|Year : 2014 | Volume
| Issue : 7 | Page : 186-194
Analysis of hepatocellular carcinoma and metastatic hepatic carcinoma via functional modules in a protein-protein interaction network
Jun Pan1, Zhijie Cong2, Ming Zhong2, Yihui Sun1
1 Department of General Surgery, Second Affiliated Hospital of Soochow University, Suzhou 215004, China
2 Department of Colorectal Surgery, Renji Hospital, School of Medical, Shanghai Jiaotong University, Shanghai 200127, Jiangsu, China
|Date of Web Publication||29-Nov-2014|
Department of General Surgery, Second Affiliated Hospital to Soochow University, 1055, Sanxiang Road, Suzhou, Jiangsu province 215004
Source of Support: None, Conflict of Interest: None
Introduction: This study aims to identify protein clusters with potential functional relevance in the pathogenesis of hepatocellular carcinoma (HCC) and metastatic hepatic carcinoma using network analysis.
Materials and Methods: We used human protein interaction data to build a protein-protein interaction network with Cytoscape and then derived functional clusters using MCODE. Combining the gene expression profiles, we calculated the functional scores for the clusters and selected statistically significant clusters. Meanwhile, Gene Ontology was used to assess the functionality of these clusters. Finally, a support vector machine was trained on the gold standard data sets.
Results: The differentially expressed genes of HCC were mainly involved in metabolic and signaling processes. We acquired 13 significant modules from the gene expression profiles. The area under the curve value based on the differentially expressed modules were 98.31%, which outweighed the classification with DEGs.
Conclusions: Differentially expressed modules are valuable to screen biomarkers combined with functional modules.
Keywords: Cluster, Gene Ontology, hepatocellular carcinoma, support vector machine
|How to cite this article:|
Pan J, Cong Z, Zhong M, Sun Y. Analysis of hepatocellular carcinoma and metastatic hepatic carcinoma via functional modules in a protein-protein interaction network. J Can Res Ther 2014;10, Suppl S3:186-94
|How to cite this URL:|
Pan J, Cong Z, Zhong M, Sun Y. Analysis of hepatocellular carcinoma and metastatic hepatic carcinoma via functional modules in a protein-protein interaction network. J Can Res Ther [serial online] 2014 [cited 2019 Aug 19];10:186-94. Available from: http://www.cancerjournal.net/text.asp?2014/10/7/186/145866
| > Introduction|| |
Hepatocellular carcinoma (HCC) develops as a consequence of underlying liver disease and is often associated with cirrhosis.  The incidence of HCC continues to increase worldwide and varies markedly in different regions. The risk factors of HCC development include hepatitis virus infection, alcoholism, smoking, food toxins (e.g. afflation), congenital liver diseases (e.g. cirrhosis), diabetes, obesity, and errors in metabolism.  Metastatic or secondary liver cancers spread from other parts or tissues of the body. Liver metastases are usually asymptomatic and nonspecific. Weight loss, fever, and anorexia are sometimes the first symptoms of primary cancer. However, early-stage HCC is difficult to diagnose. The prognosis of HCC patients remains poor because of tumor recurrence or tumor progression, and effective adjuvant therapies remain lacking to date.  Imaging techniques are usually used for diagnosis. , Meanwhile, tumor biomarkers assist the screening, diagnosis, and prognosis of HCC. For example, GOLPH 3 and α-methylacyl-CoA racemase are potential prognostic biomarkers for liver cancer, , whereas Nanog and CD146 are potential biomarkers for liver metastasis in colorectal cancer. , However, the sensitivity and specificity of a single biomarker are inadequate for the clinical diagnosis or prognosis of HCC or liver metastasis. The tumor marker alpha fetoprotein is used as a supplement in HCC history progression but not in early disease detection.  Therefore, understanding the molecular carcinogenic mechanism of HCC is crucial.
