中国科技论文在线 http://www.paper.edu.cn
GPSM2 Signal Transduction Computational Network Analysis in Human No-Tumor Hepatitis/Cirrhosis and # 5 Hepatocellular Carcinoma Transformation ZHUANG Jing, HUANG Juxiang, WANG Lin** (School of Electronic Engineering, Beijing University of Posts and Telecommunications, 100876) Abstract: LGN protein (GPSM2) is involved in transcription or cell division presented in several papers. However, how the molecular network and interpretation concerning GPSM2 signal 10 transduction between no-tumor hepatitis/cirrhosis and hepatocellular carcinoma (HCC) transformation remains to be elucidated. Here we constructed and analyzed significant higher expression gene GPSM2 activated & inhibited upstream and downstream signal transduction network from HCC vs no-tumor hepatitis/cirrhosis pateints (viral infection HCV or HBV) in GEO Dataset by using gene regulatory network inference method based on linear programming and decomposition procedure, under covering 15 GPSM2 pathway and matching signal transduction enrichment analysis by the CapitalBio MAS 3.0 integrated of public databases including Gene Ontology, KEGG, BioCarta, GenMapp, Intact, UniGene, OMIM, etc. By compared the different activated & inhibited GPSM2 network with GO analysis between no-tumor hepatitis/cirrhosis and HCC transformation, our result showed GPSM2 signal transduction network: (1) more nucleus and cytoplasm but less extracellular space protein binding in 20 no-tumor hepatitis/cirrhosis; (2) more growth factor activity but less cytoplasm enzyme activator activity in HCC; (3) less activation & more inhibition molecular numbers in no-tumor hepatitis/cirrhosis but more activation & less inhibition in HCC. Therefore, we inferred (4) GPSM2 signal transduction network stronger transcription but weaker cell differentiation as a result increasing cytoplasm protein translation in no-tumor hepatitis/cirrhosis; (5) stronger cell proliferation but weaker 25 regulation of muscle contraction as a result inceasing nuclear cell division in HCC. Key words: LGN protein (GPSM2) computational network; signal transduction; human hepatocellular carcinoma; transformation; no-tumor hepatitis/cirrhosis
0 Introduction 30 LGN protein (GPSM2) is one of our identified significant high expression genes (fold change ≥2) in human hepatocellular carcinoma compared with no-tumor hepatitis/cirrhosis from GEO Dataset GSE10140-10141 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE10140, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE10141). GPSM2 cellular component localizes in cytoplasm, and molecular function comprises GTPase activator activity, identical 35 protein binding, and biological process contains signal transduction, G-protein coupled receptor protein signaling pathway (GO (http://www.geneontology.org)). GPSM2 is related to enzyme activator activity, gtpase activator activity (GenMAPP (http://www.genmapp.org/)). From the above analysis, GPSM2 is involved in signal transduction which is very important for no-tumor hepatitis/cirrhosis and hepatocellular carcinoma (HCC) transformation. The relationships between 40 hepatitis/cirrhosis or hepatocellular carcinoma with signal transduction are as follows: Signal transduction cascades and hepatitis B and C related hepatocellular carcinoma [1]; 14-3-3zeta up-regulates hypoxia-inducible factor-1alpha in hepatocellular carcinoma via activation of PI3K/Akt/NF-small ka, CyrillicB signal transduction pathway[2]; Genistein inhibits tumor invasion by suppressing multiple signal transduction pathways in human hepatocellular carcinoma cells [3]; 45 Polycyclic aromatic hydrocarbons induce migration in human hepatocellular carcinoma cells (HepG2) through reactive oxygen species-mediated p38 MAPK signal transduction [4]; Pterostilbene inhibited tumor invasion via suppressing multiple signal transduction pathways in Foundations: National Natural Science Youth Fundation of China (No. 81501372) Brief author introduction: ZHUANG Jing (1994-), Female, Master student, Bioinformatics Correspondance author: HUANG Juxiang (1986-), Female, Associate professor, Bioinformatics. E-mail: [email protected]
- 1 - 中国科技论文在线 http://www.paper.edu.cn
human hepatocellular carcinoma cells [5]; The A3 adenosine receptor agonist CF102 induces apoptosis of hepatocellular carcinoma via de-regulation of the Wnt and NF-kappaB signal 50 transduction pathways [6]; Radiation-enhanced hepatocellular carcinoma cell invasion with MMP-9 expression through PI3K/Akt/NF-kappaB signal transduction pathway [7]; Oncogenic signal transduction and therapeutic strategy for hepatocellular carcinoma [8]. However, the distinct molecular network and interpretation concerning GPSM2 signal transduction network in no-tumor hepatitis/cirrhosis and HCC transformation remains to be elucidated. 55 HCC is one of the most common causes of cancer-related death. So to develop novel drugs in HCC has become a challenge for biologists. Here we constructed and analyzed significant higher expression gene GPSM2 activated & inhibited signal transduction network from HCC vs no-tumor hepatitis/cirrhosis pateints (viral infection HCV or HBV) in GEO Dataset by gene regulatory network inference method based on linear programming and decomposition procedure, on the 60 condition that our GPSM2 network covered GPSM2 pathway and matched signal transduction enrichment analysis by the CapitalBio MAS 3.0 based on the integration of public databases including Gene Ontology, KEGG, BioCarta, GenMapp, Intact, UniGene, OMIM, etc. The mechanisms that shut off a signal are as important as the mechanisms that turn it on. GRNInfer tool can identified molecular activation and inhibition relationships based on a novel 65 mathematic method called GNR (Gene Network Reconstruction tool) using linear programming and a decomposition procedure for inferring gene networks. The method theoretically ensures the derivation of the most consistent network structure with respect to all of the datasets, thereby not only significantly alleviating the problem of data scarcity but also remarkably improving the reconstruction reliability [9]. 70 In this paper, by compared the different activated & inhibited molecules and numbers of GPSM2 network with GO analysis between no-tumor hepatitis/cirrhosis and HCC transformation, we need to do (1) prediction of different stronger and weaker cellular component, molecular function and biological process between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation, (2) computation of different more and less activation 75 and inhibition numbers between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation, (3) different inferences between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation, (4) different evidences from literatures to support the inferences respectively. 1 Materials and Methods
80 1.1 Microarray Data We used microarrays containing 6144 genes from 25 no-tumor hepatitis/cirrhosis and 25 HCC patients in GEO Dataset GSE10140-10141 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE10140, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE10141). We preprocessed raw 85 microarray data as log2. 1.2 Gene Selection Algorithms Potential HCC molecular markers were identified using significant analysis of microarrays (SAM) [10]. We normalized data by log2, selected two classes paired and minimum fold change ≥2, and chose the significant high expression value genes of HCC compared with no-tumor 90 hepatitis/cirrhosis genes under the false-discovery rate and q-value were 0%. The q-value is like the well-known p-value, but adapted to multiple-testing situations.
