Q817G mutation in phosphodiesterase type 5 (PDE-5): conformational analysis and dissociation profile of the inhibitor Tadalafil

Ivan Pires de Oliveira1*, Caroline Honaiser Lescano2* and Gilberto De Nucci

ABSTRACT

Phosphodiesterase type 5 (PDE-5) is an important enzyme involved in the hydrolysis of cyclic guanosine monofosfate (cGMP) to guanosine monophosphate (GMP). The inhibition of this protein leads to the accumulation of cGMP in cells with various biological and therapeutic effects. Several PDE-5 inhibitors exist, with tadalafil being one of the most commonly studied and used in clinical therapy. In this study, we applied Molecular Dynamics (MD) simulations coupled to the ABF (Adaptive Biasing Force) method to study the effect of the mutation on the Gln817 residue (Q817G). The results of the free energy profiles made clear that the affinity of the inhibitor for PDE-5 is This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between dependent on the amino acid residue Gln817. The hydrogen bond made between the side chain of glutamine and the indole ring of Tadalafil results in the stabilization of the ligand in the catalytic site. Despite the prominent role of this interaction, it is important to highlight the contribution of other residues of the catalytic domain for the stabilization of the compound, due to the set of polar, hydrophobic and electrostatic interactions performed by specific amino acid residues.

KEYWORDS: Erectile dysfunction; PDE-5 inhibitors; Molecular Dynamic simulations; Adaptive Biasing Force.

1. INTRODUCTION

Cyclic guanosine monofosfate (cGMP) and phosphodiesterase 5 (PDE-5) are key signaling regulators in many physiological processes, which range from the relaxation of smooth muscle,[1] the inhibition of platelet adhesion and aggregation,[2] natriuresis,[3] sperm motility,[4] and the secretion of ions[5] to cellular growth and apoptosis.[6–8] PDE-5 modulates the intracellular concentrations of cyclic guanosine monophosphate and guanosine monophosphate (GMP), given its biological role in the hydrolysis of cGMP to [9] Accumulated cGMP acts directly through three main cell targets: (i) the cGMPdependent protein kinases;[10] (ii) channels controlled by cyclic nucleotides; (iii) phosphodiesterases regulated by cGMP.[11] The discussions in this work will cover only this last topic, because given the importance of cGMP in the intracellular signaling pathways, its elevation allows for its identification as a potential therapeutic target.
It is known that increased levels of cellular cGMP occur through the inhibition of PDE-5, with abundant expression in the human cavernous body in relation to other tissues. A factor that contributes to the clinical efficacy of PDE-5 inhibitors in the treatment of erectile dysfunction.[12] In addition, PDE-5 inhibitors are used in the treatment of pulmonary hypertension and lower urinary tract symptoms secondary to benign prostatic hyperplasia (Figure 1).[1–6] These molecules have structural similarities with cGMP and compete for bonds in the catalytic site of PDE-5.[13] The affinity and selectivity of inhibitors are therefore determined by the direct contact between certain amino acid residues of the catalytic domain and the molecular structure of the inhibitors, according to intermolecular, protein-inhibitor and specific (polar, electrostatic, hydrophobic and hydrogen bond) interactions.
The catalytic domain has a total volume of about 330 Å3, and a depth of ~10 Å. This domain is composed of four subsites: M site (metal-binding site), Q pocket (core pocket), H pocket (hydrophobic pocket) and L region (lid region).[13] This PDE-5 domain has some key residues for the stabilization and selectivity of the bonding of the cGMP substrate and the inhibitors Vardenafil, Sildenafil and Tadalafil at the catalyst domain of PDE5, as shown previously.[14] Biochemical studies performed with the PDE-5 human recombinant enzyme suggest that the Gln817 residue is very important for the affinity with the substrate, based on the kinetic bond analysis. The side chain of Gln817 forms a bidentate hydrogen bond with the N-1 proton and the C-6 carbonyl with the oxygen of guanosine.[13–15]
The mutagenesis analysis directed at the Gln817 residue indicated the importance of this residue of PDE5, which was critical for the bonding of cGMP and inhibitors, but not in the discrimination between cAMP and cGMP in the site.[14] It is interesting to note that the affinity of Tadalafil for the catalytic site is also attributed to the interactions of the well-oriented-hydrogen-bond type with Gln817.[13] There is not a lot of compelling theoretical information in the literature about the real contribution of this interaction to the stabilization of Tadalafil in the catalytic pocket, however. In this sense, this work investigated the role of the glutamine 817 of PDE-5 in the affinity of the inhibitor Tadalafil, applying a methodology based on Molecular Dynamics simulations coupled to the ABF (Adaptive Biasing Force) method, enabling a sampling of the energy profile of the inhibitor’s dissociation. The presented data set should provide molecular foundations for the understanding of the inhibitory character of Tadalafil and assist in the planning of new PDE-5 inhibitors.

