between different groups. Error bars correspond to SEM. Cell proliferation assay Significant P values in the text correspond to <0.05 ( ), Cells were trypsinized and plated in triplicate into 96-well <0.01 ( ), or <0.001 ( ). plates (day 0). Cell viability was measured by treating cells with MTT. Absorbance measurements were recorded using a Spec- Results traMax 340 (Molecular Devices) at 570 nm on the indicated Meta-analysis of five adenocarcinoma gene expression days (Cell Proliferation Kit I; Roche). All data were normalized datasets identifies 11 significantly overexpressed genes to experimental day 0. in adenocarcinoma We identified and manually curated seven published and Fluorescence-activated cell sorting analysis publicly available adenocarcinoma datasets (9, 13–15, 17–19). Cells were scraped from the plate and single cell suspensions Only datasets with accessible unprocessed raw data and con- were made by passing cells through 28G1/2 insulin syringes taining both normal and adenocarcinoma samples were used (BD) 10 times. Cells were then washed in PBS and resuspended for further analysis. We identified five datasets that met these in PF10 (10% FBS in PBS), stained at 4C for 30 minutes in the criteria (743 adenocarcinoma, 127 normal; Supplementary dark with rabbit anti-PTK7 (GTX104510; GeneTex) or IgG2a Table S1; refs. 9, 14, 17–19). We applied a recently proposed control (Clone 188B isotype:mouse, BD). For Annexin-V stain- meta-analysis approach (22) to these five datasets as outlined

2894 Cancer Res; 74(10) May 15, 2014 Cancer Research

PTK7 Lung Adenocarcinoma Meta-analysis

A Download 870 ADC B Bhattacharjee et al. and normal samples in Curation of public 5 data sets from GEO genomic data ADC Normal Manual data curation ARHGEF16 BLNK CACNB3 Data normalization (gcRMA, LOESS) COL5A2 Data EPHB2 preprocessing KCNK1 Probe-to-gene mapping MPZL1 PTK7 RUNX1 Meta-analysis by Meta-analysis by SLC2A5 combining effect size combining P values SULT1C2 FDR < 0.00001% FDR < 0.00001%

–4–2 0 2 4 Shedden et al. Landi et al. Discovery bioinformatics data analysis In silico bioinformatics

Leave-one-out meta-analysis ARHGEF16 BLNK FDR < 0.00001 CACNB3 Preliminary candidate genes COL5A2 (11 genes) EPHB2 Validation in 3 data sets KCNK1 consisting of 237 samples MPZL1 (11 genes) PTK7 RUNX1 Specificity to ADC vs. other NSCLC subtypes SLC2A5 SULT1C2 Validation in 6 data sets consisting of 932 samples –5 0 5 –5 0 5 (7 genes) Stearman et al. Su et al. Validated candidate genes

PTK7 ARHGEF16 BLNK CACNB3 Validation by IHC in COL5A2 patient derived xenografts EPHB2 and tissue microarray consisting of 32 samples KCNK1 MPZL1

Biologic validation PTK7 RUNX1 In vitro validation In vivo validation in ADC cell lines SLC2A5 SULT1C2

–4 –2 0 2 4 –4 –2 0 2 4

Figure 1. Meta-analysis of NSCLC gene expression datasets identified 11 significantly overexpressed genes in NSCLC. A, meta-analysis and functional validation pipeline. In silico bioinformatics data analysis workflow consisted of the curation of five publicly available datasets, data preprocessing, discovery meta-analysis, independent validation, and subtype analysis followed by validation in primary human patient samples and in vitro and in vivo functional validation. B, the 11 genes cluster adenocarcinoma (ADC) versus normal across each dataset. Each column is a sample and each row is the expression level of a gene. The color scale represents the raw Z-score ranging from blue (low expression) to yellow (high expression). Dendrograms above each heatmap correspond to the Pearson correlation-based hierarchical clustering of samples by expression of the 11 genes. in (Fig. 1A). Because an experiment with a large number of the datasets used for analysis, would constitute a robust gene samples can have significant influence on meta-analysis expression signature of adenocarcinoma across multiple inde- results, we performed a leave-one-out analysis, in which we pendent cohorts. excluded one dataset at a time and applied our meta-analysis Because of the heterogeneity of gene expression within and approach on the remaining datasets (see Materials and Meth- among cancer samples and datasets, we used a very stringent ods). We hypothesized that the minimal set of genes identified threshold (FDR 1 10 5) for both meta-analysis methods to as significantly overexpressed by both methods, irrespective of select overexpressed genes in adenocarcinoma. This approach

www.aacrjournals.org Cancer Res; 74(10) May 15, 2014 2895

Chen et al.

