Molecular Informatics, 35:238-252, 2016.   

Constructing and validating 3D-pharmacophore models to a set of MMP-9 inhibitors for designing novel anti-melanoma agents

Turra KM, Pineda Rivelli DP, Barros SBM, Pasqualoto KFM*



A receptor-independent (RI) four-dimensional structure-activity relationship (4D-QSAR) formalism was applied to a set of sixty-four b-N-biaryl ether sulfonamide hydroxamate derivatives, previously reported as potent inhibitors against matrix metalloproteinase subtype 9 (MMP-9). MMP-9 belongs to a group of enzymes related to the cleavage of several extracellular matrix components and has been associated to cancer invasiveness/metastasis. The best RI 4D-QSAR model was statistically significant (N = 47; r2 = 0.91; q2 = 0.83; LSE = 0.09; LOF = 0.35; outliers = 0). Leave-N-out (LNO) and y-randomization approaches indicated the QSAR model was robust and presented no chance correlation, respectively. Furthermore, it also had good external predictability (82 %) regarding the test set (N = 17). In addition, the grid cell occupancy descriptors (GCOD) of the predicted bioactive conformation for the most potent inhibitor were successfully interpreted when docked into the MMP-9 active site. The 3D-pharmacophore findings were used to predict novel ligands and exploit the MMP-9 calculated binding affinity through molecular docking procedure.

DOI: 10.1002/minf.201600004



The rational design of novel drug candidates comprises the application of  in silico approaches, particularly, when the three-dimensional (3D) information regarding a desirable molecular target and/or its ligand, or substrate, or even inhibitor, can be assessed in structural databases, such as Protein Data Bank, PDB (Berman et al., 2010). The use of computer-aided drug design (CADD) strategies in the early phase of drug development can avoid the synthesis of thousands of compounds, driving the efforts to more promising compounds (having suitable pharmacodynamics and pharmacokinetics features as well as low toxicity), reducing the number of biological assays to be performed and, consequently, decreasing the use of animal experimentation. In addition, the time and costs involved in the entire process can also be reduced, and the chances of success may significantly increase to reach the final product.

We applied the receptor-independent (RI) four-dimensional (4D) quantitative structure-activity relationship (QSAR) formalism (Hopfinger et al., 1997) to a set of compounds (b-N-biaryl ether sulfonamide hydroxamate derivatives; N = 64), previously reported as in vitro matrix metalloproteinases (MMP) inhibitors (Yang et al., 2008a, 2008b). The experimental inhibition data against the MMP-9 subtype were used as dependent variable for building the QSAR prediction models. For the first time, a detailed quantitative structure-activity analysis regarding the MMP-9 inhibition was performed.

The MMP-9 protein is overexpressed in several tumors, including melanoma (Campbell et al., 2010; Leifler et al., 2013; Sarvaiya et al., 2013), and seems to be related to tumor growth and progression since it plays a role in the regulation of tumor angiogenesis (Campbell et a., 2010). In this regard, the designing of novel MMP-9 inhibitors could be an interesting therapeutic alternative to metastatic melanoma, for instance.

The optimum mathematical model for predicting the MMP-9 inhibition (N = 47) was statistically validated and used to calculate the inhibitory activity (pIC50calc values) of novel designed compounds (chemical library), before being synthesized and experimentally tested. Then, compounds having higher pIC50calc values would be more likely promising as MMP-9 inhibitors. Furthermore, the descriptors in the QSAR model allow us to properly define the molecular modifications (nature of the substituent group and spatial coordinates) which may either improve or impair the desirable biological response.

Seven novel ligands were designed based on the QSAR model descriptors and had its inhibitory activity against MMP-9 calculated. After that, those ligands were docked into the MMP-9 binding site in order to exploit their calculated binding affinity and compare to the pIC50calc values obtained with the prediction model. The Cartesian coordinates of the complex MMP-9/ligand 7MR ((2R)-2-amino-3,3,3-trifluoro-n-hydroxy-2- {[(4-phenoxyphenyl)sulfonyl]methyl}propanamide) from X-ray diffraction, deposited in PDB (PDB ID 2OW1; Tochowicz et al., 2007), were used as template (Fig. 1A).

Regarding the calculated binding affinity, which is related to the Gibbs free energy (DG), lower energy values (more negative values) indicate that the formation of the enzyme-ligand complex is more energetically favorable. Among the seven novel ligands designed based on the QSAR model descriptors, the ligand L1 (Fig. 1B) presented higher pIC50calc value and better calculated binding affinity (more negative energy value).