2. MATERIALS AND METHODS

2.1 Conventional Molecular Dynamic (MD) simulations

The Molecular Dynamics simulations were performed using the crystal structure of human phosphodiesterase type 5, PDE-5, (PDB ID: 1XOZ). This protein was resolved with resolution of 1.37 Å, with a total of 364 residues of amino acids and two ions, Zn2+ and Mg2+ on the protein structure. For the simulations the ligand was maintained in the structure, exactly in the same position that was crystallized.[16] The enzyme was solvated with water (TIP3 model),[17] sodium and chloride ions were added to neutralize the system and mimic approximate physiological ionic strengths. The boxes simulations were built using Packmol.[18,19] All simulations were performed with NAMD[20] and figures were produced using VMD[21] program. In this work, we used isothermal-isobaric (NPT) conditions at 310.15K and 1 atm. A Langevin barostat was used with a period of 200 ps and a decay time of 100 ps to control pressure. Temperature was controlled using Langevin dynamics with a perturbation frequency of 10 ps-1. The CHARMM36 force-field was used for the protein, ions and for ligand.[22,23] Tadalafil molecule parametrization was performed with the CGenFF platform.[24,25] The mutation of the PDE-5 protein was performed manually, editing the parameters files and changing the residues Glutamine (Gln817) to a Glycine (Gly817), by convention expressed by Q817G. The conventional MD simulations were performed as follows: 1) Keeping all the protein atoms fixed in space, 10000 steps of Conjugate-Gradient (CG) energy minimization were performed, followed by 0.2 ns to allow solvent relaxation. 2) Fixing only the Cα protein atoms, the energy of the system was minimized by 500 CG steps, and relaxed by 0.2 ns of MD simulation 3) All constraints were removed, and 2 independent production simulations of 80 ns were performed for each system, totalizing 160 ns of MD simulation.

2.2 Protein-ligand free energy profiles

To obtain the protein-ligand free energy profiles we used the ABF (Adaptive Biasing Force) method, implemented in NAMD program. This method is centered on the application of external force to the simulation system, computed according to counteract the forces that act along a reaction coordinate (ξ). The reaction coordinate should be representative to the observable phenomenon, such as distance between secondary structures for the study of aperture of the catalytic site of enzymes,[26] or distance between protein and ligand atoms, adopted in this work.[27–29] The proteinligand reaction coordinate was defined as the distance between C atoms of residues Tyr612, Ala767, Ile768, Val782, Gln817 (PDE5NAT) or Gly817 (PDE5MUT) and Phe820, and a central carbon atom of Tadalafil, as shown in Figure 2 (highlighted by red spheres). This reaction coordinate was used because these residues make up the catalytic site of the PDE-5, as shown previously.[13] The distance between these atoms was sampled in the 3.0 Å to 20.0 Å interval with bins of 0.1 Å. A force constant of 10.0 kcal mol-1 Å-2 was used at the reaction coordinate boundaries. A minimum sampling of 5000 simulation steps at each bin was used to estimate the average force, before the application of the ABF force. Fifteen independent 20 ns of simulation of each system were performed to obtain converged free energy profiles, totalizing 600 ns of ABF simulation.
Structural analyses such as hydrogen bonds, root mean square deviation (RMSD), and atomic distances were performed with the tools available in the MDAnalysis suite[30] and VMD packages.[21] Details of all systems built, including conventional and coupled to the ABF method simulations, are shown in Table 1.

