Journal of the National Cancer Institute Advance Access originally published online on August 8, 2007
JNCI Journal of the National Cancer Institute 2007 99(16):1257-1269; doi:10.1093/jnci/djm083
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Published by Oxford University Press 2007.
ARTICLES |
Use of a Cytokine Gene Expression Signature in Lung Adenocarcinoma and the Surrounding Tissue as a Prognostic Classifier
Affiliations of authors: Laboratory of Human Carcinogenesis, Center for Cancer Research (MS, NY, EDB, KAZ, AB, KK, LEM, XWW, CCH) and Cancer Prevention Fellowship Program, Division of Cancer Prevention (KAZ), National Cancer Institute, National Institutes of Health, Bethesda, MD; Biology Division, National Cancer Center Research Institute, Tokyo, Japan (SM, JY); Cancer Genomics Project, National Cancer Center Research Institute, Tokyo, Japan (TS); Department of Pathology, Hamamatsu University School of Medicine, Hamamatsu, Japan (HS); Department of Pulmonary Medicine/Infection and Oncology, Nippon Medical School, Tokyo, Japan (MS, AG, SK)
Correspondence to: Curtis C. Harris, MD, Laboratory of Human Carcinogenesis, National Cancer Institute, National Institutes of Health, 37 Convent Dr, Bldg 37, Rm 3068, Bethesda, MD 20892-4258 (e-mail: curtis_harris{at}nih.gov).
 |
ABSTRACT |
|---|
 Top Abstract Context and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
Background: A 17-cytokine gene expression signature in noncancerous hepatic tissue from patients with metastatic hepatocellular carcinoma (HCC) was recently found to predict HCC metastasis and recurrence. We examined whether the cytokine gene expression profile of noncancerous lung tissue could predict the metastatic capability of adjacent lung adenocarcinoma.
Methods: We analyzed a 15–cytokine gene expression profile in noncancerous lung tissue and corresponding lung tumor tissue from 80 US lung adenocarcinoma patients using real-time quantitative reverse transcription–polymerase chain reaction. We then used unsupervised hierarchical clustering and Prediction Analysis of Microarray classification to test the prognostic ability of the 15–cytokine gene profile in the US patients and in an independent validation set comprising 50 Japanese patients with stage I disease. Survival was analyzed by the Kaplan–Meier method using the log-rank test, and univariate and multivariable Cox proportional hazards modeling were used to analyze the association of clinical variables with patient survival. All statistical tests were two-sided.
Results: A 15–cytokine gene signature in noncancerous lung tissue primarily reflected the lymph node status of 80 lung adenocarcinoma patients, whereas the gene signature of the corresponding lung tumor tissue was associated with prognosis independent of lymph node status. Cytokine Lung Adenocarcinoma Survival Signature of 11 genes (CLASS-11), a refined 11-gene signature, accurately classified patients, including those with stage I disease, according to risk of death from adenocarcinoma. CLASS-11 prognostic classification was statistically significantly associated with survival and was an independent prognostic factor for stage I patients (hazard ratio for death in the high-risk CLASS-11 group compared with the low-risk CLASS-11 reference group = 7.46, 95% confidence interval = 2.14 to 26.05; P = .002). CLASS-11 also classified patients in the validation set according to risk of recurrence.
Conclusion: CLASS-11, which consists of genes for pro- and anti-inflammatory cytokines, identifies stage I lung adenocarcinoma patients who have a poor prognosis.
Prior knowledge The finding that a unique 17-cytokine gene expression signature in noncancerous hepatic tissue from patients with metastatic hepatocellular carcinoma predicts hepatocellular carcinoma metastasis and recurrence suggests that gene expression changes in surrounding noncancerous tissue as well as in tumors can be used as biomarkers to predict prognosis and metastasis in other cancers. Study design Molecular profiling study of the prognostic ability of a cytokine gene profile in lung adenocarcinoma patients. Contribution A unique 11-cytokine gene signature that was based on expression profiling of both noncancerous lung tissue and lung tumors accurately classified stage I lung adenocarcinoma patients according to their risk of death. Implications This cytokine gene signature may be useful for determining which lung adenocarcinoma patients are at high risk of death and thus may benefit from adjuvant therapy. Limitations Patients included in the validation set had shorter follow-up and were of different ethnicity than patients in the training set.
|
Lung cancer is the leading cause of cancer death both in the United States and worldwide (1). Patients with early-stage non–small-cell lung cancer (NSCLC) who undergo curative resection still have a substantial risk of developing metastases. The 5-year survival rates for patients with stage IA and IB NSCLC are only 67% and 57%, respectively (2). The identification of sensitive and specific biomarkers predictive of unfavorable prognosis could have a clinically significant impact on NSCLC treatment strategies to aid in the selection of patients for further therapy. The cellular and molecular mechanisms for metastasis of adenocarcinoma of the lung, the predominant histologic subtype of NSCLC (1,3), remain to be elucidated. Several molecular signatures of lung adenocarcinoma that are associated with metastasis and survival have been reported (4–14).
During metastasis, cancer cells undergo a multistep process that includes invasion, entry into the circulatory system, arrest in a distant site, proliferation, and induction of angiogenesis (15). However, the molecular and microenvironmental factors that contribute to metastasis are poorly understood. Recently, gene expression profiling studies in several cancer tissue types have reported molecular signatures that are associated with metastasis (4,16,17). These mRNA expression profiling studies have demonstrated that the metastasis-associated genes of solid tumors, including lung cancers, hepatocellular carcinomas (HCCs), and breast cancers, can be found in early-stage primary tumors. These discoveries suggest that metastatic potential might be preprogrammed in some primary tumors by the oncogenic events that initiate the tumors.
We recently identified a unique 17-gene expression signature in noncancerous hepatic tissue from patients with metastatic HCC that predicts HCC metastasis and recurrence (18). The prognostic signature consisted mainly of cytokine genes that are expressed in type 1 and type 2 helper T cells (TH1 cells and TH2 cells, respectively), which are involved in inflammatory and immune responses (19). Tumor cells that produce immunosuppressive cytokines can escape the host immune response (20). These findings suggest that gene expression changes in surrounding noncancerous tissue as well as in tumors can be used as biomarkers to predict cancer prognosis and metastasis.
