International Journal of Molecular Sciences

Article Comparative Study of Interactions between Human cGAS and Inhibitors: Insights from and MM/PBSA Studies

Xiaowen Wang 1,2 and Wenjin Li 1,*

1 Institute for Advanced Study, Shenzhen University, Shenzhen 518060, China; [email protected] 2 College of Physics and Optoelectronic Engineering, Shenzhen University, Shenzhen 518060, China * Correspondence: [email protected]; Tel.: +86-755-2694-2336

Abstract: Recent studies have identified cyclic GMP-AMP synthase (cGAS) as an important target for treating autoimmune diseases, and several inhibitors of human cGAS (hcGAS) and their structures in complexation with hcGAS have been reported. However, the mechanisms via which these inhibitors interact with hcGAS are not completely understood. Here, we aimed to assess the performance of molecular mechanics/Poisson–Boltzmann solvent-accessible surface area (MM/PBSA) in evaluating the binding affinity of various hcGAS inhibitors and to elucidate their detailed interactions with hcGAS from an energetic viewpoint. Using molecular dynamics (MD) simulation and MM/PBSA approaches, the estimated free energies were in good agreement with the experimental ones, with a Pearson’s correlation coefficient and Spearman’s rank coefficient of 0.67 and 0.46, respectively. In per-residue energy decomposition analysis, four residues, K362, R376, Y436, and K439 in hcGAS were found to contribute significantly to the binding with inhibitors via hydrogen bonding, salt π π···π ···π ···π ···π  bridges, and various interactions, such as stacking, cation , hydroxyl , and alkyl  interactions. In addition, we discussed other key interactions between specific residues and ligands,

Citation: Wang, X.; Li, W. in particular, between H363 and JUJ, F379 and 9BY, and H437 and 8ZM. The sandwiched structures Comparative Study of Interactions of the inhibitor bound to the guanidinium group of R376 and the phenyl ring of Y436 were also between Human cGAS and Inhibitors: consistent with the experimental data. The results indicated that MM/PBSA in combination with Insights from Molecular Dynamics other virtual screening methods, could be a reliable approach to discover new hcGAS inhibitors and and MM/PBSA Studies. Int. J. Mol. thus is valuable for potential treatments of cGAS-dependent inflammatory diseases. Sci. 2021, 22, 1164. https:// doi.org/10.3390/ijms22031164 Keywords: cGAS; inhibitor; MM/PBSA; MD simulation; sandwiched structures

Academic Editor: A. Phillip West Received: 22 December 2020 Accepted: 21 January 2021 1. Introduction Published: 25 January 2021 Free DNA in the cytosol is detected by a type of nucleotidyltransferase called cyclic GMP-AMP (cGAMP) synthase (cGAS) [1]. Upon binding to DNA, cGAS is activated and Publisher’s Note: MDPI stays neutral produces cGAMP from GTP and ATP. The second messenger, cGAMP, activates and binds with regard to jurisdictional claims in the stimulator of interferon genes, which subsequently induces the secretion of type I published maps and institutional affil- iations. interferons and triggers the downstream innate immune response [2–5]. While recognition of pathogen DNA is essential for host defense against infections, aberrant activation of cGAS may trigger autoimmune diseases such as systemic lupus erythematosus and Aicardi– Goutières syndrome in the presence of self-DNA [6,7], which may be displaced nuclear or mitochondrial DNA generated as by-products of cellular damage. Hence, cGAS is a Copyright: © 2021 by the authors. vital drug target for treating autoimmune diseases and for preventing autoinflammation in Licensee MDPI, Basel, Switzerland. therapeutic strategies against cancers. This article is an open access article Experimental studies have shown that knockout of the cGAS gene in Trex1−/− mice distributed under the terms and conditions of the Creative Commons dramatically reduces tissue inflammation and hinders autoantibody production [8,9]. Thus, Attribution (CC BY) license (https:// inhibition of cGAS activity is a promising therapeutic strategy for autoimmune diseases. creativecommons.org/licenses/by/ Recently, several inhibitors of human (h) and mouse (m) cGAS have been reported [10–13]. 4.0/). All of them bind to the catalytic center of cGAS and suppress its nucleotidyltransferase

Int. J. Mol. Sci. 2021, 22, 1164. https://doi.org/10.3390/ijms22031164 https://www.mdpi.com/journal/ijms Int. J. Mol. Sci. 2021, 22, 1164 2 of 16 Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 2 of 16

