manuscript doi: 10.5599/admet.0000.0000 47 ADMET & DMPK 5(1) (2017) 47-56; doi: 10.5599/admet.5.1.327 Open Access : ISSN : 1848-7718 http://www.pub.iapchem.org/ojs/index.php/admet/index Original scientific paper The inhibition of adenylate kinase by 2,4-thiazolidinedione evaluated by protein-ligand docking Mihaela Ileana Ionescu Iuliu Hatieganu University of Medicine and Pharmacy, Department of Microbiology, 6 Pasteur, Cluj-Napoca, Romania Corresponding Author: E-mail: mionescu@umfcluj.ro Received: July 14, 2016; Revised: February 16, 2017; Published: March 25, 2017 Abstract Due to its crucial role in nucleotide metabolism, adenylate kinase deserves a special attention in screening of potential inhibitors. Herein, we report the assessment of the relative orientation of the ligand 2,4- thiazolidinedione to adenylate kinase crystallized in closed conformation. Protein-ligand docking was performed to estimate the binding energy and inhibition constant of 2,4-thiazolidinedione to the adenylate kinases’ active sites from different organisms. Our results revealed the best orientation of 2,4- thiazolidinedione is with Gram-positive and acid fast bacteria adenylate kinase – Ki = 0.76±0.1 mM and binding energy -4.26±0.08 kcal/mol. Human adenylate kinases display unfavourable interactions, the binding affinity fluctuating among Ki=0.84 mM and 8.8 mM (3.88±3.51); the energy binding -3.56±0.57. From the three human adenylate kinases analysed, only isoenzyme 2 shows a binding conformation similar to its counterpart from E. coli. Adenylate kinase - this small enzyme needed for survival of every organisms - interacts differently with 2,4-thiazolidinedione, this selectivity being the most important evidence of the present study. Keywords molecular docking; 2, 4-thiazolidinedione Introduction Bacterial resistance to antibiotics is an evolving problem of medical practice [1,2]. Discovery of new antibacterial agents and/or reevaluating of different targets are impetuously needed [3], [4]. These kinds of studies are usually constructed by comparison of well characterized pathogens. Escherichia coli (E. coli) is an opportunistic Gram-negative pathogen, widely isolated from nosocomial infections [5]. Streptococcus pneumoniae (S. pneumoniae) – a Gram-positive cocci – is an extracellular organism responsible for pneumoniae, otitis media, meningitis and even sepsis in very young children and adults above 65 years [6,7]. Tuberculosis treatment remains one of the most challenging problems for clinicians; multidrug- resistant Mycobacterium tuberculosis (M. tuberculosis) strains considerably diminish treatment options [8]. Adenylate kinase (AK) – ATP: AMP phosphotransferase; EC 2.7.3.4 – belong to the nucleotide kinase super family which catalyzes conversion between adenylate nucleotide by the following reaction: Mg 2+ ATP + AMP ↔ Mg 2+ ADP + ADP. AK is widely studied due to its implication in nucleotide metabolism and to its ubiquities – it is described in Archaea, prokaryotes and eukaryotes. Some AKs are very well characterized, http://www.pub.iapchem.org/ojs/index.php/admet/index mailto:mionescu@umfcluj.ro Mihaela I. Ionescu ADMET & DMPK 5(1) (2017) 47-56 48 many of them have the crystal structures solved and biochemical parameters established. AK is a flexible protein, which can adopt different conformations during catalytic process [9,10]. AK is, due to its universality and its flexibility, an attractive target for screening of new