Drug–target network and polypharmacology studies of a Traditional Chinese Medicine for type II diabetes mellitus

Drug–target network and polypharmacology studies of a Traditional Chinese Medicine for type II diabetes mellitus

Computational Biology and Chemistry 35 (2011) 293–297 Contents lists available at ScienceDirect Computational Biology and Chemistry journal homepage...

1MB Sizes 4 Downloads 106 Views

Computational Biology and Chemistry 35 (2011) 293–297

Contents lists available at ScienceDirect

Computational Biology and Chemistry journal homepage: www.elsevier.com/locate/compbiolchem

Brief communication

Drug–target network and polypharmacology studies of a Traditional Chinese Medicine for type II diabetes mellitus Jiangyong Gu a , Hu Zhang a,b , Lirong Chen a , Shun Xu b , Gu Yuan a , Xiaojie Xu a,∗ a Beijing National Laboratory for Molecular Sciences, State Key Lab of Rare Earth Material Chemistry and Applications, College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, PR China b Department of Chemistry, Zhengzhou University, Zhengzhou 450052, PR China

a r t i c l e

i n f o

Article history: Received 4 November 2010 Received in revised form 18 June 2011 Accepted 3 July 2011 Keywords: Type II diabetes Network analysis Virtual screening Traditional Chinese Medicine Drug–target network Polypharmacology

a b s t r a c t Many Traditional Chinese Medicines (TCMs) are effective to relieve complicated diseases such as type II diabetes mellitus (T2DM). In this work, molecular docking and network analysis were employed to elucidate the action mechanism of a medical composition which had clinical efficacy for T2DM. We found that multiple active compounds contained in this medical composition would target multiple proteins related to T2DM and the biological network would be shifted. We predicted the key players in the medical composition and some of them have been reported in literature. Meanwhile, several compounds such as Rheidin A, Rheidin C, Sennoside C, procyanidin C1 and Dihydrobaicalin were notable although no one have reported their pharmacological activity against T2DM. The association between active compounds, target proteins and other diseases was also discussed. © 2011 Elsevier Ltd. All rights reserved.

1. Introduction Diabetes mellitus has become a serious health problem since it has affected more than 285 million people according to the statistical data of the International Diabetes Federation (http://www.diabetesatlas.org) and more than 90% of these diabetic patients have type II diabetes mellitus which is mostly caused by insulin resistance. However, T2DM concerns with hereditary factors, insulin resistance (related to free fatty acids, inflammatory cytokines, adipokines, etc.) and ␤-cell dysfunction (Stumvoll et al., 2005). Even more important, T2DM would share some target proteins with cardiovascular diseases underlying biological pathways and processes (DeSouza and Fonseca, 2009), a positive correlation between the risks of T2DM and CVD does exits (Mellbin et al., 2010). Although there are many drugs for T2DM, none of them can cure it (Morral, 2003; Stumvoll et al., 2005). T2DM is so complicated that magic bullets which selectively target at one protein could not pull the whole biological network (interactions between proteins, nucleic acid and small molecules) back to healthy state (Berger and Iyengar, 2009; Janga and Tzakos, 2009). However, network-based approach is becoming more and more powerful to deal with complex systems (Barabasi, 2009; Girvan and Newman, 2002; Janga and Tzakos, 2009; Kitano, 2004; Luni et al., 2010). Recently, Janga and

∗ Corresponding author. Tel.: +86 10 62757456. E-mail address: [email protected] (X. Xu). 1476-9271/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiolchem.2011.07.003

Contreras-Moreira (2010) elucidated the power of network-based approaches in identifying disease markers and drug discovery. Typically, the drug–target network which links drugs and protein targets was used to interpret the mechanism of drug action (Zhu et al., 2009) and explore polypharmacology (Berger and Iyengar, 2009; Gu et al., 2009; Janga and Tzakos, 2009; Vogt and Mestres, 2010; Yildirim et al., 2007) and predict new targets for drugs (Klipp et al., 2010); the protein–disease network was employed to discover the association between diseases and target proteins (Barabasi and Oltvai, 2004; Goh et al., 2007; Janga and ContrerasMoreira, 2010; Klipp et al., 2010; Vogt and Mestres, 2010; Yildirim et al., 2007). Some Traditional Chinese Medicines which contain many active compounds might target at multiple proteins in the biological network and then the biological system would attain new equilibrium in order to reduce the harmful impact (Gu et al., 2009; Janga and Tzakos, 2009). Recently, Tasly Pharmaceutical Corporation (Tianjin, China) developed a medical composition (Tangminling Pills) which was very effective for relieving T2DM (Tong et al., 2009). This medical composition comprises eleven Chinese herb medicines: Trichosanthes kirilowii, Bupleurum longiradiatum, Citrus sinensis, Rheum officinale, Pinellia ternata, Scutellaria baicalensis, Coptis chinensis, Paeonia sterniana, Prunus mume, Astragalus membranaceus and Crataegus pinnatifida. 676 molecules contained in the medical composition were retrieved and docked to 37 T2DM-related proteins and then network analysis was conducted to elucidate the action mechanism. We found that more than one hundred active