Figure 1. Molecular docking procedure (package CLC Drug Discovery Workbench v. 2.4, QIAGEN, Aarhus, 2014): (A) Re-docking of the PDB ligand (light blue: 7MR re-docked, -72.94 kcal/mol; light violet: 7MR from the crystal structure, -69.71 kcal/mol) in the MMP-9 binding site in order to optimize the calculation conditions (root mean square deviation, RMSD = 0.15 Å); (B) Docking of the ligand L1 (yellow; -61.64 kcal/mol; RMSD = 0.07 Å) in the MMP-9 binding site.


Then, ligand L1 was used as template to design more compounds and small molecular modifications were carried out in order to exploit the related changes to the calculated binding affinity values. The docking score and RMSD values for the ligands L8 to L13 are presented in Fig.2. L10 and L13 (Fig. 3) showed docking score values a little bit better (more negative energy values) than the value found for L1, meaning there was an improvement in the binding affinity. Otherwise, L12 presented the lowest calculated binding affinity (energy value less negative) of this subset, suggesting that the substituent groups added in the aromatic rings (addition of Cl and CH3) have impaired the performance in the binding process.



figure2_tableFigure 2. Docking score and RMSD values found for ligands L8 to L13.



Figure 3. Ligands L10 (pink) and L13 (light blue) docked into the MMP-9 binding site (package CLC Drug Discovery Workbench v. 2.4, QIAGEN, Aarhus, 2014).


The ligands L10 and L13 established interactions with the residues His401, Tyr423, Val398, Leu188, Ala189, and Gln402 in the MMP-9 binding site. However, the L10 orientation (accommodation) did not allow any interaction with the zinc metal into the MMP-9 active site.




Figure 4. Binding site interactions for the ligands L10 (pink) and L13 (light blue). The coordination with the Zn2+ metal (444A) is pointed out for the oxygen (carbonyl group) of ligand L13. The distance between the two atoms is 2.39 Å (green line). The protein structure is hidden (Discovery Studio Visualizer v4.0, Accelrys Software Inc., 2005-2013).


The binding mode exploitation at molecular level provides information to drive the development process of novel potentially drug candidates. Thus, the subset of compounds more promising regarding the MMP-9 calculated binding affinity will be synthesized and experimentally tested (in vitro and in vivo models) in order to validate the in silico approach, reinforcing its importance in the whole R&D process. In addition, we will also pursue the structurally optimization of a new lead compound to proceed the development phases as potentially anti-melanoma drug candidate.



H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H., Weissig, I. N., Shindyalov, P. E. Bourne, Nucleic Acids Res. 2000, 28, 235–242.
A. J. Hopfinger, S. Wang, J. S. Tokarski, B. Jin, M. Albuquerque, P. J. Madhav, C., Duraiswami, J. Am. Chem. Soc. 1997, 119, 10509 –10524.
S. M. Yang, R. H. Scannevin, B. Wang, S. L. Burke, L. J. Wilson, P., Karnachi, K. J. Rhodes, B. Lagu, W. V. Murray, Bioorg. Med. Chem. Lett. 2008a, 18, 1135 –1139.
S. M. Yang, R. H. Scannevin, B. Wang, S. L. Burke, Z. Huang, P., Karnachi, L. J., Wilson, K. J. Rhodes, B. Lagu, W. V. Murray, Bioorg. Med. Chem. Lett. 2008b, 18, 1140–1145.
N. E. Campbell, L. Kellenberger, J. Greenaway, R.A., Moorehead, N. M. Linnerth-Petrik, J. Petrik. J. Oncology, vol. 2010, Article ID 586905, 13 pages, 2010. Doi:10.1155/2010/586905
K. S. Leifler, S. Svensson, A. Abrahamsson, C. Bendrik, J. Robertson, J. Gauldie, A.-K. Olsson, C. Dabrosin. J. Immunol. 2013, 190, 4420-4430.
P. J. Sarvaiya, D. Guo, I. Ulasov, P. Gabikian, M. S. Lesniak, Oncotarget. 2013, 4, 2171–2185.
A. Tochowicz, K. Maskos, R. Huber, R. Oltenfreiter, V. Dive, A., Yiotakis, M. Zanda, T. Pourmotabbed, W. Bode, P. Goettig, J. Mol. Biol. 2007, 371, 989–1006.



Kerly Fernanda Mesquita Pasqualoto, Ph.D.

Coordinator of the Computer-Aided Drug Design Division

ALCHEMY – Innovation, Research & Development

Prof. Lineu Prestes Ave., 2242

CIETEC/IPEN – University of São Paulo-USP

São Paulo, SP, 05508-000, Brazil



Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann