Original Article Xiao-qu Zhu1, Expression profiling based on Mei-lan Hu2, Feng Zhang3, graph-clustering approach to determine Yu Tao 4, Chun-ming Wu1, Shang-zhu Lin1, colon cancer pathway Fu-le He2 1Department of ABSTRACT Gastroenterology and Hepatology, Wenzhou Context: Colorectal cancer is the second leading cause of cancer deaths worldwide. DNA microarray-based technologies allow Hospital of Traditional simultaneous analysis of expression of thousands of genes. Chinese Medicine, 27 Dashimen Xinhe Road, Aim: To search for important molecular markers and pathways that hold great promise for further treatment of patients with colorectal Wenzhou, cancer. Zhejiang 325000, 2 Materials and Methods: Here, we performed a comprehensive gene-level assessment of colorectal cancer using 35 colorectal cancer Department of Traditional Chinese and 24 normal samples. Medicine, Hangzhou Results: It was shown that AURKA, MT1G, and AKAP12 had a high degree of response in colorectal cancer. Besides, we further First People’s Hospital, 261 Huansha Road, explored the underlying molecular mechanism within these different genes. Hangzhou, Conclusions: The results indicated calcium signaling pathway and vascular smooth muscle contraction pathway were the two significant Zhejiang 310006, 3 pathways, giving hope to provide insights into the development of novel therapeutic targets and pathways. Postgraduate student of 2011 grade, The First Clinical Medical College of KEY WORDS: Colon cancer, expression profiles, graph cluster, significant pathways Zhejiang Chinese Medical University, 548 Binwen Road, Hangzhou, INTRODUCTION main challenge is to determine which treatment is Zhejiang 310053, most likely to benefit an individual patient. 4Department of Colorectal cancer (CRC) is one of the most frequent Oncology, Wenzhou Hospital of Traditional malignancies in western countries and the third DNA microarray technology offers the ability to Chinese Medicine, most common cause of cancer-related deaths compare gene expression at a genome-wide level 27 Dashimen Xinhe worldwide, despite remarkable progress being made and to explore the transcriptional programs that Road, Wenzhou, Zhejiang 325000, [1] are turned on or off in tumors during progression in surgical techniques and therapeutic options. China from normal through premalignant stages to [6] CRC arises as a consequence of the accumulation cancer. However, extracting such gene sets For correspondence: of genetic and epigenetic alterations.[2] Till date, information from large data sets derived from Dr. Fu-le He, heterogeneous biological samples has proven to Department of inactivation of the tumor suppressor genes, Traditional Chinese Adenomatous polyposis coli (APC) and p53, and be difficult. Pathway analysis programs such as Medicine, Hangzhou graph clustering can be of help.[7] Therefore, in activation of the oncogene, Kirsten-ras (K-ras), are First People’s Hospital, our study, we aimed to yield sets of significant, 261 Huansha Road, thought to be particularly important determinants differentially expressed genes (DEGs) by DNA Hangzhou, of colorectal tumor initiation and progression.[3,4] Zhejiang 310006, microarray analysis. From these confirmed gene Besides, CRC development is also an epigenetic China. sets, relevant pathways’ networks could be E-mail: hfl2000@sina. gene inactivation mechanism by DNA methylation reconstructed, ultimately leading to a more com of its promoter region. CpG island methylation reliable understanding of the underlying biology Access this article online affects a number of genes in colon cancer, and the mechanism in CRC.[8] significance of the epigenetic alterations in the Website: www.cancerjournal.net DOI: 10.4103/0973-1482.119351 pathogenesis of colon cancer has been reported MATERIALS AND METHODS PMID: **** widely, such as the netrin-1 receptors are aberrantly Quick Response Code: methylated in primary CRC and are significantly Microarray analysis was performed between 35 correlated with Dukes’ stage C.[5] CRC and 24 normal samples to identify differential genes. The microarray expression data can be Numerous new therapies hold great promise for accessed through the Gene Expression Omnibus the treatment of patients with brain cancer, but the under accession number GSE23878 (http://www. Journal of Cancer Research and Therapeutics - July-September 2013 - Volume 9 - Issue 3 467 Zhu, et al.: Expression profiling determines colon cancer pathway ncbi.nlm.nih.gov/geo/) which is based on the Affymetrix Human Genome U133 Plus 2.0 Array. For the GSE23878 dataset, the limma method[9] was used to identify DEGs. The original expression data sets from all conditions were processed into expression estimates using the Robust Multichip Average (RMA) method with the default settings implemented in bioconductor, and then the linear model was constructed. The DEGs only with the fold change >2 and P value <0.05 were selected. For demonstrating the potential connection, the Spearman rank correlation (r) was used for comparative target genes’ correlations. The significance level was set at r > 0.9 and local false discovery rate (fdr)[10] <0.05. All statistical tests were performed with the R program (http://www.r-project.org/). Figure 1: Expression profiles of 59 DEGs To identify co-expressed groups, we used DPClus,[7] a graph- clustering algorithm that can extract densely connected nodes as a cluster. It is based on density and periphery tracking of clusters. DPClus is freely available from http://kanaya.naist. jp/DPClus/. In this study, we used the overlapping mode with the DPClus settings. We set the parameter settings of cluster property (cp); density values were set to 0.5[11] and minimum cluster size was set to 5. The pathway[12] database records networks of molecular interactions in the cells and variants of them specific to particular organisms (http://www.genome.jp/kegg/). The Gene Ontology (GO)[13] project is a major bioinformatics initiative with the aim of standardizing the representation of gene and gene product attributes across species and databases. Figure 2: Graph clustering of correlated modules in CRC (threshold r ≥ 0.9). Using the DPClus algorithm we extracted 9 clusters in CRC. The internal nodes of the clusters are connected by green edges; [14] InterPro is an integrated database of predictive protein neighboring clusters are connected by red edges “signatures” used for the classification and automatic annotation of proteins and genomes. Table 1: List of enriched KEGG pathways in cluster2 and 9 The DAVID[15] was used to identify over-represented pathways, detected by DPClus InterPro domains, and GO terms in biological process based Cluster Term P value Benjamini on hypergeometric distribution test. P value <0.05 was the Cluster 2 hsa04270: Vascular 0.001422 0.009911 threshold for the analysis. smooth muscle contraction RESULTS Cluster 9 hsa04020: Calcium 0.034612 0.161486 signaling pathway We obtained publicly available microarray data sets GSE23878 from Gene Expression Omnibus (GEO). After microarray analysis, 1383 DEGs with the fold change >2 and P value <0.05 were selected. At r ≥ 0.9, DPClus[7] identified nine clusters in the correlation network for CRC; they ranged in size from 5 to 12 genes [Figure 2]. To get the relationships among DEGs, the co-expressed To assess the significance of the clusters, we used the over- value (r > 0.9 and fdr < 0.05) was chosen as the threshold. represented KEGG pathways (so-called Kyoto Encyclopedia Finally, 324 relationships among 59 DEGs were constructed a of Genes and Genomes (KEGG) enrichment analysis) in the correlation network. The expression profiles of the 59 DEGs clusters. The results of graph clustering with KEGG enrichment are seen in Figure 1. analysis are presented in Table 1. 468 Journal of Cancer Research and Therapeutics - July-September 2013 - Volume 9 - Issue 3 Zhu, et al.: Expression profiling determines colon cancer pathway Table 2: List of enriched GO terms in clusters 1–9 detected by DPClus Cluster Term P valuemin Benjamini Cluster 1 GO:0006955~immune response 3.40E-07 6.81E-07 Cluster 2 GO:0031032~actomyosin structure 0.008254 0.467374 organization Cluster 3 GO:0010038~response to metal ion 0.018981 0.331313 Cluster 4 NA - - Cluster 5 GO:0007017~microtubule-based process 0.001032 0.079304 Cluster 6 NA - - Cluster 7 NA - - Cluster 8 GO:0030198~extracellular matrix organization 0.030402 0.981364 Cluster 9 NA - - GO, gene ontology; NA, no term was detected Table 3: List of enriched InterPro domains in clusters 1–9 encoded protein is found at the centrosome in interphase cells detected by DPClus and at the spindle poles in mitosis. AURKA overexpression Cluster Term P valuemin Benjamini was detected in colorectal carcinogenesis, especially in Cluster 1 IPR003597: 2.04E-11 1.83E-10 chromosomal instable carcinomas.[6] Immunoglobulin C1-set Cluster 2 NA - - MT1G gene belongs to the MT family encoding a class of Cluster 3 IPR018064: 1.56E-20 4.67E-20 metal-binding proteins involved in several cellular processes, Metallothionein, including potent antioxidant function against various types vertebrate, metal-binding site of oxidative damage as well as regulation of zinc and copper Clusters 4–8 NA - - homeostasis, and their expression is often dysregulated in Cluster 9 IPR013098: 0.048692 0.776316 human tumors. In a human colon cancer cell line, the level Immunoglobulin I-set of MT achieves its maximum near the G1/S stage of the cell NA, No term was detected cycle, suggesting that MT plays a physiological role in cell proliferation. MT expression was stronger in CRC at early stage than in advanced carcinomas.[16] Furthermore, MT1G genes The enrichment analysis method yields two significant were found to be up-regulated in CRC cell lines HCT116 and pathways, vascular smooth muscle contraction pathway and SW620 cells after f-adiponectin treatment and were suggested calcium signaling pathway [Table 1]. to have an anticarcinogenic role in colorectal carcinogenesis based on previous reports.[17] Several GO categories were enriched among these genes in the correlation network, including immune response, actomyosin AKAP12 gene encodes a member of the A-kinase anchor protein structure organization, response to metal ion, etc.