identified 14 genes significantly and highly overexpressed in linear regression model using the geometric mean had the area adenocarcinoma across all five datasets. In addition, we esti- under receiver operating characteristic (ROC) curve ranging mated between-studies heterogeneity in effect sizes. Three from 0.809 to 0.993 (Fig. 2B). However, because of a very small genes (ARHGAP8, CEACAM1, and DDR1) had significant number of normal samples in the Takeuchi and colleagues between-studies heterogeneity (P 0.05) and were removed dataset, high area under the curve (AUC) value should be from the further analysis because we did not have sufficient interpreted with caution. Thus, publicly available datasets can information available to further explore heterogeneity in these consistently identify a small set of genes as overexpressed in genes (Supplementary Table S1 and Supplementary Fig. S1). human adenocarcinoma. However, as only adenocarcinoma As expected, hierarchical clustering of the five discovery data- samples were used in this analysis, it is possible that these 11 sets using the remaining 11 genes distinguished normal genes are upregulated in other cancers compared with normal lung from adenocarcinoma samples (Fig. 1B). We further tissue. Therefore, we sought to determine whether the 11-gene validated the 11 genes in three additional independent datasets signature was specific for adenocarcinoma or whether it was (16, 20, 21; 146 adenocarcinoma, 91 normal; Supplementary also upregulated in other NSCLC subtypes. Table S2). All genes were significantly overexpressed in the validation datasets at FDR 5% with at least one meta-analysis Six of the 11 genes are overexpressed in adenocarcinoma method, and all except three (CACNB3, SLC2A5, and SULT1C2) compared with other NSCLC subtypes were significantly overexpressed as assessed by both methods We analyzed an additional eight gene expression datasets (Supplementary Fig. S2 and Supplementary Table S3). Geo- consisting of 1,040 human NSCLC samples (16, 20, 40–43; metric mean of the 11 overexpressed genes was significantly 687 adenocarcinoma, 353 SCC; Supplementary Table S4). Six higher for adenocarcinoma compared with normal samples in of the 11 genes were significantly overexpressed in adenocar- each of the three validation datasets (Fig. 2A). A univariate cinoma compared with SCC, whereas two genes were

A Hou et al. Takeuchi et al. Lo et al. Normal Normal Normal P = 1.05E–14 P = 4.8E–04 P = 3.64E–03 ADC ADC 8 ADC 6.0

0.6 7 5.5 6

5.0 0.4 5 Geometric mean Geometric mean Geometric mean

4.5 4

0.2 4.0 3 Normal ADC Normal ADC Normal ADC B Hou et al. Takeuchi et al. Lo et al. 1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

0.25 0.25 0.25 True positive rate (Sensitivity) True positive rate (Sensitivity) True positive rate (Sensitivity)

0.00 AUC = 0.94 (95% CI, 0.96 − 0.91) 0.00 AUC = 0.97 (95% CI, 0.99 − 0.95) 0.00 AUC = 0.81 (95% CI, 0.9 − 0.73) 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 False positive rate (1−specificity) False positive rate (1−specificity) False positive rate (1−specificity)

Figure 2. Geometric mean of the 11 overexpressed genes were signiﬁcantly higher in adenocarcinoma (ADC) samples than normal samples in three independent data sets. A, violin plots summarizing the geometric mean of the 11-gene signature in normal and adenocarcinoma. Each dot is a sample, and black horizontal line is the mean of geometric mean in a given sample group. B, performance of a univariate classiﬁcation model for predicting adenocarcinoma versus normal based on gene expression of the 11-gene signature across the independent datasets measured by ROC curves.

2896 Cancer Res; 74(10) May 15, 2014 Cancer Research

PTK7 Lung Adenocarcinoma Meta-analysis

significantly underexpressed in adenocarcinoma compared nificantly overexpressed in NSCLC samples compared with with SCC (Supplementary Table S5). One of the six overex- normal samples irrespective of subtypes in all discovery (Sup- pressed gene, SULT1C2, had significant between-studies het- plementary Fig. S1) and validation datasets (Supplementary erogeneity (P 0.05) in the eight datasets. Therefore, it was Fig. S2). Furthermore, PTK7 is specifically overexpressed in removed from further analysis. Geometric mean of the remain- adenocarcinoma compared with SCC (Supplementary Fig. ing five overexpressed genes was significantly higher for ade- S3C). Therefore, we focused on further investigations on PTK7. nocarcinoma compared with SCC in seven of the eight datasets (Supplementary Fig. S3A). The geometric mean of the 5-gene Overexpression of PTK7 in primary human and mouse signature in adenocarcinoma was also significantly higher than lung adenocarcinoma specimens that in LCC in the two additional datasets (46 LCC; Supple- Increased expression of PTK7 (Fig. 3A) was assessed by mentary Fig. S3A). A univariate linear regression model using immunohistochemical staining. First, two patient-derived the geometric mean had the area under ROC curve ranging xenografts of human lung adenocarcinoma were analyzed and from 0.603 to 0.766 for discriminating between adenocarcino- noted to stain positively for PTK7 to varying intensities ma and other NSCLC subtypes (Supplementary Fig. S3B). (Fig. 3B). Staining patterns were heterogeneous, with both In summary, a meta-analysis of 13 independent NSCLC nuclear and cytoplasmic staining present within each speci- datasets consisting of 2,026 human lung samples (1,430 men. Next, a tissue microarray consisting of human samples adenocarcinoma, 353 SCC, 46 LCC, and 197 normal) iden- across various stages and grades of adenocarcinoma was used tified five genes (ARHGEF16, CACNB3, MPZL1, PTK7,and to determine PTK7 expression. Fourteen of the 20 adenocar- RUNX1)thataresignificantly and consistently overexpressed cinoma samples stained positively for PTK7, of which five in adenocarcinoma across all datasets, and are able to stained strongly and nine stained weakly to moderately distinguish adenocarcinoma samples from normal lung (Fig. 3C; Supplementary Fig. S4A and Supplementary Table samples or other NSCLC subtypes. Consistent overexpres- S6). In contrast, all normal lung specimens were negative for sion of these genes despite the potential presence of biologic epithelial staining of PTK7 (Supplementary Fig. S4A). We found and technological confounding factors, including different no correlation between PTK7 staining and stage or grade, cohorts, treatments, and microarray technologies strongly although this may be due to limited sample size. Notably, suggests that at least some of these genes may be function- samples with weak staining tended to have more cytoplasmic ally important in adenocarcinoma and may play a role in the rather than nuclear PTK7 (Supplementary Table S6). Thus, pathogenesis of this disease. PTK7 seems to be overexpressed in most human patient primary adenocarcinoma samples and a significant subset The receptor tyrosine kinase, PTK7, is overexpressed in demonstrates high levels of staining. adenocarcinoma Consistent with these results, a well-established mouse PTK7 belongs to the receptor tyrosine kinase family. It is model of adenocarcinoma (47) also exhibited increased mRNA located on chromosomal locus 6p21, which is within the top 4% transcript levels of Ptk7 (Supplementary Fig. S4C) and positive of all regions of most frequent copy number gain in an PTK7 protein staining by immunohistochemistry (Supplemen- independent lung cancer dataset (ref. 44, analyzed using tary Fig. S4D) relative to normal lung. These results suggest Oncomine; ref. 45, data not shown). In addition, 6p21 has been that PTK7 is overexpressed in both human and mouse ade- associated with poor survival in NSCLC (46). PTK7 was sig- nocarcinoma specimens.