In recent years, advanced genomic, proteomic, and metabolomic technologies have been applied to identify reliable and novel HCC biomarkers.  High-throughput biological technologies such as microarray platform and protein-protein interaction (PPI) network are new approaches of developing HCC biomarkers. ,, However, biomarker identification methods that strictly use gene expression profiles cannot show correlations between different genes within the biomarker gene set. , Furthermore, different researchers have obtained different gene profiles and a few common genes for similarly diagnosed patients.  Therefore, many studies used computerized approaches based on datasets covering both human PPI networks and cancer gene expression profiles. Network-based analyses of living cells have been employed to characterize specific physiological functions, signaling and metabolic networks, and genes with clinical significance. ,
In the present study, we identified differentially expressed modules (DEMs) between HCC and metastatic hepatic carcinoma (MHCC), and screened module biomarkers on the basis of a PPI network. This study provided insights into the mechanism of HCC metastasis and identified biomarkers that may serve as drug targets to treat early-onset MHCC in humans.
| > Materials and methods|| |
The gene expression profile of the hepatic carcinoma is obtained from the GSE3500 dataset in the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo). Data from 207 samples were analyzed, including 111 HCC samples, 10 MHCC samples, and 86 normal liver tissues. These data were based on 13 platforms (GPL2648, GPL2649, GPL2831, GPL2868, GPL2906, GPL2935, GPL2938, GPL2948, GPL3007, GPL3008, GPL3009, GPL3010, and GPL3011).
All patients participating in this study provided informed consent prior to surgery. All tissues were surgically resected, snap frozen in liquid nitrogen for half an hour after the resection, and then stored at −80°C. In most cases, both tumor and adjacent nontumor tissues were collected. Total RNA was extracted with RNeasy kit (Qiagen) RNeasy Kit is for purification of total RNA from animal cells, animal tissues, bacteria, and yeast, and for RNAcleanup, and mRNA was isolated from total RNA using the FastTrack (Invitrogen) or Poly (A) Pure (Ambion) mRNA Isolation KitsFast which make mRNA isolation from small samples mRNA purification kit. mRNA was directly purified with the FastTrack (Invitrogen) kit. We obtained 10,413 genes after converting ID_REF into gene symbols. The samples were divided into three groups that comprise hepatic carcinoma samples (H), metastatic hepatic carcinoma samples (MH), and normal liver tissues (N).
| > Protein-protein interaction network|| |
The Human Protein Reference Database (HPRD, http://www. hprd.org/) is a centralized platform to visually depict and integrate information on domain architecture, posttranslational modifications, interaction networks, and disease association for each protein in the human proteome. All information in HPRD has been manually extracted from the literature by expert biologists who have read, interpreted, and analyzed the published data. We downloaded the protein interaction data from HPRD to construct a human PPI network using Cytoscape (http://cytoscape.org/). This software allows the visualization and integration of complex networks to any type of attribute data.
| > Differentially expressed modules|| |
Network modules can be extracted through different methods, such as MINE and MCODE of Cytoscape plugins. MINE was created to identify high-quality modules of gene products within highly interconnected biological networks. The algorithm allows a high degree of flexibility and user-customization of results with a few adjustable parameters. MINE outperforms several popular clustering algorithms in identifying modules with high modularity and obtains satisfactory overall recall and precision of functional annotations in PPI networks from both Saccharomyces cerevisiae and Caenorhabditis elegans. MCODE is a Cytoscape plugin that detects clusters (highly interconnected regions) in a network. Clusters differ in different network types. For instance, the clusters in a PPI network are often protein complexes and parts of pathways, whereas the clusters in a protein similarity network represent protein families. MCODE is a popular clustering method that uses vertex weighting to grow clusters from a starting vertex of high local weight by iteratively adding neighboring vertices with similar weights. In this study, we applied both plugins to obtain the modules using a cut-off value of >2 for the connectivity degree of nodes (proteins in the network).
We calculated the module scores in all the samples using the following formula:
Where M k is the score of module k, n is the number of genes in module k, and Z ij is the expression score of gene i in sample j.
Basing on the above algorithm, we obtained module activity matrixes, which indicate the different expression levels between the H and MH samples. We utilized the Wilcoxon rank-sum test to measure the module significance between HCC and MHCC, and identify DEMs.
| > Results|| |
0 Differentially expressed genes
We obtained 1336 differentially expressed genes (DEGs) after comparing the MH and N samples. Moreover, 1624 and 759 DEGs were available to compare the H samples with the N and MH samples, respectively. H-MH and HN shared 436 DEGs, H-MH and MH-N shared 650 DEGs, and H-MH and H-N shared 188 DEGs [Figure 1]. To determine functional similarities and differences, we annotated the functions of these DEGs based on Gene Ontology (GO, cellular component, molecular function, and biological process) and KEGG pathway. The three types of DEGs shared 133 genes, of which 16 were related to cancer. Functional enrichment analysis revealed the presence of the three types of DEGs in the extracellular region, membrane-enclosed lumen, envelope, and organelle. These common DEGs participate in catalytic, electron carrier, antioxidant, and enzyme regulator activities, which are involved in cellular component biogenesis, stimulus-response, localization establishment, immunity, and metabolism. They also participate in PPAR signaling, synthesis and degradation (valine, leucine, and isoleucine degradation; ketone body synthesis and degradation; and terpenoid backbone biosynthesis), metabolism pathways (retinol, tryptophan, butanoate, fatty acid, and pyruvate metabolism), and one carbon pool by folate. The biological processes in which the DEGs of H-MH participate were shared by the DEGs of the H-N and MH-N samples. The H-N and MH-N samples shared 636 DEGs. Except for the common biological processes in which all the three types of DEGs are involved, the DEGs of the H-N and MH-N samples are involved in death, growth, cellular processes, and cellular component biogenesis. Specifically, the DEGs of HCC principally dysregulate the following processes: Viral reproduction, locomotion, binding, reproduction, and development. The DEGs of H-MH and H-N significantly promote transporter and binding activities, respectively. Functional enrichment analysis revealed that the DEGs for HCC and MHCC have significantly different roles in transporter activity. The diverse pathways among the three types of DEGs are shown in [Table 1]. Approximately, 57.9% of the DEGs of MHCC are involved in metabolic pathways, such as ascorbate and aldarate metabolism. These results provided insights into HCC mechanism.
|Figure 1: The relationship of differentially expressed genes among H-MH, H-N and MH-N|
Click here to view
|Table 1: Different enriched pathways among DEGs of H-MH, H-N, and MH-N samples |
Click here to view
Human protein interaction network
We mapped the three types of DEGs and cancer genes into the PPI network, which consisted of 9613 nodes and 36940 interaction pairs, to differentiate HCC and MHCC. We calculated the important topological features of the three types of DEGs and cancer genes. These features include average shortest path, clustering coefficient, degree distribution, and closeness centrality [Figure 2]. All the topological features were similar except for liver cancer genes, which had the highest degree and the shortest path. The degree of liver cancer genes was 70, and the closeness centrality was 0.3. The biological network showed high aggregation, reflecting the intense modularization of gene networks.
|Figure 2: The statistic of topological features for the differentially expressed genes of H-MH, H-N and MH-N. (a) Average shortest path, (b) closeness centrality (c) clustering coefficient (d) degree distribution|
Click here to view
MINE and MCODE were applied to cluster the gene sets and obtain the specific function modules. Finally, 614 and 111 clusters were acquired (node score cutoff: 0.2, K-Core: 2, maximum depth from seed: 100, false degree cut-off: 2). We constructed the expression profiles of module genes and acquired the significant DEMs between H-N and H-MH using the Wilcoxon rank-sum test (P < 0.05). Finally, we acquired 60 DEMs with MINE and 13 DEMs with MCODE. To assess the classification efficiency, we compared three features, namely, the different expression genes between HCC and MHCC, the DEMs of MCODE, and the DEMs of MINE. The area under the curve values for the three features was 72.52%, 98.31%, and 91.5% [Figure 3].