294

J. Gu et al. / Computational Biology and Chemistry 35 (2011) 293–297

2008) and KEGG pathway (Kanehisa et al., 2010) were downloaded from RCSB protein data bank (Berman et al., 2000). These crystal structures were loaded into Cerius2 and a flexible docking between them and 676 compounds was conducted by the LigandFit module. The non-default parameters for docking were listed below: Monte Carlo method was used for sampling conformations, 15,000 trials per compound; site partitioning was used to ensure the full access of potential docking orientation in the active site; 1.0 nm was the cut-off distance for Non-bonded interaction and Dreiding force field was used for the grid energy calculations and scoring the docking. 2.3. Network analysis In this work, we constructed two networks by Cytoscape 2.8.1 (Smoot et al., 2011): one is the drug–target network (D–T network) and another is drug–drug association network (D–D network). The D–T network (Fig. 2A) was constructed by linking the compounds and proteins if the docking score of a compound and a target protein was in the top twenty (top 3% in all compounds). The D–D network (Fig. 2B) was built by linking two compounds if they shared one or more target proteins. 3. Results and discussion

Fig. 1. The protocol of virtual screening and network analysis.

compounds might interact with target proteins. The whole biological network would be stabilized in a healthy state or be pulled back to normal when it is in an unhealthy state. 2. Methods 2.1. Distribution of molecules in chemical space The schematic protocol of this method is shown in Fig. 1. First, 676 compounds (Table S1) which were contained in eleven herbs were retrieved from Beilstein database and Chinese Herbal Drug Database (Qiao et al., 2002). Cerius2 4.10 was employed to add hydrogen atoms and optimize the conformation of these molecules. The molecular descriptors of these compounds were calculated in the QSAR module of Cerius2 on the workstation of IBM IntelliStation Z Pro and several important molecular descriptors are listed in Table 1. Then principal component analysis (PCA) was also conducted by QSAR module to visualize the distribution (Fig. S1). 2.2. Virtual screening The crystal structures of protein–ligand complex of 37 proteins which were related to T2DM according to DrugBank (Wishart et al., Table 1 The mean, median, maximum and minimum of key molecular descriptors of 676 compounds contained in the medical composition. Descriptors

Mean

Max

Min

Median

Molecular weight Number of chiral centers No. of H-bond acceptors No. of H-bond donors Number of rotatable bonds Surface area Vm AlogP98 Rule of 5 violations