inhibitors. Thiazolidinediones (TZDs) derivates – synthetic agonists for the peroxisome proliferator-activating receptor- gamma receptor – were demonstrated as potential inhibitors of S. pneumoniae out-growth [11]. As we are interested in inhibition of AK by thiazolidine derivates [12], the present study is focused on 2,4- thiazolidinedione interaction with the active site of bacterial and human AKs. Materials and Methods Data Collection and Design The sequences and annotations for the proteins examined were collected from NCBI Protein Database (http://www.ncbi.nlm.nih.gov/protein). AKs X-ray 3D structures were downloaded from the Protein Data Bank (http://www.rcsb.org/pdb). AKs co-crystallized with the substrates or substrates analogs like the inhibitor P1,P5-di(adenosine-5')pentaphosphate (Ap5A) (C20H29N10O22P5), were stored for further procedures. The structure of the ligand 2,4-thiazolidinedione (C3H3NO2S) was retrieved from the Open Chemistry Database – National Center for Biotechnology Information PubChem Compound Database; CID=5437, https://pubchem.ncbi.nlm.nih.gov/compound/5437. This open database provided chemicals as SDF files. Protein-ligand docking To analyze the binding of 2,4-thiazolidinedione, we performed protein-ligand docking by AutoDock4, a widely used non-commercial program, which provided useful instruments for accurate prediction of binding interaction [13–15]. The AutoDockTools (http://mgltools.scripps.edu/downloads) offer an excellent interface for non-bioinformaticians. Interpreting the docking data was possible by employing of free 3D molecular visualization tools – BIOVIA Discovery Studio Visualizer 2016 (DS Visualizer 2016); (http://accelrys.com/products/discovery-studio/visualization-download.php). AKs included in the present study were presented in the Table 1. Firstly, from the 3D structure, substrate or substrate analogs (Ap5A) and all solvent molecules were removed and a closed conformation was obtained, which allows an algorithm based on a rigid protein structure. In our case, this docking algorithm was sufficient to obtain reliable data. The Lamarckian Genetic Algorithm (LGA), with a maximum of 2,500,000 energy evaluations, was choose as searching parameter [16]. Multiple sequence alignment Multiple sequence alignment was performed using Clustal Omega program (http://www.ebi.ac.uk/Tools/msa/clustalo), an open access tool provided by the European Bioinformatics Institute – EMBL-EBI [21,22]. Statistical evaluations: Data were expressed as mean ± SD. Inclusion criteria: AKs from any microorganism or from human origin with crystal structure determined by X-ray crystallography co-crystallized with substrate or substrate analogs, so a closed conformation is provided. Exclusion criteria: AKs with no crystal structure recorded in protein databases or 3D structure in open conformation. http://www.ncbi.nlm.nih.gov/protein http://www.rcsb.org/pdb http://pubchem.ncbi.nlm.nih.gov/search/#collection=compounds&query_type=mf&query=C3H3NO2S&sort=mw&sort_dir=asc https://pubchem.ncbi.nlm.nih.gov/compound/5437 http://mgltools.scripps.edu/downloads http://accelrys.com/products/discovery-studio/visualization-download.php http://www.ebi.ac.uk/Tools/msa/clustalo ADMET & DMPK 5(1) (2017) 47-56 Adenylate kinase inhibition by 2,4-thiazolidinedione doi: 10.5599/admet.5.1.327 49 Table 1 The main characteristics of the AKs used in protein-ligand docking ACCESSION NUMBER Source co-crystallized with substrate