A B Figure 3. PTK7 is overexpressed in PTK7 PDX 1 PDX 2 human adenocarcinoma. A, forest Bhattacharjee et al. Shedden et al. plot of PTK7 expression across all Landi et al. discovery and validation meta- Stearman et al. Su et al. analysis datasets. The x-axis is the Takeuchi et al. standardized mean difference Hou et al. Lo et al. between adenocarcinoma (ADC) and normal on a log2 scale. Thus, a Summary ﬁ value of 1 signi es a 2-fold −1.0 −0.5 0.0 0.5 1.0 1.5 2.0 difference in gene expression Standardized mean difference (log2 scale) between cancer and normal. B, representative images of PTK7 C Normal Grade 1 Grade 2 Grade 2–3 expression in two patient-derived xenograft (PDX) specimens. C, representative images of PTK7 staining of normal lung and adenocarcinoma spanning grades 1 to 3. Scale bars, 50 mm.

www.aacrjournals.org Cancer Res; 74(10) May 15, 2014 2897

Chen et al.

Loss of PTK7 in lung cancer cell lines decreases cell dependency on PTK7 was not determined by its basal expres- viability and induces apoptosis sion (Supplementary Fig. S5B; P ¼ 0.51) or mutations of To determine whether PTK7 could play a functional role in any known cancer genes (Supplementary Fig. S5C). However, the maintenance of adenocarcinoma, two shRNAs directed cell lines with chromosomal gains of the locus harboring against PTK7 were used to decrease expression in a panel of PTK7 tended to be more sensitive to PTK7 knockdown (P ¼ lung cancer cell lines (Supplementary Fig. S5A). Importantly, 0.047; Fig. 4B). PTK7 knockdown had no effect on normal lung epithelial cells To further delineate how PTK7 affects cell viability, (Fig. 4A). However, PTK7 depletion using two independent we chose the three cell lines most sensitive to PTK7 knock- shRNAs led to a signiﬁcant decrease in viability in a subset of down (H1299, H2009, and H23) for further analysis. Sub- NSCLC cell lines (Fig. 4A, Supplementary Table S7). The stantial PTK7 knockdown was validated at the RNA level by

A 2.0 B shPTK7-1 1.5 R = 0.67 1.5 shPTK7-2 1.0 1.0 H23 H1299 0.5 H2009 0.5 Relative viability Relative PTK7 sensitivity 0.0 0.0 (viability after PTK7 KD) 0246 H23 H1299 H2009 H358 H727 H1650 H2087 H460 H1792 H1568 H441 H1975 A549 H2126 PTK7 copy number Normal lung epithelial cells CD E H1299 H1299 H1299

shGFP shPTK7-1 shPTK7-2 shGFP shPTK7-1 shPTK7-2

PTK7 shGFP shPTK7-2 0.5 1.4 11.6 shPTK7-1 Count IgG2a Δ

(% of max) PARP PI 1.3 4.8 4.3 β-Actin 103 105 103 105 103 105 PE Annexin-V H2009 H2009 H2009

shGFP shPTK7-1 shPTK7-2 shGFP shPTK7-1 shPTK7-2

2.5 6.0 8.1 PTK7 IgG2a Count shPTK7-2 shGFP shPTK7-1

(% of max) PI Δ 3.6 3.7 4.0 PARP 3 5 3 5 3 5 10 10 10 10 10 10 β-Actin PE Annexin-V H23 H23 H23

shGFP shPTK7-1 shPTK7-2 shGFP shPTK7-1 shPTK7-2

4.9 9.4 18.2 PTK7

Count IgG2a shPTK7-1 shPTK7-2 shGFP

(% of max) PI ΔPARP 0.6 2.1 3.2 3 5 3 5 3 5 β 10 10 10 10 10 10 -Actin PE Annexin-V