395.71 4.5 7.4 4.1 8.8 432.0 351.1 2.2 0.99

1707.22 33 47 28 46 1503.3 1292.4 12.8 4

615.48 0 0 0 0 84.3 55.8 −5.0 0

60.05 2 6 3 7 406.6 318.6 2.1 0

Table 1 shows clearly that most compounds contained in the medical composition have good drug-like properties since most of them do not violate Lipinski’s rule of five (Lipinski et al., 1997). For example, the mean of number of H-bond Acceptors and number of H-bond Donors is 7.4 and 4.1, respectively. The mean of violations of rule of five is less than 1. Moreover, the distribution of these molecules in chemical space can be visualized by PCA which could reduce dimension. The first two and first three principal components were plotted in Fig. S1. It shows that these molecules are widely distributed in the chemical space and so it provides opportunities to find active compounds. We compared the distribution between these molecules from the medical composition and known drugs for T2DM from DrugBank (Wishart et al., 2008). The large portion of overlap (Fig. S1C) indicated that the nature of these compounds were very similar to that of known drugs for T2DM. In order to evaluate the accuracy of this docking, we checked the hit rate (the ratio of active compounds to top twenty compounds) for each target protein. Generally, the hit rate of virtual screening is 35% (Doman et al., 2002). Herein the hit rate of for the different proteins varies from 30% to 75% (Table S3). For example, in the top twenty compounds of the virtual screening of insulin receptor and PPAR␥, there were 14 and 10 active compounds (Table S4) whose biological activity for T2DM or other related diseases were reported in the literature in the two sets, respectively. Meanwhile, the hit rate depends heavily on the accuracy of the crystal structure of protein and protein type. Jones et al. found that more false ˚ structure positives were generated if poor resolution (below 2.5 A) was adopted (Jones et al., 1997). Plewczynski et al. (2011) advised that the molecular weight of ligand should not exceed 1000 amu and it was better to use crystal structures rather than NMR structures. However, docking is an effective method to help us find active compounds. Since this medical composition has good effects on type II diabetic patients according to clinical trials, the action mechanism is most probably due to the interactions between the compounds and target proteins related to T2DM. However, T2DM is a complex disease and it concerns with many genes and gene products (Stumvoll et al., 2005). It is most likely that these gene and gene products make up a great and interlinked network so that these building blocks of the body could have function as a whole. Meanwhile, when a drug which targets at a protein in this network, it would

J. Gu et al. / Computational Biology and Chemistry 35 (2011) 293–297 Table 2 Network properties of the D–T network and D–D network.

D–T network D–D network

Average degree

Network density

Network centralization

Cluster coefficient

7.63 48.8

0.040 0.313

0.065 0.443

0.0 0.75

have some effects on the body through this network. More importantly, multiple drugs may target at several proteins and then the whole network would be affected. The general network properties of the D–T network and D–D network (Fig. 2) are listed in Table 2. Typically, the D–T network is not so dense and most molecules target at few proteins. In order to find community structures in D–T network (Fig. 2A), a hierarchical clustering was conducted by clusterMaker (http://www.cgl.ucsf.edu/cytoscape/cluster/clusterMaker.html) which was a plugin of Cytoscape 2.8.1 (Smoot et al., 2011). There are four main branches in the node tree (Fig. S2) and then we found three major clusters and a small cluster (Fig. 2C) in D–T network by using k-means method in igraph-0.5.4 (modularity and q was 0.5 and 0.1, respectively) (Csardi and Nepusz, 2006). The third cluster (clus3) the smallest cluster which only comprises Glucokinase and its potential drugs. The proteins in each other clusters would be highly relevant to each other. For example, Glycogen synthase kinase-3 beta and Protein Kinase C are both important proteins in glycogen synthesis (cluster 1). Glucagon-like peptide-1 receptor

295

(GLP1R) and Insulin degrading enzyme (IDE) are also relevant in the D–T network (cluster 2). Actually, GLP1R can bind its agonist glucagon-like peptide-1 to increase insulin secretion (Runge et al., 2008). However, IDE is a protease which can cleave insulin to maintain the homeostasis of insulin (Camberos and Cresto, 2007). These proteins are also related to other diseases. For example, Protein Kinase C and Peroxisome proliferator-activated receptors are important targets of cardiovascular disease and thrombosis; Tumor necrosis factor ␣ plays a very important role in the developmental process of multiple cancers. Those molecules contained in Tangminling Pills targeted at these proteins and so would have some effects on related diseases. Tasly Pharmaceutical Corporation is developing new indications of Tangminling Pills now. Fig. 2B is so complicated that we cannot obtain visual information. However, network analysis did a good job. We analyzed the degree and betweenness of the compounds in these two networks (Table S5) by using the software CentiBin 1.4.2 (Junker et al., 2006). Generally, those nodes (compounds) which have higher degree would have larger betweenness according to Table S5 in both D–D and D–T network and then these compounds would be more important (Goni et al., 2008; Jeong et al., 2001). That is, these compounds might exert some influence on the body if they are taken into the body since they play important roles in the biological network. There are 12 and 10 known active compounds in the top 20 molecules which have highest degree and largest betweenness in the D–T network, respectively. The number of known active compounds in the top 20 molecules which have highest degree and

Fig. 2. The drug–target association network (A) and drug–drug association network (B). The white and red ellipse represent target proteins related to T2DM and compounds contained in the medical composition, respectively. (C) Three major clusters and a small cluster in D–T network by using k-means method in igraph-0.5.4, detail information of each cluster can be found in supplementary information.