3. RESULTS AND DISCUSSION

3.1 Conformational mobility dependent on the mutation

The intermolecular interactions of the inhibitor with the amino acid residues in the catalytic site result in its spatial stabilization, as shown in Figure 3. The analysis of conformational changes in the simulation time clearly indicates a very similar mobility profile of the inhibitor in both systems, PDE5NAT and PDE5MUT (see Figure 3A), and one can even observe a virtual overlap of the structures at different times of the trajectory (panels I, II and III). Similarly, PDE-5 had low mobility (RMSD < 1.8 Å), indicating no significant conformational change (Figure 3B). In this sense, we can infer two conclusions when observing the results of the conventional MD simulations: (i) the positioning of Tadalafil is not significantly affected by the mutation promoted in Gln817 (Q817G); (ii) PDE-5 maintains its three-dimensional structure without change of the conformational state, including in the catalytic site. These results suggest that despite the mutation, no disturbance is observed for both ligand and protein. The structural composition of the catalytic domain (M site, Q pocket, H pocket and L region) can aid in the understanding of these results.[13] According to Card et al., there are approximately 32 residues that comprise the catalytic site of PDE-5 and interact with ligands (H613, H617, H653, D654, H657, N662, M681, E682, D724, L725, D764, Y612, L765, A767, I768, Q775, I778, A779, V782, A783, L804, I813, M816, Q817, F820, G659, N661, E785, F786, Q789, T802, M805 ).[16] We can therefore suggest that any disturbance promoted by mutation Q817G may be offset by other residues that comprise the catalytic pocket (compensatory effect), stabilizing the ligand in an alternative way. Quantitative aspects that differ in the two systems will be discussed later with the application of the ABF method.

3.2 Glx-Tadalafil hydrogen bonds

The role of the residue Gln817 in the stabilization of the inhibitor was investigated based on its interactions with the type of hydrogen bond. It is suggested that this type of interaction is one of the mechanisms responsible for keeping Tadalafil allocated in the catalytic pocket.[13,14,31] In this case, the inhibitor can be stabilized by hydrogen bonds of the nitrogen of the indole ring with the side chain of glutamine 817, with the inhibitor acting as a donor and Gln817 as an acceptor of the bond, as shown in Figure 4A.
It is interesting to observe that the hydrogen bonds are highly oriented, as shown in Figure 4B. In this study, a hydrogen bond was considered when the distance between the heavy atoms (Tad-N-H···O-Prot) 3.0 Å with an angle of 180o ± 20o. With this definition, it is possible to observe that hydrogen bonds are observed in approximately 59% of the simulation time (highlighted by the blue arrow in Figure 4B). If we consider the upper limits, both for the distance between the heavy atoms and the variation of the angle, then we can observe a rapid convergence for approximately 90% of the simulation time in which Tadalafil and PDE-5 have a hydrogen bond. These results reinforce the discussions in the previously published work by Sung et al., clearly showing the orientation of this protein-ligand bond.[13] Similarly, this residue has been shown to be important for the affinity with other inhibitors, such as Sildenafil (Zoraghi et al. suggests they are bidented)[15], Vardenafil and IBMX and cGMP and cAMP substrates,[14,32] via interaction by hydrogen bond.
The interpretation of the result, shown through hydrogen bonds, suggests that the side chain of Gln817 is of significant importance in the maintenance of Tadalafil in the catalytic site and changes in this micro-system should therefore decrease the Tadalafil-protein affinity. As previously shown in Figure 3, however, the protein mutation Q817G does not reveal major conformational impacts in the protein and the ligand. The effects promoted by the mutation must be therefore be more related with the binding energy than the disposition of the ligand in the catalytic domain. In fact, the mutation of Gln817 directly affects the affinity of the ligand, as shown experimentally by Zoraghi et al. through the calculations of ΔGmut, [15] discussed here later from the molecular point of view.
We know that the Q817G mutation did not affect the mobility of the inhibitor Tadalafil in the catalytic site (see Figure 3A). Similarly, we can show that the residue of Glx (generic representation for Gln and Gly) also doesn't have its mobility changed by the mutation, keeping its spatial position close to the starting one, as shown in Figure 5. In other words, the indole ring of Tadalafil did not affect the disposition of Glx817. This residue remained at ~5.5 Å (distance N-C) in both simulated systems, PDE5NAT and PDE5MUT (Figure 5A and 5B).
Although the systems don't show a clear distinction in terms of conformational mobility, we cannot ignore the importance of the hydrogen bond for the stabilization of the inhibitor. Recently, Huang et al. showed that the atomic distance between [PD5Gln817-OE1]···[Tadalafil-N1] has values very close to those calculated in this work. Huang et al. calculated distances of 3Å, similar to those shown by the red line in Figure 5A (~2.5 Å ≤ distances ≤ 3.0 Å).[31] It is interesting to note that Wang et al. identified values very close to these for the distances between GlnOE1-Nitrogen for the inhibitor Sildenafil, a value between 2.9 and 3.2 Å.[32] It is clear that this is an important intermolecular interaction for the stabilization of ligands in the catalytic domain to PDE5, due to well-oriented hydrogen bonds.