- 2 - 中国科技论文在线 http://www.paper.edu.cn
1.3 Unsupervised Clustering Significant higher expression genes from no-tumor hepatitis/cirrhosis vs HCC was done unsupervised clustering using cluster 3.0(http://bonsai.ims.tokyo.ac.jp/~mdehoon/software/cluster). 95 The steps are as follows: Step 1 Loading and filtering 100% data ; Step 2 Normalizing log transform data for adjusting data; Step 3 Choosing gene culster and array cluster; Step 4 Choosing average linkage of hierarchical clustering Step 5 Doing TreeView. 1.4 Molecule Annotation System (MAS) Molecule Annotation System (MAS) (http://bioinfo.capitalbio.com/mas3/) is a web-based 100 software toolkit for a whole data mining and function annotation solution to extract & analyze biological molecules relationships from public databases of biological molecules and signification. MAS uses relational database of biological networks created from millions of individually modeled relationships between genes, proteins, complexes, cells, disease and tissues. MAS allows a view on your data, integrated in biological networks according to different kinds of biological 105 context. This unique feature results from multiple lines of evidences which are integrated in MASCORE. MAS helps to understand relationship of gene expression data through the given molecular symbols list, and provides thorough, unbiased and visible results. The primary databases of MAS integrated various well-known biological resources such as Gene Ontology, KEGG, BioCarta, GenMapp, HPRD (http://www.hprd.org/), MINT 110 (http://mint.bio.uniroma2.it/mint/Welcome.do), BIND (http://www.blueprint.org/), Intact (http://www.ebi.ac.uk/intact/), UniGene (www.ncbi.nlm.nih.gov/UniGen), OMIM (http://www.ncbi.nlm.nih. gov/entrez/query.fcgi?db=OMIM). MAS offers various query entries and graphics result. The system represents an alternative approach to mining and catching on biological signification for high-throughput array data. 115 The same used algorithm is P Qvalue in GO and pathway of module. In the actual analysis, we need further screening process for acertain P value threshold to obtained the significance of GO / pathway of the false positive rate, named False Discovery Rate (FDR). So It is more convenient to give a Q value for each P value to reflect the FDR as P value threshold. The Q threshold value is supposed to be 0.05. If it is less than 0.05, then it means the p value as the 120 significance level of the false positive rate is lower. The p value is smaller indicating that protein (or gene) in the GO (or pathway) is more significant enrichment. The algorithm is used to calculate the p value in GO and Pathway of module is hypergeometric distribution. M N − M m−1 i n − i p = 1− i=0 N n (1) 125 N refers to the number of genes(genome-wide gene universe) of a species included in the MAS library. m(the intersection of M and n ) is the number of genes contained in the specific Pathway. p0=m/M, p1=n/N. P value is the probability value of the false negative null hypothesis H0: p0 = p1. When P value is less than 0.05, it means differentially expression genes significantly enriched in this pathway.