296

J. Gu et al. / Computational Biology and Chemistry 35 (2011) 293–297

Fig. 2. (Continued )

largest betweenness in the D–D network is 10 and 12, respectively. It indicates that the hit rate of finding active compounds by using network based analytical methods is more than 50%. These known active compounds (Table 3) in the four sets are nearly the same and they would be key compounds which play a role in relieving T2DM. Meanwhile, those molecules whose activity have not been reported would be potential active compounds and worth testing. The action mechanism of this medical composition is that these active compounds might target at the whole network rather than target at one protein and then it provides a new approach

for drug discovery since it may reduce the side effect and drug resistance. Two compounds (Danshensu (312) and Acetylcholine (411)) were clearly worth considering since they formed a bridge between two clusters in Fig. 2B. However, their biological functions for T2DM have not been reported and they may be novel lead compounds. In conclusion, we constructed drug–target association network and drug–drug association network which can help to elucidate the action mechanism. Meanwhile, we use the D–T network to interpret the pharmacology of the medical composition, that is the active

J. Gu et al. / Computational Biology and Chemistry 35 (2011) 293–297

297

Table 3 Key players in the medical composition according to network analysis of D–D network and D–T network. D–T network

D–D network

Index

Known

Chemical name

Index

Known

Chemical name

21 190 298 262 10 304 213 300 194 209

Yes Yes No Yes Yes No No No Yes No

Isostrictinin 1,6-Di-O-galloylglucose Rheidin A Rheindianthron Tellimagrandin II Sennoside C Procyanidin B-2 8-C-␤-d-glucopyranoside Rheidin C 3-O-galloylprocyanidin B-1 Procyanidin B-3 7-O-␤-d-glucopyranoside

226 292 18 394 202 303 15 268 255 354

No No Yes Yes No No Yes Yes Yes No

Piceatannol 3-O-␤-d-(6 -O-galloyl) Palmidin A 5-Desgalloylstarchyurin Procyanidin C1 (+)-Catechin 3 -O-␤-d-glucopyranoside Sennidin C 1,2,6-Trigalloylglucopy-ranose Sennoside A Anthraglycoside B Dihydrobaicalin

compounds contained in the medical composition target at multiple proteins related to T2DM and then the whole network is affected and new equilibrium would be reached. Acknowledgment This work was supported by National Key Special Project of Science and Technology for Innovation Drugs (Grant No. 2008ZX09401-006). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.compbiolchem.2011.07.003. References Barabasi, A.L., 2009. Scale-free networks: a decade and beyond. Science 325, 412–413. Barabasi, A.L., Oltvai, Z.N., 2004. Network biology: understanding the cell’s functional organization. Nat. Rev. Genet. 5, 101–115. Berger, S.I., Iyengar, R., 2009. Network analyses in systems pharmacology. Bioinformatics 25, 2466–2472. Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov, I.N., Bourne, P.E., 2000. The protein data bank. Nucleic Acids Res. 28, 235–242. Camberos, M.D., Cresto, J.C., 2007. Insulin-degrading enzyme hydrolyzes ATP. Exp. Biol. Med. 232, 281–292. Csardi, G., Nepusz, T., 2006. The igraph software package for complex network research. InterJournal Complex Systems, 1695–1703. DeSouza, C., Fonseca, V., 2009. Therapeutic targets to reduce cardiovascular disease in type 2 diabetes. Nat. Rev. Drug Discov. 8, 361–367. Doman, T.N., McGovern, S.L., Witherbee, B.J., Kasten, T.P., Kurumbail, R., Stallings, W.C., Connolly, D.T., Shoichet, B.K., 2002. Molecular docking and highthroughput screening for novel inhibitors of protein tyrosine phosphatase-1B. J. Med. Chem. 45, 2213–2221. Girvan, M., Newman, M.E.J., 2002. Community structure in social and biological networks. Proc. Natl. Acad. Sci. U.S.A. 99, 7821–7826. Goh, K.I., Cusick, M.E., Valle, D., Childs, B., Vidal, M., Barabasi, A.L., 2007. The human disease network. Proc. Natl. Acad. Sci. U.S.A. 104, 8685–8690. Goni, J., Esteban, F.J., de Mendizabal, N.V., Sepulcre, J., Ardanza-Trevijano, S., Agirrezabal, I., Villoslada, P., 2008. A computational analysis of protein–protein interaction networks in neurodegenerative diseases. BMC Syst. Biol. 2, 52. Gu, J.Y., Yuan, G., Zhu, Y.H., Xu, X.J., 2009. Computational pharmacological studies on cardiovascular disease by Qishen Yiqi Diwan. Sci. China Ser. B 52, 1871–1878. Janga, S.C., Contreras-Moreira, B., 2010. Dissecting the expression patterns of transcription factors across conditions using an integrated network-based approach. Nucleic Acids Res. 38, 6841–6856.