3.3 Network of interactions on the catalytic domain

The Gln817 mutation affects the affinity of inhibitors for the catalytic site, as previously demonstrated for IBMX, Silfenafil, Vardenafil and Tadalafil.[15] Values of ΔGmut ~9.3 kcal/mol are observed, however, revealing that the remaining residues work to keep the ligand fixed in the catalytic domain.[15] In other words, the inhibitory nature of Tadalafil can be better understood if we look at the range of interactions with other amino acid residues that compose the catalytic domain. Table 2 shows the residues that are often found very close to the inhibitor Tadalafil (distance ≤ 3 Å) during the trajectories of the MD simulations (the values between brackets indicate the percentage of the total time of the trajectories). It is interesting to observe that, unlike what is proposed by other authors,[16] the residues His617, His653, Asn662, Met681, Glu682, Asp724, Asp764, Leu765, Gly659, Asn661, Glu785, Gln789, Thr802 and Met805 were not identified as important to stabilize Tadalafil (considering the strict distance ≤ 3 Å). On the other hand, we stress the importance of all other remaining residues, highlighted in Table 2, which must perform a network of hydrophobic, polar, electrostatic and hydrogen bond interactions in order to accommodate the inhibitor within the catalytic domain.
The analyses of the MD simulations allow us to identify how the residues behave in the catalytic pocket in order to accommodate Tadalafil after the disturbance caused by the mutation of glutamine (Q817G) and the consequent loss of the protein-inhibitor hydrogen bond. Figure 6 shows in detail the residues involved in this process. Clearly, some residues lose affinity for the inhibitor, especially Leu725, Ala767 and Met816 (see Figure 6D). The compensatory effect is clearly identified with residues Gln775, Ile813 and the emergence of new interactions with the charged residues His613, Asp654 and His657 (see Figure 6B). It is interesting to observe that the contact with residue Glx817 is very little affected (100% to 97.1%), and it is frequently observed near the inhibitor. However, the nature of the interaction is clearly distinct, as shown in section 3.2 (Figure 5) and discussed below.
The inhibitory character of Tadalafil from the molecular point of view should not only be explained by its ability to establish a hydrogen bond between its indole ring and the side chain of Gln817.[13–15] The most correct way to address this affinity should be based on the network of interactions with the side chains of other residues, or even the ability to interact with the protein backbone of PDE-5. Figure 7 illustrates two important observations: (i) polar interaction possibilities with Gly817 (distance 1, d1); (ii) polar interactions of the indole ring with Gln775 (distance 2, d2). Briefly, we may suggest that the loss of the hydrogen bond of the inhibitor with the side chain of Gln817 (discussed in section 3.2) is partly compensated by polar interactions with the backbone of the residue Gly817 (Figure 7) and the favoring with some residues of the catalytic site (Figure 6B).