130 1.5 Network Establishment of Candidate Genes The entire network was constructed using GRNInfer and GVedit tools (http://www.graphviz.org/About.php). GRNInfer is a novel mathematic method called GNR (Gene
- 3 - 中国科技论文在线 http://www.paper.edu.cn
Network Reconstruction tool) based on linear programming and a decomposition procedure for inferring gene networks. The method theoretically ensures the derivation of the most consistent 135 network structure with respect to all of the datasets, thereby not only significantly alleviating the problem of data scarcity but also remarkably improving the reconstruction reliability. The following equation (2) represents all of the possible networks for the same dataset. −1 T T T J = (X '−A)U V + YV = J + YV (2) Where is an n×n Jacobian matrix or connectivity matrix, X= J = (Jij )nn = f (x) / x
140 (x(t1),…,x(tm)), A=(a(t1),…,a(tm) and X’=(x’(t1),…,x’(tm)) are all n×m matrices with T T x’i(tj)=[xi(tj+1)-xi(tj)]/[tj+1-tj] for i=1,…,n; j=1,…,m. X(t)=(x1(t),…,xn(t)) ∈Rn, a=(a1…,an) ∈Rn,
xi(t) is the expression level (mRNA concentrations) of gene i at time instance t. y=(yij) is an n×n -1 matrix, where yij is zero if ej≠0 and is otherwise an arbitrary scalar coefficient. ∧ =diag (1/ei)
and 1/e is set to be zero if ei=0. U is a unitary m×n matrix of left eigenvectors, ∧=diag (e1,…,en) 145 is a diagonal n×n matrix containing the n eigenvalues and VT is the transpose of a unitary n×n matrix of right eigenvectors [9]. We established network based on the fold change ≥2 distinguished genes and selected parameters as lambda 0.0 because we used one dataset and tried several thresholds 1, 0.5, 0.1, 1.0e-3, 1.0e-4, 1.0e-6, 1.0e-7, 1.0e-8, 1.0e-9. Lambda was a positive parameter, which balanced the matching and sparsity terms in the objective function. Using 150 different thresholds, we could predict various networks with different edge density. 2 Results
2.1 Candidate Novel Activated & Inhibited Genes of GPSM2 Upstream and Downstream Network in Human No-tumor Hepatitis/Cirrhosis and HCC Transformation by GRNInfer 155 We obtained 240 significant high expression molecules (fold change ≥2) from 6144 genes of 25 HCC vs 25 no-tumor hepatitis/cirrhosis in the same GEO Dataset GSE10140-10141( see supplements). We did GRNInfer as follows: Step 1: Format Microarray Datasets into desired datafile. Step 2: Open the datafile by click the 'open' button and browse the datafile location. Step 3: Choose proper parameters. There are two paramters you can choose to provide by altering the 160 default value: Lambda: This parameter is used in the inferring algorithm to adjust the spars structure of the network. The default value is 0.0. Threshold: This parameter is used in the control the output file GRN.dot, which can be visulized by the neato tool of software Graphviz. The Threshold paramter make the edge whose strength of link is smaller than Threshold not shown in the network graph. The smaller this parameter, the more edges in the network graph. We tried 165 GRNInfer in several thresholds 1, 0.5, 0.1, 1.0e-3, 1.0e-4, 1.0e-6, 1.0e-7, 1.0e-8, 1.0e-9. At last we selected threshold 1.0e-9 because its result covered GPSM2 pathway by the CapitalBio MAS 3.0 software from the published data. Step 4: Computing by click the 'Infer' button when the datafile and paramters are ready. Step 5: Checking the results. We identified candidate genes of GPSM2 upstream and downstream network from our constructed total network between no-tumor 170 hepatitis/cirrhosis and HCC transformation by GRNInfer separately, as shown in Tab. 1.
