Whole Transcriptome Analysis of Pre-invasive and Invasive Early Squamous Lung Carcinoma in Archival Laser Microdissected Samples

Abstract

Background: Preinvasive squamous cell cancer (PSCC) are local transformations of bronchial epithelia that are frequently observed in current or former smokers. Their different grades and sizes suggest a continuum of dysplastic change with increasing severity, which may culminate in invasive squamous cell carcinoma (ISCC). As a consequence of the difficulty in isolating cancerous cells from biopsies, the molecular pathology that underlies their histological variability remains largely unknown.

Method: To address this issue, we have employed microdissection to isolate normal bronchial epithelia and cancerous cells from low- and high-grade PSCC and ISCC, from paraffin embedded (FFPE) biopsies and determined gene expression using Affymetric Human Exon 1.0 ST arrays. Tests for differential gene expression were performed using the Bioconductor package limma followed by functional analyses of differentially expressed genes in IPA.

Results: Examination of differential gene expression showed small differences between low- and high-grade PSCC but substantial changes between PSCC and ISCC samples (184 vs 1200 p-value <0.05, fc ±1.75). However, the majority of the differentially expressed PSCC genes (142 genes: 77%) were shared with those in ISCC samples. Pathway analysis showed that these shared genes are associated with DNA damage response, DNA/RNA metabolism and inflammation as major biological themes. Cluster analysis identified 12 distinct patterns of gene expression including progressive up or down-regulation across PSCC and ISCC. Pathway analysis of incrementally up-regulated genes revealed again significant enrichment of terms related to DNA damage response, DNA/RNA metabolism, inflammation, survival and proliferation. Altered expression of selected genes was confirmed using RT-PCR, as well as immunohistochemistry in an independent set of 45 ISCCs.

Conclusions: Gene expression profiles in PSCC and ISCC differ greatly in terms of numbers of genes with altered transcriptional activity. However, altered gene expression in PSCC affects canonical pathways and cellular and biological processes, such as inflammation and DNA damage response, which are highly consistent with hallmarks of cancer.

Keywords: Exon arrays; Gene expression profiling; Invasive squamous cell cancer; Microdissection; Preinvasive squamous cell cancer.

Fig. 1

Sample classification, general variability of normalized gene expression and differential gene expression in the selected FFPE PSCC and ISCC samples. a Sample classification according to independent reviews by three expert histopathologists. Each of the first three columns in the heatmap corresponds to one of the three involved pathologists (R1, R2 and R3) and rows to the selected samples. The color of each heatmap cell indicates the corresponding sample classification (see legend, white = no review for technical reasons). The FC column represents the final classification based on two concordant reviews. b Plot shows the result of a Principal component analysis (PCA) of normalized gene expression data obtained from all the included samples. ISCC and PSCC samples are clearly separated. Ellipses delineate 95% confidence level for each sample class respectively. c PCA plot of normalized gene expression data from low- and high-grade PSCC. Samples are widespread and show no clear separation. d PCA plot of normalized gene expression data from ISCC and high-grade PSCC. TN0, TNx and Nx ISCC samples are widespread and do not show recognizable differences. High-grade PSCC are clearly separated from ISCC samples. e Venn diagram illustrate the numbers of differentially expressed genes in ISCC and PSCC, including the intersection between both. f The Venn diagram represents the numbers of differentially expressed genes in high-grade and low-grade PSCC, including the intersection between both

Fig. 2

Positional enrichment and cluster analysis differentially expressed genes in PSCC and ISCC. a Genome wide positional enrichment of aberrant transcription in PSCC and ISCC. Tracks of the circos plot from outside to inside: chromosome ideogram (except Y chromosome), in which red lines demarcates the centromere; heatmap of enriched aberrant transcription in ISCC; heatmap of enriched aberrant transcription in ISCC. In both heatmaps gene up-regulation is coded in red and gene down-regulation in blue. b Cluster analysis of genes differentially expressed in low-grade PSCC, high-grade PSCC or ISCC; (i) heatmap of standardized gene expression; (ii) z-scores and (iii) histograms of standardized expression of the selected genes across normal bronchial epithelia (C), low-grade PSCC (L), high-grade PSCC (H) and ISCC (I); (iv) profiles of aberrant gene expression for the 12 identified clusters. The value in brackets represents the number of genes assigned to each of the 12 clusters

Fig. 3

Ingenuity pathway analysis (IPA) of aberrant transcription in PSCC and ISCC. (i) The first heatmap column shows all IPA pathways that are significantly affected in ISCC (p-value < =0.05). The remaining 3 columns indicate the association of these IPA pathways with differentially expressed genes in all PSCC, high-grade PSCC or low-grade PSCC. (ii) This heatmap summarise the connection between IPA pathways significantly related to gene expression changes in ISCC (p-value < =0.05) and genes sorted by cluster analysis. Note that genes from cluster 8 were not linked to any IPA pathways associated with ISCC

Fig. 4

Validation of the array data obtained from FFPE PSCC and ISCC samples. a The bar plot compares the fold changes of mRNA levels for four individual genes in ISCC tumours and paired controls obtained either by Human Exon arrays or RT-PCR. b Anti-PTTG1 staining in FFPE ISCC (Tumour) and paired bronchial mucosa (Normal). In ISCC anti-PTTG1 appears moderately in the cytoplasm and strongly in about ~70% of nuclei. The used ISCC is unrelated to the samples for the gene expression study. c H-scores of anti-PTTG1 immunostaining in a set of 45 independent ISCC (Tumour) and paired bronchial epithelia (Normal). Cytoplasmic and nuclear PTTG1 levels are significantly raised in ISCC and therefore cross-validate the gene expression data. Nuclear staining is increased 4x in ISCC whereas the cytoplasmic was raised 1.5x