In this study, we sought to clarify whether the gene expression profile of noncancerous lung tissue could predict the metastatic capability of adjacent lung adenocarcinoma by focusing on the expression profiles of 18 cytokine genes in paired noncancerous and tumor tissues from 80 patients with lung adenocarcinoma. We ultimately identified an 11-gene signature, Cytokine Lung Adenocarcinoma Survival Signature of 11 genes (CLASS-11), that we tested for its ability to predict lymph node metastasis and disease prognosis in an independent group of lung cancer patients.
|
Methods |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
Clinical Samples
We evaluated 138 pairs of noncancerous lung tissue and corresponding primary lung tumor tissues from lung adenocarcinoma patients who had undergone surgical resection from 1988 to 1999 at the University of Maryland Medical Center, Sinai Hospital of Baltimore, Baltimore County General Hospital, St Agnes Hospital, and Mercy Medical Center (all in Baltimore, MD). Five-year survival information after surgery was available for all patients. All tissue was freshly collected during surgery, snap-frozen, and stored at –80 °C. The noncancerous lung tissue was obtained from a minimum of 2 cm away from the tumor to assure that the distant noncancerous lung tissues were free from cancerous cells. Peripheral portions of resected lung tumors were sectioned, stained with hematoxylin–eosin, and evaluated by the study pathologist (H. Sugimura) to confirm that the lung tumors met the histologic criteria for adenocarcinoma according to the 2004 World Health Organization (WHO) classification (1). We excluded patients who had presented with histologic subtypes other than adenocarcinoma (adenosquamous and sarcomatoid or plemorphic carcinoma with adenocarcinoma component; n = 5) or whose samples had less than 50% tumor cell content on the slides (n = 1) or were defined by quantitative reverse transcription–polymerase chain reaction (qRT–PCR) exclusion criteria as described below (n = 52). The remaining 80 pairs of noncancerous lung tissue and corresponding primary lung tumor tissues from US lung adenocarcinoma patients were used to identify a gene signature: 53 patients had stage I disease, 20 had stage II disease, six had stage III disease, and one had stage IV disease according to WHO TNM (tumor–node–metastasis) staging (2). Normal lung tissue from four cancer-free patients who had undergone surgical resection at the University of Maryland Medical Center from 1996 to 1999 was used as a reference group for each tissue sample; two of the four patients had a smoking history of 20 or more pack-years (one had a bronchial carcinoid, the other had a hamartoma), and the other two patients had a smoking history of less than 20 pack-years (one had a bronchial carcinoid, the other had a fibrous soft tissue tumor).
We used an independent validation set of tumor and corresponding noncancerous tissue samples from 50 Japanese patients with stage I adenocarcinoma who had undergone surgical resection from 1999 to 2003 at the National Cancer Center Hospital (Tokyo) and Hamamatsu University School of Medicine Hospital (Hamamatsu). Information on patient survival and recurrence during 3 years of follow-up was available for all 50 cases (Supplementary Table 1, available online). Institutional review board approval and written informed consent from all patients were obtained at each collection site.
RNA Isolation and Quantitative Reverse Transcription–Polymerase Chain Reaction Analysis
Total RNA was isolated from the lung tumor and noncancerous lung tissues from the 80 US patients, the lung tumor and noncancerous lung tissues from the 50 Japanese patients, and the normal lung tissue from the four cancer-free patients with the use of TRIzol reagent (Invitrogen, Carlsbad, CA), according to the manufacturer’s instructions. Total RNA (3 µg) was converted to complementary DNA (cDNA) with the use of random hexamers and a SuperScript III First-Strand Synthesis kit (Invitrogen), according to the manufacturer's instructions. The cDNAs were then used for qRT–PCR analysis of an 18-gene expression profile. The expression profiles of 12 cytokine genes (i.e., interleukin 1
[IL-1], interleukin 1
[IL-1], interleukin 2 [IL-2], interleukin 4 [IL-4], interleukin 5 [IL-5], interleukin 8 [IL-8], interleukin 10 [IL-10], interleukin 12 p35 [IL-12p35], interleukin 12 p40 [IL-12p40], interleukin 15 [IL-15], interferon gamma [IFN-
] and tumor necrosis factor- [TNF-]) were quantified with the use of TaqMan Cytokine Gene Expression Plates (Applied Biosystems, Foster City, CA), according to the manufacturer's instructions. We also analyzed the expression of the interleukin 6 (IL-6), major histocompatibility complex (MHC) class II antigen, DR alpha (HLA-DRA), MHC class II antigen, DP alpha 1 (HLA-DPA1), annexin A1 (ANXA1), platelet proteoglycan (PRG1), and colony-stimulating factor 1 (CSF1) genes by using TaqMan Gene Expression Assays (Applied Biosystems). Reactions were performed with the use of a PRISM 7700 Sequence Detector System (Applied Biosystems). Human 18S ribosomal RNA (rRNA) labeled with VIC reporter dye (Applied Biosystems) was used as an endogenous control. Gene expression was quantified using the comparative method (2–
CT), where CT = threshold cycle, CT = (CT cytokine – CT 18S rRNA) – (CT reference – CT 18S rRNA), as previously described (21). We excluded the gene expression data for 52 US patients because the average CT values for the cytokine genes were greater than 35 cycles (22). The raw data are available at http://www3.cancer.gov/intra/LHC/lhcpage.htm.
Determination of Tissue Cytokine Concentrations
Protein lysates were prepared by homogenizing 50 mg of lung tumor tissue or noncancerous tissues in 500 µL of tissue homogenizing buffer (50 mM Tris–HCl pH 7.6, 150 mM NaCl, 0.1% sodium dodecyl sulfate, 1% Nonidet P-40, and 0.5% sodium deoxycholate). The homogenates were kept on ice for 30 minutes, then centrifuged at 13000g for 30 minutes. The supernatant was collected and the protein concentration of the supernatant was measured by a Bradford assay (Bio-Rad, Hemel Hempstead, U.K.). We determined the protein concentrations of six cytokines (IL-8, IL-1, IL-2, IL-10, IFN-, and TNF-) by using cytokine assay kits (Meso Scale Discovery, Gaithersburg, MD) according to the manufacturer's instructions. Results are expressed as picograms of cytokine per milligrams of total protein.
Statistical Analysis
Unsupervised hierarchical clustering analysis was performed with the use of Cluster View and Tree View programs (http://genexpress.stanford.edu/tutorials/cluster_view.html and http://genexpress.stanford.edu/tutorials/tree_view.html, respectively; Stanford University, Palo Alto, CA). For class prediction based on the qRT–PCR profiling, we used Prediction Analysis of Microarray (PAM), an algorithm for class prediction from gene expression data developed by Tibshirani et al. (23), in which classification is based on the nearest shrunken centroid, coupled with 10-fold cross-validation (Supplementary Methods, available online). PAM has been used to classify several types of tumors on the basis of their gene expression profiles (10,18,23). This method is also efficient in finding genes for classification and prediction of cancer. We randomly assigned the 80 patients to training (n = 40) and test (n = 40) sets for cross-validation analysis. These two cohorts had similar clinical profiles (Supplementary Table 2, available online). Kaplan–Meier survival analysis was used to compare the survival of 79 of the 80 US patients in the training set (one patient died during the surgery; that patient's survival time was 0 days). The resulting survival curves were compared with one another using the Cox–Mantel log-rank test. The survival curves for the smaller Japanese cohort comprising the validation set were compared using the Wilcoxon log-rank test, which gives more weight to differences between the survival curves at earlier times. Cox proportional hazards modeling (both univariate and multivariable tests) was used to analyze the effect of six clinical variables on patient survival (i.e., age, sex, race, smoking history, tumor differentiation, and TNM stage). Both of the final models met the proportional hazards assumption as determined by Schoenfeld residuals. Kaplan–Meier survival analysis was performed using Prism 4 software (GraphPad, Aurona, CO) and verified with the use of Stata software (version 9.1; Stats Corp., College Station, TX). Cox proportional hazards modeling was performed using Stata 9.1. The Wilcoxon matched pairs signed rank test (in Prism 4 software) was used to compare the protein expression between tumor and noncancerous tissues. All statistical tests were two-sided, and statistical significance was defined as P less than .05.
