Gene Signatures Stratify Computed Tomography Screening Detected Lung Cancer in High-Risk Populations

Abstract

Background: Although screening programmes of smokers have detected resectable early lung cancers more frequently than expected, their efficacy in reducing mortality remains debatable. To elucidate the biological features of computed tomography (CT) screening detected lung cancer, we examined the mRNA signatures on tumours according to the year of detection, stage and survival.

Methods: Gene expression profiles were analysed on 28 patients (INT-IEO training cohort) and 24 patients of Multicentre Italian Lung Detection (MILD validation cohort). The gene signatures generated from the training set were validated on the MILD set and a public deposited DNA microarray data set (GSE11969). Expression of selected genes and proteins was validated by real-time RT-PCR and immunohistochemistry. Enriched core pathway and pathway networks were explored by GeneSpring GX10.

Findings: A 239-gene signature was identified according to the year of tumour detection in the training INT-IEO set and correlated with the patients' outcomes. These signatures divided the MILD patients into two distinct survival groups independently of tumour stage, size, histopathological type and screening year. The signatures can also predict survival in the clinically detected cancers (GSE11969). Pathway analyses revealed tumours detected in later years enrichment of the PI3K/PTEN/AKT pathway, with up-regulation of PDPK1, ITGB1 and down-regulation of FOXO1A. Analysis of normal lung tissue from INT-IEO cohort produced signatures distinguishing patients with early from late detected tumours.

Interpretation: The distinct pattern of "indolent" and "aggressive" tumour exists in CT-screening detected lung cancer according to the gene expression profiles. The early development of an aggressive phenotype may account for the lack of mortality reduction by screening observed in some cohorts.

Keywords: Aggressive; CT screening; Gene signature; Indolent; Lung cancer; PI3K/PTEN/AKT signalling pathway.

Fig. 1

Unsupervised hierarchical clustering of 17 tumours detected in years 1 and 2 (T) and 11 cases detected in years 3, 4 and 5 (T*) using 239 differentially expressed genes.

Fig. 2

Survival analysis of 24 MILD patients grouped according to the signature of CT-year of screening. These survival curves are based on (a) overall survival analysis and (b) disease free survival analysis. The (two-sided) p value is from by Log-rank (Mantel–Cox) test.

Fig. 3

Overall survival analysis of in silico data (GSE11969) considering (a) all the 79 patients or (b) the 40 stage I alone, grouped according to the signature of CT-year of screening. The (two-sided) p value is from by Log-rank (Mantel–Cox) test.

Fig. 4

The 239-gene of differentially expressed genes according to the CT screening year were imported into the GeneSpring GX10 for searching the common regulators of these genes. The connection between these genes was built up and unlinked nodes (genes) were removed. Blue lines and squares signify that a defined regulatory relationship exits between genes. Grey lines and squares signify that a putative regulatory relationship between genes has been identified but not biochemically defined. +, positive regulation; −, negative regulation.

Fig. 5

Pathway enrichment analysis reveals PI3K/PTEN/AKT signalling and apoptosis pathway as the most significantly related pathway. ITGB1 has higher level of expression in the late year tumours while FOXO1A has higher levels in the tumours detected at first two years.