|Figure 3: The receiver operating characteristic (ROC) curves of support vector machine classifier. (a) The ROC curve when the differentially expressed modules (DEMs) extracted by MCODE as the character of classification. (b) The ROC curve when the differentially expressed genes between hepatocellular carcinoma and metastatic hepatic carcinoma as the character of classification. (c) The receiver operating characteristic curve when the DEMs extracted by MINE as the character of classification|
Click here to view
| > Discussion|| |
0 Differentially expressed gene
Comparison of the H-N and MH-N samples showed that most of the common DEGs were expressed in the same direction, while 43 of the DEGs were expressed in different directions. The common DEGs between the H-MH and MH-N samples showed the same expression trend. A total of 99 DEGs showed a diverse regulation trend between the H-MH and H-N samples, which were located in the cell and organelle. These DEGs are involved in localization, Wnt signaling pathway, chronic myeloid leukemia, catalytic activity, and electron carrier activity. Particularly, the Wnt/β-catenin pathway is activated in at least 1/3 of the H samples, and a significant number of these samples have mutations in the β-catenin gene. Recent studies have revealed that targeting the Wnt/β-catenin signaling pathway in liver cancer stem cells and HCC cell lines with FH535 is effective at several levels. , Network analysis revealed that the DEGs of the H-MH and H-N samples shared similar topological properties. The degree and clustering coefficient of the DEGs in H-N was higher than those of the DEGs in H-MH [Figure 2]c and d. This result indicates that the DEGs in H-N assume biological functions in the form of modules.
Nonspecific differentially expressed modules
The 13 DEMs derived by MCODE were used to differentiate MHCC from HCC [Figure 4]. Approximately, 84.6% of the significant DEMs, except for modules 94 and 95, contained cancer genes or DEGs. The three genes in module 94 were myosin VIIA (MYO7A), Usher syndrome 1C (USH1C), and cadherin-related 23 (CDH23). MYO7A is a member of the myosin gene family. Myosins are mechanochemical proteins characterized by the presence of a motor domain, an action-binding domain, a neck domain that interacts with other proteins, and a tail domain that serves as an anchor. Defects in this gene are associated with the mouse shaker-1 phenotype and the human Usher syndrome 1B, which are characterized by deafness, reduced vestibular function, and retinal degeneration in humans.  USH1C encodes a scaffold protein that functions in the assembly of Usher protein complexes; defects in this gene are caused by USH1C and nonsyndromic sensorineural deafness autosomal recessive type 18.  CDH23 is a member of the cadherin superfamily and encodes calcium dependent cell-cell adhesion glycoproteins, which are involved in stereocilia organization and hair bundle formation. The gene is located in a region containing the human deafness loci DFNB12 and USH1D. USH1D and nonsyndromic autosomal recessive deafness DFNB12 are caused by allelic mutations of this cadherin-like gene. Upregulation of this gene may also be associated with breast cancer.  The mechanism by which the three genes regulate the metastasis of cancer genes remains unclear to date. However, we speculate that that these genes disrupt the normal biological function as a module to induce cancer development.