3.4 PDE-5···Tadalafil dissociation profile

The ABF method has been successfully employed to quantify the energy profile of protein-ligand dissociations in different systems, with the ligand being a simple molecule or a peptide.[33,34] In this work, the contribution of residue Gln817 to the affinity of Tadalafil for the protein was assessed by applying the ABF methodology so that dissociation profiles were obtained with PDE-5 in its native (PDE5NAT) and mutant (PDE5MUT) forms. The comparison of the ΔG profiles clearly reveals the contribution associated with the residue of Gln817 to the affinity of the inhibitor, shown in Figure 8. One can see that in the minimum of ΔG, with ξ in the order of 4.3 Å, the difference between the valleys is ~5.2 kcal/mol. This is therefore the energy contribution associated with glutamine compared with the mutated glycine residue (Q817G). This contribution is significant, but we also observed a profile that indicates the high affinity for the catalytic domain of PDE-5 even with the mutant protein. This reinforces the importance of the network of interactions with other residues for the maintenance of the inhibitor in the catalytic cavity. In fact, Zoraghi et al. compared mutation Q817A and identified values of ΔΔG ~2.7 kcal/mol (ΔΔG = ΔGwildtype - Δmutated), showing the energy contribution of the hydrogen bond (made with Gln817) for the affinity PDE-5···Tadalafil. On the other hand, the network of interactions with other residues is also extremely important, with the difference in energy (ΔGwildtype) for the protein in native state being 12.1 kcal/mol compared to the mutant PDE-5 (ΔGmut) of 9.3 kcal/mol.[15] Moreover, the mathematical integration leads to the comparison of absolute values for each dissociation profile shows a difference approximately 34.7 kcal/mol (from 205.5 kcal/mol to 170.8 kcal/mol). Despite the theoretical and experimental values are hardly correlated, the contribution of mutation is clear. For instance, the PDE5 mutation promotes decreasing in the potency of the inhibition of Tadalafil, according IC50 values, from 2.2 nM to 210 nM,[15] leading an increase in the IC50 of ~208 nM. Moreover, the mutation changes the theoretical free energies by 34.7 kcal/mol, thus a proportion of decay at ~6 nM by each kcal/mol. Considering both ΔΔG, from experimental and ABF method, we can estimate a relative quantities of each chemical species with temperature at 303 K and 310.15 K, respectively. The results suggest Kexp ≈ 0.01 and Kteo << 0.01, showing a major proportion of the complex protein-inhibitor with wild enzyme. As expected the simulations indicate a more significant contribution of Gln817 than experimental results, due to higher ΔΔG* value.

3.5 Protein-inhibitor energy interaction

The ΔG dissociation profiles, shown above, make clear the importance of the Gln817 side chain for the stabilization of inhibitors. We can assess how the interaction energies of PDE-5 with Tadalafil are affected by the Q817G mutation (shown in Figure 9). Figures 9A and 9B show how the protein-inhibitory interaction energies vary at simulation time (electrostatic, Van der Walls and total energies). Figure 9C shows that the total energy is affected by the mutation, with an energy variation of ~ 10 kcal/mol (PDE5NAT: maximum peak at ~-61 kcal/mol; PDE5MUT: peak at ~-51 kcal/mol). We can observe in Figure 9D that the energies associated with Van der Walls type interactions are not affected by the mutation, with a peak at ~-45 kcal/mol. On the other hand, the displacement at the peak of the total energy (Figure 9C) can be explained by the electrostatic interactions (Figure 9E). In summary, we can infer that the Q817G mutation directly affected the form (and intensity) with which the inhibitor binds in the catalytic pocket, specifically reducing the electrostatic interactions with residues, especially due to the absence of the hydrogen bond performed by side chain of Gln817. This disturbance in electrostatic interactions was successfully captured by the ABF method, according to the ΔG profiles discussed previously. The importance of Gln817 can be extrapolated to others different ligands, as such previously discussed by B. Sung et al.[13] For instance, the interactions with cGMP Sildenafil occur via bidentate hydrogen bonds performed between pyrazolopyrimidinone group and -amide group of Gln817. Clearly, this interaction should be critical for another set of ligands with this same chemical group, such as Vardenafil and Udenafil.[35,36] In this sense, it is evident the importance of this residue to stabilize not only Tadalafil, but also a series of PDE5 binding compounds, as suggested in this study.