or substrate analogs No. of residues References Bacterial source 3HPQ E. coli Ap5A 214 [17] 1ANK E. coli AMP, AMPPNP 214 [18] 4W5J S. pneumoniae Ap5A, Mg 2+ 217 [19] 1ZIP B. stearothermophilus Ap5A, Mn 2+ , and Mg 2+ 217 [9] 2CDN M. tuberculosis two ADP molecules and Mg 2+ 201 [20] 2RGX Aquifex aeolicus Ap5A, Zn 2+ 206 [10] Homo sapiens 2C95 (Chain A – AK1) Bis(Adenosine)-5'-Tetraphosphate; malonate ion 196 Unpublished 2C9Y (Chain A – AK2) Bis(Adenosine)-5'- Tetraphosphate; Ethylene Glycol 242 Unpublished 2BWJ (Chain A – AK5) AMP, Cl - , SO4 2- 199 Unpublished Results and Discussion Protein-ligand docking AKs from different sources were subjected to docking with 2,4-thiazolidinedione. It is already established that AK is a natural flexible protein, binding and releasing of its substrates are followed by motions of the AMP-binding domain and LID domain. As a result of sophisticated conformational transitions a so-called closed conformation results during catalytic process [23,24]. Starting from these premises, rigid protein structures were considered for further docking algorithm. For each AK, ten conformations were obtained as a result of docking process. The conformation with the smallest binding energy (∆G) and the smallest inhibition constant (Ki) was considered the best. As it is shown in the Table 2, regarding AKs from Gram-positive and acid fast species, ∆G values are almost the same with a mean of -4.26±0.08 kcal/mol, and Ki are clustered around the mean 0.76±0.1 mM. On the other hand, the human counterpart AKs exhibit more diverse values. Human AK isoenzyme 2 (2C9Y) – mitochondrial origin – exhibits values similar to enzymes of bacterial sources. Active sites residues which interact with the ligand 2,4-thiazolidinedione are shown in the Table 3. The most striking observation is related to AKs of human origin, as it is illustrated in the Figures 1 and 2. The AK5 (2BWJ) – expressed exclusively in the brain [25] – is clearly a poor target, none of the residues from the active site being involved in 2,4-thiazolidinedione binding. Further, in the case of AK2 (2C9Y), the ligand forms standard hydrogen bonds with two nonpolar residues from AMP-binding region (Gly100 and Phe101) and with Lys28 from ATP-binding domain. Besides, from the AKs examined in the present study, AK2 and one AK from E. coli 3HPQ are the only ones which bind 2,4-thiazolidinedione with P-loop domain residues (Lys28 and Lys13, respectively). Notably, the same oxygen atom (O2) (Scheme 1) of the ligand forms hydrogen bond with lysine but different atoms are involved – NZ regarding 2C9Y and CE for 3HPQ (Figures 3 and 4). On the contrary, apart from the residues involved in AMP-binding, AK1 shows only a glycine from NMP-binding domain which binds the ligand. This might be a reasonable explanation of almost twofold value Ki of AK1 (1.99 mM) compared with AK2 (0.84 mM). A unique detail is observed in the case of Gram-positive and acid fast AKs – a favorable interaction between Pi-sulfur of the ligand and the nonpolar residue Phe35 from NMP-binding domain (Figures 5 and Mihaela I. Ionescu ADMET & DMPK 5(1) (2017) 47-56 50 6) might play an important role in stabilization of AK-2,4-thiazolidinedione complex [26]. Multiple sequence alignment exhibits a conservative replacement of Phe35 with leucine on other AKs. Moreover, in A. aeolicus’ AK, van der Waals’ forces between Leu35 and the only sulfur atom of the ligand are noticed. Table 2. Energy