Figure 4. Disruption of PTK7 in three NSCLC cell lines leads to a decrease cell viability and increase in apoptosis. A, knockdown of PTK7 with two independent hairpins across a panel of NSCLC cell lines and assessment of cell viability by MTT. B, scatter plot of PTK7 sensitivity versus copy number in the panel of cell lines. PTK7 sensitivity was calculated by averaging the relative viability across both hairpins after PTK7 knockdown. Copy number at the 6p21 PTK7 locus was queried in the Sanger Catalogue of Somatic Mutations in Cancer (COSMIC) database. R, the linear correlation coefﬁcient. C, PTK7 knockdown validation by FACS. D, FACS analysis of Annexin-V–positive cells. E, cleaved DPARP immunoblotting and b-actin loading control.

2898 Cancer Res; 74(10) May 15, 2014 Cancer Research

PTK7 Lung Adenocarcinoma Meta-analysis

qRT-PCR (Supplementary Fig. S5D), coinciding with a cell lines (Supplementary Fig. S6A–S6F), arguing against an off- decrease in protein expression (Fig. 4C). Knockdown of target effect. PTK7 in all three cell lines led to an increase in apoptosis Upon phosphorylation by MKK7, JNK translocates into the as detected by Annexin-V staining (Fig. 4D), as well as by nucleus, leading to the heterodimerization of c-jun with c-fos cleaved PARP immunoblotting (Fig. 4E). Thus, knockdown to form the AP-1 transcriptional complex. Gene expression of PTK7 in several adenocarcinoma cell lines in vitro results profiling following PTK7 knockdown in the H1299 and H2009 in decreased cell viability and increased apoptosis where- cell lines (Supplementary Fig. S6G) revealed an overrepresen- as normal lung epithelial cells are unaffected. These results tation of AP-1 binding sites associated with the differentially suggest that PTK7 may be involved in signaling cascades expressed genes (Supplementary Fig. S6H). Thus, in some that regulate survival of a subset of adenocarcinoma. adenocarcinoma cell lines, loss of PTK7 leads to dysregulation of MKK7–JNK and possibly other signaling pathways that Disrupted PTK7 expression activates MKK7–JNK ultimately lead to decreased proliferation and increased signaling apoptosis. During normal development PTK7 is thought to play a role in mediating planar cell polarity (26, 27). However, it remains Xenotransplantation of lung cancer cells depleted of unclear why disruption of this signaling pathway is important PTK7 reduces tumor burden in cancer cells as it is not evident why planar cell polarity The above studies demonstrate that PTK7 is required for cell mechanisms would be involved in oncogenesis. To understand viability in a subset of adenocarcinoma cell lines in vitro. To test the signaling pathways that are perturbed upon PTK7 disrup- whether PTK7 is important for tumor maintenance in vivo,we tion, we surveyed the extracellular signal–regulated kinase examined the effect of PTK7 depletion on xenograft adeno- (ERK), AKT, JNK, and p38MAPK pathways by immunoblotting carcinoma tumor growth. Xenotransplantation of H1299 (Fig. 5). We found no significant change in ERK and AKT (Fig. 6A and B), H2009 (Fig. 6C and D), and H23 (Fig. 6E and signaling, and no consistent change in p38 phosphorylation F) cell lines infected with shRNA hairpins against PTK7 into status across the three cell lines. However, phosphorylation of immunocompromised mice resulted in a significant decrease MKK7 at Ser171/Thr275, two residues crucial for MKK7 activ- in tumor growth relative to control shRNA-infected cells. PTK7 ity, was clearly increased after PTK7 knockdown. Consistent loss was confirmed by immunohistochemistry for the two with the activation of MKK7, phosphorylation of JNK, a down- PTK7-specific hairpins relative to control (Fig. 6G). The control stream target of MKK7, was also increased in response to PTK7 experiment of infecting normal lung epithelial cell lines was knockdown in the three cell lines tested with two different not performed, as these cell lines do not readily grow in vivo. shRNAs. Activation of MKK7 was also observed when pooled PTK7 depletion led to a significant decrease in pHH3, a siRNAs were used to deplete PTK7 expression across all three marker of mitosis (Fig. 6G and H), and a concomitant increase in CC3, a marker of apoptosis (Fig. 6G and I). Taken together, these studies suggest that PTK7 expression H1299 H2009 H23 may be required for tumor maintenance in vivo for at least a subset of adenocarcinoma.

Discussion shGFP shPTK7-1 shPTK7-2 shGFP shPTK7-1 shPTK7-2 shGFP shPTK7-1 shPTK7-2 pERK The goal of this study was to identify candidate driver genes in lung adenocarcinoma through a meta-analysis of NSCLC ERK gene expression data and to functionally validate their biologic pAKT (T308) signiﬁcance. The meta-analysis method described here was AKT designed to account for biases inherent in single studies and pMKK7 nominate commonly overexpressed genes. The explosion of large amounts of gene expression data from many different MKK7 platforms in multiple independent cohorts has enabled the use

pJNK of in silico tests across numerous patient populations, many more so than a clinical trial could accomplish. With the JNK increasingly widespread availability of gene expression data, pJUN a paradigm shift toward inclusion of a "pre-validation" meta- p-p38 analysis step in biologic studies may accelerate translational research. p38 PTK7 seems to have a context-speciﬁc value as a prognostic β -Actin marker. In gastric cancers, it is a favorable marker of differ- entiated cancers (28), but in AML and esophageal cancer, PTK7

Figure 5. PTK7 regulates MKK7–JNK signaling. Assessment of the ERK, is a poor prognostic marker associated with reduced disease- AKT, JNK, and p38 signaling pathways by immunoblotting in three free survival (29, 30). We were unable to ﬁnd an association NSCLC cell lines. between high PTK7 expression and patient survival in NSCLC

www.aacrjournals.org Cancer Res; 74(10) May 15, 2014 2899

Chen et al.