Tab. 1 Candidate upstream and downstream genes of GPSM2 network between no-tumor hepatitis/cirrhosis and HCC transformation by GRNInfer respectively. Upstream
- 4 - 中国科技论文在线 http://www.paper.edu.cn
human no-tumor hepatitis/cirrhosis Human hepatocellular carcinoma NEK2, NUSAP1, CDKN3, DLG7, LCN2, SFRP4, NUSAP1, CDKN3, DLG7, LCN2, SFRP4, RRM2, RRM2, TROAP, HIST1H3H, TAGLN2, MYBL2, TROAP, HIST1H3H, TAGLN2, MYBL2, TK1, PRCC, TK1, PRCC, E2F1, BRCA1, CENPF, SCML2, E2F1, BRCA1, CENPF, SCML2, BIRC5, TOP2A, BIRC5, TOP2A, SPINK1, MDK, GNG10, FOXM1, SPINK1, MDK, GNG10, FOXM1, TRAF2, SPON2, TRAF2, SPON2, RAB3B, ECT2, ENAH, TTK, RAB3B, KIAA0859, ENAH, PIGC, TTK, CDC20, CDC20, PLK4, TSTA3, VDR, MELK, MAP4K4, TSTA3, VDR, MCM2, XRCC2, MELK, MAP4K4, MAP2, AMELY, CDC6, SEMA3B, TUBG1, CCNE2, MAP2, AMELY, CDC6, SEMA3B, KIAA0513, KIAA0513, CSTF2, CPD, IGF2BP3, NCAPH, CSTF2, TNFRSF9, IGF2BP3, NCAPH, YWHAE, SBF1, SBF1, PVRL2, ORC1L, EIF1AX, ZWINT, GDPD5, ORC1L, EIF1AX, ZWINT, GDPD5, ARHGDIG, GRM1, ARHGDIG, MCM4, LLGL2, BUB1B, LLGL2, BUB1B, GALK1, RBCK1, DDX10, C9orf127, GALK1, ACTN2, C9orf127, ACTG2, CLIC1, RNF185, CCNA2, LTBP2, SORT1, ZIC2, CDH13, CCNA2, RBM34, SORT1, ZIC2, CDH13, GPC3, GPC3, ST6GALNAC, CYP21A2, CEBPA, PRKCG, ST6GALNAC, CYP21A2, AKR1B10, CEBPA, HIST1H2AG, UBE2C, CCNB2, CST6, NAT9, HOXD4, PRKCG, HIST1H2AG, UBE2C, CCNB2, CST6, CNTNAP2, MUTYH, PSMC3IP, SFTPA2, CDC2, CD34, MYH6, HOXD4, CAMK1, SFTPA2, ALK, CD34, NFKBIB, TP53I11, CHST1, NOTCH3, HMGB2, NFKBIB, TP53I11, CHST1, NOTCH3, FOLR1, FOLR1, BLVRA, LGALS3, RFC4, MKRN3, CHAF1A, LGALS3, RFC4, MKRN3, CHAF1A, KPNA2, KPNA2, ITGA2, CELSR2, SLC16A3, GNAZ, KCTD2, ITGA2, EPHA4, CELSR2, SLC16A3, GNAZ, TPSD1, GJA5, BAP1, CAD, HIST1H2BJ, PPP1R12B, KCTD2, TPSD1, GJA5, RRP1B, BAP1, CAD, REG3A, ESM1, WDR1, TRIP13, GML, CKS1B, HIST1H2BJ, PPP1R12B, REG3A, TRIP13, GML, ROBO1, FGF9, SYN2, MYCN, ELAVL3, AFP, CCL20, CKS1B, ROBO1, ORC6L, FGF9, SYN2, ELAVL3, LEF1, SERPINB2, IRF5, BCAT1, PRKG2, SLC6A12, AFP, CCL20, CBX5, LEF1, SERPINB2, IRF5, REG1A, C4orf8, TPST2, S100P, NR5A1, MAP2K6, ISG20, BCAT1, SLC6A12, REG1A, C4orf8, TPST2, HIST1H2AD, MMP9, CORO2A, PTHLH, SQLE, S100P, NR5A1, MAP2K6, CHL1, MMP9, PTHLH, PHLDA2, CSPG4, CCNB1, CRYGA, STX1A, UNG, SQLE, PHLDA2, CSPG4, CCNB1, CRYGA, SULT1C2, MAOA, FKBP1B, HOXA5, GAS7, STX1A, UNG, MYOM1, MAOA, ZNF43, FKBP1B, CYP51A1, EYA1, PRSS1, CHRNA4, KATNB1, HOXA5, CYP51A1, EYA1, CHRNA4, KATNB1, MAN2A1, ALDH3A1, F13A1, STMN1, FLJ33790, MAN2A1, ALDH3A1, F13A1, VCAN, EFNA1, EFNA1, PLA2G1B, KCNQ3, PAGE4, TCAP, CDKN2C, PLA2G1B, KCNQ3, PAGE4, TCAP, CDKN2C, NQO1, CYP17A1, KIAA0101, TSHB, KLRC3, DKK1, NQO1, CYP17A1, KIAA0101, TSHB, KLRC3, ESPL1, MRPL49, RIMS3, NKX2-5, PTHR2, NTN1, DKK1, ESPL1, MRPL49, RIMS3, NKX2-5, PTHR2, CTHRC1, TBL3, MMP11, PROK1, TSR1, ADAMDEC1, NTN1, CTHRC1, MMP11, TSR1, ADAMDEC1, SCGB1D2, PCOLCE2, NUP62, NINJ2, LAPTM4B, SCGB1D2, NUP62, NINJ2, LAPTM4B, DMN, RABGGTA, LOX, SOX2, NRXN3, GPSM2 RABGGTA, LOX, SOX2, SSTR5, NRXN3, GPSM2 Downstream human no-tumor hepatitis/cirrhosis Human hepatocellular carcinoma NEK2, NUSAP1, CDKN3, LCN2, SFRP4, RRM2, NEK2, NUSAP1, CDKN3, DLG7, SFRP4, RRM2, TAGLN2, MYBL2, TK1, BRCA1, CENPF, SCML2, TROAP, HIST1H3H, MYBL2, TK1, PRCC, E2F1, BIRC5, TOP2A, SPINK1, MDK, FOXM1, SPON2, BRCA1, CENPF, SCML2, BIRC5, TOP2A, SPINK1, ECT2, KIAA0859, TTK, CDC20, PLK4, TSTA3, MDK, FOXM1, TRAF2, SPON2, RAB3B, ECT2, MAPT, VDR, MCM2, CIAO1, MELK, MAP4K4, KIAA0859, ENAH, PIGC, SLC4A3, TTK, CDC20, CCNE2, MAP2, AMELY, CDC6, SEMA3B, PLK4, TSTA3, MAPT, MCM2, CIAO1, XRCC2, CSTF2, CPD, TNFRSF9, IGF2BP3, NCAPH, CCNE2, MAP2, AMELY, CDC6, SEMA3B, KIAA0513, YWHAE, PVRL2, ORC1L, EIF1AX, ZWINT, CSTF2, TNFRSF9, IGF2BP3, NCAPH, ORC1L, GDPD5, GRM1, ARHGDIG, MCM4, LLGL2, EIF1AX, ZWINT, GDPD5, GRM1, ARHGDIG, MCM4, ACTN2, DDX10, ACTG2, CLIC1, RNF185, LLGL2, BUB1B, GALK1, ACTN2, RBCK1, DDX10, CCNA2, RBM34, SORT1, ZIC2, ST6GALNAC, C9orf127, PTTG1, CLIC1, RNF185, CCNA2, LTBP2, CYP21A2, AKR1B10, CEBPA, PRKCG, UBE2C, SORT1, ZIC2, CDH13, GPC3, CYP21A2, AKR1B10, CCNB2, CST6, MYH6, CNTNAP2, MUTYH, CEBPA, PRKCG, HIST1H2AG, CCNB2, CST6, NAT9, PSMC3IP, SFTPA2, ALK, CDC2, NFKBIB, MYH6, HOXD4, CAMK1, CNTNAP2, PSMC3IP, TP53I11, CHST1, NOTCH3, HMGB2, FOLR1, CDC2, CD34, NFKBIB, TP53I11, NOTCH3, HMGB2, LGALS3, RFC4, MKRN3, KPNA2, ITGA2, FOLR1, LGALS3, RFC4, MKRN3, KPNA2, ITGA2, EPHA4, CELSR2, GNAZ, KCTD2, MAPK3, BAP1, EPHA4, CELSR2, GNAZ, KCTD2, MAPK3, RRP1B, CAD, HIST1H2BJ, PPP1R12B, REG3A, ESM1, BAP1, CAD, HIST1H2BJ, PPP1R12B, REG3A, ESM1, WDR1, LYPD3, TRIP13, GML, CKS1B, ROBO1, WDR1, LYPD3, TRIP13, CKS1B, ROBO1, ORC6L, FGF9, MYCN, AFP, CCL20, CBX5, SERPINB2, FGF9, SYN2, MYCN, AFP, CCL20, CBX5, LEF1, IRF5, BCAT1, PRKG2, SLC6A12, REG1A, C4orf8, SERPINB2, IRF5, ISG20, PRKG2, C4orf8, TPST2, S100P, MMP9, CORO2A, PTHLH, SQLE, S100P, NR5A1, MAP2K6, CHL1, HIST1H2AD, MMP9, PHLDA2, CSPG4, CCNB1, CRYGA, STX1A, CORO2A, PTHLH, SQLE, PHLDA2, CCNB1, CRYGA, UNG, SULT1C2, MYOM1, HOXA5, GAS7, STX1A, UNG, SULT1C2, MYOM1, FKBP1B, HOXA5, CYP51A1, EYA1, CHRNA4, KATNB1, MAN2A1, GAS7, EYA1, PRSS1, KATNB1, MAN2A1, OCRL, OCRL, ALDH3A1, VCAN, STMN1, EFNA1, F13A1, VCAN, STMN1, FLJ33790, EFNA1, PLA2G1B, KCNQ3, PAGE4, CDKN2C, NQO1, KIAA0101, KCNQ3, PAGE4, CDKN2C, NQO1, CYP17A1, TSHB, KLRC3, DKK1, ESPL1, MRPL49, RIMS3, KIAA0101, TSHB, KLRC3, DKK1, ESPL1, MRPL49,
- 5 - 中国科技论文在线 http://www.paper.edu.cn
NTN1, CTHRC1, ADAMDEC1, MS4A2, RIMS3, NKX2-5, NTN1, CTHRC1, TBL3, MMP11, SCGB1D2, PCOLCE2, NUP62, NINJ2, RABGGTA, PROK1, TSR1, ADAMDEC1, MS4A2, SCGB1D2, LOX, SOX2, SSTR5, NRXN3, GPSM2 PCOLCE2, NUP62, NINJ2, LAPTM4B, DMN, RABGGTA, LOX, SOX2, SSTR5, NRXN3, GPSM2 175 2.2 GPSM2 Signal Transduction Network in no-tumor hepatitis/cirrhosis and HCC Transformation by GRNInfer We identified the novel molecules of GPSM2 signal transduction network in no-tumor hepatitis/cirrhosis and HCC transformation, under covering GPSM2 pathway and matching signal transduction enrichment analysis by the CapitalBio MAS 3.0 software from the published data. 180 Upstream MDK, GNG10, VDR, CAMK1, GNAZ, CHL1, CSPG4, CHRNA4, PLA2G1B activated GPSM2, and SSTR5, SFRP4, PPP1R12B, FGF9, CCL20, NR5A1, MAP2K6, GRM1, CLIC1, TRAF2, NFKBIB, EPHA4 inhibited GPSM2; Downstream GPSM2 activated CCL20, VDR, GRM1, NFKBIB, PPP1R12B and inhibited SFRP4, MDK, CLIC1, CNTNAP2, EPHA4, GNAZ, FGF9, PRKG2, CSPG4, CHRNA4, OCRL, SSTR5 in no-tumor hepatitis/cirrhosis, as 185 shown in Fig. 1.
Fig. 1 Activated & inhibited upstream and downstream network of GPSM2 signal transduction in no-tumor hepatitis/cirrhosis by GRNInfer. Arrowhead represents activation relationship, empty cycle represents inhibition relationship. 190 Upstream SFRP4, MDK, TRAF2, CNTNAP2, NFKBIB, GNAZ, FGF9, CCL20, PRKG2, CHRNA4 activated GPSM2, and PLA2G1B, GNG10, NR5A1, MAP2K6, CSPG4, PPP1R12B, VDR inhibited GPSM2; Downstream GPSM2 activated SFRP4, MDK, TRAF2, GRM1, CAMK1, EPHA4, FGF9, CCL20, NR5A1, OCRL, PLA2G1B and inhibited SSTR5, CLIC1, MAP2K6, 195 CHL1, GNAZ, PPP1R12B, CNTNAP2, NFKBIB, PRKG2 in HCC, as shown in Fig. 2.
- 6 - 中国科技论文在线 http://www.paper.edu.cn
Fig. 2 Activated & inhibited upstream and downstream network of GPSM2 signal transduction in HCC by GRNInfer. Arrowhead represents activation relationship, empty cycle represents inhibition relationship.