activity. RU.521 is a promising inhibitor (IC50 = 0.11 µM) of mcGAS that can regulate thethroughput levels of screening interferon; study however, showed it binds that th toe hcGAS compounds with lowG108 affinity (containing (IC50 =a 2.94pyrazoleµM) ingroup) spite and of approximately G105 (containing 60% a sequence2-amino pyridine identity ring) between were hcGAS novel andspecific mcGAS inhibitors [14]. Aof high-throughputhcGAS, which targeted screening the study identical showed activation that the loop compounds region G108containing (containing R376 aand pyrazole Y436, group)similar andto the G105 binding (containing pathways a 2-amino of 2,3-cG pyridineAMP [12]. ring) While were G150 novel and specific G108 showed inhibitors high of hcGAS,binding whichaffinities targeted in complexes the identical with hcGAS, activation with loop IC region50 values containing of 10.2 nM R376 and and 27.5 Y436, nM, similarrespectively, to the they binding were pathways both inactivated of 2,3-cGAMP during [12 assessment]. While G150 with and mcGAS G108 showed[12]. Another high bindinghigh-affinity affinities inhibitor, in complexes PF–06928215, with which hcGAS, bind withs to IC the50 hcGASvalues active of 10.2 sites, nM was and shown 27.5 nM, to respectively,have a dissociation they were constant both ( inactivatedKd) of 200 nM during (IC50 = assessment 4.9 µM) using with high-throughput mcGAS [12]. Another screen- high-affinitying assays [11]. inhibitor, Based on PF–06928215, the inactive whichPF–06928215 binds toin thecell-based hcGAS cGAS, active three sites, potent was shown com- topounds, have a18, dissociation S2, and S3 constant (IC50 = 29.88 (Kd) ± of 3.20, 200 nM13.1 (IC± 0.09,50 = 4.9andµ 4.9M) ± using 0.26 µM), high-throughput with highly screeningconsistent assaysbinding [11 modes,]. Based were on theidentified inactive by PF–06928215 Zhao and co-workers in cell-based [13]. cGAS, More threeinformation potent compounds,regarding various 18, S2, hcGAS and S3 (ICinhibito50 = 29.88rs as ±therapeutic3.20, 13.1 ±targets0.09, andof interest 4.9 ± 0.26 areµ presentedM), with highly in Ta- consistentble S1. binding modes, were identified by Zhao and co-workers [13]. More information regardingTo complement various hcGAS and validate inhibitors the as results therapeutic obtained targets using of an interest experimental are presented approach, in Tablecomputational S1. modeling, molecular dynamics (MD) simulations, and molecular mechan- ics/Poisson–BoltzmannTo complement and solvent-accessible validate the results surface obtained area using (MM/PBSA) an experimental can be approach, used for com-ana- putationallyzing the modeling,real-time dynamic molecular behavior dynamics of (MD)biomolecules simulations, and predicting and molecular the binding mechanics/ free Poisson–Boltzmannenergy. These methods solvent-accessible have been widely surface us areaed for (MM/PBSA) elucidating can the be binding used for mechanism analyzing theand real-time interaction dynamic modes behaviorof protein-inhibitor of biomolecules systems and [15–17]. predicting Recently, the binding virtual free screening energy. Thesewas performed methods for have the been first widelytime, combined used for wi elucidatingth MD and the modeling binding methods, mechanism to identify and in- teractionnew hcGAS-binding modes of protein-inhibitor inhibitors [13]. However, systems complete [15–17]. understanding Recently, virtual regarding screening the wasmo- performedlecular mechanism for the first of hcGAS-inhibitor time, combined withinteractions, MD and especially modeling from methods, the viewpoint to identify of new the hcGAS-bindingdynamics and energy inhibitors of the [13 interactions,]. However, completeis still lacking. understanding regarding the molecular mechanismIn this ofstudy, hcGAS-inhibitor we aimed to interactions, assess the especiallyperformance from of the MM/PBSA viewpoint in of evaluating the dynamics the and energy of the interactions, is still lacking. binding affinity of 10 hcGAS inhibitors (listed in Figure 1 and Table S1). In addition, we In this study, we aimed to assess the performance of MM/PBSA in evaluating the aimed to elucidate the binding modes of hcGAS with these inhibitors in combining MD binding affinity of 10 hcGAS inhibitors (listed in Figure1 and Table S1). In addition, we simulations and per-residue energy decomposition analysis and compare these observa- aimed to elucidate the binding modes of hcGAS with these inhibitors in combining MD sim- tions with those obtained using experimental methods. In combination with existing vir- ulations and per-residue energy decomposition analysis and compare these observations tual screening approaches, our study could provide theoretical guidance in the search for with those obtained using experimental methods. In combination with existing virtual new cGAS inhibitors and thus be helpful in the treatment of autoimmune diseases. screening approaches, our study could provide theoretical guidance in the search for new cGAS inhibitors and thus be helpful in the treatment of autoimmune diseases.

FigureFigure 1.1.Chemical Chemical structures structures of of the the hcGAS-binding hcGAS-binding inhibitors inhibitors used used in this in studythis study with theirwith proteintheir data (PDB) bank names.(PDB) Tricyclicnames. Tricyclic pyridoindole pyridoindole core for JUMcore for (pyrazole JUM (pyrazole ring) and ring) JUJ (2-aminoand JUJ (2-amino pyridine pyridine ring), diazolo-pyrimidine ring), diazolo-pyrimidine aromatic aromatic core for KHM,core for KKP, KHM, and KKP, KKM, and tetrazolo-pyrimidine KKM, tetrazolo-pyrimidine aromatic core aromatic for 8ZM core and for 9BV 8ZM (thiazole and 9BV ring), (thiazole bi (1, 2, ring), 4-triazole) bi (1, core2, 4-triazole) for ER9, core for ER9, tricyclic benzimidazole core for 9BY, and pyridine and pyrimidine rings for 9BS are shown. tricyclic benzimidazole core for 9BY, and pyridine and pyrimidine rings for 9BS are shown. 2. Results and Discussion

Int. J. Mol. Sci. 2021, 22, 1164 3 of 16

2. Results and Discussion 2.1. Structural Analysis Before performing the MM/PBSA analysis, it was necessary to assess the convergence of the sampled structures during MD simulations [18]. The root mean square deviation (RMSD) of the 10 hcGAS-inhibitor complexes as a function of time are plotted in Figure S1. Simultaneously, Table1 lists the average RMSD values for isolated hcGAS and inhibitors, including the zinc-thumb domain and whole systems. As shown in Figure S1, the RMSDs of all the hcGAS proteins and inhibitors fluctuated within 0.3 and 0.15 nm, respectively. In particular, 9BS-bound hcGAS showed higher fluctuation with an average RMSD of 0.274 ± 0.030 nm than 9BY and 9BV-bound hcGAS (Figure S1a). The inhibitors 9BS and 9BV displayed more accurate structural overlap with the experimental ones compared to other inhibitors, with RMSD values of 0.017 ± 0.046 and 0.014 ± 0.049 nm, respectively; 9BY showed larger RMSD distribution of 0.089 ± 0.029 (Figure S1b). JUM and ER9 show relatively larger RMSD fluctuations in the range of 0.1−0.15 nm within 7 ns than the other inhibitors (Figure S1b), while reaching final equilibrium with the average RMSD values of 0.097 ± 0.044 and 0.116 ± 0.056 nm, respectively (Table1). The zinc-thumb domain plays a key role in maintaining cGAS recognition, which was evident from the RMSD values lesser than 0.05 nm for all systems (see Table1). Overall, these plots suggested that the inhibitors stayed in the preferred positions in the binding pockets and that the entire hcGAS-inhibitor complex remained well balanced during the MD simulations. To further confirm that the protein-ligand complexes were converged within the 20 ns MD simulations, one of the three parallel simulations for each inhibitor was extended to 50 ns, and comparable RMSDs were observed as to the ones within the previous 20 ns simulations (Figure S2).

Table 1. Averaged root mean square deviation (RMSD) (nm) evaluations for hcGAS, inhibitor, zinc-thumb domain, and whole structures in 10 hcGAS-inhibitor systems.

Inhibitor Name hcGAS Inhibitor Zinc-Thumb Whole PDB ID JUM 0.243 ± 0.023 0.097 ± 0.044 0.021 ± 0.009 0.243 ± 0.023 6MJU JUJ 0.250 ± 0.021 0.070 ± 0.018 0.027 ± 0.009 0.250 ± 0.021 6MJW KHM 0.239 ± 0.028 0.065 ± 0.017 0.051 ± 0.028 0.243 ± 0.027 6NAO 8ZM 0.256 ± 0.019 0.037 ± 0.027 0.035 ± 0.014 0.256 ± 0.019 5V8O KKP 0.243 ± 0.029 0.046 ± 0.020 0.028 ± 0.010 0.243 ± 0.029 6NFG KKM 0.241 ± 0.023 0.069 ± 0.021 0.033 ± 0.014 0.241 ± 0.023 6NFO ER9 0.253 ± 0.020 0.116 ± 0.056 0.041 ± 0.017 0.266 ± 0.032 6LRL 9BY 0.244 ± 0.025 0.089 ± 0.029 0.017 ± 0.012 0.245 ± 0.023 5VDW 9BS 0.274 ± 0.030 0.017 ± 0.046 0.025 ± 0.009 0.275 ± 0.030 5VDU 9BV 0.242 ± 0.024 0.014 ± 0.049 0.033 ± 0.010 0.242 ± 0.024 5VDV

2.2. Binding Free Energy MM/PBSA analysis of three parallel trajectories was used for determining the free energies of hcGAS binding with various inhibitors (see Table2). The binding free energies were also estimated from each simulation separately (Tables S2–S4), and they were strongly correlated with the one averaged from all three simulations (see Figure S3), demonstrating the consistency of the results from independent simulations. The reported experimental data on IC50 and Kd of the inhibitors were used to evaluate the performance of MM/PBSA calculations. For conveniently comparing the binding affinities between the complexes, we classified the hcGAS-binding inhibitors into three categories (see Table S1): Tricyclic pyridoindole core of JUM and JUJ (I) [12], diazolo-/tetrazolo-pyrimidine aromatic core of KHM, 8ZM, KKP, and KKM (II) [11], and the other three inhibitors, including various heterocycles of 9BY, 9BS, and 9BV (III) [10]. It is noteworthy that inhibitor ER9 contains a bi(1,2,4-triazole) core, which is a novel inhibitor obtained from virtual screening based on the inhibitor KHM [13], and was hence placed in category II for analysis. Int. J. Mol. Sci. 2021, 22, 1164 4 of 16

Table 2. The averaged binding free energies, ∆Gbind (kJ/mol), with the standard error of the mean for 10 hcGAS-inhibitor systems obtained from MM/PBSA calculations, along with other components. Here, ∆Gpb = ∆Eele + ∆Gpb/solv and ∆Gnp = ∆EvdW + ∆Gnp/solv.

Inhibitor ∆EvdW ∆Eele ∆Gpb/solv ∆Gnp/solv ∆Gpb ∆Gnp ∆Gbind IC50 (µM) Kd (µM) JUM −150.2 ± 2.2 −40.5 ± 9.6 134.6 ± 9.8 −17.3 ± 0.2 94.1 ± 9.7 −167.5 ± 1.2 −73.4 ± 5.3 0.0275 [12] − JUJ −159.0 ± 3.0 −56.6 ± 8.5 130.2 ± 8.5 −18.4 ± 0.2 73.7 ± 8.5 −177.4 ± 1.6 −103.8 ± 5.5 0.0102 [12] − KHM −155.7 ± 3.5 −144.0 ± 10.3 164.1 ± 8.9 −16.5 ± 0.3 20.1 ± 9.6 −172.2 ± 1.9 −152.2 ± 5.5 2.0/4.9 [11] 0.2 [11] 8ZM −124.1 ± 2.2 −26.7 ± 4.5 68.6 ± 5.2 −12.0 ± 0.1 41.9 ± 4.8 −136.1 ± 1.1 −94.3 ± 3.4 78 [11] 171 [11] KKP −145.6 ± 2.7 −52.9 ± 6.2 92.6 ± 6.4 −14.8 ± 0.2 39.7 ± 6.3 −160.3 ± 1.5 −120.6 ± 5.6 69/125 [11] 78 [11] KKM −147.6 ± 4.1 −87.6 ± 10.1 129.6 ± 7.0 −15.5 ± 0.3 42.0 ± 8.5 −163.1 ± 2.2 −121.2 ± 6.5 8.1/17.5 [11] 2.7 [11] ER9 −134.0 ± 3.2 −81.8 ± 5.7 176.1 ± 7.4 −13.8 ± 0.2 94.3 ± 6.6 −147.9 ± 1.7 −53.8 ± 4.9 13.1 [13] − 9BY −132.6 ± 2.8 −117.4 ± 5.5 148.8 ± 6.0 −12.7 ± 0.2 31.4 ± 5.8 −145.3 ± 1.5 −114.1 ± 4.3 − 236 ± 19 [10] 9BS −92.9 ± 2.7 −35.1 ± 4.1 73.3 ± 5.0 −10.5 ± 0.2 38.2 ± 4.6 −103.4 ± 1.5 −65.3 ± 3.6 − 64 ± 3 [10] 9BV −110.1 ± 1.9 −14.3 ± 4.5 94.9 ± 4.4 −9.9 ± 0.2 80.6 ± 4.5 −120.0 ± 1.0 −39.2 ± 3.6 − 80 ± 4 [10]

As shown in Table2, the binding free energies predicted using MM/PBSA are in accordance with the experimentally determined affinities. In terms of the IC50 values, the binding affinity of JUJ-bound hcGAS was stronger than that of JUM-bound hcGAS (0.0102 and 0.0275 µM, respectively). This was well predicted from the calculated binding free energies of −103.8 ± 5.5 kJ/mol and −73.4 ± 5.3 kJ/mol. MM/PBSA also predicted the order of the binding free energy of the inhibitors with diazolo-/tetrazolo-pyrimidine aromatic core to be KHM > KKM > KKP > 8ZM, which is consistent with the IC50 values of 2.0, 8.1, 69.0, and 78.0 µM, respectively. Although the IC50 (13.1 µM) of ER9 in the enzyme activity assay was stronger than that of KKP (69 µM) determined using a fluorescence polarization assay [11,13], it is difficult to compare their binding affinities owing to the use of different approaches. In addition, the calculated binding free energies of 9BS and 9BV bound to hcGAS were −65.3 ± 3.6 and −39.2 ± 3.6 kJ/mol, respectively, with dissociation constant Kd of 64 ± 3 and 80 ± 4, respectively. The experimental data suggested that the affinity of 9BY was weaker than those of 9BS and 9BV, although the calculated binding free energy of the hcGAS–9BY interaction (−114.1 ± 4.3 kJ/mol) was predicted to be more than that of 9BS and 9BV. We further evaluated the correlation of the binding free energy between the pIC50/∆Gexp and MM/PBSA calculation. It should note that the IC50 or Kd values for these three groups of ligands were obtained from different experimental methods, more specifically, high-throughput screening assay for ligand group I [12], fluorescence polarization assay for ligand group II excluding ER9 [11], enzyme activity assay for ER9 [13], and a novel SPR-based enzymatic assay for ligand group III [10]. Since the value of IC50 was dependent on the enzyme concentration used in the experiments, the correlation between pIC50 and the predicted ∆GMM/PBSA was thus discussed separately for inhibitors in different groups. As shown in Figure2a, the predicted binding free energy for groups I and II negatively correlated with the experimental values, especially for the ligand in group II excluding ER9 a good correlation coefficient of 0.86 was obtained. As shown in Figure2b, the correlation between ∆Gexp and ∆GMM/PBSA for groups II and III was as high as 0.61. Note that the RMSD of both ER9 and 9BY (Figure S1) were significantly larger than the ones of other inhibitors, which might affect the reliability of their predicted ∆GMM/PBSA and resulted in an underestimated correlation. Thus, the above results suggested that MM/PBSA could be a promising and cheap approach to estimate the binding affinity of potential inhibitors of hcGAS. The averaged binding free energies at different time scales from three parallel trajec- tories for the 10 hcGAS-inhibitor systems obtained using MM/PBSA analysis are shown in Table3. Compared to the experimental binding affinities, most calculated binding free energies (∆Gbind) are generally overestimated, similar to the results of other studies on protein-ligand interactions [19,20]. The Spearman’s and Pearson’s correlation coefficients are also provided here to evaluate the ranking of the binding free energies and their cor- relation with experimental data [21,22]. As listed in Table3, the Pearson’s correlation coefficient (r) was 0.52 at the early simulation stage (0−4 ns), with Spearman’s rank corre- lation coefficient (ρ) of 0.32, while the studies on dynamics predicted relatively stronger correlations and fluctuation in the vicinity of 0.66 for r. Furthermore, the rankings of the Int. J. Mol. Sci. 2021, 22, 1164 5 of 16

binding free energies remained constant at ρ = 0.46 after 8 ns. In addition, we inspected the various energy distributions as a function of time, as shown in Figure S4. The components ∆EvdW and ∆Gnp/solv were generally stabilized during the entire trajectories. However, the electrostatic and polar solvation energies showed larger fluctuations in hcGAS binding with JUJ, KHM, and KKM than the other inhibitors, leading to binding free energies with stronger disturbance. Despite the limitations of MM/PBSA studies, which were mainly due to force field accuracy, polar contribution to solvation, entropy estimation, and sampling Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 5 of 16 conditions, this method was used as a potential tool for predicting the relative binding free energy of protein-ligand interactions [23,24].

