Purpose/Objectives. Primary small cell esophageal carcinoma (SCEC) represents a rare and aggressive malignancy without any prospective clinical trial or established treatment strategy at present. Although previous studies have indicated similarities between SCEC and small cell lung cancer (SCLC) in terms of their clinical manifestations, pathology, and morphology, very little genetic information is available on this highly malignant tumor. At present, patients with SCEC are staged and treated according to the guidelines established for SCLC. However, early recurrence and distant metastasis are common, and long-time survivors are rare. Current options available for patients with relapsed SCEC are fairly unsatisfactory, and their prognosis is generally poor. Novel therapeutic approaches against SCEC are therefore urgently needed and require a deeper understanding of the underlying genetic mechanisms. The current investigation aims to characterize the gene expression profile and copy number variations (CNVs) in SCEC to clarify molecular markers and pathways that may possess clinical significance. Materials/Methods. De novo expression array was carried out on three matched sets of primary SCEC and adjacent normal tissue samples procured from the institutional tissue bank, utilizing the Affymetrix HG U133 Plus 2.0 Array. After individual tissue normalization, the statistical software GeneSpring GX 12.5 was used to determine differentially expressed genes (DEGs) in the tumors relative to their paired normal tissues. Gene enrichments in addition to functional annotation and gene interaction networks were performed using DAVID 6.8 and STRING 10.0, respectively. A gene alteration was determined to be recurrent if it was observed in at least 2 samples. Chromosomes X and Y were not included in calculations as gender differences are a known source of analysis bias. The DEGs of at least one SCEC sample could be mapped to the CNV regions (fold change (FC) ≥ 2 and false discovery rate (FDR) < 0.01) after gene expression profiling by RefSeq Transcript ID. These overlapped genes were subjected to the functional annotation using DAVID 6.8. In order to elucidate the effect of CNV on mRNA expression, we integrated the genome-wide copy number data and gene expression in 3 paired samples. CNV-associated gene expression aberration (CNV-FC) was calculated for the recurrent DEGs using previously published integrated microarray data as reference. Pearson’s correlation coefficient was employed to determine if there was a statistical correlation between the gene expression log2 ratios and their copy numbers using the SPSS 19.0 software. Genes that possessed CNV-FC ≥ 2 and r≥0.6 (p<0.05) were determined to be genes potentially associated with cancer. Results. High-quality DNA and total RNA were first extracted from both SCEC and normal tissues. Microarray data showed significant upregulation in WNT gene sets and downregulation in the PTEN and notch gene sets in SCEC. Functional annotation showed that genes associated with DNA replication, mitosis, cell cycle, DNA repair, telomere maintenance, RB, and p53 pathways were significantly altered in SCEC compared to corresponding noncancerous tissues (Benjamini p<0.05). Thirteen recurrent CNVs were found in all SCEC samples by array CGH. Chromosomal regions with gain were located in 14q11.2, and regions with loss were located in 4q22.3-23.3, 3q25.31-q29, 5p15.31-15.2, 8q21.11-24.3, and 9p23-13.1 in all samples. In two samples, the 14q11.2-32.33 region was amplified, whereas 3p26.3-25.3, 4p16.3-11, 4q11-22.3, 4q23-25, 8p23.3, and 16p13.3 were deleted. We further identified 306 genes that consistently differed in copy number and expression (194 upregulated and 112 downregulated) between the SCEC and noncancerous tissues in all three samples. These genes were significantly enriched with those involved in cell cycle, mitosis, DNA repair, P53 pathway, and RB pathway, according to their functional annotation. These 306 DEGs also included network genes of the above pathways such as NUF2, CCNE2, NFIB, ETV5, KLF5, ATAD2, NDC80, and ZWINT. In addition, 39 individual DEGs demonstrated a minimum 2-fold copy number-associated expression change (median: 5.35, 95% CI: 4.53–16.98) and Pearson’s correlation coefficient ≥ 0.6 (p<0.05), of which PTP4A3 showed the highest correlation (CNV-FC = 21362.13; Pearson’s correlation coefficient = 0.9983; p=0.037). Two distinct groups of genes belonging to each SCEC and nonmalignant tissues were observed upon unsupervised two-way (genes and samples) hierarchical clustering. Conclusions. The current investigation is the first to produce data regarding the genomic signature of SCEC at the transcription level and in relation to CNVs. Our preliminary data indicate possible key roles of WNT and notch signaling in SCEC and overexpressed PTP4A3 as a potential therapeutic target. Further validation of our findings is warranted.
An Improved Pearson’s Correlation Proximity-Based Hierarchical Clustering for Mining Biological Association between Genes