Transcriptomic architecture of the adjacent airway field cancerization in non-small cell lung cancer

Abstract

Background: Earlier work identified specific tumor-promoting abnormalities that are shared between lung cancers and adjacent normal bronchial epithelia. We sought to characterize the yet unknown global molecular and adjacent airway field cancerization (FC) in early-stage non-small cell lung cancer (NSCLC).

Methods: Whole-transcriptome expression profiling of resected early-stage (I-IIIA) NSCLC specimens (n = 20) with matched tumors, multiple cytologically controlled normal airways with varying distances from tumors, and uninvolved normal lung tissues (n = 194 samples) was performed using the Affymetrix Human Gene 1.0 ST platform. Mixed-effects models were used to identify differentially expressed genes among groups. Ordinal regression analysis was performed to characterize site-dependent airway expression profiles. All statistical tests were two-sided, except where noted.

Results: We identified differentially expressed gene features (n = 1661) between NSCLCs and airways compared with normal lung tissues, a subset of which (n = 299), after gene set enrichment analysis, statistically significantly (P < .001) distinguished large airways in lung cancer patients from airways in cancer-free smokers. In addition, we identified genes (n = 422) statistically significantly and progressively differentially expressed in airways by distance from tumors that were found to be congruently modulated between NSCLCs and normal lung tissues. Furthermore, LAPTM4B, with statistically significantly increased expression (P < .05) in airways with shorter distance from tumors, was upregulated in human immortalized cells compared with normal bronchial epithelial cells (P < .001) and promoted anchorage-dependent and -independent lung cancer cell growth.

Conclusions: The adjacent airway FC comprises both site-independent profiles as well as gradient and localized airway expression patterns. Profiling of the airway FC may provide new insights into NSCLC oncogenesis and molecular tools for detection of the disease.

Figure 1.

Schematic of the transcriptomic analysis of the airway field cancerization (FC) in non–small cell lung cancer (NSCLC). FC samples comprised of matched lung tumors, multiple cytologically controlled normal airways with varying proximity from tumors, and uninvolved normal lung tissues were obtained from 20 resected early-stage (I–IIIA) NSCLC specimens. Gene expression profiling of all the samples (n = 194) was performed to characterize global FC profiles (differentially expressed in the same direction between both tumors and airways compared with normal lung tissues) in the normal-appearing airway adjacent to NSCLCs (site-independent analysis). Adjacent FC profiles were then analyzed by gene set enrichment analysis (GSEA) in a set of large airways of smokers (n = 129) with and without lung cancer (17) to identify FC profiles that can distinguish lung cancer among smokers. Global expression profiling was also used to delineate FC profiles that are modulated in the airway with respect to proximity from corresponding NSCLCs (site-dependent analysis). Quantitative real-time polymerase chain reaction (PCR) was used to confirm the differential expression of select airway FC markers.

Figure 2.

Identification of adjacent airway field cancerization (FC) profiles in non–small cell lung cancer (NSCLC). A) Hierarchical clustering of gene features statistically significantly differentially expressed between both NSCLCs and airways compared with normal lung tissues (n = 1661). Columns represent samples (n = 194 samples from 20 case patients), and rows represent gene features (red = upregulated; blue = downregulated). B) Functional pathways analysis using ingenuity pathways analysis (IPA) of the differentially expressed genes. Statistical significance of the identified overrepresented canonical pathways is indicated by the negative log of the P values. Functional pathways and interaction network analysis by IPA depicting the most statistically significantly (P < .001) modulated pathway (C) and the gene network with highest number of differentially expressed and closely related (G-protein coupled receptors) interacting genes (D); red = higher experession; green = lower expression. Genes selected for subsequent confirmation by quantitative real-time polymerase chain reaction are highlighted by black margins.

Figure 3.

Analysis of adjacent airway field cancerization (FC) profiles in large airways of patients with suspected lung cancer. Gene set enrichment analysis (GSEA) was performed, as described in the Supplementary Methods (available online), to identify genes that were upregulated and downregulated in the adjacent airway FC that were also concordantly enriched between genes differentially expressed between histologically normal airways of smokers with and without lung cancer (17). Hierarchical cluster analysis using the 299 leading edge genes (n = 59 upregulated and 240 downregulated) was performed side by side in the adjacent airway FC (A) and in airways of smokers with and without lung cancer (B). Columns represent samples, and rows represent gene features (red = upregulated; blue = downregulated). P value showing statistical significance of separate clustering of airways of smokers with lung cancer from those of healthy smokers was obtained by the two-sided Fisher exact test.