|
Results |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
Hierarchical Clustering Analysis of Paired Noncancerous Lung Tissue and Corresponding Lung Adenocarcinoma
To investigate the role of the lung environment in promoting metastasis, we first analyzed the expression profiles of 18 cytokine genes in noncancerous lung tissue samples from 80 US lung adenocarcinoma patients. Seventeen of these genes were previously shown to be part of a unique inflammation/immune response–related signature in noncancerous hepatic tissue from HCC patients (18). This signature also included the IL-6 gene, which was not part of the HCC signature; IL-6 is a multifunctional cytokine that is expressed predominantly in tumor-infiltrating lymphocytes and normal bronchial epithelial cells in lung cancer (24,25), and high circulating levels of IL-6 are associated with disease progression and poor survival in lung cancer patients (26,27). We then eliminated the IL-4, IL-5, and IL-12p40 genes from the 18-gene profile because their expression was detected in only 24 (30%), 20 (25%), and 26 (33%) noncancerous lung tissue samples, respectively. We selected the 15 genes whose expression was detected in more than 70% of the noncancerous lung tissue samples for further analysis.
Unsupervised hierarchical clustering analysis of the 80 noncancerous lung tissue samples, which was based on the similarities in expression patterns of the 15-gene panel, revealed two distinct main clusters, namely cluster A (n = 31 samples) and cluster B (n = 49 samples) (Fig. 1, A). An examination of the relationship between the two clusters and patient and tumor characteristics revealed statistically significant differences between cluster A and cluster B with respect to lymph node metastasis status (N0 versus N1–2; P = .002, Fisher's exact test) and TNM stage (stage I versus stages II–IV; P = .002, Fisher's exact test) (Table 1). Only one patient with distant metastasis (M1) was included among the 80 US patients. Therefore, the 15–cytokine gene signature was not influenced by distant metastasis status (M0 versus M1). These results suggested that the 15–cytokine gene signature of noncancerous tissue reflected primarily the lymph node status (i.e., metastasis versus no metastasis) of these 80 patients. However, there was no statistically significant difference in overall survival between the two clusters of patients in the two clusters (Table 1 and Fig. 1, D).
|
|
We next analyzed the 15–cytokine gene expression profiles of the corresponding lung tumor tissue from the 80 patients. Unsupervised hierarchical clustering analysis separated these 80 lung cancer tissues into two distinct clusters, cluster C and cluster D, each of which contained 40 tumor samples (Fig. 1, B). There were no statistically significant differences between clusters C and D with respect to lymph node metastasis status or TNM stage (Table 1). However, survival status (
5 years versus <5 years) was statistically significantly different between clusters C and D (P = .01; Table 1), as were the Kaplan–Meier curves for overall survival (P = .009) (Fig. 1, E). Next, we separated the 80 US lung adenocarcinoma patients into four groups based on the cluster patterns of the noncancerous lung tissue and the lung tumor tissue of each patient (Fig. 1, C). Group 1 (n = 19) comprised patients whose noncancerous lung tissue was in cluster A (i.e., the nonmetastatic cluster) and whose tumor tissue was in cluster D (i.e., the good prognosis cluster); these patients were categorized as being at a low risk of death. In contrast, groups 2–4 (n = 61) comprised patients whose noncancerous lung tissue was in cluster B (i.e., the metastatic cluster) and/or whose tumor tissue was in cluster C (i.e., the poor prognosis cluster); these patients were categorized as being at a high risk of death. Kaplan–Meier survival analysis revealed that the 5-year overall survival rate for patients in groups 2–4 was statistically significantly less than that of patients in group 1 (32% versus 79%, difference = 47%, 95% confidence interval [CI] = 25% to 69%, P = .001) (Fig. 1, F). When we restricted the analysis to patients with stage I adenocarcinoma (n = 53), the 5-year overall survival rate for patients in groups 2–4 (n = 36) was still statistically significantly less than that of patients in group 1 (n = 17) (39% versus 82%, difference = 43%, 95% CI = 19% to 67%, P = .006) (Fig. 1, G). These results suggest that the combination of the 15–cytokine gene profile of the noncancerous lung tissue with that of lung tumor tissue from the same individual is strongly associated with disease outcome, even in stage I cases.
Association Between 15–Cytokine Gene Expression Profile and Lymph Node Metastasis and Prognosis in Lung Adenocarcinoma Patients
We next used the PAM cross-validation algorithm to test the ability of the 15–cytokine gene profile of noncancerous tissue to classify patients for lymph node metastasis (N0 versus N1–2). We randomly distributed the 80 US patients in training (n = 40) and test (n = 40) sets for this cross-validation analysis. Overall, the PAM classification of the 15–cytokine gene profile using a threshold value of 0.00 correctly identified the lymph node status of 32 (80%) of 40 cases in the training set (Supplementary Table 3, A, available online). Furthermore, patients with lymph node metastasis (n = 24) expressed statistically significantly higher levels of five cytokine genes (IL-10, IL-8, IL-6, IL-2, and IFN-) than cases with no nodal metastasis (n = 56) (Table 2). Of those five genes, those encoding cytokines IL-10, IL-8, and IL-6 were the top three genes for discriminating cases with lymph node metastasis from cases without metastasis by PAM ranking.
|
We next examined whether PAM classification of the 15–cytokine gene profile from corresponding lung tumor tissue could classify patients according to 5-year survival. Overall, PAM classification of the 15–cytokine gene profile using a threshold value of 0.00 correctly classified outcomes in 28 (70%) of the 40 cases in the training cohort (Supplementary Table 3, A, available online). Of the 15 cytokine genes, nine (IL-10, IL-8, TNF-
, IL-1, IL-15, IL-2, IFN-, IL-1, and IL-12p35) were expressed at statistically significantly lower levels (P<.05 to P<.001) in the lung tumor tissue of patients who survived for at least 5 years (n = 34) than in patients who survived for fewer than 5 years (n = 56) (Table 2). Of the nine genes, IL-8, TNF-, and IFN- had the highest fold differences in expression between the group of patients who survived for at least 5 years and the group of patients who survived for fewer than 5 years (4.2, 3.9, and 5.4, respectively). In the PAM analysis, these genes were the top three genes that contributed to survival by PAM ranking. Construction of the CLASS-11 Prognosis Predictor
We next identified the smallest number of cytokine genes whose expression could accurately classify lung adenocarcinoma patients according to lymph node status. Using 10-fold cross-validation by PAM, both the cross-validated and test errors were minimized near a shrinkage value of 0.18 (Table 2; Supplementary Fig. 1, A, available online). The value 0.18 yielded a subset of 11 selected genes. The top 11 genes (IL-10, IL-8, IL-6, TNF-, IL-1, IL-15, IL-2, IFN-, IL-1, IL-12p35, and CSF1) according to the PAM ranking were subsequently found to be the minimum number necessary for lymph node classification. Overall, PAM classification of these 11 genes correctly identified the lymph node status of 32 (80%) of the 40 patients in the training set (Supplementary Table 3, A, and Supplementary Fig. 1, C, available online). In addition, these 11 genes were also the optimal number of genes for prognostic classification (at a shrinkage value of 0.39) (Table 2; Supplementary Fig. 1, B, available online). Overall, PAM classification of the 11 genes correctly classified prognosis in 29 (73%) of the 40 patients in the training set (Supplementary Table 3, B, and Supplementary Fig. 1, D, available online). We named this gene expression signature the Cytokine Lung Adenocarcinoma Survival Signature of 11 genes (CLASS-11).