|Figure 4: 13 significant differentially expressed modules. *Color: The purple indicates the cancer gene; Red presents the common genes among cancer genes, differentially expressed genes (DEGs) of H-MH, H-N and MH-N; the orange presents the DEGs of H-N; Lt Orange Brown is the DEGs of MH-N. *Node shape: "V" presents the down regulation genes. Triangle presents the up regulation genes|
Click here to view
Differentially expressed module markers related to cancer
Five cancer gene-related modules (i.e. modules 4, 24, 25, and 35) were used to analyze module functions. We used module 4 with 28 genes as an example for the analysis. Among the 28 genes, 20 are cancer genes and 1 is an upregulation gene (SHB). Specifically, MET was common among the cancer genes and DEGs of H-MH, H-N, and MH-N. All the genes in the module were enriched for protein functioning in a common biological process as annotated by GO (hypergenometric test with a false discovery rate [FDR] of 0.05). The module genes are synapse components that participate in molecular transducer, catalytic, and binding activities [Table 2]. These genes are involved in immunity, multi-organismal processes, death, and reproduction. The module genes activate the pathways of different cancer types, such as colorectal cancer. The ErbB signaling pathway, Janus kinase (JAK)-signal transducer and activator of transcription (STAT) signaling pathway, and so on may influence the occurrence of HCC. Recent studies have demonstrated unique clustered genetic profiles during the daily progression and transdifferentiation of hepatic stellate cells (HSCs); JAK-STAT signaling may be critical in the early stages of transdifferentiation and in promoting tumorigenesis in HCC.  However, the significance of JAK/STAT signaling to HSC transdifferentiation has been determined by treating cells with a JAK2 inhibitor.  Basing on the gene interaction in module 4, we found that JAK2 was a cancer gene and interacted with half of the module genes. JAK2 is a protein tyrosine kinase involved in a specific subset of cytokine receptor signaling pathways; it is also constitutively associated with the prolactin receptor and is required for responses to gamma interferon.  JAK2 is important in myeloid cell proliferation and differentiation, and fusion JAK2 mutations have been described in acute megakaryocytic leukemia and acute leukemia/chronic myeloid malignancies.  Furthermore, JAK2 is a member of the Janus family of nonreceptor tyrosine kinases (JAK1, JAK2, JAK3, and tyrosine kinase 2) that transduce signals downstream of types I and II cytokine receptors via STAT. Tefferi A et al.  indicated that both JAK1 and JAK3 mutations occur in common human cancers. Thus, we infer that the module related to JAK2 affects tumor metastasis from other tissues to the liver.
Function enrichment of differentially expressed modules
Network modules can be utilized to map associated biological functions in a sensitive manner through network clusters using stringent downstream analysis.  To identify the functions of these module genes, we annotated the genes of the upper five modules into DAVID using statistical tests, including hypergenometric test and Benjamin and Hochberg FDR correction. Statistical significance was considered at P < 0.05. Module 3 comprises nine genes, eight of which belong to the COP9 signalosome (CSN). CSN is a protein complex that is conserved throughout eukaryotes; this complex is critical to the development of all multicellular organisms. Recent studies have suggested that CSN sustains the activity of SCF and other cullin-based ubiquitin ligases.  In module 3, COPS5, COPS2, and GPS1 are downregulated in H-N while COPS7A is upregulated in MH-N. The diverse regulation in HCC and MHCC promotes the metastasis of tumor cells. Cullin-5 (CUL5) is a tumor suppressor gene that promotes cell colony formation and inhibits cell cycle arrest.  Moreover, silencing CUL5 reduces cellular sensitivity, and endogenous CUL5 suppresses epithelial cell transformation by several pathways, including inhibition of Src-Cas-induced ruffling through SOCS6. , In the cellular component, the genes in module 3 are principally localized in the nucleus, signalosome, protein complex, and macromalecular complex. The genes in module 71, which are principally detected in the mitochondrial inner membrane presequence translocase complex, participate in protein localization in the mitochondrion and protein targeting to the mitochondrion. The genes in module 85 are located in the nuclear pore; these genes are involved in the establishment of RNA localization and nucleic acid transport. Located in the juxtapanodal region of the axon, the genes in module 95 regulate transmembrane and ion transport processes. Although the prediction was based on DEMs in the PPI network, the present detailed analysis was still limited. That is, verifying the functional significance of the modules by utilizing the trials is complex [Figure 5], [Figure 6].
|Figure 5: The cellular component for module 3, 71, 85 and 95. *The deeper color present the more significant node|
Click here to view
|Figure 6: The biological process for module 71, 85, 95 and 99. *The deeper color present the more significant node|
Click here to view
| > Conclusions|| |
Some features of the gene expression patterns were associated with specific phenotypic and genotypic tumor characteristics, including growth rate, vascular invasion, and p53 overexpression. Combining the gene expression profiles, we used a network biological approach to predict the biomarkers that distinguish HCC from MHCC by identifying DEMs. Deriving the genes in functional modules as biomarkers is more effective than single DEGs.
| > References|| |
Huang XB, Li J, Zheng L, Zuo GH, Han KQ, Li HY, et al.