200 We further computed activation and inhibition numbers of GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation. Our result showed GPSM2 signal transduction network less activation & more inhibition molecular numbers in no-tumor hepatitis/cirrhosis but more activation & less inhibition in HCC, as shown in Tab. 2. Tab. 2 Activated & inhibited upstream and downstream molecular numbers of GPSM2 signal transduction in 205 no-tumor hepatitis/cirrhosis and HCC by GRNInfer. con represents control (human no-tumor hepatitis/cirrhosis). exp: experiment (HCC patients). GPSM2 upstream Term con con exp exp (activation) (inhibition) (activation) (inhibition) Signal Transduction 9 12 10 7 GPSM2 downstream Term Con Con Exp Exp (activation) (inhibition) (activation) (inhibition) Signal Transduction 5 12 11 9 3 Discussion Our aim is to compare and analyze novel GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation for potential novel markers to prognosis and 210 therapy of HCC. In this paper, by compared the different activated & inhibited molecules and numbers of GPSM2 network with GO analysis between no-tumor hepatitis/cirrhosis and HCC transformation, what were (1) different stronger and weaker cellular component, molecular function and biological process between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation? (2) different more and less activation and inhibition 215 numbers between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation? (3) different inferences between GPSM2 signal transduction network of no-tumor hepatitis/cirrhosis and HCC transformation? (4) different evidences from literatures to support the inferences respectively?
- 7 - 中国科技论文在线 http://www.paper.edu.cn
To verify whether our identified genes can separate 2 samples groups (hepatocellular 220 carcinoma vs no-tumor tissue), we did unsupervised clustering. Our result showed most control group (no-tumor hepatitis/cirrhosis) clustered together and most experiment group also together. Two group were observed red colors reflecting all the upregulation and no missing data in both no-tumor hepatitis/cirrhosis and HCC (we did not showed this figure because it was used before). Our unsupervised clustering with expression levels of the identified genes tested separation of 2 225 samples groups (hepatocellular carcinoma vs no-tumor tissue). To confirm our prediction for covering the published data, we setup GPSM2 interaction and pathway in HCC using the CapitalBio MAS 3.0 software for standard comparison with our GPSM2 network. We identified GPSM2 interaction and pathway from our total established network by inputting significant high expression genes (fold change ≥2) to the CapitalBio MAS 230 3.0 software based on the integration of public databases including Gene Ontology, KEGG, BioCarta, GenMapp, Intact, UniGene, OMIM, etc. GPSM2 interaction molecules included GNAI2, HRAS, GPSM2, GNAI1, INSC, GNAO1, GNAI3, CAA77669, GLIS2, NUMA1, RAPS, GPSM2 pathway molecules consisted of PPP1R12B. By similarity comparison, our high expression molecules of HCC did not contain GPSM2 interaction proteins whereas completely covered 235 GPSM2 pathway proteins. To establish candidate novel genes of GPSM2 upstream and downstream network covering GPSM2 pathway, we tried GRNInfer in several thresholds 1, 0.5, 0.1, 1.0e-3, 1.0e-4, 1.0e-6, 1.0e-7, 1.0e-8, 1.0e-9. At last we selected threshold 1.0e-9 to setup GPSM2 network between no-tumor hepatitis/cirrhosis and HCC transformation in order to cover GPSM2 pathway from 240 public databases (Tab. 1). To construct novel GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation respectively by GRNInfer, our GPSM2 network need to match signal transduction enrichment analysis. We identified signal transduction enrichment from our total established enrichment results by inputting significant high expression genes (fold change ≥2) to 245 the CapitalBio MAS 3.0 software based on the integration of public databases including Gene Ontology, KEGG, BioCarta, GenMapp, Intact, UniGene, OMIM, etc. We determined molecules of signal transduction enrichment including SFRP4, MDK, GNG10, TRAF2, TANK, GPSM2, VDR, GRM1, CLIC1, CAMK1, CNTNAP2, NFKBIB, GNAZ, PPP1R12B, FGF9, CCL20, PRKG2, NR5A1, MAP2K6, CHL1, CSPG4, CHRNA4, OCRL, PLA2G1B, MS4A1, SSTR5, EPHA4. 250 Based on signal transduction enrichment, we predicted novel activated & inhibited upstream and downstream network of GPSM2 signal transduction between no-tumor hepatitis/cirrhosis and HCC transformation by GRNInfer separately, as shown in Fig. 1, 2. We further compared and interpret the different molecules of GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation considering activation and inhibition relationship. 255 Firstly, we identified the different novel molecules of GPSM2 signal transduction network in no-tumor hepatitis/cirrhosis compared with HCC. The different upstream GNG10, VDR, CAMK1, CHL1, CSPG4, PLA2G1B activated GPSM2, and SFRP4, TRAF2, GRM1, CLIC1, NFKBIB, EPHA4, FGF9, CCL20, SSTR5 inhibited GPSM2; the different downstream GPSM2 activated VDR, NFKBIB, PPP1R12B and inhibited SFRP4, MDK, EPHA4, FGF9, CSPG4, CHRNA4, 260 OCRL in no-tumor hepatitis/cirrhosis. In order to further interpret molecular mechanism of GPSM2 signal transduction network in no-tumor hepatitis/cirrhosis, we analyzed GO of the different activated and inhibited molecules from no-tumor hepatitis/cirrhosis. For example, activated NFKBIB in no-tumor hepatitis/cirrhosis cellular component included nucleus, cytoplasm, cytosol, and molecular function comprised transcription coactivator activity, signal transducer
- 8 - 中国科技论文在线 http://www.paper.edu.cn
265 activity, protein binding, and biological process was relevant to transcription, signal transduction. Inhibited SFRP4 in HCC cellular component consisted of extracellular region, extracellular space, and molecular function comprised protein binding, and biological process was involved in signal transduction, embryo implantation, Wnt receptor signaling pathway, cell differentiation. Our result demonstrated GPSM2 signal transduction network more nucleus and cytoplasm but less 270 extracellular space protein binding in no-tumor hepatitis/cirrhosis. Therefore, we inferred GPSM2 signal transduction network stronger transcription but weaker cell differentiation as a result increasing cytoplasm protein translation in no-tumor hepatitis/cirrhosis. We also found evidences from literatures to support the inference [11, 12]. Secondly, we identified the different novel molecules of GPSM2 signal transduction network 275 in HCC compared with no-tumor hepatitis/cirrhosis. The different upstream SFRP4, TRAF2, CNTNAP2, NFKBIB, FGF9, CCL20, PRKG2 activated GPSM2, and GNG10, VDR, CSPG4, PLA2G1B inhibited GPSM2; the different downstream GPSM2 activated SFRP4, MDK, TRAF2, CAMK1, EPHA4, FGF9, NR5A1, OCRL, PLA2G1B and inhibited NFKBIB, PPP1R12B, MAP2K6, CHL1 in HCC. In order to further interpret molecular mechanism of GPSM2 signal 280 transduction network in HCC, we analyzed GO of the different activated and inhibited molecules from HCC. For example, activated MDK in HCC cellular component included extracellular region, and molecular function comprised growth factor activity, heparin binding, and biological process was relevant to signal transduction, development, nervous system development, cell proliferation, response to wounding, cell differentiation, adrenal gland development. Inhibited PPP1R12B in 285 HCC cellular component consisted of cytoplasm, and molecular function comprised enzyme activator activity, and biological process was involved in regulation of muscle contraction, signal transduction. Our result indicated GPSM2 signal transduction network more growth factor activity but less cytoplasm enzyme activator activity in HCC. Therefore, we deduced GPSM2 signal transduction network stronger cell proliferation but weaker regulation of muscle contraction as a 290 result inceasing nuclear cell division in HCC. We also found evidences from literatures to support the inference [13-17]. We further computed activation and inhibition numbers of GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation. Our result showed GPSM2 signal transduction network less activation & more inhibition molecular numbers in 295 no-tumor hepatitis/cirrhosis but more activation & less inhibition in HCC. Our computation is consistent with our prediction because cytoplasm protein translation appears less activation and more inhibition molecular pattern compared with nuclear proliferation more activation and less inhibition. 4 Conclusions 300 In conclusion, our high expression molecules of HCC did not contain GPSM2 interaction proteins whereas completely covered GPSM2 pathway proteins by similarity comparison. We identified the different novel activated & inhibited upstream and downstream genes of GPSM2 signal transduction network between no-tumor hepatitis/cirrhosis and HCC transformation, on the condition that our GPSM2 network covered GPSM2 pathway and matched signal transduction 305 enrichment analysis by the CapitalBio MAS 3.0 software from the published data. By compared the different activated & inhibited GPSM2 network with GO analysis between no-tumor hepatitis/cirrhosis and HCC transformation, our result showed GPSM2 signal transduction network: (1) more nucleus and cytoplasm but less extracellular space protein binding in no-tumor hepatitis/cirrhosis; (2) more growth factor activity but less cytoplasm enzyme activator activity in
- 9 - 中国科技论文在线 http://www.paper.edu.cn
310 HCC; (3) less activation & more inhibition molecular numbers in no-tumor hepatitis/cirrhosis but more activation & less inhibition in HCC. We inferred (4) GPSM2 signal transduction network stronger transcription but weaker cell differentiation as a result increasing cytoplasm protein translation in no-tumor hepatitis/cirrhosis; (5) stronger cell proliferation but weaker regulation of muscle contraction as a result inceasing nuclear cell division in HCC. Therefore, it is very useful 315 to identify GPSM2 signal transduction network for the understanding of molecular mechanism between no-tumor hepatitis/cirrhosis and HCC transformation. Acknowledgements This work is supported by the grants from National Natural Science Youth Fundation of China (81501372). 320 References
[1] Branda, M. and J.R. Wands. Signal transduction cascades and hepatitis B and C related hepatocellular carcinoma [J]. Hepatology, 2006, 43 (5): 891-902. [2] Tang, Y., P. Lv, Z. Sun, L. Han, B. Luo, and W. Zhou. 14-3-3zeta up-regulates hypoxia-inducible factor-1alpha 325 in hepatocellular carcinoma via activation of PI3K/Akt/NF-small ka, CyrillicB signal transduction pathway [J]. Int J Clin Exp Pathol, 2015, 8 (12): 15845-15853. [3] Wang, S.D., B.C. Chen, S.T. Kao, C.J. Liu, and C.C. Yeh. Genistein inhibits tumor invasion by suppressing multiple signal transduction pathways in human hepatocellular carcinoma cells [J]. BMC Complement Altern Med, 2014, 14 26. 330 [4] Song, M.K., Y.J. Kim, M. Song, H.S. Choi, Y.K. Park, and J.C. Ryu. Polycyclic aromatic hydrocarbons induce migration in human hepatocellular carcinoma cells (HepG2) through reactive oxygen species-mediated p38 MAPK signal transduction [J]. Cancer Sci, 2011, 102 (9): 1636-1644. [5] Pan, M.H., Y.S. Chiou, W.J. Chen, J.M. Wang, V. Badmaev, and C.T. Ho. Pterostilbene inhibited tumor invasion via suppressing multiple signal transduction pathways in human hepatocellular carcinoma cells [J]. 335 Carcinogenesis, 2009, 30 (7): 1234-1242. [6] Bar-Yehuda, S., S.M. Stemmer, L. Madi, D. Castel, A. Ochaion, S. Cohen, F. Barer, A. Zabutti, G. Perez-Liz, L. Del Valle, and P. Fishman. The A3 adenosine receptor agonist CF102 induces apoptosis of hepatocellular carcinoma via de-regulation of the Wnt and NF-kappaB signal transduction pathways [J]. Int J Oncol, 2008, 33 (2): 287-295. 340 [7] Cheng, J.C., C.H. Chou, M.L. Kuo, and C.Y. Hsieh. Radiation-enhanced hepatocellular carcinoma cell invasion with MMP-9 expression through PI3K/Akt/NF-kappaB signal transduction pathway [J]. Oncogene, 2006, 25 (53): 7009-7018. [8] Tanaka, S., K. Sugimachi, S. Maehara, N. Harimoto, K. Shirabe, and J.R. Wands. Oncogenic signal transduction and therapeutic strategy for hepatocellular carcinoma [J]. Surgery, 2002, 131 (1 Suppl): S142-7. 345 [9] Wang, Y., T. Joshi, X.S. Zhang, D. Xu, and L. Chen. Inferring gene regulatory networks from multiple microarray datasets[J]. Bioinformatics, 2006, 22 (19): 2413-20. [10] Storey., J.D. A direct approach to false discovery rates.[J]. J. Roy. Stat. Soc., Ser. B, 2002, 64: 479-498. [11] Blumer, J.B., J. Chandler, and S.M. Lanier. Two related Heterotrimeric G protein regulators, AGS3 and LGN, exhibit distinct differences in tissue distribution and subcellular location[J]. Faseb Journal, 2002, 16 (4): 350 A577-A577. [12] Blumer, J.B., L.J. Chandler, and S.M. Lanier. Expression analysis and subcellular distribution of the two G-protein regulators AGS3 and LGN indicate distinct functionality - Localization of LGN to the midbody during cytokinesis[J]. Journal of Biological Chemistry, 2002, 277 (18): 15897-15903. [13] Blumer, J.B., R. Kuriyama, T.W. Gettys, and S.M. Lanier. The G-protein regulatory (GPR) motif-containing 355 Leu-Gly-Asn-enriched protein (LGN) and Gi alpha 3 influence cortical positioning of the mitotic spindle poles at metaphase in symmetrically dividing mammalian cells[J]. European Journal of Cell Biology, 2006, 85 (12): 1233-1240. [14] Fukukawa, C., K. Ueda, T. Nishidate, T. Katagiri, and Y. Nakamura. Critical Roles of LGN/GPSM2 Phosphorylation by PBK/TOPK in Cell Division of Breast Cancer Cells[J]. Genes Chromosomes & Cancer, 49 360 (10): 861-872. [15] Mochizuki, N., G. Cho, B. Wen, and P.A. Insel. Identification and cDNA cloning of a novel human mosaic protein, LGN, based on interaction with G(alpha i2)[J]. Gene, 1996, 181 (1-2): 39-43. [16] Morin, X., F. Jaouen, and P. Durbec. Control of planar divisions by the G-protein regulator LGN maintains progenitors in the chick neuroepithelium[J]. Nature Neuroscience, 2007, 10 (11): 1440-1448. 365 [17] Tall, G.G. and A.G. Gilman. Resistance to inhibitors of cholinesterase 8A catalyzes release of G alpha i-GTP and nuclear mitotic apparatus protein (NuMA) from NuMA/LGN/G alpha i-GDP complexes[J]. Proceedings of the National Academy of Sciences of the United States of America, 2005, 102 (46): 16584-16589.
- 10 - 中国科技论文在线 http://www.paper.edu.cn
370 GPSM2在非肿瘤肝炎肝硬化组织和肝 癌之间的信号转导计算网络分析 庄婧，黄菊香，王琳 （北京邮电大学电子工程学院，100876） 摘要：许多论文中提到LGN蛋白（GPSM2）参与转录或细胞分裂。然而，非肿瘤肝炎/肝硬 375 化和肝癌（HCC）转化之间的GPSM2信号转导的分子网络和解释仍有待阐明。本文采用GEO 数据集通过基于GO，KEGG，GenMAPP，BioCarta，Intact，UniGene，OMIM和Disease等的 整合分析的分子注释系统MAS使GPSM2的网络覆盖并匹配其通路进行信号转导的富集度分 析，然后通过GRNInfer算法对比分析了肝癌样本与非肿瘤肝炎及肝硬化样本，构 建了GPSM2 的激活和抑制的上下游信号转导网络，最后通过GO分析比较了非肿瘤肝炎/肝硬化和肝癌之 380 间转化的GPSM2激活与抑制网络的不同。根据实验结果，得出GPSM2信号转导网络的五点 结论：（1）在非肿瘤肝炎/肝硬化中，显示更多的细胞核和细胞质以及较少的细胞外空隙蛋 白质结合；（2）在肝癌中表现出较强的生长因子活性和较弱的胞浆酶活化剂活性；（3）在 非肿瘤肝炎/肝硬化中体现较少的激活分子数量和较多抑制分子数量，然而在肝癌中显示较 多的激活分子和较少的抑制分子数量。（4）因此，我们推断在非肿瘤肝炎/肝硬化中，GPSM2 385 信号转导网络具备更强大的转录功能但较弱的细胞分化特性，从而能增加细胞质蛋白质的转 化；（5）在肝癌中体现更强的细胞增殖同时较弱的肌肉收缩的规律，从而提高核细胞分裂。 关键词：生物医学工程；LGN蛋白（GPSM2）计算网络；信号转导；肝癌；非肿瘤肝炎肝 硬化组织 中图分类号：R735.7
- 11 -