Article Figures & Data
Figures
- Figure 1.
Schematic diagram of genomic and transcriptional analyses performed in the study. A, genomic analyses. A 1,398-sample cohort was assembled, from which smoking-related CNAs were identified. These alterations together with reported smoking-related signatures were used in supervised classification analyses to assess the predictive power in classification of smoking history. B, transcriptional analyses. An 841-sample gene expression discovery cohort was used to search for gene signatures able to predict smoking history through supervised classification analysis. Moreover, the discovery cohorts together with the TCGA validation cohort was used in unsupervised class discovery to determine the impact of patient smoking status on the global transcriptional landscape, and the relationship of smoking status with transcriptional subgroups in lung adenocarcinoma.
- Figure 2.
CNAs in adenocarcinoma stratified by smoking history. A, pattern of gross CNAs in the 1,398-sample cohort measured as fraction of the genome altered by copy-number gain or loss (CN-FGA) in never-smokers versus smokers, and never-smokers/current smokers/former smokers. B, CN-FGA for individual cohorts in the 1,398-sample cohort (see Supplementary Table S1) stratified into never-smokers or smokers, showing differences between individual cohorts in which group displayed most CNAs. P values were calculated using the Student t test, requiring ≥4 patients in each tested group. CLCGP: The Clinical Lung Cancer Genome Project.
- Figure 3.
Supervised classification of smoking status based on CNAs and transcriptional patterns. A, results of genomic classification based on regions from Massion et al. (15), Thu et al. (12), Weir et al. (10), Broet et al. (11), Planck et al. (29), and significant CNAs between never-smokers/current smokers/former smokers in the current study (Table 2) for prediction of never-smokers/smokers (NS/S), never-smokers/current smokers (NS/CS), and never-smokers/former smokers (NS/FS). Only PAM-based models showed as these had the best performance. Each combination was repeated up to 10 times with different 50%/50% training and test sample cohort compositions to obtain an average balanced accuracy across test sets (bars) together with an SD estimate. B, classification of never-smokers and smokers based on transcriptional patterns. Bars, the mean balanced accuracy with SD in the test sets for each training set (x axis) across all 34 investigated models. In total, 170 classifier tests were made for each training cohort (34 models applied to 5 test sets) displayed as individual points. C, classification of never-smokers and current smokers based on transcriptional patterns, displayed as in B. For each training set (x-axis), 29 models were trained and applied to three test sets totaling to 87 tests per training cohort (points).
- Figure 4.
Transcriptional patterns in adenocarcinomas stratified by smoking status. A, consensus clustering was performed in six adenocarcinoma cohorts (k = 3 clusters, expression SD >0.5 as prefilter). For each cohort, the cluster with the highest number of never-smokers was identified (never-smoker–enriched cluster). Next, all cohorts were pooled to a meta-cohort (n = 841 samples). The heatmap shows mean z-score transformed values for different features (rows) for respective group (columns). The z-score transformation allows a common heatmap scale to be applied to all features. Red, higher expression/frequency values of a feature; blue, lower values. The heatmap shows the consistency between never-smokers (NS) and smokers (S) within non–never-smoker-enriched or never-smoker–enriched clusters, and the strong differences between never-smoker–enriched and non–never-smoker-enriched cases for the different features. P values computed using Wilcoxon or Fisher test, referring to group comparisons of all samples, never-smokers only, smokers only. ns, nonsignificant. B, overall survival (OS) (censored at 5 years) for smokers (top) and never-smokers (bottom) in never-smoker–enriched and non–never-smoker-enriched clusters from A. C, characterization of consensus clusters from analysis of 435 TCGA cases. Cluster 2 represents the never-smoker–enriched cluster (see also Supplementary Fig. S3). Heatmap shows mean z-score values for different features (rows) for respective group (columns) as in A. For mutations, the percentage of mutated cases in each group is shown. Oncogene drivers include mutations in EGFR, KRAS, ERBB2, BRAF, and gene fusions involving ALK, RET, and ROS1. Percentage of amplified cases refers to number of cases in each group with >1 high-level amplification in the focal CNA regions reported by Planck et al. (29). P values calculated using the Fisher or Kruskal–Wallis test similar to A. D, overall survival for all TCGA patients in consensus clusters (top) and smokers specifically (bottom) in clusters from C. For never-smokers, CCL2 patients showed borderline nonsignificant association with better overall survival (log-rank P = 0.07).
Tables
- Table 1.