Bioinformatics analysis reveals potential candidate drugs for HCC. Pathol Oncol Res 2013;19:251-8.
El-Serag HB, Rudolph KL. Hepatocellular carcinoma: Epidemiology and molecular carcinogenesis. Gastroenterology 2007;132:2557-76.
He TL, Zheng KL, Li G, Song B, Zhang YJ. Identification of typical miRNAs and target genes in hepatocellular carcinoma by DNA microarray technique. Eur Rev Med Pharmacol Sci 2014;18:108-16.
Ogura T, Hara K, Hijioka S, Mizuno N, Imaoka H, Niwa Y, et al.
Can endoscopic ultrasound-guided fine needle aspiration offer clinical benefit for tumors of the ampulla of vater? - An initial study. Endosc Ultrasound 2012;1:84-9.
Costache MI, Iordache S, Karstensen JG, Saftoiu A, Vilmann P. Endoscopic ultrasound-guided fine needle aspiration: From the past to the future. Endosc Ultrasound 2013;2:77-85.
JianXin J, Cha Y, ZhiPeng L, Jie X, Hao Z, Meiyuan C, et al.
GOLP3 is a predictor of survival in patients with hepatocellular carcinoma. Clin Invest Med 2014;37:E233.
Xu B, Cai Z, Zeng Y, Chen L, Du X, Huang A, et al.
a-Methylacyl-CoA racemase (AMACR) serves as a prognostic biomarker for the early recurrence/metastasis of HCC. J Clin Pathol 2014;67:974-9.
Tian B, Zhang Y, Li N. CD146 protein as a marker to predict postoperative liver metastasis in colorectal cancer. Cancer Biother Radiopharm 2013;28:466-70.
Xu F, Dai C, Zhang R, Zhao Y, Peng S, Jia C. Nanog: A potential biomarker for liver metastasis of colorectal cancer. Dig Dis Sci 2012;57:2340-6.
Plentz RR, Boozari B, Malek NP. Guideline compliant diagnostics of hepatocellular carcinoma. Radiologe 2014;54:660-3.
Chaiteerakij R, Addissie BD, Roberts LR. Update on Biomarkers of Hepatocellular Carcinoma. Clin Gastroenterol Hepatol 2013.
Hoshida Y, Villanueva A, Kobayashi M, Peix J, Chiang DY, Camargo A, et al.
Gene expression in fixed tissues and outcome in hepatocellular carcinoma. N Engl J Med 2008;359:1995-2004.
Hoshida Y, Nijman SM, Kobayashi M, Chan JA, Brunet JP, Chiang DY, et al.
Integrative transcriptome analysis reveals common molecular subclasses of human hepatocellular carcinoma. Cancer Res 2009;69:7385-92.
Budhu A, Chen Y, Kim JW, Forgues M, Valerie K, Harris CC, et al.
Induction of a unique gene expression profile in primary human hepatocytes by hepatitis C virus core, NS3 and NS5A proteins. Carcinogenesis 2007;28:1552-60.
Wang YC, Chen BS. A network-based biomarker approach for molecular investigation and diagnosis of lung cancer. BMC Med Genomics 2011;4:2.
Lu H, Meng Q, Wen Y, Hu J, Zhao Y, Cai L. Increased EHD1 in non-small cell lung cancer predicts poor survival. Thorac Cancer 2013;4:422-32.
Ein-Dor L, Zuk O, Domany E. Thousands of samples are needed to generate a robust gene list for predicting outcome in cancer. Proc Natl Acad Sci U S A 2006;103:5923-8.