FigureFigure 2. 2.Correlations Correlations between between available available pIC pIC5050/experimental/experimental binding free energiesenergies andand thethe averagedaveraged bindingbinding freefree energiesenergies estimatedestimated usingusing thethe MM/PBSAMM/PBSA method for ligand (a) inin groupsgroups II andand II,II, andand ((bb)) inin groupsgroups IIII andand III.III. TheThe threethree ligand ligand categoriescategories are are distinguished distinguished by by red red dots dots for for group group I, blueI, blue dots do forts for group group II, andII, and green green dots dots for groupfor group III, respectively.III, respectively. In (a In), (a), the black line fits to all the data points of group II, while the red line to the ones for the group II but ER9. In (b), the the black line fits to all the data points of group II, while the red line to the ones for the group II but ER9. In (b), the black black line fits to all the data points. line fits to all the data points. The averaged binding free energies at different time scales from three parallel trajec- Table 3. Comparison of experimentaltories forbinding the 10 energyhcGAS-inhibitor (∆Gexp) and systems the estimated obtained binding using free MM/PBSA energy, ∆G bindanalysis(kJ/mol), are atshown different time scales obtainedin from Table MM/PBSA 3. Compared analysis, to alongthe experimental with Spearman’s binding rank correlationaffinities, coefficientmost calculated (ρ) and Pearsonbinding free correlation coefficient (r). energies (∆Gbind) are generally overestimated, similar to the results of other studies on pro- tein-ligand interactions [19,20]. The Spearman’s and Pearson’s correlation coefficients are Inhibitor ∆Gexp 0–4ns 4–8ns 8–12ns 12–16ns 16–20ns also provided here to evaluate the ranking of the binding free energies and their correla- JUM NA a −51.8 −73.5 −73.2 −76.7 −84.5 JUJ NA a tion with −experimental89.9 data−90.1 [21,22]. As listed−117.0 in Table 3, the− Pearson’s114.7 correlation−110.4 coeffi- KHM −38.5 cient (r) was−155.1 0.52 at the early−140.8 simulation stage−165.9 (0‒4 ns), with− Spearman’s160.8 rank− correlation150.9 8ZM −21.6 coefficient− (94.1ρ) of 0.32, while− 93.6the studies on −dynamics96.0 predicted−89.8 relatively stronger−95.3 corre- KKP −23.6 lations and−132.2 fluctuation in the−124.8 vicinity of 0.66− 107.5for r. Furthermore,−120.3 the rankings− of106.6 the bind- KKM −32.0 ing free energies−115.4 remained− 124.3constant at ρ =− 0.46121.7 after 8 ns. In− 122.4addition, we inspected−120.2 the a − − − − − ER9 NA various energy49.4 distributions52.9 as a function of time,64.0 as shown in 58.7Figure S4. The components49.9 9BY −20.8 −119.1 −114.4 −116.0 −120.5 −110.8 9BS −24.1 ∆EvdW and− ∆62.4Gnp/solv were generally−67.2 stabilized− 64.9during the entire−57.6 trajectories. However,−66.8 the 9BV −23.5 electrostatic−29.9 and polar solvation−35.5 energies sh−owed46.2 larger fluctuations−41.0 in hcGAS−45.1 binding Pearson’s r with JUJ, KHM,0.52 and KKM than 0.53 the other inhibitors, 0.70 leading to 0.62 binding free energies 0.67 with Spearman’s ρ stronger disturbance.0.32 Despite 0.50 the limitation 0.46s of MM/PBSA studies, 0.46 which were 0.46 mainly a Not available. The experimentaldue binding to force free energy,field accuracy,∆Gexp, was polar obtained contribution using the Equation to solvation,∆Gexp = − RTln(1/entropyKd estimation,). The values of and the sam- inhibitors JUM, JUJ, and ER9 arepling not available, conditions, as Kd values this method are not known. was used as a potential tool for predicting the relative bind- ing free energy of protein-ligand interactions [23,24]. 2.3. Energy Decomposition Analysis Table 3. Comparison of experimentalEnergy binding decompositions energy (∆Gexp for) and 10 hcGAS-inhibitorthe estimated binding interactions free energy, are ∆G shownbind(kJ/mol), in Table at 2. different time scales obtained from MM/PBSA analysis, along with Spearman’s rank correlation coefficient (ρ) and Pearson vdW (∆EvdW), electrostatic energy (∆Eele), and nonpolar solvation energy (∆Gnp/solv) con- correlation coefficient (r). tributed to attractive interactions, while polar solvation energy (∆Gpb/solv) contributed Inhibitor repulsively∆Gexp to the0–4ns total binding 4–8ns free energies. 8–12ns Similar to 12–16nsthe findings of Zhao 16–20ns and co- JUM workersNA a[ 13], the–51.8 total nonpolar –73.5 energies (∆Gnp –73.2) were stronger –76.7 than the total polar –84.5 energies JUJ NA a –89.9 –90.1 –117.0 –114.7 –110.4 KHM –38.5 –155.1 –140.8 –165.9 –160.8 –150.9 8ZM –21.6 –94.1 –93.6 –96.0 –89.8 –95.3 KKP –23.6 –132.2 –124.8 –107.5 –120.3 –106.6 KKM –32.0 –115.4 –124.3 –121.7 –122.4 –120.2 ER9 NA a –49.4 –52.9 –64.0 –58.7 –49.9 9BY –20.8 –119.1 –114.4 –116.0 –120.5 –110.8 9BS –24.1 –62.4 –67.2 –64.9 –57.6 –66.8 9BV –23.5 –29.9 –35.5 –46.2 –41.0 –45.1

Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 6 of 16

Pearson’s r 0.52 0.53 0.70 0.62 0.67 Spearman’s ρ 0.32 0.50 0.46 0.46 0.46 a Not available. The experimental binding free energy, ∆Gexp, was obtained using the Equation ∆Gexp = ‒RTln(1/Kd). The values of the inhibitors JUM, JUJ, and ER9 are not available, as Kd values are not known.

2.3. Energy Decomposition Analysis Energy decompositions for 10 hcGAS-inhibitor interactions are shown in Table 2. vdW (∆EvdW), electrostatic energy (∆Eele), and nonpolar solvation energy (∆Gnp/solv) contrib- Int. J. Mol. Sci. 2021, 22, 1164 uted to attractive interactions, while polar solvation energy (∆Gpb/solv) contributed repul-6 of 16 sively to the total binding free energies. Similar to the findings of Zhao and co-workers [13], the total nonpolar energies (∆Gnp) were stronger than the total polar energies (∆Gpb), indicating that total nonpolar components in the hcGAS-inhibitor interactions contributed (more∆Gpb ),significantly indicating that than total total nonpolar polar interactions. components In in terms the hcGAS-inhibitor of attractive contributions, interactions vdW con- tributedand electrostatic more significantly interactions than contributed total polar interactions.the most to binding In terms free of attractive energies, contributions, especially for vdWthe inhibitors and electrostatic KHM, 9BY, interactions KKM, and contributed ER9. For the the most hcGAS-9BV to binding and free hcGAS-8ZM energies, especially interac- for the inhibitors KHM, 9BY, KKM, and ER9. For the hcGAS-9BV and hcGAS-8ZM interac- tions, the vdW contributions exceeded ‒110 kJ/mol, while the electrostatic interactions tions, the vdW contributions exceeded −110 kJ/mol, while the electrostatic interactions provided moderate contributions under ‒30 kJ/mol, comparable to the nonpolar solvation provided moderate contributions under −30 kJ/mol, comparable to the nonpolar solvation energies. The nonpolar energy of 9BY (∆Gnp = ‒145.3 ± 1.5 kJ/mol) established stronger energies. The nonpolar energy of 9BY (∆Gnp = −145.3 ± 1.5 kJ/mol) established stronger affinitiesaffinities thanthan thosethose ofof 9BS9BS and and 9BV 9BV mentioned mentioned above. above.