ABG H1299 H1299 Figure 6. PTK7 depletion in three

) PTK7 pHH3 CC3 3 2,000 shLUC NSCLC cell lines decreases tumor shPTK7-1 shLUC 1,500 shPTK7-2 burden in a xenotransplantation 1,000 assay. A, tumor volume over time ** shPTK7-1 shLUC of xenografted H1299 with control 500 *** shRNA against luciferase (shLUC), 0 Tumor volume (mm volume Tumor 0 5 10 15 20 25 30 shPTK7-2 or two independent hairpins Days 1 cm against PTK7. Error bars represent C D SEM and signiﬁcant P values H2009 H2009 shPTK7-1 correspond to <0.01 ( )or<0.001 )

3 400 shLUC ( ) when compared with the shPTK7-1 300 shPTK7-2 shLUC shLUC control. Values are mean 200 SEM (n ¼ 4–8). B, representative ex ** shPTK7-1 vivo images of H1299 tumors. C

100 *** shPTK7-2 and D, same ﬁgures for H2009 cell 0 Tumor volume (mm volume Tumor 0 5 10 15 20 25 30 35 40 shPTK7-2 line. E and F, same ﬁgures for H23 Days 1 cm cell line. G, staining of PTK7, F HI* E pHH3, and CC3 by * ** H23 H23 60 150 immunohistochemistry in the )

3 150 sshLUChLUC * sshPTK7-1hPTK7-1 shLUC 40 100 H1299 xenografts. H, sshPTK7-2hPTK7-2 100 quantification of pHH3-positive 20 50 cells/F.O.V. shPTK7-1 (no growth) cells/F.O.V. cells per field of view (n ¼ 6). I, CC3-positive CC3-positive pHH3-positive pHH3-positive 50 0 0 quantification of CC3-positive *** shPTK7-2 fi n ¼ 0 cells per eld of view ( 6). Tumor volume (mm volume Tumor 010 20 30 40 50 60 70 80 1 cm shLUC shLUC Days shPTK7-1shPTK7-2 shPTK7-1shPTK7-2

gene expression datasets with clinical annotations (data not the membrane (53). In contrast with HER3, no known ligands shown), although one report suggested a prognostic signifi- are known to bind PTK7 in the human setting, and it is unclear cance in NSCLC (48). whether PTK7 plays a role as a signal modulator, scaffold Functional studies in vitro identified several NSCLC cell lines protein, or another role. that were highly sensitive to depletion of PTK7. Reduced In vertebrates, the PCP pathway regulates convergent exten- expression of PTK7 led to increased cell death both in vitro sion movements and neural tube closure. PCP also is involved and in vivo. This increase in cell death was accompanied by in regulating orientation of stereociliary bundles of sensory increased signaling through the JNK pathway. However, JNK hair cells in the inner ear. Studies of Ptk7 C-terminal knockout inhibition alone did not rescue cell viability after PTK7 loss mice firmly establish this pseudokinase as a regulator of PCP in (data not shown), suggesting that PTK7 loss alters other mammals (27). The precise signaling role of PTK7 in develop- signaling pathways and that JNK activation, while present, is ment is not clear, although the drosophila homologue off-track not sufficient to induce apoptosis. Further studies will be (otk) may act as a coreceptor for plexins to mediate semaphor- needed to identify other signaling pathways downstream of ing signaling (54). Most recently, PTK7/OTK has been sug- PTK7. Importantly, in NSCLC, PTK7 loss did not consistently gested to act by inhibiting canonical Wnt signaling (55). alter signaling through the ERK, AKT, or p38MAPK pathways. Although loss of PTK7 has been associated with activation of Approximately 10% of the human kinome consists of pro- WNT in Xenopus, we did not detect consistent activation of teins classified as pseudokinases on the basis of alterations in WNT signaling in the absence of PTK7 (data not shown). Thus, residues that are thought to be critical to kinase catalytic the effect of PTK7 loss on lung cancer pathogenesis is likely to activity. The pseudokinase PTK7 is most closely related to be independent of the regulation of WNT signaling. In sum- active tyrosine kinases such as Ros, ALK, and LTK (49). Despite mary, the data presented here describe PTK7 as a functionally the absence of a catalytic domain, increasing evidence suggests relevant protein in NSCLC. We demonstrate that its disruption a key role for pseudokinases in many biologic processes (50). in vitro and in vivo decreases tumor cell growth. These obser- The best-described pseudokinase with an oncogenic role is vations provide a framework for further studies to characterize HER3, which is a member of the EGF receptor family. After PTK7 as a potentially therapeutic target in adenocarcinoma. ligand binding, HER3 heterodimerizes with HER2, resulting in the interaction of the pseudokinase domain of HER3 and the Disclosure of Potential Conflicts of Interest active kinase domain of HER2, leading to the autophosphor- No potential conflicts of interest were disclosed. ylation of HER2 and activation of downstream phosphoinosi- tide 3-kinase and ERK signaling pathways (51, 52). Pseudoki- nases can also function as scaffold proteins. For example, KSR Authors' Contributions Conception and design: R. Chen, P. Khatri, A.J. Butte, E.A. Sweet-Cordero proteins interact with protein kinases of the mitogen-activated Development of methodology: P. Khatri, M. Polin, J. Shrager, A.J. Butte, protein kinase pathway to localize the signaling components to E.A. Sweet-Cordero