Figure 4.

Analysis of airway expression profiles by distance from corresponding non–small cell lung cancers (NSCLCs). Ordinal logistic regression analysis of airways was performed, as described in the Supplementary Methods (available online), and identified 422 gene features with statistically differential expression in the airway with respect to tumor proximity (false discovery rate < 5%). A) Clustering analysis was performed as described in the Supplementary Sweave Reports (available online), and airway samples were arranged by difference in the expression values of the site-dependent genes between the upregulated and downregulated gene clusters. Columns represent samples, and rows represent gene features (red = upregulated; blue = downregulated). B) Functional pathways analysis by ingenuity pathways analysis of the site-dependent differentially expressed genes. Statistical significance of the identified overrepresented canonical pathways is indicated by the negative log of the P values. The site-dependent effect in the adjacent field cancerization (FC) was quantified as described in the Supplementary Methods and Sweave Reports (available online). Box plots depicting site-dependent FC score in airways (C) and between corresponding paired NSCLCs and normal lung tissue after statistical analysis by one-sided t tests (D). Heavy lines indicate medians, and whiskers indicate maximum and minimum FC scores. Airway distance from tumors is numerically indicated with a range of 1 (closest) to 5 (farthest).

Figure 5.

Quantitative real-time polymerase chain reaction (QRTPCR) analysis of airway field cancerization (FC) markers. Expression of TGFBR2 (A), VIPR1 (B), NETO2 (C), and LAPTM4B (D) was analyzed by QRTPCR in 18 of 20 NSCLC FC case subjects studied with sufficient RNA from airway samples left over after expression profiling. Expression of the indicated genes is depicted by group (NSCLCs, airways, and normal lung tissues; upper panels) and across airway samples based on distance from corresponding NSCLCs (1 = airway closest to tumors; 5 = airways relatively farthest from tumors; middle panels). Relative mRNA expression was assessed by QRTPCR, normalized to that of TBP, and quantified using the 2^-ΔΔCT relative quantification method as detailed in the Supplementary Methods (available online). PCR reactions for each FC sample were carried out in duplicate. Boxes indicate ± standard error of the mean; error bars indicate standard deviation. Statistical significance of differences in expression among NSCLC, airway, and normal lung groups was assessed by the Kruskal–Wallis test (upper panels), and statistical significance of differences among different airways was assessed by analysis of variance (middle lanes). Correlation between expression of the indicated genes quantified by microarray and QRTPCR analyses was statistically assessed by Spearman rank (lower panels). LAPTM4B = lysosomal protein transmembrane 4 beta; NETO2 = neuropilin (NRP) and tolloid (TLL)-like 2; TGFBR2 = transforming growth factor beta receptor II; VIPR1 = vasoactive intestinal peptide receptor 1; TBP = TATA box binding protein. *P < .05; ** P < .001; NS, not significant. All statistical tests were two-sided.

Figure 6.

Effect of RNA interference-mediated knockdown of LAPTM4B on lung cancer cell anchorage-dependent and -independent growth and colony formation. Calu-6 lung cancer cells were stably transfected with empty vectors, vectors containing scrambled short hairpin RNA (shRNA), as well as vectors with LAPTM4B-specific shRNA sequences as described in the Methods section. A) Quantitative real-time polymerase chain reaction analysis depicting statistically significantly reduced LAPTM4B relative expression in two Calu-6 sublines (shLAPTM4B-1 and shLAPTM4B-2) stably transfected with two different LAPTM4B-specific shRNA sequences compared with Calu-6 cells stably transfected with empty and scrambled shRNA-containing vectors. B) The indicated stably transfected Calu-6 cells were seeded in triplicates (5×10⁴ cells/well) in 12-well plates for 72 hours, after which the number of cells in each well was computed by the trypan blue exclusion method. Cells were seeded in triplicates at a seeding density of 250 cells per well in six-well culture plates or 150 cells per well on soft agar for assessment of anchorage-dependent (C) and -independent (D) colony formation, respectively. Cell colonies were then quantified as described in the Methods section. All experiments were done in triplicate. Representative images of cell colonies are depicted in the lower panels and were obtained with a phase-light microscope. Error bars indicate standard deviation. *P < .05; **P < .001 by the two-sided Student t test.