Zhuang L, Wu Y, Han J, Ling X, Wang L, Zhu C, et al.
A network biology approach to discover the molecular biomarker associated with hepatocellular carcinoma. Biomed Res Int 2014;2014:278956.
Zhang L, Guo Y, Li B, Qu J, Zang C, Li F, et al.
Identification of biomarkers for hepatocellular carcinoma using network-based bioinformatics methods. Eur J Med Res 2013;18:35.
Gedaly R, Galuppo R, Daily MF, Shah M, Maynard E, Chen C, et al.
Targeting the Wnt/ß-catenin signaling pathway in liver cancer stem cells and hepatocellular carcinoma cell lines with FH535. PLoS One 2014;9:e99272.
Galuppo R, Maynard E, Shah M, Daily MF, Chen C, Spear BT, et al.
Synergistic inhibition of HCC and liver cancer stem cell proliferation by targeting RAS/RAF/MAPK and WNT/ß-catenin pathways. Anticancer Res 2014;34:1709-13.
MYO7A myosin VIIA [ Homo sapiens (human) ].Gene Database updated on 9-Nov-2014.Available from:http://www.ncbi.nlm.nih.gov/gene/4647.
USH1C Usher syndrome 1C (autosomal recessive, severe) [ Homo sapiens (human) ]Gene Database updated on 9-Nov-2014.Available from:http://www.ncbi.nlm.nih.gov/gene/10083.
CDH23 cadherin-related 23 [ Homo sapiens (human) ].Gene Database updated on 9-Nov-2014.Available from:http://www.ncbi.nlm.nih.gov/gene/64072.
Wilson GS, Tian A, Hebbard L, Duan W, George J, Li X, et al.
Tumoricidal effects of the JAK inhibitor Ruxolitinib (INC424) on hepatocellular carcinoma in vitro
. Cancer Lett 2013;341:224-30.
Lakner AM, Moore CC, Gulledge AA, Schrum LW. Daily genetic profiling indicates JAK/STAT signaling promotes early hepatic stellate cell transdifferentiation. World J Gastroenterol 2010;16:5047-56.
Brooks AJ, Dai W, O'Mara ML, Abankwa D, Chhabra Y, Pelekanos RA, et al.
Mechanism of activation of protein kinase JAK2 by the growth hormone receptor. Science 2014;344:1249783.
Tefferi A. JAK and MPL mutations in myeloid malignancies. Leuk Lymphoma 2008;49:388-97.
Stevens A, De Leonibus C, Hanson D, Dowsey AW, Whatmore A, Meyer S, et al.
Network analysis: A new approach to study endocrine disorders. J Mol Endocrinol 2014;52:R79-93.
Cope GA, Deshaies RJ. COP9 signalosome: A multifunctional regulator of SCF and other cullin-based ubiquitin ligases. Cell 2003;114:663-71.
Ma C, Qi Y, Shao L, Liu M, Li X, Tang H. Downregulation of miR-7 upregulates Cullin 5 (CUL5) to facilitate G1/S transition in human hepatocellular carcinoma cells. IUBMB Life 2013;65:1026-34.
Teckchandani A, Laszlo GS, Simó S, Shah K, Pilling C, Strait AA, et al.
Cullin 5 destabilizes Cas to inhibit Src-dependent cell transformation. J Cell Sci 2014;127:509-20.
Samant RS, Clarke PA, Workman P. E3 ubiquitin ligase Cullin-5 modulates multiple molecular and cellular responses to heat shock protein 90 inhibition in human cancer cells. Proc Natl Acad Sci U S A 2014;111:6834-9.
Luo T, et al
., Molecular mechanisms of cellular mechanosensing. Nat Mater, 2013. 12: p. 1064-71.
Huang Y, et al
, Gd complexes of diethylenetriaminepentaacetic acid conjugates of low-molecular-weight chitosan oligosaccharide as a new liver-specific MRI contrast agent. Magn Reson Imaging, 2013. 31: p. 604-9.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2]