2900 Cancer Res; 74(10) May 15, 2014 Cancer Research

PTK7 Lung Adenocarcinoma Meta-analysis

Acquisition of data (provided animals, acquired and managed patients, Grant Support provided facilities, etc.): R. Chen, P. Khatri, P.K. Mazur, M. Polin, Y. Zheng, C.D. R. Chen was funded by the Stanford Graduate Fellowship and Cancer Hoang, J. Shrager, Y. Xu, E.A. Sweet-Cordero Biology Training Grant. P. Khatri, A. Butte, and E.A. Sweet-Cordero were Analysis and interpretation of data (e.g., statistical analysis, biostatistics, funded by the US National Cancer Institute (R01 CA138256). P.K. Mazur and computational analysis): R. Chen, P. Khatri, M. Polin, D. Vaka, S. Vicent, S. Vicent were funded by the Tobacco-Related Disease Research Program of E.A. Sweet-Cordero California (TRDRP). P.K. Mazur was also funded by the Stanford University Writing, review, and/or revision of the manuscript: R. Chen, P. Khatri, Dean's Fellowship and the Child Health Research Institute and Lucile Y. Zheng, J. Shrager, S. Vicent, A.J. Butte, E.A. Sweet-Cordero Packard Foundation for Children's Health Postdoctoral Fellowship. Y. Zheng Administrative, technical, or material support (i.e., reporting or orga- was funded by the American Lung Association. nizing data, constructing databases): P. Khatri, J. Shrager, S. Vicent The costs of publication of this article were defrayed in part by the payment Study supervision: R. Chen, P. Khatri, A.J. Butte, E.A. Sweet-Cordero of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Acknowledgments The authors thank Julien Sage, Laura Attardi, Steve Artandi, Monte Winslow, Received September 25, 2013; revised January 21, 2014; accepted February 27, and all members of the Sweet-Cordero laboratory for helpful discussions. 2014; published OnlineFirst March 20, 2014.

References 1. Howlader N, Noone AM, Krapcho M, Neyman N, Aminou R, Altekruse 15. Ding L, Getz G, Wheeler DA, Mardis ER, McLellan MD, Cibulskis K, SF, et al. (eds). SEER cancer statistics review, 1975–2009 (Vintage et al. Somatic mutations affect key pathways in lung adenocarcinoma. 2009 Populations). Bethesda, MD: National Cancer Institute [cited Nature 2008;455:1069–75. 2012 Aug 13]. Available from: http://seer.cancer.gov/csr/1975_2009_ 16. Hou J, Aerts J, den Hamer B, van Ijcken W, den Bakker M, Riegman P, pops09/index.html. et al. Gene expression-based classification of non-small cell lung 2. Keedy VL, Temin S, Somerfield MR, Beasley MB, Johnson DH, carcinomas and survival prediction. PloS ONE 2010;5:e10312. McShane LM, et al. American Society of Clinical Oncology provisional 17. Landi MT, Dracheva T, Rotunno M, Figueroa JD, Liu H, Dasgupta A, clinical opinion: epidermal growth factor receptor (EGFR) mutation et al. Gene expression signature of cigarette smoking and its role in testing for patients with advanced non-small-cell lung cancer consid- lung adenocarcinoma development and survival. PloS ONE 2008;3: ering first-line EGFR tyrosine kinase inhibitor therapy. J Clin Oncol e1651. 2011;29:2121–7. 18. Stearman RS, Dwyer-Nield L, Zerbe L, Blaine SA, Chan Z, Bunn PA Jr, 3. Soda M, Choi YL, Enomoto M, Takada S, Yamashita Y, Ishikawa S, et al. Analysis of orthologous gene expression between human pul- et al. Identification of the transforming EML4-ALK fusion gene in non- monary adenocarcinoma and a carcinogen-induced murine model. small-cell lung cancer. Nature 2007;448:561–6. Am J Pathol 2005;167:1763–75. 4. Pao W, Hutchinson KE. Chipping away at the lung cancer genome. Nat 19. Su LJ, Chang CW, Wu YC, Chen KC, Lin CJ, Liang SC, et al. Selection Med 2012;18:349–51. of DDX5 as a novel internal control for Q-RT-PCR from microarray data 5. Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, et al. using a block bootstrap re-sampling scheme. BMC Genomics 2007; Repeated observation of breast tumor subtypes in independent 8:140. gene expression data sets. Proc Natl Acad Sci U S A 2003; 20. Takeuchi T, Tomida S, Yatabe Y, Kosaka T, Osada H, Yanagisawa K, 100:8418–23. et al. Expression profile-defined classification of lung adenocarcinoma 6. West M, Blanchette C, Dressman H, Huang E, Ishida S, Spang R, shows close relationship with underlying major genetic changes and et al. Predicting the clinical status of human breast cancer by using clinicopathologic behaviors. J Clin Oncol 2006;24:1679–88. gene expression profiles. Proc Natl Acad Sci U S A 2001;98: 21. Lo FY, Chang JW, Chang IS, Chen YJ, Hsu HS, Huang SF, et al. The 11462–7. database of chromosome imbalance regions and genes resided in lung 7. Rhodes DR, Barrette TR, Rubin MA, Ghosh D, Chinnaiyan AM. Meta- cancer from Asian and Caucasian identified by array-comparative analysis of microarrays: interstudy validation of gene expression genomic hybridization. BMC Cancer 2012;12:235. profiles reveals pathway dysregulation in prostate cancer. Cancer Res 22. Khatri P, Roedder S, Kimura N, De Vusser K, Morgan AA, Gong Y, et al. 2002;62:4427–33. A common rejection module (CRM) for acute rejection across multiple 8. Hoshida Y, Nijman SM, Kobayashi M, Chan JA, Brunet JP, Chiang DY, organs identifies novel therapeutics for organ transplantation. J Exp et al. Integrative transcriptome analysis reveals common molecular Med 2013;210:2205–21. subclasses of human hepatocellular carcinoma. Cancer Res 2009; 23. Grassot J, Gouy M, Perriere G, Mouchiroud G. Origin and molecular 69:7385–92. evolution of receptor tyrosine kinases with immunoglobulin-like 9. Shedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, Gerald domains. Mol Biol Evol 2006;23:1232–41. WL, et al. Gene expression-based survival prediction in lung adeno- 24. Park SK, Lee HS, Lee ST. Characterization of the human full-length carcinoma: a multi-site, blinded validation study. Nat Med 2008; PTK7 cDNA encoding a receptor protein tyrosine kinase-like molecule 14:822–7. closely related to chick KLG. J Biochem 1996;119:235–9. 10. Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature 25. Mossie K, Jallal B, Alves F, Sures I, Plowman GD, Ullrich A. Colon of metastasis in primary solid tumors. Nat Genet 2003;33:49–54. carcinoma kinase-4 defines a new subclass of the receptor tyrosine 11. Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, kinase family. Oncogene 1995;11:2179–84. et al. Large-scale meta-analysis of cancer microarray data identifies 26. Shnitsar I, Borchers A. PTK7 recruits dsh to regulate neural crest common transcriptional profiles of neoplastic transformation and migration. Development 2008;135:4015–24. progression. Proc Natl Acad Sci U S A 2004;101:9309–14. 27. Lu X, Borchers AG, Jolicoeur C, Rayburn H, Baker JC, Tessier-Lavigne 12. Segal E, Friedman N, Koller D, Regev A. A module map showing M. PTK7/CCK-4 is a novel regulator of planar cell polarity in verte- conditional activity of expression modules in cancer. Nat Genet brates. Nature 2004;430:93–8. 2004;36:1090–8. 28. Lin Y, Zhang LH, Wang XH, Xing XF, Cheng XJ, Dong B, et al. PTK7 as a 13. Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE, novel marker for favorable gastric cancer patient survival. J Surg Oncol et al. Gene-expression profiles predict survival of patients with lung 2012;106:880–6. adenocarcinoma. Nat Med 2002;8:816–24. 29. Shin WS, Kwon J, Lee HW, Kang MC, Na HW, Lee ST, et al. Oncogenic 14. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, et al. role of protein tyrosine kinase 7 in esophageal squamous cell carci- Classification of human lung carcinomas by mRNA expression profil- noma. Cancer Sci 2013;104:1120–6. ing reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U 30. Prebet T, Lhoumeau AC, Arnoulet C, Aulas A, Marchetto S, Audebert S, S A 2001;98:13790–5. et al. The cell polarity PTK7 receptor acts as a modulator of the

www.aacrjournals.org Cancer Res; 74(10) May 15, 2014 2901

Chen et al.