We next tested the prognostic ability of CLASS-11 for the 40 patients in the test set. PAM classification revealed that CLASS-11 correctly predicted both lymph node status and prognosis in 31 (78%) of the 40 patients (Supplementary Table 3 and Supplementary Figs. 1, E and F, available online). We performed PAM classification of CLASS-11 for all 80 US adenocarcinoma cases. Consistently, CLASS-11 correctly predicted lymph node status in 63 (79%) of the 80 patients and prognosis in 60 (75%) of the 80 patients (data not shown).
Based on the results of the clustering analyses of noncancerous tissue and the lung tumor tissues, we next used a combination of the two PAM classifiers (i.e., the lymph node status classifier and the prognosis classifier) to improve the prognostic classification by CLASS-11 for potential future clinical application (Fig. 2). CLASS-11 classified the 40 cases in the training set into one of two survival risk groups: the low-risk-of-death group or high-risk-of-death group. The low-risk-of-death group (n = 12 cases) comprised patients whose noncancerous lung tissue was classified by PAM as no lymph node metastasis (i.e., NON) and whose tumor tissue was classified by PAM as good prognosis (i.e., GOOD). The high-risk-of-death group (n = 28 cases) comprised patients whose noncancerous lung tissue was classified by PAM as lymph node metastasis (i.e., MET) and/or whose tumor tissue was classified by PAM as poor prognosis (i.e., POOR). The 5-year survival rates were 75% (95% CI = 41% to 91%) for the low-risk-of-death group and 30% (95% CI = 14% to 47%) for the high-risk-of-death group. Kaplan–Meier survival analysis showed that the high-risk group had statistically significantly worse survival than the low-risk group (P = .02) (Fig. 3, A). Subsequent evaluation of the predictive power of CLASS-11 in the test set classified the 40 cases into a high-risk-of-death group (n = 27 cases) or the low-risk-of-death group (n = 13 cases). The high-risk group had statistically significantly worse survival than the low-risk group (P<.001) (Fig. 3, B). The 5-year survival rates were 85% (95% CI = 51% to 96%) in the low-risk-of-death group and 22% (95% CI = 9% to 39%) in the high-risk-of-death group.
|
|
For the 80 US patients included in this study, those classified by CLASS-11 as being at a high risk of death (n = 55) had statistically significantly worse survival than those classified by CLASS-11 as being at a low risk of death (n = 25) (P<.001) (Fig. 3, C). In addition, among the 53 patients with stage I adenocarcinoma, those classified by CLASS-11 as being at a high risk of death (n = 34) had statistically significantly worse survival than those classified by CLASS-11 as being at a low risk of death (n = 19) (P<.001) (Fig. 3, D). Among the stage I patients, the 5-year survival rates were 84% (95% CI = 59% to 95%) for those at a low risk of death and 35% (95% CI = 20% to 51%) for those at a high risk of death. Notably, CLASS-11 correctly predicted poor survival in 24 (89%) of the 27 stage I cases in the high-risk-of-death group.
In a sensitivity analysis, we excluded the three cases of bronchioloalveolar adenocarcinoma from the Kaplan–Meier survival analysis of the stage I cases because this histologic subtype can be associated with a more favorable prognosis (28,29). Among the remaining stage I patients (n = 50), those classified by CLASS-11 as being at a high risk of death (n = 32) had statistically significantly worse survival than those classified by CLASS-11 as being at a low risk of death (n = 18) (P<.001) (data not shown). The prognostic ability of CLASS-11 was also examined for differences with respect to stage. Among patients with stage IA adenocarcinoma (n = 28), there was a non–statistically significant difference in survival between the groups at high and low risks of death (P = .07) (Supplementary Fig. 2, A, available online). Among patients with stage IB (n = 25) or stage II (n = 20) adenocarcinoma, those classified by CLASS-11 as being at a high risk of death had statistically significantly poorer survival than those classified by CLASS-11 as being at a low risk of death (P<.001 and P = .01, respectively) (Supplementary Fig. 2, B and C, available online).
Finally, we investigated whether the prognostic ability of CLASS-11 was affected by underlying clinical covariates by performing univariate and multivariable Cox proportional hazards survival analyses. The univariate analysis of the 80 US patients revealed that TNM stage (hazard ratio [HR] for death in the stage II–IV group compared with the stage I reference group = 2.58, 95% CI = 1.42 to 4.67; P = .002) and CLASS-11 prognostic classification (HR for death in the high-risk CLASS-11 group compared with the low-risk CLASS-11 reference group = 6.23, 95% CI = 2.44 to 15.86; P<.001) were both statistically significant predictors of survival (Table 3). The multivariable analysis of the 80 US patients, which adjusted for TNM stage, showed that the CLASS-11 prognostic classification was statistically significantly associated with survival and was an independent prognostic factor for lung adenocarcinoma (HR for death in the high-risk CLASS-11 group compared with the low-risk CLASS-11 reference group = 5.94, 95% CI = 2.32 to 15.17; P<.001). We also performed univariate and multivariable survival analyses for the 53 stage I patients. The univariate analysis revealed that tumor differentiation (HR for death in the moderate/poorly differentiated group compared with the well-differentiated reference group = 3.51, 95% CI = 1.05 to 11.74; P = .04), TNM stage (HR for death in the stage II–IV group compared with the stage I reference group = 2.76, 95% CI = 1.21 to 6.25; P = .02), and CLASS-11 prognostic classification (HR for death in the high-risk CLASS-11 group compared with the low-risk CLASS-11 reference group = 6.54, 95% CI = 1.95 to 21.96; P = .002) were statistically significantly associated with survival (Table 3). The multivariable Cox proportional hazards model of all stage I patients, which controlled for race, tumor differentiation, and TNM stage, also revealed that the CLASS-11 prognostic classification was statistically significantly associated with survival and was an independent prognostic factor for stage I cases (HR for death in the high-risk CLASS-11 group compared with the low-risk CLASS-11 reference group = 7.46, 95% CI = 2.14 to 26.05; P = .002). When we excluded the three bronchoalveolar adenocarcinoma cases from both the univariate and multivariate analyses, the results remained statistically significant (data not shown). Therefore, the presence of this histologic subtype did not influence the prediction of survival by CLASS-11.