2.4.2.4. SingleSingle ResidueResidue EnergyEnergy AnalysisAnalysis TheThe bindingbinding freefree energyenergy ofof singlesingle aminoamino acidsacids inin eacheach systemsystem waswas determineddetermined toto identifyidentify the the residuesresidues that that contributed contributed considerably considerably to to the the bindingbinding and and thethe similaritiessimilarities and and differencesdifferences inin thethe bindingbinding ofof thethe 1010 inhibitorsinhibitors totothe the hcGAS hcGAS protein. protein. WeWe evaluatedevaluated thethe residuesresidues distributed distributed within within 0.6 0.6 nm nm fromfrom anyany atomatom inin thethe inhibitor,inhibitor, thethe per-residueper-residue binding binding freefree energiesenergies ofof whichwhich areare listedlisted inin Table Table S5. S5. Among Among these these amino amino acids, acids, we we selected selected and and focusedfocused on on the the residues residues that that contributed contributed significantly significantly to to the the binding binding free free energy energy (Figure (Figure3). K362,3). K362, R376, R376, Y436, Y436, and and K439 K439 contributed contributed attractively attractively to the to hcGAS-inhibitor the hcGAS-inhibitor interactions, interac- excepttions, except for the for binding the binding of K362 of and K362 K439 and with K439 ER9. with In ER9. addition, In addition, E383 participated E383 participated in more in repulsivemore repulsive interactions interactions with most with inhibitors most inhibitors than the than other the residues, other residues, with a small with binding a small freebinding energy free of energy 1.60 kcal/mol of 1.60 kcal/mol with ER9 with (see ER9 Table (see S5).Table L377 S5). inL377 hcGAS in hcGAS also contributedalso contrib- repulsivelyuted repulsively to its bindingto its binding with variouswith various inhibitors, inhibitors, especially especially in the in JUJ, the 8ZM,JUJ, 8ZM, KKP, KKP, 9BY, 9BS,9BY, and 9BS, 9BV-binding and 9BV-binding systems. systems. Compared Compared to others, to others, only the only inhibitors the inhibitors JUJ, 9BY, JUJ, and 9BY, 8ZM and contributed8ZM contributed distinctly distinctly to binding to binding with H363, with H363, F379, andF379, H437. and H437.

FigureFigure 3.3.Binding Binding free free energies energies of of the the residues residues with with major majo contributionsr contributions within within 6 Å 6 ofÅ inhibitorsof inhibitors in allin hcGAS-inhibitorall hcGAS-inhibitor interactions. interactions.

2.5.2.5. BindingBinding ModesModes ofof thethe InhibitorsInhibitors withwith hcGAShcGAS FigureFigure3 3clearly clearly indicates indicates that that the the polar polar amino amino acids acids K362, K362, R376, R376, and and K439, K439, and and the the aromaticaromatic aminoamino acidacid Y436,Y436, interactedinteracted primarilyprimarily withwith thethe inhibitors,inhibitors, whichwhich werewere inin goodgood agreementagreement with with the the experimental experimental results results [ 10[10–13–13].]. Hence, Hence, we we focused focused on on the the binding binding modes modes of these four key residues with various inhibitors in the active site based on MM/PBSA and MD analyses.

2.5.1. Role of K362 The predicted binding free energies of K362 with the inhibitors JUJ, KHM, 8ZM, KKP, KKM, 9BY, and 9BS were significantly more attractive than –8.0 kJ/mol (see Figure3 and Table S5). Among them, maximum interaction was observed between K362 and KHM, with a binding free energy of −24.90 kJ/mol. Figure4 shows that this strong attractive energy was mainly due to the salt bridge interaction formed by the carboxyl group of KHM and the side chain of K362, which was in agreement with the experimental observation regarding the vital interaction between the carboxylate headpiece of KHM and the polar residue K362, the latter playing a key role in maintaining the cGAS linear intermediate [11]. Second Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 7 of 16

of these four key residues with various inhibitors in the active site based on MM/PBSA and MD analyses.

2.5.1. Role of K362 The predicted binding free energies of K362 with the inhibitors JUJ, KHM, 8ZM, KKP, KKM, 9BY, and 9BS were significantly more attractive than –8.0 kJ/mol (see Figure 3 and Table S5). Among them, maximum interaction was observed between K362 and KHM, with a binding free energy of ‒24.90 kJ/mol. Figure 4 shows that this strong attractive en- Int. J. Mol. Sci. 2021, 22, 1164 ergy was mainly due to the salt bridge interaction formed by the carboxyl group of KHM7 of 16 and the side chain of K362, which was in agreement with the experimental observation regarding the vital interaction between the carboxylate headpiece of KHM and the polar residue K362, the latter playing a key role in maintaining the cGAS linear intermediate in ranking was the binding free energy of KKM (−16.84 kJ/mol), where the O−H group of [11]. Second in ranking was the binding free energy of KKM (‒16.84 kJ/mol), where the KKM was suggested to form one H-bonded contact with the −H group of the side chain O‒H group of KKM was suggested to form one H-bonded contact with the C‒H group of of K362 (see Figure4), although this interaction was not reported in a recent structural the side chain of K362 (see Figure 4), although this interaction was not reported in a recent study [11]. Here, another novel C−H ··· π interaction was predicted between the side structural study [11]. Here, another novel C‒H · · · π interaction was predicted between chain of K362 and the pyridine ring of JUJ, with a binding free energy of −11.61 kJ/mol. the side chain of K362 and the pyridine ring of JUJ, with a binding free energy of ‒11.61 Although K362 in the hcGAS-KKP complex was predicted to have a favorable binding kJ/mol. Although K362 in the hcGAS-KKP complex was predicted to have a favorable energy of −11.84 kJ/mol, no contact between K362 and KKP was formed in either the binding energy of ‒11.84 kJ/mol, no contact between K362 and KKP was formed in either simulation structure or the one [11], which was similar to the interactions between the simulation structure or the crystal one [11], which was similar to the interactions be- K362 and additional inhibitors JUM, 8ZM, ER9, 9BY, 9BS, and 9BV (see Figure S5). Among tweenthese inhibitors, K362 and additional only 8ZM, inhibitors 9BY, 9BS, andJUM, 9BV 8ZM, were ER9, in loose9BY, 9BS, contact and with 9BV the(see polar Figure group S5). Amonghead of these K362. inhibitors, only 8ZM, 9BY, 9BS, and 9BV were in loose contact with the polar group head of K362.