binding and inhibition constant (Ki) of the best conformation of the complex AK- 2, 4-thiazolidinedione AK ∆G, Kcal/mol Ki, mM Scheme 1. The numbering of 2,4-thiazolidinedione atoms Gram-negative bacteria 3HPQ -4.0 1.17 1ANK -4.18 0.86 2RGX -4.22 0.81 Mean ± SD -4.13±0.09 0.94±0.16 Gram-positive and acid fast bacteria 1ZIP -4.36 0.64 4W5J -4.27 0.77 2CDN -4.16 0.89 Mean ± SD -4.26±0.08 0.76±0.1 Homo sapiens 2C95 -3.68 1.99 2C9Y -4.19 0.84 2BWJ -2.8 8.8 Mean ± SD -3.56±0.57 3.88±3.51 Total Mean ± SD -3.98 ± 0.46 1.86 ± 2.48 Table 3. Residues of substrate binding domains, according to Protein Database (www.rcsb.org/pdb/home/home.do) details Code color for interactions with 2,4-thiazolidinedione: in green are shown conventional hydrogen bonds; blue – van der Waals’ forces, brown – carbon hydrogen bonds; orange – Pi-sulfur, red – unfavorable acceptor-acceptor AK NMP-binding domain AMP binding LID domain ATP binding (P-loop) 3HPQ Ser30-Val60 Lys57-Val59, Gly85-(Phe86, Pro87)- Arg88, Thr31, Arg36, Gln92, Arg167 Gly122-Asp159 Gly10-(Lys13)-Thr15, Val132- Tyr133, Arg119, Arg123, Lys200 1ANK Ser30-(Leu58)- Val59 Lys57-(Leu58)-Val59, Gly85-(Phe86)- Arg88, Thr31, Arg36, Gln92, Thr156, Arg167 Gly122-Asp159 Gly10-Thr15, Val132-Tyr133, Arg119, Arg123, Lys200 2RGX Ser30-(Thr31/Thr31, Leu35, Leu58)-Val59 Glu57-Val59, Gly82, Phe83, Pro84, Arg85, Thr31/Thr31, Gln89, Arg131 Gly123-Asp153 Gly10-Thr15, Val133-Tyr134, Arg120, Arg124, Lys189 1ZIP Ser30-(Thr31, Phe35, Leu58)-Val59 Asp57-(Leu58)-Val59, Gly85-Arg88, Ser30, Arg36, Gln92, Gln159, Arg171 Gly126-Asp163 Gly10-Thr15, Thr136-Tyr137, Arg127, Gln199 4W5J Ser30-(Phe35)- Val59 Glu57-(Leu58)-Val59, Gly85-(Gly86, Tyr87)-Arg88, Thr31, Arg36, Gln93, Arg156, Arg167 Gly127-Asp159 Gly10-Thr15, Thr137-Phe138, Arg128, Gln195 2CDN Ser30-(Phe35, Leu58)- Val59 Asp57-(Leu58)-Val59, Gly85-(Tyr86)- Arg88, Thr31, Arg36, Gln92, Arg129, Arg140 Gly126-Asp132 Gly10-Thr15, Arg127, Gly166 2C95 Ser38-(Gly40)- Val67 Gln65-Val67, Gly94-Arg97, Thr39, Arg44, Gln101, Arg138, Arg149 Lys131-Asp141 Gly18-Thr23, Arg132, Gly177 2C9Y Ala45-Val74 Lys72-Val74, Gly100, Phe101, Pro102, Arg103, Thr46, Arg51, Gln107, Arg175, Arg186 Gly141-Asp178 Gly25-(Lys28)-Thr30, Ser151- Tyr152, Arg138, Arg142, Gln214 2BWJ Ser41-Val70 Asp68-Val70, Gly97-Arg100, Thr42, Arg47, Arg104, Arg152 Gln136-Asp144 Gly21-Thr26, Arg135, Gly180 http://www.rcsb.org/pdb/home/home.do ADMET & DMPK 5(1) (2017) 47-56 Adenylate kinase inhibition by 2,4-thiazolidinedione doi: 10.5599/admet.5.1.327 51 The interactions of 2,4-thiazolidinedione with AKs of Gram-negative bacteria A. aeolicus – a hyperthermophilic species – and two structures from E. coli are shown in the Figures 3 and 4. In the latter cases, even though the aminoacid sequences and the structures are identical, different ligand orientations were observed. 2C95 2C9Y 2BWJ Figure 1. Human AKs docked with 2,4-thiazolidinedione. 2C95-AK1; 2C9Y-AK2; 2BWJ-AK5 Code color for interactions: in green are shown conventional hydrogen bonds; yellow – carbon hydrogen bonds; red – unfavorable bump 2C95 2C9Y 2BWJ Figure 2. Interaction 2,4-thiazolidinedione with of human AKs 3HPQ 1ANK 2RGX Figure 3. Interaction 2,4-thiazolidinedione with of AKs from Gram-negative microorganisms Mihaela I. Ionescu ADMET & DMPK 5(1) (2017) 47-56 52 3HPQ 1ANK 2RGX Figure 4. AKs from Gram-negative microorganisms docked with 2,4-thiazolidinedione. 