Janga, S.C., Tzakos, A., 2009. Structure and organization of drug–target networks: insights from genomic approaches for drug discovery. Mol. Biosyst. 5, 1536–1548. Jeong, H., Mason, S.P., Barabasi, A.L., Oltvai, Z.N., 2001. Lethality and centrality in protein networks. Nature 411, 41–42. Jones, G., Willett, P., Glen, R.C., Leach, A.R., Taylor, R., 1997. Development and validation of a genetic algorithm for flexible docking. J. Mol. Biol. 267, 727–748. Junker, B.H., Koschutzki, D., Schreiber, F., 2006. Exploration of biological network centralities with CentiBiN. BMC Bioinformatics 7, 219. Kanehisa, M., Goto, S., Furumichi, M., Tanabe, M., Hirakawa, M., 2010. KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res. 38, D355–D360. Kitano, H., 2004. Biological robustness. Nat. Rev. Genet. 5, 826–837. Klipp, E., Wade, R.C., Kummer, U., 2010. Biochemical network-based drug–target prediction. Curr. Opin. Biotechnol. 21, 511–516. Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 1997. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 23, 3–25. Luni, C., Shoemaker, J., Sanft, K., Petzold, L., Doyle, F., 2010. Confidence from uncertainty – a multi-target drug screening method from robust control theory. BMC Syst. Biol. 4, 161. Mellbin, L.G., Anselmino, M., Ryden, L., 2010. Diabetes, prediabetes and cardiovascular risk. Eur. J. Cardiovas. Prev. Rehabil. 17, S9–S14. Morral, N., 2003. Novel targets and therapeutic strategies for type 2 diabetes. Trends Endocrinol. Metab. 14, 169–175. Plewczynski, D., Lazniewski, M., Augustyniak, R., Ginalski, K., 2011. Can we trust docking results? Evaluation of seven commonly used programs on PDBbind database. J. Comput. Chem. 32, 742–755. Qiao, X.B., Hou, T.J., Zhang, W., Guo, S.L., Xu, S.J., 2002. A 3D structure database of components from Chinese traditional medicinal herbs. J. Chem. Inform. Comput. Sci. 42, 481–489. Runge, S., Thogersen, H., Madsen, K., Lau, J., Rudolph, R., 2008. Crystal structure of the ligand-bound glucagon-like peptide-1 receptor extracellular domain. J. Biol. Chem. 283, 11340–11347. Smoot, M.E., Ono, K., Ruscheinski, J., Wang, P.L., Ideker, T., 2011. Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics 27, 431–432. Stumvoll, M., Goldstein, B.J., van Haeften, T.W., 2005. Type 2 diabetes: principles of pathogenesis and therapy. Lancet 365, 1333–1346. Tong, X.L., Zhu, Y.H., Zhou, S.P., 2009. A pharmaceutical composition for treating diabetes, China patent, CN101357174. Vogt, I., Mestres, J., 2010. Drug–target networks. Mol. Inform. 29, 10–14. Wishart, D.S., Knox, C., Guo, A.C., Cheng, D., Shrivastava, S., Tzur, D., Gautam, B., Hassanali, M., 2008. DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res. 36, D901–D906. Yildirim, M.A., Goh, K.I., Cusick, M.E., Barabasi, A.L., Vidal, M., 2007. Drug–target network. Nat. Biotechnol. 25, 1119–1126. Zhu, M., Gao, L., Li, X., Liu, Z.C., Xu, C., Yan, Y.Q., Walker, E., Jiang, W., Su, B., Chen, X.J., Lin, H., 2009. The analysis of the drug-targets based on the topological properties in the human protein–protein interaction network. J. Drug Target. 17, 524–532.