Co-regulation analysis of closely linked genes identifies a highly recurrent gain on chromosome 17q25.3 in prostate cancer

Abstract

Background: Transcriptional profiling of prostate cancer (PC) has unveiled new markers of neoplasia and allowed insights into mechanisms underlying this disease. Genomewide analyses have also identified new chromosomal abnormalities associated with PC. The combination of both classes of data for the same sample cohort might provide better criteria for identifying relevant factors involved in neoplasia. Here we describe transcriptional signatures identifying distinct normal and tumoral prostate tissue compartments, and the inference and demonstration of a new, highly recurrent copy number gain on chromosome 17q25.3.

Methods: We have applied transcriptional profiling to tumoral and non-tumoral prostate samples with relatively homogeneous epithelial representations as well as pure stromal tissue from peripheral prostate and cultured cell lines, followed by quantitative RT-PCR validations and immunohistochemical analysis. In addition, we have performed in silico colocalization analysis of co-regulated genes and validation by fluorescent in situ hybridization (FISH).

Results: The transcriptomic analysis has allowed us to identify signatures corresponding to non-tumoral luminal and tumoral epithelium, basal epithelial cells, and prostate stromal tissue. In addition, in silico analysis of co-regulated expression of physically linked genes has allowed us to predict the occurrence of a copy number gain at chromosomal region 17q25.3. This computational inference was validated by fluorescent in situ hybridization, which showed gains in this region in over 65% of primary and metastatic tumoral samples.

Conclusion: Our approach permits to directly link gene copy number variations with transcript co-regulation in association with neoplastic states. Therefore, transcriptomic studies of carefully selected samples can unveil new diagnostic markers and transcriptional signatures highly specific of PC, and lead to the discovery of novel genomic abnormalities that may provide additional insights into the causes and mechanisms of prostate cancer.

Figure 1

Graphical representation of microarray data analysis. A) Unrooted dendrogram showing the clustering of samples into four different classes: normal samples (blue), tumoral samples (red), stromal samples (yellow) and cultured cell lines (green). B) Unsupervised hierarchical cluster representation using microarray data for the 318 FADA-selected genes. T, tumoral samples; N, normal samples; S, pure stromal samples; CL, cell lines.

Figure 2

Quantitative RT-PCR validation of selected genes. A) Hierarchical clusters built using microarray data corresponding to the 45 genes selected for Q-PCR validation. T, N, S and CL are as described in Fig. 1. B) Hierarchical sample and gene cluster generated with Q-PCR data for genes showing concordance with microarray data (either upregulated or downregulated in both determinations) in 4 normal-tumoral non-microdissected matched samples and 7 normal-tumoral microdissected matched samples. C) Graphical representation of the Q-PCR tumoral/normal ratios for the 26 genes showing the best concordance with microarray data (9 overexpressed and 17 underexpressed in tumoral vs. normal samples). Shown are the averages for each of the selected genes of the Q-PCR tumoral/normal ratios for microdissected epithelia (light gray bars) and for whole, non-microdissected tissues (dark gray bars).

Figure 3

Immunohistochemical analysis of selected markers. A, D): Overexpression of myosin VI (MYO6) in tumoral (t) and PIN glands, as compared to normal (n) glands (inset, higher magnification showing a luminal reinforcement). B, E): Overexpression of multidrug resistance-associated protein 4 (ABCC4) in tumoral and PIN glands (inset in B, higher magnification showing a membrane staining pattern). C) Ephrin type-A receptor 2 precursor (EPHA2) was not detected in tumoral glands, while it was expressed in a subpopulation of basal cells in normal glands (n). F) Higher magnification showing a cytoplasmic and membrane staining pattern of EPHA2. Bars correspond to 100 μm.

Figure 4

Identification of a highly recurrent chromosomal gain on 17q25.3 in prostate tumors by transcript co-regulation analysis. A) Ideogram of chromosome 17 detailing the genes located on 17q25.3 with a corregulated overexpression in prostate cancer, as determined by our microarray analysis (red: genes selected by FADA as overexpressed in tumor samples; black: genes not selected by FADA but present in the Human Genome Focus microarrays; grey: genes not present in Human Genome Focus microarrays). B) Heat map showing the relative expression levels of the genes on 17q25.3 shown in (A). The order of the genes is from centromeric (top) to telomeric (bottom). Fluorescent in situ hybridization of prostate tissue samples (C, D and E) hybridized with the centromeric CEP17 probe (green) and the 17q25.3-specific BAC clone RP11-165M24 (red). Selected representative regions of prostate carcinoma (C) and PIN (D) areas from the same sample, and a lymph node metastasis (E), illustrating copy number gains in 17q25.3. The figures shown are also representative of samples with a 17q25.3 gain.