3HPQ-E. coli; 1ANK- E. coli; 2RGX- A. aeolicus. Code color for interactions: in green are shown conventional hydrogen bonds; yellow – carbon hydrogen bonds; orange – Sulfur-X; red – unfavorable acceptor-acceptor One explanation could reside on the fact that the two enzymes being co-crystallized with different substrate analogs (Table 1). It is well known that AK is a flexible enzyme, the LID domain and NMP-binding domain being highly dynamic. Therefore, the transition between open and closed states is characterized by important conformational fluctuations (dynamics). Studies performed with AK mutated in the residues which do not alter its structure, demonstrated that the binding interactions of different ligands affect the dynamics of AK, without accompanying structural changes of the enzyme [23,27,28]. A. aeolicus’s AK (2RGX) shows particularly comparable docking with S. pneumoniae (4W5J) and M. tuberculosis (2CDN) counterparts. The exception is Phe35 residue which is not present in Gram-negative AKs. On the contrary, 2RGK binds the sulfur from 2,4-thiazolidinedione structure to Thr31 residue. Moreover, glutamine, a polar residue strictly conserved in AKs – the last residue of the consensus sequence motif [LIVMFYWCA] – [LIVMFYW] (2)–D–G–[FY]–P–R–X(3) –[NQ] – promotes hydrogen bonds with O2 / O3 in A. aeolicus, E. coli (1ANK), AK1, M. tuberculosis, and Gram-positive AKs. On these six AKs, sulfur atom contributes to protein-ligand complex stabilization also. Contrary to AKs of Gram-positive and acid fast bacteria where Pi-sulfur interactions were observed, hydrogen bonds were noticed for the rest. Regarding 1ANK and AK1, the valine residue Val59, respectively Val67, strictly conserved in AKs, participates in hydrogen bonding with sulfur atom. ADMET & DMPK 5(1) (2017) 47-56 Adenylate kinase inhibition by 2,4-thiazolidinedione doi: 10.5599/admet.5.1.327 53 4W5J 1ZIP 2CDN Figure 5. Interaction 2,4-thiazolidinedione with of AKs from Gram-positive and acid fast microorganisms 4W5J 1ZIP 2CDN Figure 6. AKs from Gram-positive and acid fast microorganisms docked with 2,4-thiazolidinedione. 4W5J- S. pneumoniae; 1ZIP- B. stearothermophilus; 2CDN- M. tuberculosis. Code color for interactions: in green are shown conventional hydrogen bonds; yellow – carbon hydrogen bonds; orange – Pi-sulfur In Gram-negative microorganisms, the ligand 2,4-thiazolidinedione binds different residues. Looking at E. coli’s AKs, we noticed that Gly85, another residue strictly conserved from the consensus sequence motif, promotes a hydrogen bonding with the sulfur in the instance of 3HPQ, but failed to establish a valid interaction in 1ANK. On the contrary, in human AK2, this residue (Gly100) interacts with the O2 atom from the 2,4-thiazolidinedione structure. As it is demonstrated in the Figures 5 and 6 and in multiple sequence alignment (Figure 7), AKs from Gram-positive and acid fast microorganisms, bind almost in the same manner the residues involved in AMP- binding (Thr31, Gln93, and Phe35), a notable difference is for AK from B. stearothermophilus In this latter instance, the smallest binding affinity is noticed – Ki = 0.64 mM. One can speculate other Gram-positive pathogens AKs’ have the same inhibition pattern. AKs of human origin display unfavorable interactions, the binding affinity fluctuating among 0.84 mM Mihaela I. Ionescu ADMET & DMPK 5(1) (2017) 47-56 54 and 8.8 mM. Mitochondrial isoenzyme 2 of AK behaves in the same manner as one E. coli’s AK - 3HPQ. As demonstrated in the Figure 7, they bind the same residues and