Figure 4. Binding of three inhibitors, JUJ, KHM, and KKM, with K362. The C atoms of inhibitors and residues are shown Figure 4. Binding of three inhibitors, JUJ, KHM, and KKM, with K362. The C atoms of inhibitors and residues are shown in in green and yellow, respectively. The N, O, and Cl atoms are shown as blue, red, and magenta, respectively. All the greenhydrogen and atoms yellow, are respectively. hidden for The clarity. N, O, The and proteins Cl atoms are are displayed shown as as blue, purple red, line and ribbon. magenta, The respectively. light pink, orange, All thehydrogen and blue atomsdashes arerepresent hidden the for C clarity.‒H· · ·π The, salt proteins bridge, and are displayedH-bonded asinteractions, purple line respectively. ribbon. The light pink, orange, and blue dashes represent the C−H···π, salt bridge, and H-bonded interactions, respectively. 2.5.2. Role of R376 2.5.2.Figure Role of 5 R376clearly shows that hcGAS R376 plays a vital role in binding to any inhibitor and tendsFigure to5 useclearly its guanidinium shows that hcGAS group R376with playsthe five- a vital or six-membered role in binding rings to any of heterocy- inhibitor clicand inhibitors, tends to use forming its guanidinium important group cation· with · ·π the interactions, five- or six-membered and affecting rings molecular of heterocyclic recog- nition,inhibitors, biological forming structures, important and cation functions···π interactions, [25,26]. For and example, affecting two molecular cation· · recognition, ·π interac- tionsbiological were structures,formed by and the functionsN atom of [ 25the,26 gu].anidinium For example, group two in cation R376··· inπ complexinteractions with were the diazolo-pyrimidineformed by the N atom aromatic of the core guanidinium of KKP and group KKM in(–24.33 R376 and in complex –26.80 kJ/mol), with the while diazolo- two Npyrimidine atoms of the aromatic guanidinium core of group KKP andin R376 KKM bound (–24.33 with and the –26.80five- and kJ/mol), six-membered while two rings N ofatoms heterocyclic of the guanidinium 9BY and JUJ group (–22.63 in R376and –14.38 bound kJ/mol), with the respectively. five- and six-membered Figure 5 also rings shows of thatheterocyclic R376 uses 9BY an and N atom JUJ (–22.63 of the and guanidinium –14.38 kJ/mol), group respectively. to form one Figure cation·5 also · ·π showsinteraction that withR376 the uses 1,2,4-triazole an N atom of core the guanidiniumof ER9 (–7.45 groupkJ/mol) to and form six-membered one cation··· πringsinteraction of 8ZM with (–19.60 the kJ/mol),1,2,4-triazole 9BS (–14.36 core of kJ/mol), ER9 (–7.45 and kJ/mol) 9BV (–9.90 and six-memberedkJ/mol), respectively. rings of These 8ZM (–19.60structural kJ/mol), inter- actions9BS (–14.36 are consistent kJ/mol), and with 9BV the (–9.90 experimental kJ/mol), respectively.observations These [10,12,13]. structural In addition, interactions hydro- are genconsistent bonding with interactions the experimental were observed observations in KHM [10,12 (–32.09,13]. In kJ/mol), addition, JUM hydrogen (–7.88 bondingkJ/mol), 8ZM,interactions KKM, wereand 9BY observed (–22.63 in kJ/mol) KHM (–32.09 in complex kJ/mol), with JUM the (–7.88 side chain kJ/mol), of R376. 8ZM, Salt KKM, bridge and interaction9BY (–22.63 was kJ/mol) only observed in complex between with the the side guanidinium chain of R376. group Salt of bridgeR376 and interaction the carboxyl was only observed between the guanidinium group of R376 and the carboxyl group of the ER9 interaction. Simultaneously, C−H ··· π interactions were also predicted to be involved in the binding of R376 with inhibitors such as JUJ, 8ZM, KKP, 9BY, and 9BV.

Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 8 of 16

Int. J. Mol. Sci. 2021, 22, 1164 group of the ER9 interaction. Simultaneously, C‒H · · · π interactions were also predicted8 of 16 to be involved in the binding of R376 with inhibitors such as JUJ, 8ZM, KKP, 9BY, and 9BV.

Figure 5. The modes of binding of the 10 inhibitors with R376. The C atoms of the inhibitors and R376 are shown in green Figure 5. The modes of binding of the 10 inhibitors with R376. The C atoms of the inhibitors and R376 are shown in green and yellow, respectively. The N, O, S, and Cl atoms are shown in blue, red, orange, and magenta, respectively. All the andhydrogen yellow, respectively.atoms are hidden The for N, clarity. O, S, and The Clproteins atoms are are displayed shown in as blue, purple red, line orange, ribbon. and The magenta,green, orange, respectively. and light pink All the hydrogendashes atomsrepresent are the hidden H-bonded, for clarity. π· · ·cation The proteins or salt bridge, are displayed and C‒H· as · purple·π interactions, line ribbon. respectively. The green, orange, and light pink dashes represent the H-bonded, π···cation or salt bridge, and C−H···π interactions, respectively. 2.5.3. Role of Y436 2.5.3. RoleThe hcGAS-bound of Y436 inhibitors contain multiple aromatic rings (see Figure 1), which led Theto direct hcGAS-bound π· · ·π stacking inhibitors interactions contain with multiple Y436 at the aromatic active site, rings as (see shown Figure in Figure1), which 6. ledSimultaneously, to direct π··· πOstacking‒H · · · π interactions interactions withwere Y436 often at observed the active when site, Y436 as shown contacted in Figure the 6 . Simultaneously,inhibitors KHM, O JUM,−H 8ZM,··· π KKP,interactions 9BY, and were 9BV. often For instance, observed Y436 when interacted Y436 contacted with three the inhibitorsaromatic rings KHM, of JUM, 9BV and 8ZM, 9BY KKP, in the 9BY, form and of 9BV.π · · · Forπ stacking instance, interactions, Y436 interacted consistent with with three aromaticexperimental rings observations of 9BV and [10]. 9BY Although in the form two of O‒πH··· · · · π interactionsstacking interactions, were present consistent in the with9BY-Y436 experimental interaction, observations the free energies [10]. Although of binding two of the O− twoH ···inhibitors,π interactions 9BV and were 9BY, presentwith inY436 the 9BY-Y436were ‒9.09 interaction,and ‒9.07 kJ/mol the free (see energies Table S5), of respectively. binding of the Second, two inhibitors,Y436 formed 9BV two and 9BY,π· · with·π stacking Y436 wereinteractions−9.09 andwith −the9.07 five- kJ/mol and six-membered (see Table S5), heterocyclic respectively. rings Second, of the in- Y436 hibitors KHM (–8.98 kJ/mol), JUJ (–6.15 kJ/mol), KKM (–13.09 kJ/mol), 8ZM (–13.03 formed two π···π stacking interactions with the five- and six-membered heterocyclic rings kJ/mol), and KKP (–8.00 kJ/mol), where a small hydrophobic pocket was observed be- of the inhibitors KHM (–8.98 kJ/mol), JUJ (–6.15 kJ/mol), KKM (–13.09 kJ/mol), 8ZM tween KHM and Y436 [11]. Among these, the structures of 8ZM, KKM, and KKP were (–13.03 kJ/mol), and KKP (–8.00 kJ/mol), where a small hydrophobic pocket was observed similar, while the free energy of KKP binding to Y436 was predicted to be ‒5.0 kJ/mol between KHM and Y436 [11]. Among these, the structures of 8ZM, KKM, and KKP were weaker than that of 8ZM and KKM binding to Y436 (see Table S5). In addition, Y436 similar, while the free energy of KKP binding to Y436 was predicted to be −5.0 kJ/mol formed π · · · π interactions with two five-membered rings of ER9 (–7.08 kJ/mol), which weaker than that of 8ZM and KKM binding to Y436 (see Table S5). In addition, Y436 was in good agreement with the experimental data [13]. Furthermore, only one π· · ·π π ··· π formedstacking was foundinteractions in the Y436-JUM/9BS with two five-membered interactions. Two rings additional of ER9 (–7.08 O‒H kJ/mol),· · · π and whichone π···π wasC‒H in · good· · π interactions agreement in with the theJUM-Y436 experimental contact, data with [13free]. energy Furthermore, of –7.00 onlykJ/mol, one were stackingalso observed. was found in the Y436-JUM/9BS interactions. Two additional O−H ··· π and one C−H ··· π interactions in the JUM-Y436 contact, with free energy of –7.00 kJ/mol, were also observed.

