Volume 26 Issue 1 - February 28, 2014 PDF
In silico identification of oncogenic potential of fyn-related kinase in hepatocellular carcinoma
Jia-Shing Chen1,  Wei-Shiang Hung1,  Hsiang-Han Chan1,  Shaw-Jenq Tsai2,3, H. Sunny Sun 1,3,*
1 Institute of Molecular Medicine, College of Medicine, National Cheng Kung University
2 Department of Physiology, College of Medicine, National Cheng Kung University
3 Bioinformatics Center, Center for Biotechnology and Biosciences, National Cheng Kung University
Font Enlarge
Cancer development is a complex and heterogeneous process. Moreover, cancer is considered to be a genetic disease and develops, at least in part, because of DNA damage1. Generally speaking, genes involved in tumorigenesis are divided into two major categories based on their actions: oncogenes and tumor suppressor genes (TSGs) 2. While oncogenes encode proteins that control cell proliferation and/or apoptosis3, a TSG is a protein that protects a cell from entering the course of carcinogenesis4. About 5–10% of human genes are predicted to contribute to oncogenesis, whereas current experimentally validated cancer genes only cover 1% of the human genome. Thus hundreds of cancer genes still need to be explored.

Identification of new TAGs by computational prediction
To search for new genes involving in carcinogenesis and facilitate cancer research, we developed a systematic workflow to predict new cancer genes and established tumor-associated gene (TAG) database5. First, known cancer genes were collected from the PubMed database. A semi-automatic information retrieving engine was designed to collect specific information of the target genes from various web resources and store in the TAG database. Second, the TAG information is analyzed by building up domain weighting profiles of their protein sequence that can specify the feature of particular type of oncoproteins. The TAG database contained 366 domains. Among these, 49 were present in both the oncogene and TSG groups. The remaining 317 domains were uniquely in either the oncogene or the TSG group. Each TAG gene in the training set had an oncogene and a TSG score; these were calculated as the sum of the oncogene weights and TSG weights of the domains in that gene. After multiplying by 100, the oncogene and TSG scores of genes in the TAG database ranged from 0.1 to 18.08 (average, 3.71) and from 0.05 to 3.21 (average, 0.86), respectively. Finally, the profile was used to identify novel TAGs by searching against currently available cDNA sequences in the human genome database. A total of 90157 human full-length cDNA (FLC) sequences were downloaded and analyzed by the procedure illustrated in Fig. 1. We found 33959 transcripts containing TAG domains with domain numbers ranging from1 to 11. This included 8958 unique gene symbols and 498 transcripts currently without symbols. The results of TAG domain analysis of FLC were filtered by the TAG training set to identify novel TAGs. Using the cutoff values for the ‘unique domain signature’ determined by the training set, and removing the current TAGs from the predicted gene list, we finally obtained 183 new TAGs that included 78 oncogenes and 105 TSGs.

Figure. 1. Schematic of workflow for the identification of new TAGs.

To reveal the molecular properties of these TAGs, we applied pathway and gene ontology analysis using the MetaCore software package. Among the 78 new oncogenes, most were functionally involved in protein kinase activity (71 objects, P-value 1.047E-105) and 60% of these proteins were specifically implicated in protein tyrosine kinase (PTK) activity (P-value 2.444E-82). Thus, these predicted oncogenes are involved in many tumorigenic pathways including calcium signaling and mitogenic signaling (Fig. 2A). Furthermore, 19 out of 26 significantly associated diseases (P-value51.0E-10) were either tumors/carcinomas or neoplasms, indicating that dysfunction of these genes indeed was associated with cancer formation (71 objects, P-value 1.047E-105) and clustered on ovarian, breast and lung cancers (Disease networks, Fig. 2B). On the other hand, the properties of predicted TSGs were diverse, except that many exhibited binding ability (84 objects, P-value 9.323E-07) and were involved in responses to stimuli (73 objects, P-value 3.257E-10). The most significant process according to the GeneGo Map categories for TSG was vascular development-angiogenesis (Fig. 2C), and dysfunction of these predicted new TSGs is known to contribute to many diseases like arteriosclerosis (21 objects, P-value 3.532E-09), and may lead to the development of cancers from various tissues including prostate, lung and colon (Fig. 2D). In summary, these data strongly support roles of the predicted TAGs in tumorigenesis.

Figure. 2. Histograms representing the top GeneGo Map Folders and Disease Networks from analysis of the gene lists of newly predicted oncogenes (A and B) and tumor suppressors (C and D). Key differentially representative processes are shown. In each histogram, the longer each bar the more significant (-log pValue), and both map folders (A and C) and disease networks (B and D) are organized according to descending significance.