have comparable binding affinity, Ki = 0.84 mM and 0.86 mM (Table 2). On the contrary, even though AK2 has energy binding and Ki comparable to Gram-positive microorganisms, 2,4-thiazolidinedione binds AMP-binding pocket and ATP-binding domain simultaneously. Considering these different binding patterns of human AKs the selectivity of 2,4- thiazolidinedione is an advantage in the context of antimicrobial treatment. Multiple sequence alignment Primary structure offers useful details; comparisons of even small identities could be exploited for finding interesting patterns. It would be of interest to speculate that enzymes with similar primary structure, behave the same with otherwise well characterized enzyme. Sometimes, observing minute details on primary structure has a major impact in decipher intriguing mechanism of a particular protein. The ligand chooses for the present work, 2,4-thiazolidinedione, has proven to bind important residues involved in AMP-binding. Figure 7. Multiple sequence alignment. In gray background is marked ATP-binding domain – the Walker A motif GXXXXGKGT/S; in blue background are shown leucine and valine residues, very well conserved in AKs family; in green background is highlighted the consensus sequence [LIVMFYWCA] – [LIVMFYW] (2)–D–G–[FY]– P–R–X(3)–[NQ], were aspartic acid and arginine, strictly conserved in AKs family, are bolded; in red are colored the residues which interact with 2,4-thiazolidinedione The main limitation of the present work is the availability of AKs’ crystal structures deposited in the public databases. Even though some databases have some weaknesses – some of them being redundant – they remain a powerful tool in fundamental research. Unfortunately, kinetic experiments were not ADMET & DMPK 5(1) (2017) 47-56 Adenylate kinase inhibition by 2,4-thiazolidinedione doi: 10.5599/admet.5.1.327 55 performed, which was another limit of the present study. But the biochemical characterization of a particular enzyme, not always deciphers all capabilities of the protein of interest. To put these findings in perspective, two approaches could be taken in consideration. Either the screening of library protein molecules [28] to find receptors to the highest affinity to 2,4-thiazolidinedione, or design novel TZDs derivates with better binding affinity. Inevitably, inhibitors used for bacterial infections could interact with human counterpart’s target. Although these evidences are not strong enough to persuade us that 2,4-thiazolidinedione is selective inhibitors for bacterial AKs, a novel perspective of AK potentiality is revealed. Results obtained in this study lead us to the conclusion that refining the structure of 2,4-thazolidinedione could be a viable idea in developing potential inhibitors for vital enzymes like AK . Conclusions This study demonstrated an effort to reveal interactions of AK of bacterial / human origin with 2,4- thiazolidinedione. The most noteworthy evidences were achieved referring to Gram-positive and acid fast bacteria’s AKs; 2,4-thiazolidinedione interacts with the same residues belonging to NMP-binding region. Further, the complex comprising AKs of S. pneumoniae and M. tuberculosis exhibit the same ligand-protein pattern. Human AKs interactions with 2,4-thiazolidinedione are divergent; a single notable connection being observed in case of mitochondrial isoenzyme 2 and its counterpart from E. coli. AK, a small enzyme needed for survival, could be an attractive target for TZDs derivates, at least in case of M. tuberculosis and Gram-positive microorganisms. Acknowledgements: The author would like to thank the referees for their courteously comments. References [1] V.D. Rosenthal, Am. J. Infect. Control 44(6) (2016) e81–e90. [2] J. Perry, N. Waglechner, G. Wright, Cold Spring Harb. Perspect. Med. 6(6) (2016) 1-9. [3] M. Jamal, T. Hussain, C. Rajanna Das, S. Andleeb, Jundishapur J. Microbiol. 