Gene Expression Profile And Copy Number Variations Of Small Cell Esophageal Carcinoma:an Bioinformatics Analyses

Purpose/Objectives:Primary small cell esophageal carcinoma(SCEC) is a rare, aggressive malignancy without any prospective clinical trial or established treatment strategy at present time. Although previous studies suggested that there were similarities between SCEC and small cell lung cancer(SCLC) in clinical manifestation and pathological morphology, genetic studies on this highly malignant tumour remained sparse. Presently, patients with SCEC were staged and treated according to the guidelines for SCLC, however, early recurrence or distant metastasis was common but long-time survivor was rare. The efficacy of second line treatment for SCEC patients with relapsed disease was fairly unsatisfactory, thus their prognosis was generally poor. Innovative treatment based on improved understanding of the genetic alterations of SCEC is urgently awaited. This study was designed to investigate the gene expression profile and copy number variations(CNVs) of SCEC, and compare it with the known data of SCLC and esophageal adenocarcinoma (EAC)/esophageal squamous carcinoma(ESCC).Materials/Methods:De novo expression array analysis was performed on three pairs of frozen tissues of primary SCEC and corresponding normal samples from institutional tissue bank using the Affymetrix HG U133 Plus 2.0 Array. These data were complemented with public domain expression data sets from the GEO repository using the same working platforms during the last 3 years, which included primary SCLC, primary EAC/ESCC, and normal lung/esophagus specimens(series GSE30219, GSE26886). After individual normalization, primary tumors were submitted to statistical analysis(GeneSpring GX 12.5) for identifying differentially expressed genes(DEGs) as compared with their paired normal tissues. Gene enrichments with functions and genes interactions were analyzed by DAVID 6.7 and STRING 9.0, respectively.The research of CNVs in SCEC was realized by array-based Comparative Genomic Hybridization(aCGH) via Agilent Human aCGH 4x180k Array platform. Quanlity control and putative CNV intervals identification in each sample were performed by CytoGenomics software 2.7.8.0. An alteration was defined as recurrent if it was present in at least 2 samples. Due to the gender differences between the arrays that could cause bias in the analysis, chromosomes X and Y were excluded from the calculation. The genes located in CNV regions of at least 1 SCEC samples were connected with DEGs[false discovery rate(FDR)<0.01 and fold change(FC)≥2] in gene expression profiling by Refseq Transcript ID. These overlapped genes were subjected to the functional annotation using DAVID 6.7. In order to elucidate the effect of CNV to mRNA expression, we performed integration of genome-wide DNA copy number and gene expression data in these 3 SCEC sample pairs. Copy number associated aberration in gene expression(CNV-FC) for recurrent genes were computed consulting to previously published studies on integrated microarray analysis. Further, Pearson correlation coefficients between copy number log2ratios and expression log2ratios for these genes were calculated in SPSS 19.0 software. The genes with CNV-FC≥ 2 meanwhile r≥0.6(p<0.05) were selected as possibly cancer-associated genes. Finally, we compared the CNVs and their involved genes of SCEC with SCLC, ESCC refer to published papers.Result:The quality of total RNA, DNA extracted from 3 SCEC and the corresponding adjacent normal samples was high. The expression data demonstrated that SCEC had more DEGs in common with SCLC than EC (829 vs 450; FDR<0.01 and FC≥2), leading to a greater correlation between SCEC and SCLC (Pearson correlation coefficient was 0.60 for SCEC vs SCLC,0.51 or 0.45 for SCEC vs ESCC or EAC, the coefficient was 0.73 for ESCC vs EAC). Similar findings were obtained by Principal Component Analysis (PCA) using all DEGs retrieved from these four groups. Functional annotation showed that a higher proportion of biological processes and pathways were in common between SCEC and SCLC, which were associated with cell cycle, mitosis, DNA replication, telomere maintenance, DNA repair, P53 and RB pathway (Benjamini p<0.05). Comparing with EAC/ESCC, SCEC shared more co-up regulated DEGs coding for the aforementioned common pathways with SCLC (584 vs 155). In addition, SCEC and SCLC possessed overlapped gene interactive network with CENPF, NEK2, KIF11, TMPO, FOXM1 as common skeleton centred by NUF2. The genes involved in the SCEC regulated network related to cell cycle, mitosis, cell cycle checkpoint, spindle organization, microtubule binding, cytoskeletal protein binding and other biological process.Thirteen recurrent CNVs were found by aCGH from the 3 pairs of SCEC samples. The gained region detected in all samples were located at 14q11.2, the region of loss detected in all samples were located at 4q22.3-23.3; 3q25.31-q29,5p15.31-15.2,8q21.11-24.3, 9p23-13.1.14q11.2-32.33 were detected with amplification in 2 samples, whereas 3p26.3-25.3,4p16.3-11,4q11-22.3,4q23-25,8p23.3,16p13.3 were deleted in 2 samples. We further identified 306 genes changed consistently between copy number and mRNA expression(194 up-regulated and 112 down-regulated). These genes significantly enriched in cell cycle, mitosis, DNA repair, P53 pathway and RB pathway, in accord with the functional annotation outcome of gene expression profiling. Importantly, most of the network genes in the gene expression profiling were included in these 306 genes, such as NUF2, CCNE2, NFIB, ETV5, KLF5, ATAD2, NDC80, ZWINT. There were 39 individual genes which were both at least 2-fold copy number associated alteration in their expression (median 5.35,95% CI:4.53-16.98) and Pearson correlation coefficient≥0.6(p<0.05). An unsupervised two-way(genes and samples) hierarchical clustering of 3 pairs of SCEC samples based on these genes revealed two distinct clusters separating SCEC samples from adjacent noncancerous samples. Of these, the gene representing the highest correlation was PTP4A3(CNV-FC=21362.13), and the Pearson correlation coefficient was 0.9983(p=0.037). Lacking aCGH data of SCLC and ESCC using the same platform as our study to date, we compared the CNVs and their involved genes of SCEC with acknowledged, recurrent CNVs and related genes of SCLC and ESCC from the published data. SCEC and SCLC shared more possibly cancer-associated genes with similar frequency than it shared with ESCC.Conclusions:This study was the first attempt to examine the genomic signatures of SCEC at the transcription level and from the CNVs perspective, with comparison to SCLC or EAC/ESCC. Our preliminary data indicated that SCEC and SCLC display notably similar patterns of gene expression and CNVs on cell cycle, mitosis, DNA replication, DNA repair and P53, RB pathway, suggesting that the two diseases might be regulated by similar transcriptional networks. Amplification and overexpression of NUF2 and PTP4A3 might play a key role in carcinogenesis and metastasis of SCEC. Further validation study is warranted.