3.6 Residues contacted in the disassociation of the ligand

The dissociation of the ligand of the catalytic site induced by the ABF method allows us to identify the residues that participate in this process. The ABF method allows the ligand to be removed from the catalytic site in a diffusive way, because the forces that act to maintain the ligand in the cavity are also counteracted by the external ABF force.[28] In this sense, we can map the residues that are involved in the inhibitor dissociation from the catalytic cavity of PDE-5, with the following being directly involved: Tyr612, Gln663, Leu725, Ala726, Ile729, Ala767, Ile768, Gln775, Ile778, Ala779, Val782, Ala783, Phe786, Phe787, Thr802, Leu804, Met805, Ile813, Met816, Glx817 (Gln or Gly), Phe820, Ala823 and Ile824, highlighted in Figure 10. The residues that aren't part of the catalytic domain (Gln663, Ala726, Ile729, Val782, Thr802, Met805, Ala823 and Ile824) are located in the region likely associated with the entry "gate" of PDE-5. It is interesting to observe that one of the residues that interacts with the disassociation ligand, Gln663, is part of the H-loop (residues from 660 to 683) of the protein. According to Wang et al. this region is linked to the opening mechanism of the catalytic site, and it should take up to four conformations, with displacements of 7-35 Å depending on the inhibitor, Sildenafil, IBMX or Icarisid II.[32]

4. CONCLUSION

Many PDE-5 inhibitors are found in the literature with different chemical structures, and these inhibitors should act in the same protein domain as PDE-5. The explanations involving the affinity of Tadalafil for the catalytic site are often associated with the ability to make a well-oriented hydrogen bond between the indole ring of the inhibitor and the side chain of Gln817. At the same time, we can identify a number of other residues that comprise the catalytic pocket and which play an Mardepodect important role in the stabilization of the ligand. In this sense, we can infer that the set of interactions involving these residues clearly should stabilize Tadalafil and other inhibitors, thus the residue of Gln817 is only one component of a range of important residues for the stabilization of ligands in the active site of PDE-5.