8(11) (2015) e25917. [4] G. Dinos, C. Athanassopoulos, D. Missiri, P. Giannopoulou, I. Vlachogiannis, G. Papadopoulos, D. Papaioannou, D. Kalpaxis, Antibiotics 5(2) (2016) 1-21. [5] J. Xia, J. Gao, W. Tang, Biosci. Trends 10(1) (2016) 14–21. [6] M.T. Htar, D. Christopoulou, H.-J. Schmitt, BMC Infect. Dis. 15 (2015) 1-10. [7] C. Trucchi, C. Paganino, F. Ansaldi, J. Prev. Med. Hyg. 56(3) (2015) E144–E149. [8] T. Maitre, A. Aubry, V. Jarlier, J. Robert, N. Veziris, and CNR-MyRMA, Med. Mal. Infect. 47(1) (2016) 3-10. [9] M.B. Berry, G.N.J. Phillips, Proteins Struct. Funct. Genet, 32(3) (1998) 276–288. [10] K.A. Henzler-Wildman, V. Thai, M. Lei, M. Ott, M. Wolf-Watz, T. Fenn, E. Pozharski, M.A. Wilson, G.A. Petsko, M. Karplus, C.G. Hübner, D. Kern, Nature 450(7171) (2007) 838–844. [11] M.E. Stegenga, S. Florquin, A.F. de Vos, T. van der Poll, Crit. Care Med. 37(2) (2009) 614–618. [12] M.I. Ionescu, A.Z. Costache, O.Oniga, H.L. Banciu, I.Lupan, I. Lupan, Rom. Rev. Lab. Med. 21(1) (2013) 93–99. [13] D.S. Goodsell, A.J. Olson, Proteins Struct. Funct. Genet. 8(3) (1990) 195–202. [14] G.M. Morris, R. Huey, W. Lindstrom, M.F. Sanner, R.K. Belew, D.S. Goodsell, A.J. Olson, J. Comput. Chem. 30(16) (2009) 2785–2791. [15] S.M. Rizvi, S. Shakil, M. Haneef, EXCLI J. 23(12) (2013) 831–857. [16] G.M. Morris, D.S. Goodsell, R.S. Halliday, R. Huey, W.E. Hart, R.K. Belew, A.J. Olson, J. Comput. Chem. 19(14) (2015) e1004586. Mihaela I. Ionescu ADMET & DMPK 5(1) (2017) 47-56 56 [17] T.P. Schrank, D.W. Bolen, V.J. Hilser, Proc. Natl. Acad. Sci. U. S. A. 106(40) (2009) 16984–16989. [18] M.B. Berry, B. Meador, T. Bilderback, P. Liang, M. Glaser, G.N. Phillips, Proteins 19(3) (1994) 183– 198. [19] T.T. Thach, S. Lee, Acta Crystallogr. Sect. F Struct. Biol. Commun. 70(11) (2014) 1468–1471. [20] M. Bellinzoni, A. Haouz, M. Graña, H. Munier-Lehmann, W. Shepard, P.M. Alzari, Protein Sci. 15(6) (2006) 1489–1493. [21] W. Li, A. Cowley, M. Uludag, T. Gur, H. McWilliam, S. Squizzato, Y.M. Park, N. Buso, R. Lopez, Nucleic Acids Res. 43 (2015) W580–584. [22] F. Sievers, A. Wilm, D. Dineen, T.J. Gibson, K. Karplus, W. Li, R. Lopez, H. McWilliam, M. Remmert, J. Söding, J.D. Thompson, D.G. Higgins, Mol. Syst. Biol. 7, (2011) 539. [23] C.W. Müller, G.E. Schulz, J. Mol. Biol. 224(1) (1992) 159–177. [24] S. Esteban-Martín, R. B. Fenwick, J. Ådén, B. Cossins, C. W. Bertoncini, V. Guallar, M. Wolf-Watz, X. Salvatella, PLoS Comput. Biol. 10(7) (2014) e1003721. [25] A.S.L. Ng, J. Kramer, A. Centurion, J. Dalmau, E. Huang, J.A. Cotter, M.D. Geschwind, J. Neuroimmunol. 15(287) (2015) 31-35. [26] A.L. Ringer, A. Senenko, C.D. Sherrill, Protein Sci. 16(10) (2007) 2216–2223. [27] V.V Hemanth, G. Rao, S. Gosavi, PLoS Comput Biol. 10(11) (2014) e1003938. [28] Y.G. Chushak, R.R. Chapleau, J.S. Frey, C.A. Mauzy, J.M. Gearhart, Toxicol. Res. 4(2) (2015) 519–526. ©2017 by the authors; licensee IAPC, Zagreb, Croatia. This article is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/) http://creativecommons.org/licenses/by/3.0/