Clinical characteristics of patients with adenocarcinoma in genomic and gene expression cohorts
Genomic cohort Okayama et al. (27) Chitale U133A (32) Chitale U133 2plus (32) Fouret et al. (13) Landi et al. (18) Shedden et al. (31) TCGA Der et al. (33) Tarca et al. (34) Usage Discovery Discovery Discovery Discovery Discovery Discovery Discovery Validation Validation Validation Data type Copy number Expression Expression Expression Expression Expression Expression Expression Expression Expression Total number of patients 1,398 226 91 102 103 58 356 435 115 70 Smoking history 1,398 226 90 102 103 58 262 423 115 70 Never-smokers 277 115 17 19 63 16 33 65 23 19 Smokers 1,121 111 73 83 40 42 229 358 92 51 Current smokers 391 — 13 12 — 24 20 102 36 40 Former smokers 567 — 60 71 — 18 209 256 56 11 Pack-years (median) 40 — 37 34 — — — 40 — — Heavy smokers (%)b 22% — 21% 28% — — — 17% — — Gender Male/female 586/695 105/121 41/50 42/60 15/84 35/23 189/166 201/234 59/56 48/22 Mutation status EGFR-mutated 205 127 15 24 49 — — a — — KRAS-mutated 327 20 11 36 17 — — a — — EGFRwt and KRASwt 564 79 65 42 33 — — a — — Stage I 739 168 53 70 57 22 224 238 83 37 II 235 58 20 10 10 21 77 100 32 33 III 272 0 18 17 32 12 51 74 — — IV 66 0 0 5 0 3 0 22 — — Platform — Affymetrix U133 2plus Affymetrix U133A Affymetrix U133 2plus Affymetrix U133 2plus Affymetrix U133 2plus Affymetrix U133A RNAseq Affymetrix U133 2plus Affymetrix U133 2plus - Table 2.
Smoking-related CNAs with >20% frequency difference between never-smokers, current smokers, and former smokers
Type Cytoband Regiona Size (Mbp) Number of genes Most altered group Focal CNAs (29) Gain 5q31.3-q32 chr5:142071001-145089001 3.02 3 Never-smoker Gain 5q33.1-q35.3 chr5:147723001-180711001 32.99 201 Never-smoker Amp_5q35.1 Gain 8q13.1-q13.2 chr8:67043001-69050001 2.01 15 Current Gain 8q13.3 chr8:72872001-73673001 0.8 3 Current Gain 8q21.11-q21.12 chr8:76325001-80141001 3.82 6 Current Gain 8q21.13-q21.3 chr8:81953001-92345001 10.39 38 Current Amp_8q21.13 Gain 8q21.3-q22.1 chr8:93353001-94154001 0.8 0 Current Gain 16p13.3-p12.1 chr16:1-24042000 24.04 225 Never-smoker Amp_16p13.13 Loss 5q12.1-q13.2 chr5:62799001-73107001 10.31 38 Current Loss 5q13.3-q35.3 chr5:76125001-180711001 104.59 559 Current Del_5q14.3 Loss 19p13.2-p12 chr19:7095000-20985000 13.89 346 Current Del_19p13.3-p13.2 Loss 22q13.1-q13.33 chr22:37498001-49690001 12.19 132 Current Del_22q13.31-q13.32 -
↵ahg18 coordinates.
-
Supplementary Methods, Figures 1 - 6, Tables 1 - 3
Files in this Data Supplement:
- Data Supplement - Supplementary Figure S1. Frequency of copy number alterations in lung adenocarcinoma groups defined by patient smoking history and genomic principal component analysis.
- Data Supplement - Supplementary Figure S2. Detailed results of supervised classification of smoking status in adenocarcinoma gene expression microarray cohorts.
- Data Supplement - Supplementary Figure S3. Distribution of never-smokers and smokers in unsupervised consensus clusters from adenocarcinoma gene expression microarray cohorts.
- Data Supplement - Supplementary Figure S4. Molecular and clinicopathological characteristics of adenocarcinoma consensus clusters in microarray and TCGA cohorts with respect to patient smoking history.
- Data Supplement - Supplementary Figure S5. Principal component analysis of the TCGA and Chitale et al. transcriptional cohorts showing the variance in gene expression associated with clinicopathological factors and gene expression phenotypes.
- Data Supplement - Supplementary Figure S6. Expression of proliferation-associated genes in adenocarcinoma subgroups defined by clinical smoking history or gene expression phenotypes (bronchioid, magnoid, squamoid), and the association of these phenotypes with smoking history.
- Data Supplement - Detailed description of data processing steps including mutation analyses, detection of CNAs, unsupervised transcriptional analyses, and supervised genomic and transcriptional analyses.
- Data Supplement - Legends for supplementary figures and tables.
- Data Supplement - Supplementary Table S1. Clinicopathological characteristics of individual adenocarcinoma cohorts in the combined 1398-sample genomic cohort.
- Data Supplement - Supplementary Table S2. Detailed clinicopathological characteristics of the 1398-sample genomic cohort and all transcriptional adenocarcinoma cohorts with respect to patient smoking history.
- Data Supplement - Supplementary Table S3. Smoking-related copy number alterations from genome-wide analysis in lung adenocarcinoma.
Volume 20, Issue 18