Fyn-related kinase (FRK) promotes Hep3B cell growth, transformation and invasive ability
Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide6 and shows high prevalence in Asia and Africa7. According to the annual report from the Bureau of Health Promotion, Department of Health, Taiwan (http://www.bhp.doh.gov.tw/bhpnet/portal/Default.aspx), HCC is the second most common cancer in the Taiwanese population but is the leading cause of cancer-related death in Taiwanese males. Although recent studies have revealed many genetic and epigenetic changes that lead to the aberrant activation of several signaling cascades, the pathogenesis of HCC is heterogeneous among patients8, the molecular mechanisms of hepatocarcinogenesis remain largely unclear. As a proof-of-concept, one predicted oncogene, FRK, which shows an aberrant digital expression pattern in liver cancer cells, was selected for further investigation. Using 68 paired hepatocellular carcinoma samples, we found that FRK was up-regulated in 52% of cases (Fig. 3A; P<0.001). Cell proliferation assay showed that FRK overexpression promoted the proliferation of Hep3B cells (Fig. 3B; P<0.05). Furthermore, soft agar assay showed that FRK overexpression promoted colony formation (P<0.01) and increased the colony size (P<0.0001) in Hep3B_FRK cells compared with the control cells (Fig. 3C). Tumorigenic assays performed in Hep3B (Fig. 3D) and HepG2 (Fig. 3E) cell lines revealed a significant correlation between the level of FRK expression and invasiveness, suggesting that FRK is a positive regulator of invasiveness in liver cancer cells.

Fig. 3. FRK plays roles in HCC tumorigenesis. (A) FRK protein expression in six paired HCC cases. HepG2 and α-tubulin were used as positive and loading controls. Numbers at the bottom of each lane represent relative expression level in tumor compared with normal tissue. (B) Relative proliferation rates of Hep3B cells plotted for those carrying control vector or overexpressing FRK (left), and those containing control sh_luc or sh_FRK (right). (C) Number (left) and size (right) of colonies in transformed Hep3B cells. Both transformed plates (upper) and barcharts (lower) are shown. (D&E) FRK promotes invasion in Hep3B and HepG2 cell lines. The invasion assays were carried out using Matrigel-coated transwells. The crystal violet-stained cells (upper) were counted and the total number of cells from five randomly selected fields on the transwell was measured; the ratio of invasive cells compared with control cells is shown (lower). (D) Left penal: Hep3B_vector and Hep3B_FRK cells. Right penal: Hep3B_sh_luc and Hep3B_sh_FRK cells. (E) Left penal: HepG2_vector and HepG2_FRK cells. Right penal: HepG2_sh_luc and HepG2_sh_FRK cells. ***P<0.001

In conclusion, the oncogenic potential of FRK was suggested by computational analysis, confirmation of the FRK effect on liver cancer provided strong support for the idea that predicted TAGs are highly likely to be involved in tumorigenesis. In addition, our data demonstrated the accuracy of computational prediction and suggested that other predicted TAGs can be potential targets for future cancer research, and we believe the outcome of this study will benefit cancer research.


  1. Vogelstein,B. and Kinzler,K.W. (2004) Cancer genes and the pathways they control. Nat. Med., 10, 789–799.
  2. Bowden,G.T. et al. (1994) Oncogene activation and tumor suppressor gene inactivation during multistage mouse skin carcinogenesis. Cancer Res., 54,1882s–1885s.
  3. Croce,C.M. (2008) Oncogenes and cancer. N. Engl. J. Med., 358, 502–511.
  4. Sherr,C.J. (2004) Principles of tumor suppression. Cell, 116, 235–246.
  5. Chan,H.H. et al. (2007) TAG: a comprehensive database for cancer research. NAR., 35, Molecular Biology Database Collection, #992.
  6. Caldwell,S. and Park,S.H. (2009) The epidemiology of hepatocellular cancer: from the perspectives of public health problem to tumor biology. J. Gastroenterol., 44 (Suppl. 19), 96–101.
  7. Bosch,F.X. et al. (2004) Primary liver cancer: worldwide incidence and trends. Gastroenterology, 127, S5–S16.
  8. Cha,C. and Dematteo,R.P. (2005) Molecular mechanisms in hepatocellular carcinoma development. Best Pract. Res. Clin. Gastroenterol., 19, 25–37.
< Previous
Next >
Copyright National Cheng Kung University