chemotherapeutic response in acute myeloid leukemia and impairs 43. Micke P, Edlund K, Holmberg L, Kultima HG, Mansouri L, Ekman S, clinical outcome. Blood 2010;116:2315–23. et al. Gene copy number aberrations are associated with survival in 31. Puppo F, Thome V, Lhoumeau AC, Cibois M, Gangar A, Lembo F, et al. histologic subgroups of non-small cell lung cancer. J Thorac Oncol Protein tyrosine kinase 7 has a conserved role in Wnt/beta-catenin 2011;6:1833–40. canonical signalling. EMBO Rep 2011;12:43–9. 44. Weir BA, Woo MS, Getz G, Perner S, Ding L, Beroukhim R, et al. 32. Meng L, Sefah K, O'Donoghue MB, Zhu G, Shangguan D, Noorali A, Characterizing the cancer genome in lung adenocarcinoma. Nature et al. Silencing of PTK7 in colon cancer cells: caspase-10-depen- 2007;450:893–98. dent apoptosis via mitochondrial pathway. PLoS ONE 2010;5: 45. Rhodes DR, Kalyana-Sundaram S, Mahavisno V, Varambally R, Yu J, e14018. Briggs BB, et al. Oncomine 3.0: genes, pathways, and networks in a 33. Zhijin W, Irizarry RA, Gentleman R, Martinez-Murillo F, Spencer F. A collection of 18,000 cancer gene expression profiles. Neoplasia model-based background adjustment for oligonucleotide expression 2007;9:166–80. arrays. J Am Stat Assoc 2004;99:909–17. 46. Kim TM, Yim SH, Lee JS, Kwon MS, Ryu JW, Kang HM, et al. Genome- 34. Storey J. A direct approach to false discovery rates. J R Stat Soc wide screening of genomic alterations and their clinicopathologic 2002;64:479–98. implications in non-small cell lung cancers. Clin Cancer Res 2005; 35. Fisher RA. Statistical methods for research workers. Edinburgh, Lon- 11:8235–42. don: Oliver and Boyd; 1925. 47. Jackson EL, Willis N, Mercer K, Bronson RT, Crowley D, Montoya R, 36. Mazur PK, Einwachter H, Lee M, Sipos B, Nakhai H, Rad R, et al. et al. Analysis of lung tumor initiation and progression using conditional Notch2 is required for progression of pancreatic intraepithelial neo- expression of oncogenic K-ras. Genes Dev 2001;15:3243–48. plasia and development of pancreatic ductal adenocarcinoma. Proc 48. Endoh H, Tomida S, Yatabe Y, Konishi H, Osada H, Tajima K, et al. Natl Acad Sci U S A 2010;107:13438–43. Prognostic model of pulmonary adenocarcinoma by expression pro- 37. Lundberg AS, Randell SH, Stewart SA, Elenbaas B, Hartwell KA, filing of eight genes as determined by quantitative real-time reverse Brooks MW, et al. Immortalization and transformation of primary transcriptase polymerase chain reaction. J Clin Oncol 2004;22: human airway epithelial cells by gene transfer. Oncogene 2002; 811–19. 21:4577–86. 49. Boudeau J, Miranda-Saavedra D, Barton GJ, Alessi DR. Emerging 38. Rubinson DA, Dillon CP, Kwiatkowski AV, Sievers C, Yang L, Kopinja J, roles of pseudokinases. Trends Cell Biol 2006;16:443–52. et al. A lentivirus-based system to functionally silence genes in primary 50. Zeqiraj E, van Aalten DM. Pseudokinases-remnants of evolution or key mammalian cells, stem cells and transgenic mice by RNA interference. allosteric regulators? Curr Opin Struct Biol 2010;20:772–81. Nat Genet 2003;33:401–6. 51. Holbro T, Beerli RR, Maurer F, Koziczak M, Barbas CF 3rd, Hynes NE. 39. Tomayko MM, Reynolds CP. Determination of subcutaneous tumor The ErbB2/ErbB3 heterodimer functions as an oncogenic unit: ErbB2 size in athymic (nude) mice. Cancer Chemother Pharmacol 1989; requires ErbB3 to drive breast tumor cell proliferation. Proc Natl Acad 24:148–54. Sci U S A 2003;100:8933–8. 40. Sato M, Larsen JE, Lee W, Sun H, Shames DS, Dalvi MP, et al. Human 52. Dar AC. A pickup in pseudokinase activity. Biochem Soc Trans 2013; lung epithelial cells progressed to malignancy through specific onco- 41:987–94. genic manipulations. Mol Cancer Res 2013;11:638–50. 53. Kolch W. Coordinating ERK/MAPK signalling through scaffolds and 41. Tang H, Xiao G, Behrens C, Schiller J, Allen J, Chow CW, et al. A 12- inhibitors. Nat Rev Mol Cell Biol 2005;6:827–37. gene set predicts survival benefits from adjuvant chemotherapy in 54. Winberg ML, Tamagnone L, Bai J, Comoglio PM, Montell D, Goodman non-small cell lung cancer patients. Clin Cancer Res 2013;19: CS. The transmembrane protein off-track associates with Plexins and 1577–86. functions downstream of Semaphorin signaling during axon guidance. 42. Botling J, Edlund K, Lohr M, Hellwig B, Holmberg L, Lambe M, et al. Neuron 2001;32:53–62. Biomarker discovery in non-small cell lung cancer: integrating gene 55. Peradziryi H, Kaplan NA, Podleschny M, Liu X, Wehner P, Borchers A, expression profiling, meta-analysis, and tissue microarray validation. et al. PTK7/Otk interacts with Wnts and inhibits canonical Wnt sig- Clin Cancer Res 2013;19:194–204. nalling. EMBO J 2011;30:3729–40.

2902 Cancer Res; 74(10) May 15, 2014 Cancer Research

A Meta-analysis of Lung Cancer Gene Expression Identifies PTK7 as a Survival Gene in Lung Adenocarcinoma

Ron Chen, Purvesh Khatri, Pawel K. Mazur, et al.

Cancer Res 2014;74:2892-2902. Published OnlineFirst March 20, 2014.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-13-2775

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2014/03/20/0008-5472.CAN-13-2775.DC1

Cited articles This article cites 53 articles, 22 of which you can access for free at: http://cancerres.aacrjournals.org/content/74/10/2892.full#ref-list-1

Citing articles This article has been cited by 19 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/74/10/2892.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/74/10/2892. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.