|
Validation of the CLASS-11 Prognosis Predictor
We examined the robustness of CLASS-11 for classifying patients into prognostic groups in an independent set of tumor and corresponding noncancerous lung tissue samples from 50 Japanese stage I adenocarcinoma patients who had undergone surgical resection from 1999 to 2003. The Japanese and US patients differed from each other in several respects; in particular, there was a higher frequency of heavy smokers (smoking history of 20 or more pack-years) and of patients with moderate or poorly differentiated carcinoma among the US patients than among the Japanese patients (P<.001 and P = .009, respectively) (Supplementary Table 1, available online). The CLASS-11 prognostic classifier classified the 50 Japanese patients into a high-risk-of-death group (n = 37 cases) and a low-risk-of-death group (n = 13 patients). The Kaplan–Meier survival curves, stratified by CLASS-11 prognostic classification, are shown in Fig. 3, E and F. Although CLASS-11 correctly classified the eight Japanese patients who had died during the 3 years of follow-up as being at a high risk of death, there was no statistically significant difference between the low-risk and high-risk groups in overall survival, perhaps because of the shorter follow-up period for this cohort (P = .08; Wilcoxon log-rank test) (Fig. 3, E). However, the group classified as being at a high risk of death (n = 37) had statistically significantly poorer disease-free survival than the group at a low risk of death (n = 13) (P = .03; Wilcoxon log-rank test) (Fig. 3, F).
Confirmation of Cytokine Gene Expression Results by Enzyme-Linked Immunosorbent Assay
We next evaluated the correlation between cytokine mRNA expression as measured by qRT–PCR and cytokine protein expression as measured by enzyme-linked immunosorbent assay. To do so, we determined the cytokine protein concentrations in pairs of noncancerous tissue and tumor tissue from 30 of the 80 US patients for whom tissue was available for this analysis. IL-8 protein was detected in all tumor and noncancerous tissues examined (Supplementary Fig. 3, A, available online), whereas five other cytokines (IL-1, IL-2, IL-10, IFN-, and TNF-) were not detectable in many of the tissues examined (data not shown). The tumor tissues expressed statistically significantly higher levels of IL-8 than the noncancerous tissues (median IL-8 concentration [interquartile range] for tumor versus noncancerous tissue: 6.45 pg/mg tissue [1.11–10.25 pg/mg tissue] versus 1.89 pg/mg tissue [1.19–4.61 pg/mg tissue], P = .003, Wilcoxon matched pairs signed rank test). There was no correlation between IL-8 mRNA expression and IL-8 protein concentration in noncancerous tissue (Spearman's rho = .10; P = .59) (data not shown). However, in tumor tissue, IL-8 protein concentration was statistically significantly correlated with IL-8 mRNA expression (Spearman's rho = .47; P = .009) (Supplementary Fig. 3, B, available online). Furthermore, Kaplan–Meier survival analysis of these 30 patients stratified by the median IL-8 concentration in tumor tissue revealed that patients whose tumors expressed IL-8 protein at levels at or greater than the median concentration (n = 16) had statistically significantly worse survival than patients whose tumors expressed IL-8 protein at levels below the median (n = 14) (P = .03) (Supplementary Fig. 3, C, available online). Thus, the increased level of IL-8 gene expression in tumor tissue was associated with elevated IL-8 protein concentration in the tumor tissue and with poor overall survival.
|
Discussion |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
In this study, we constructed a classification algorithm based on analyses of noncancerous and tumor tissues (CLASS-11) to predict the prognosis of lung adenocarcinoma patients with stage I. Several mRNA and microRNA microarray–based studies have demonstrated that lung tumor tissues have unique molecular profiles that are associated with metastasis and survival in patients with NSCLC (5–14). For example, in one study (7), an mRNA signature of 50 genes correctly predicted the survival of patients with stage I lung adenocarcinoma, and in another study (13), hsa-mir-155 and hsa-let7a-2 microRNAs were associated with prognosis independent of tumor stage in lung adenocarcinoma. In addition, mRNA expression profiling has been used to show that distinct subclasses of lung adenocarcinoma and neuroendocrine lung tumors are associated with poor outcome (5,6,8,9,12,14). Most recently, an mRNA expression profile was shown to identify a subset of stage IA NSCLC patients who were at a high risk of recurrence (14).
The inflammatory status of noncancerous tissue surrounding a tumor may play an important role in promoting tumor progression and metastasis. We reported previously that a cytokine gene signature of surrounding noncancerous tissue from HCC patients can predict metastasis and recurrence of HCC (18). In this study, we demonstrated a similar phenomenon in lung adenocarcinoma. The expression profile of CLASS-11, which consists of 11 pro- and anti-inflammatory cytokine genes, in noncancerous tissue reflected lymph node metastasis, which is associated with poor prognosis. TH1 cells produce proinflammatory cytokines (e.g., IL-2, IFN-, and TNF-) as part of the cell-mediated immune response, and TH2 cells regulate humoral immunity by expressing anti-inflammatory cytokines (e.g., IL-4 and IL-10). A previous report (25) indicated that tumor tissue from NSCLC patients has higher expression of TH2 cytokine genes than of TH1 cytokine genes. Higher serum levels of TH2 cytokines have also been associated with a poor prognosis in patients with NSCLC (30,31). Our results suggest that the propensity of a tumor to metastasize may depend on genetic and epigenetic changes that affect cytokine gene expression in the tumor and on the immunologic status of the host.
Identification of molecular markers that are associated with lymph node metastasis in lung cancer patients may provide information for future therapies for lung cancer. In recent analyses (10,32), gene expression profiles in tumor tissue predicted lymph node status of patients with primary lung adenocarcinoma. Studies of histologic characteristics of lung cancer (33) and other malignancies (34) have described an association between lymph node metastasis and lymph node immunoreactivity.
These reports have demonstrated that long-term survival is directly associated with the development of cellular immunity and inversely associated with the development of a humoral immune response produced by TH2 cytokines in the regional lymph nodes. Lymphocytic infiltration of primary lung tumor is also associated with lymph node metastasis (33,35,36). It has also been reported that up to one-third of the total tumor-infiltrating lymphocyte population in lung tumors are immunosuppressive CD4+CD25+ regulatory T cells (37). In addition, the extent of infiltration of these regulatory T cells in the tumor has been found to play a role in the outcome for NSCLC patients (38). NSCLC cells themselves have been reported to produce as well as to induce TH2 cytokines (25). Thus, it will be interesting to use immunohistochemical analysis and laser capture microdissection technique to delineate the roles of various immune cells in contributing to the molecular pathogenesis of lung cancer. The information on immune cell status may be useful in stratifying patients for adjuvant therapy trials, including trials of immunotherapy.
In this study, CLASS-11 predicted disease prognosis independent of lymph node status. Results of a recent study (39) have also suggested that lymph node involvement and length of survival represent distinct biologic processes in breast cancer. In that study, the subset of genes whose expression was associated with lymph node metastasis (which included chemokines, chemokine receptors, and IFN-associated genes) was distinct from the genes whose expression predicted survival (which included genes involved in cell cycle control and cell signaling and growth factor receptor genes). In our analyses of lung tumor tissue, IL-8 and TNF- were the top two genes for predicting prognosis by PAM ranking. It is interesting that the proteins expressed by IL-8 and TNF- may have angiogenic activities in several cancers including NSCLC (40,41). Therefore, elevated levels of IL-8 and TNF- in tumor tissue from adenocarcinoma patients could enhance angiogenesis and the occurrence of microinvasion. IL-8 (also known as CXCL8) also functions as a positive autocrine growth factor for NSCLC (42–44). In addition, a recent study demonstrated that IL-8 is a key member of the mRNA signature for a field effect in lung cancer (45). Consequently, it seems possible that high expression levels of IL-8 and TNF- in tumors might increase tumor growth rate, resulting in poor survival independent of lymph node status.
Our study is limited by the short follow-up period of Japanese patients when compared with the US patients, which limits our ability to assess how accurately CLASS-11 predicts the 5-year survival of the Japanese patients. American and Japanese patients also may have different genetic alterations in lung adenocarcinoma that could contribute differently to clinical outcomes (46,47). For example, epidermal growth factor receptor gene mutations that are associated with increased sensitivity of lung cancer to drugs were statistically significantly more frequent among patients from Japan than among patients from the United States (46,47). To our knowledge, differences in cytokine gene expression profiles between American and Japanese patients have not been previously reported. In this study, the robustness of CLASS-11 was further supported by its ability to predict high risk of recurrence in Japanese stage I cases. Therefore, CLASS-11 is likely to be a unique signature in predicting prognosis of adenocarcinomas. In addition, several studies (28,29) have reported associations between histologic parameters, such as histologic subtypes and grades, and prognosis for lung cancer patients. For example, bronchioloalveolar carcinoma patients been reported to have a favorable prognosis (28,29). In our study, the 5-year overall survival was still statistically significantly different between low-risk and high-risk groups in the 50 US patients without bronchioloalveolar carcinoma.
In conclusion, our data suggest that a unique cytokine gene expression signature of noncancerous lung tissue and corresponding tumor tissue in lung adenocarcinoma predicts metastasis and disease progression. These findings suggest that the lung tumor and its surrounding lung environment interact. In addition, the CLASS-11 algorithm, which was based on the gene expression profiles of both noncancerous and tumor tissue, was useful for identifying stage I adenocarcinoma patients who are at a high risk of death. The CLASS-11 algorithm might eventually guide personalized therapies. Several randomized trials (48–52) have reported that adjuvant chemotherapy improves survival among patients with resected NSCLC. Adjuvant chemotherapy after surgery for stage II–IIIA adenocarcinoma is now considered to be the standard of care based on the results of three trials (49–51). In addition, two randomized trials (48,52) of postoperative adjuvant chemotherapy versus surgery alone in patients with stage IB adenocarcinoma showed that adjuvant chemotherapy statistically significantly improved survival. In this context, CLASS-11 may be useful for predicting which stage IB and II patients are at a high risk of death and may, therefore, benefit from adjuvant therapy.
|
Funding |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
Intramural Research Program of National Institutes of Health; National Cancer Institute; Center for Cancer Research.
|
NOTES |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
The study sponsors had no role in the design of the study, data collection, analysis and interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication.
We thank Drs Ewy A. Mathe and Rosemary I. Braun for helpful discussion, Dorothea Dudek-Creaven for editorial assistance, and Drs Paul Goldsmith and Michele Gunsior for determining cytokine protein concentrations. We also thank Drs Raymond T. Jones, Andrew Borkowski, and Mark J. Krasna at the University of Maryland and Baltimore Veterans Administration for sample collection and pathology report and Audrey Salabes for interviewing the lung cancer patients.
|
REFERENCES |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionFundingReferencesNotes |
|---|
(1) Travis WD, Brambilla E, Harris CC, Muller-Hermelink HK. World Health Organization Classification of Tumours, Pathology and Genetics: tumours of the lung, pleura, thymus and heart. (2004) Lyon (France): IARC Press.
(2) Mountain CF. Revisions in the International System for Staging Lung Cancer. Chest (1997) 111:1710–7.[CrossRef][Web of Science][Medline]
(3) Devesa SS, Bray F, Vizcaino AP, Parkin DM. International lung cancer trends by histologic type: male:female differences diminishing and adenocarcinoma rates rising. Int J Cancer (2005) 117:294–9.[CrossRef][Web of Science][Medline]
(4) Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasis in primary solid tumors. Nat Genet (2003) 33:49–54.[CrossRef][Web of Science][Medline]
(5) Garber ME, Troyanskaya OG, Schluens K, Petersen S, Thaesler Z, Pacyna-Gengelbach M, et al. Diversity of gene expression in adenocarcinoma of the lung. Proc Natl Acad Sci USA (2001) 98:13784–9.
(6) Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, et al. Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci USA (2001) 98:13790–5.
(7) Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE, et al. Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med (2002) 8:816–24.[Web of Science][Medline]
(8) Miura K, Bowman ED, Simon R, Peng AC, Robles AI, Jones RT, et al. Laser capture microdissection and microarray expression analysis of lung adenocarcinoma reveals tobacco smoking- and prognosis-related molecular profiles. Cancer Res (2002) 62:3244–50.
(9) Endoh H, Tomida S, Yatabe Y, Konishi H, Osada H, Tajima K, et al. Prognostic model of pulmonary adenocarcinoma by expression profiling of eight genes as determined by quantitative real-time reverse transcriptase polymerase chain reaction. J Clin Oncol (2004) 22:811–9.
(10) Xi L, Lyons-Weiler J, Coello MC, Huang X, Gooding WE, Luketich JD, et al. Prediction of lymph node metastasis by analysis of gene expression profiles in primary lung adenocarcinomas. Clin Cancer Res (2005) 11:4128–35.
(11) Meyerson M, Carbone D. Genomic and proteomic profiling of lung cancers: lung cancer classification in the age of targeted therapy. J Clin Oncol (2005) 23:3219–26.
(12) He P, Varticovski L, Bowman ED, Fukuoka J, Welsh JA, Miura K, et al. Identification of carboxypeptidase E and -glutamyl hydrolase as biomarkers for pulmonary neuroendocrine tumors by cDNA microarray. Hum Pathol (2004) 35:1169–209.[CrossRef]
(13) Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M, et al. Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell (2006) 9:189–98.[CrossRef][Web of Science][Medline]
(14) Potti A, Mukherjee S, Peterson R, Dressman HK, Bild A, Koontz J, et al. A genomic strategy to refine prognosis in early-stage non-small-cell lung cancer. New England J Med (2006) 355:570–80.