REFERENCES

[1] E. J. Tsai, D. A. Kass, Cyclic GMP signaling in cardiovascular pathophysiology and therapeutics. Pharmacol. Ther. 2009, 122, 216–238.
[2] C.B. Mendes-Silverio, L. O. S. Leiria, R. P. Morganti, G. F. Anhê, S. Marcondes, F. Z. Mónica, et al., PLoS One 2012, 7.
A. Sindic, J. R. Hirsch, A. Velic, H. Piechota, E. Schlatter, Kidney Int 2005, 67, 1420.
B. Willipinski-Stapelfeldt, J. Lübberstedt, S. Stelter, K. Vogt, A. K. Mukhopadhyay,
D. Müller, Mol. Hum. Reprod. 2004, 10, 543.
[3] N. Arshad, S. S. Visweswariah, In FEBS Letters, 2012, Vol. 586, pp. 2835–2840.
[4] B. Zhu, S. Strada, Curr. Top. Med. Chem. 2007, 7, 437.
[5] K. Bian, F. Murad, sGC-cGMP signaling: target for anticancer therapy. Adv. Exp. Med. Biol. 2014, 814, 5–13.
[7] A. Das, D. Durrant, C. Mitchell, E. Mayton, N. N. Hoke, F. N. Salloum, et al., Proc. Natl. Acad. Sci. 2010, 107, 18202.
[8] J. D. Corbin, S. H. Francis, D. J. Webb, Urology 2002, 60, 4.
[10] S. H. Francis, J. L. Busch, J. D. Corbin, Pharmacol. Rev. 2010, 62, 525.
[11] F. Murad, N. Engl. J. Med. 2006, 355, 2003.
[12] A. Morelli, S. Filippi, R. Mancina, M. Luconi, L. Vignozzi, M. Marini, et al., Endocrinology 2004, 145, 2253.
[13] B. Sung, K. Hwang, Y. Jeon, J. Lee, Y. Heo, J. Kim, et al., Nature 2003, 425, 98.
[14] R. Zoraghi, J. D. Corbin, S. H. Francis, J. Biol. Chem. 2006, 281, 5553.
[15] R. Zoraghi, S. H. Francis, J. D. Corbin, Biochemistry 2007, 46, 13554.
[16] G. L. Card, B. P. England, Y. Suzuki, D. Fong, B. Powell, B. Lee, et al., Structure 2004, 12, 2233.
[17] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein, J. Chem. Phys. 1983, 79, 926.
[18] L. Martinez, R. Andrade, E. G. Birgin, J. M. Martinez, J. Comput. Chem. 2009, 30, 2157.
[19] J. M. Martínez, L. Martínez, J. Comput. Chem. 2003, 24, 819.
[20] J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, et al., J. Comput. Chem. 2005, 26, 1781.
[21] W. Humphrey, A. Dalke, K. Schulten, J Mol Graph 1996, 14, 27.
[22] O. Guvench, S. S. Mallajosyula, E. P. Raman, E. Hatcher, K. Vanommeslaeghe, T. J. Foster, et al., J. Chem. Theory Comput. 2011, 7, 3162.
[23] J. Huang, S. Rauscher, G. Nawrocki, T. Ran, M. Feig, B. L. de Groot, et al., Nat. Methods 2016, 14, 71.
[24] K. Vanommeslaeghe, A. D. MacKerell, J. Chem. Inf. Model. 2012, 52, 3144.
[25] K. Vanommeslaeghe, E. P. Raman, A. D. MacKerell, J. Chem. Inf. Model. 2012, 52, 3155.
[26] I. P. de Oliveira, G. E. Jara, L. Martinez, Phys. Chem. Chem. Phys. 2017, 19, 31499.
[27] A. Pohorille, C. Jarzynski, C. Chipot, J. Phys. Chem. B 2010, 114, 10235.
[28] J. Hénin, G. Fiorin, C. Chipot, M. L. Klein, J. Chem. Theory Comput. 2010, 6, 35.
[29] E. Darve, D. Rodríguez-Gómez, A. Pohorille, J. Chem. Phys. 2008, 128.
[30] L. Martínez, 2017.
[31] Y. Y. Huang, Z. Li, Y. H. Cai, L. J. Feng, Y. Wu, X. Li, et al., J. Chem. Inf. Model. 2013, 53, 3044.
[32] H. Wang, Y. Liu, Q. Huai, J. Cai, R. Zoraghi, S. H. Francis, et al., J. Biol. Chem. 2006, 281, 21469.
[33] M. Lapelosa, Biophys. Chem. 2018, 232, 22.
[34] Q. Bai, Y. Shen, N. Jin, H. Liu, X. Yao, Biochim. Biophys. Acta – Gen. Subj. 2014, 1840, 2128.
[35] J. D. Corbin, A. Beasley, M. A. Blount, S. H. Francis, In Neurochemistry International, 2004, Vol. 45, pp. 859–863.
[36] M. C. Cho, J.-S. Paick, Ther. Clin. Risk Manag. 2014, 10, 341.