(15) Hanahan D, Weinberg RA. The hallmarks of cancer. Cell (2000) 100:57–70.[CrossRef][Web of Science][Medline]
(16) van 't Veer LJ, Dai H, Van de Vijver MJ, He YD, Hart AA, Mao M, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature (2002) 415:530–6.[CrossRef][Medline]
(17) Ye QH, Qin LX, Forgues M, He P, Kim JW, Peng AC, et al. Predicting hepatitis B virus-positive metastatic hepatocellular carcinomas using gene expression profiling and supervised machine learning. Nat Med (2003) 9:416–23.[CrossRef][Web of Science][Medline]
(18) Budhu A, Forgues M, Ye QH, Jia LH, He P, Zanetti KA, et al. Prediction of venous metastases, recurrence and prognosis in hepatocellular carcinoma based on a unique immune response signature of the liver microenvironment. Cancer Cell (2006) 10:99–111.[CrossRef][Web of Science][Medline]
(19) Abbas AK, Murphy KM, Sher A. Functional diversity of helper T lymphocytes. Nature (1996) 383:787–93.[CrossRef][Medline]
(20) Lewis CE, Pollard JW. Distinct role of macrophages in different tumor microenvironments. Cancer Res (2006) 66:605–12.
(21) Bustin SA. Absolute quantification of mRNA using real-time reverse transcription polymerase chain reaction assays. J Mol Endocrinol (2000) 25:169–93.[Abstract]
(22) Karlen Y, McNair A, Perseguers S, Mazza C, Mermod N. Statistical significance of quantitative PCR. BMC Bioinformatics (2007) 20:131.
(23) Tibshirani R, Hastie T, Narasimhan B, Chu G. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA (2002) 99:6567–72.
(24) Takizawa H, Ohtoshi T, Ohta K, Yamashita N, Hirohata S, Hirai K, et al. Growth inhibition of human lung cancer cell lines by interleukin 6 in vitro: a possible role in tumor growth via an autocrine mechanism. Cancer Res (1993) 53:4175–81.
(25) Asselin-Paturel C, Echchakir H, Carayol G, Gay F, Opolon P, Grunenwald D, et al. Quantitative analysis of Th1, Th2 and TGF-beta1 cytokine expression in tumor, TIL and PBL of non-small cell lung cancer patients. Int J Cancer (1998) 77:7–12.[CrossRef][Web of Science][Medline]
(26) De Vita F, Orditura M, Auriemma A, Infusino S, Roscigno A, Catalano G. Serum levels of interleukin-6 as a prognostic factor in advanced non-small cell lung cancer. Oncol Rep (1998) 5:649–52.[Web of Science][Medline]
(27) Yanagawa H, Sone S, Takahashi Y, Haku T, Yano S, Shinohara T, et al. Serum levels of interleukin 6 in patients with lung cancer. Br J Cancer (1995) 71:1095–8.[Web of Science][Medline]
(28) Noguchi M, Morikawa A, Kawasaki M, Matsuno Y, Yamada T, Hirohashi S, et al. Small adenocarcinoma of the lung. Histologic characteristics and prognosis. Cancer (1995) 75:2844–52.[CrossRef][Web of Science][Medline]
(29) Zell JA, Ou SH, Ziogas A, Anton-Culver H. Epidemiology of bronchioloalveolar carcinoma: improvement in survival after release of the 1999 WHO classification of lung tumors. J Clin Oncol (2005) 23:8396–405.
(30) Neuner A, Schindel M, Wildenberg U, Muley T, Lahm H, Fischer JR. Prognostic significance of cytokine modulation in non-small cell lung cancer. Int J Cancer (2002) 101:287–92.[CrossRef][Web of Science][Medline]
(31) De Vita F, Orditura M, Galizia G, Romano C, Roscigno A, Lieto E, et al. Serum interleukin-10 levels as a prognostic factor in advanced non-small cell lung cancer patients. Chest (2000) 117:365–73.[CrossRef][Web of Science][Medline]
(32) Kikuchi T, Daigo Y, Katagiri T, Tsunoda T, Okada K, Kakiuchi S, et al. Expression profiles of non-small cell lung cancers on cDNA microarrays: identification of genes for prediction of lymph-node metastasis and sensitivity to anti-cancer drugs. Oncogene (2003) 22:2192–205.[CrossRef][Web of Science][Medline]
(33) Di Giorgio A, Mingazzini P, Sammartino P, Canavese A, Arnone P, Scarpini M. Host defense and survival in patients with lung carcinoma. Cancer (2000) 89:2038–45.[CrossRef][Web of Science][Medline]
(34) Hunter RL, Ferguson DJ, Coppleson LW. Survival with mammary cancer related to the interaction of germinal center hyperplasia and sinus histiocytosis in axillary and internal mammary lymph nodes. Cancer (1975) 36:528–39.[CrossRef][Web of Science][Medline]
(35) Nakamura H, Ishiguro K, Mori T. Different immune functions of peripheral blood, regional lymph node, and tumor infiltrating lymphocytes in lung cancer patients. Cancer (1988) 62:2489–97.[CrossRef][Web of Science][Medline]
(36) Lee TK, Horner RD, Silverman JF, Chen YH, Jenny C, Scarantino CW. Morphometric and morphologic evaluations in stage III non-small cell lung cancers. Prognostic significance of quantitative assessment of infiltrating lymphoid cells. Cancer (1989) 63:309–16.[CrossRef][Web of Science][Medline]
(37) Woo EY, Yeh H, Chu CS, Schlienger K, Carroll RG, Riley JL, et al. Cutting edge: regulatory T cells from lung cancer patients directly inhibit autologous T cell proliferation. J Immunol (2002) 168:4272–6.
(38) Okita R, Saeki T, Takashima S, Yamaguchi Y, Toge T. CD4+CD25+ regulatory T cells in the peripheral blood of patients with breast cancer and non-small cell lung cancer. Oncol Rep (2005) 14:1269–73.[Web of Science][Medline]
(39) Huang E, Cheng SH, Dressman H, Pittman J, Tsou MH, Horng CF, et al. Gene expression predictors of breast cancer outcomes. Lancet (2003) 361:1590–6.[CrossRef][Web of Science][Medline]
(40) Smith DR, Polverini PJ, Kunkel SL, Orringer MB, Whyte RI, Burdick MD, et al. Inhibition of interleukin 8 attenuates angiogenesis in bronchogenic carcinoma. J Exp Med (1994) 179:1409–15.
(41) Yao PL, Lin YC, Wang CH, Huang YC, Liao WY, Wang SS, et al. Autocrine and paracrine regulation of interleukin-8 expression in lung cancer cells. Am J Respir Cell Mol Biol (2005) 32:540–7.
(42) Pold M, Zhu LX, Sharma S, Burdick MD, Lin Y, Lee PP, et al. Cyclooxygenase-2-dependent expression of angiogenic CXC chemokines ENA-78/CXC ligand (CXCL) 5 and interleukin-8/CXCL8 in human non-small cell lung cancer. Cancer Res (2004) 64:1853–60.
(43) Zhu YM, Webster SJ, Flower D, Woll PJ. Interleukin-8/CXCL8 is a growth factor for human lung cancer cells. Br J Cancer (2004) 91:1970–6.[CrossRef][Web of Science][Medline]
(44) Wislez M, Fujimoto N, Izzo JG, Hanna AE, Cody DD, Langley RR, et al. High expression of ligands for chemokine receptor CXCR2 in alveolar epithelial neoplasia induced by oncogenic kras. Cancer Res (2006) 66:4198–207.
(45) Spira A, Beane JE, Shah V, Steiling K, Liu G, Schembri F, et al. Airway epithelial gene expression in the diagnostic evaluation of smokers with suspect lung cancer. Nat Med (2007) 13:361–6.[CrossRef][Web of Science][Medline]
(46) Paez JG, Janne PA, Lee JC, Tracy S, Greulich H, Gabriel S, et al. EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy. Science (2004) 304:1497–500.
(47) Shigematsu H, Lin L, Takahashi T, Nomura M, Suzuki M, Wistuba II, et al. Clinical and biological features associated with epidermal growth factor receptor gene mutations in lung cancers. J Natl Cancer Inst (2005) 97:339–46.
(48) Kato H, Ichinose Y, Ohta M, Hata E, Tsubota N, Tada H, et al. A randomized trial of adjuvant chemotherapy with uracil-tegafur for adenocarcinoma of the lung. N Engl J Med (2004) 350:1713–21.
(49) Arriagada R, Bergman B, Dunant A, Le Chevalier T, Pignon JP, Vansteenkiste J. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med (2004) 350:351–60.
(50) Winton T, Livingston R, Johnson D, Rigas J, Johnston M, Butts C, et al. Vinorelbine plus cisplatin vs. observation in resected non-small-cell lung cancer. N Engl J Med (2005) 352:2589–97.
(51) Douillard JY, Rosell R, De Lena M, Carpagnano F, Ramlau R, Gonzales-Larriba JL, et al. Adjuvant vinorelbine plus cisplatin versus observation in patients with completely resected stage IB-IIIA non-small-cell lung cancer (Adjuvant Navelbine International Trialist Association [ANITA]): a randomised controlled trial. Lancet Oncol (2006) 7:719–27.[CrossRef][Web of Science][Medline]
(52) Roselli M, Mariotti S, Ferroni P, Laudisi A, Mineo D, Pompeo E, et al. Postsurgical chemotherapy in stage IB nonsmall cell lung cancer: long-term survival in a randomized study. Int J Cancer (2006) 119:955–60.[CrossRef][Web of Science][Medline]
(53) Vahakangas KH, Bennett WP, Castren K, Welsh JA, Khan MA, Blomeke B, et al. p53 and K-ras mutations in lung cancers from former and never-smoking women. Cancer Res (2001) 61:4350–6.
Manuscript received January 8, 2007; revised June 8, 2007; accepted June 26, 2007.
Related Article in JNCI
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2007 99: 1213.
This article has been cited by other articles:
|
|
 |
|
  F. Colotta, P. Allavena, A. Sica, C. Garlanda, and A. Mantovani Cancer-related inflammation, the seventh hallmark of cancer: links to genetic instability Carcinogenesis, July 1, 2009; 30(7): 1073 - 1081. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. C. Borczuk, R. L. Toonkel, and C. A. Powell Genomics of Lung Cancer Proceedings of the ATS, April 15, 2009; 6(2): 152 - 158. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. A. Howrylak, T. Dolinay, L. Lucht, Z. Wang, D. C. Christiani, J. M. Sethi, E. P. Xing, M. P. Donahoe, and A. M. K. Choi Discovery of the gene signature for acute lung injury in patients with sepsis Physiol Genomics, April 10, 2009; 37(2): 133 - 139. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Verri, L. Roz, D. Conte, T. Liloglou, A. Livio, A. Vesin, A. Fabbri, F. Andriani, C. Brambilla, L. Tavecchio, et al. Fragile Histidine Triad Gene Inactivation in Lung Cancer: The European Early Lung Cancer Project Am. J. Respir. Crit. Care Med., March 1, 2009; 179(5): 396 - 401. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Khawam, J. Giron-Michel, Y. Gu, A. Perier, M. Giuliani, A. Caignard, A. Devocelle, S. Ferrini, M. Fabbi, B. Charpentier, et al. Human Renal Cancer Cells Express a Novel Membrane-Bound Interleukin-15 that Induces, in Response to the Soluble Interleukin-15 Receptor {alpha} Chain, Epithelial-to-Mesenchymal Transition Cancer Res., February 15, 2009; 69(4): 1561 - 1569. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Enewold, L. E. Mechanic, E. D. Bowman, Y.-L. Zheng, Z. Yu, G. Trivers, A. J. Alberg, and C. C. Harris Serum Concentrations of Cytokines and Lung Cancer Survival in African Americans and Caucasians Cancer Epidemiol. Biomarkers Prev., January 1, 2009; 18(1): 215 - 222. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Mantovani Cancer-related Inflammation: The Seventh Hallmark of Cancer ASCO Educational Book, January 1, 2009; 2009(1): 723 - 726. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Steiling, J. Ryan, J. S. Brody, and A. Spira The Field of Tissue Injury in the Lung and Airway Cancer Prevention Research, November 1, 2008; 1(6): 396 - 403. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Song, R. Chaerkady, A. C. Tan, E. Garcia-Garcia, A. Nalli, A. Suarez-Gauthier, F. Lopez-Rios, X. F. Zhang, A. Solomon, J. Tong, et al. Antitumor activity and molecular effects of the novel heat shock protein 90 inhibitor, IPI-504, in pancreatic cancer Mol. Cancer Ther., October 1, 2008; 7(10): 3275 - 3284. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. S. Herbst, J. V. Heymach, and S. M. Lippman Lung Cancer N. Engl. J. Med., September 25, 2008; 359(13): 1367 - 1380. [Full Text] [PDF]
|
|
|
|
 |
|
 M. Skrzypski, E. Jassem, M. Taron, J. J. Sanchez, P. Mendez, W. Rzyman, G. Gulida, D. Raz, D. Jablons, M. Provencio, et al. Three-Gene Expression Signature Predicts Survival in Early-Stage Squamous Cell Carcinoma of the Lung Clin. Cancer Res., August 1, 2008; 14(15): 4794 - 4799. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Wilkening, B. Tavelin, F. Canzian, K. Enquist, R. Palmqvist, A. Altieri, G. Hallmans, K. Hemminki, P. Lenner, and A. Forsti Interleukin promoter polymorphisms and prognosis in colorectal cancer Carcinogenesis, June 1, 2008; 29(6): 1202 - 1206. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Badoual, G. Bouchaud, N. E. H. Agueznay, E. Mortier, S. Hans, A. Gey, F. Fernani, S. Peyrard, P. L. -Puig, P. Bruneval, et al. The Soluble {alpha} Chain of Interleukin-15 Receptor: A Proinflammatory Molecule Associated with Tumor Progression in Head and Neck Cancer Cancer Res., May 15, 2008; 68(10): 3907 - 3914. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Dubey and C. A. Powell Update in Lung Cancer 2007 Am. J. Respir. Crit. Care Med., May 1, 2008; 177(9): 941 - 946. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||