Int. J. Mol. Sci. 2021, 22, 1164 9 of 16 Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 9 of 16

Figure 6. Binding of the 10 inhibitors used in this study with Y436. The C atoms of inhibitors and Y436 are shown in green Figureand 6.yellow,Binding respectively. of the 10 inhibitorsThe pink and used white in this dashes study represent with Y436. the π The· · ·π C stacking atoms of and inhibitors O‒H· · ·π and interactions, Y436 are shownrespectively. in green andThe yellow, light respectively.pink dashes in The JUM pink and and JUJ whiteshow dashesthe C‒H· represent · ·π interactions. the π··· Theπ stacking schemes and of the O− atomH··· andπ interactions, protein colors respectively. are the Thesame light as pink in Figure dashes 5. in JUM and JUJ show the C−H···π interactions. The schemes of the atom and protein colors are the same as in Figure5. 2.5.4. Role of K439 2.5.4. RoleThe binding of K439 modes of K439 with the 10 inhibitors are shown in Figure S6. No evident contacts,The binding such as modes hydrogen of K439 bonding with or the π‒ 10interactions, inhibitors are were shown observed in Figure in K439 S6. Nobinding evident contacts,with the suchinhibitors, as hydrogen which was bonding consistent or π −wiinteractions,th experimental were observations observed in[10–13]. K439 How- binding withever, the comparable inhibitors, free which energies was consistentof binding with K439 experimental were observed observations for KHM [10 and–13 KKM]. How- ever,(–12.0 comparable kJ/mol), JUM, free JUJ, energies 8ZM, ofKKP, binding 9BY, withand 9BS K439 (–5.42 were to observed–8.86 kJ/mol), for KHM and ER9 and and KKM 9BV (–2.0 kJ/mol) (Table S5), indicating the importance of K439 in interactions with the (−12.0 kJ/mol), JUM, JUJ, 8ZM, KKP, 9BY, and 9BS (−5.42 to −8.86 kJ/mol), and ER9 and inhibitors, especially KHM and KKM, during binding. 9BV (−2.0 kJ/mol) (Table S5), indicating the importance of K439 in interactions with the Aromatic amino acids, H363, F379, and H437, which evidently affected binding with inhibitors, especially KHM and KKM, during binding. the individual inhibitors, were also considered (see Figure 3), although these interactions Aromatic amino acids, H363, F379, and H437, which evidently affected binding with were not observed experimentally [10–13]. In particular, the binding free energy contri- the individual inhibitors, were also considered (see Figure3), although these interactions bution of JUJ-bound H363 to the complex was the largest, predicted to be ‒4.23 kJ/mol (see were not observed experimentally [10–13]. In particular, the binding free energy contri- Table S5). This strong interaction was mainly derived from the hydrogen bonding formed − butionby the ofH363 JUJ-bound backbone H363 and tothe the amino complex pyridine was ring the of largest, JUJ (Figure predicted 7a). Compared to be 4.23 to those kJ/mol (seewith Table the S5).other This inhibitors, strong interactionF379 provided was another mainly distinct derived binding from the free hydrogen energy bonding(‒5.18 formedkJ/mol) by upon the H363binding backbone with 9B andY. The the backbone amino pyridine N atom ringof F379 of JUJ was (Figure involved7a). in Compared the H- tobonded those withcontact the with other the inhibitors, carboxyl group F379 of provided 9BY (Figure another 7b). In distinct addition, binding in the free hcGAS– energy (−8ZM5.18 interaction kJ/mol) upon shown binding in Figure with 7c, 9BY.a hydrogen The backbone bond was N formed atom ofbetween F379 was the side involved chain in theof H-bondedH437 and the contact tetrazolo-pyrimidine with the carboxyl of 8ZM, group leading of 9BY to a (Figurelarger binding7b). In free addition, energy inof the‒ hcGAS–8ZM5.07 kJ/mol, whereas interaction the binding shown inof FigureH437 to7 c,JUJ a and hydrogen ER9 contributed bond was to formed moderate between binding the sidefree chain energies of H437 of ‒2.99 and and the ‒ tetrazolo-pyrimidine2.57 kJ/mol, respectively. of 8ZM, leading to a larger binding free en- ergy of −5.07 kJ/mol, whereas the binding of H437 to JUJ and ER9 contributed to moderate binding free energies of −2.99 and −2.57 kJ/mol, respectively.

Int. J. Mol. Sci. 2021, 22, 1164 10 of 16

Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 10 of 16

Int. J. Mol. Sci. 2021, 22, x FOR PEER REVIEW 10 of 16

Figure 7. The binding details of (a) H363 and JUJ, (b) F379 and 9BY, and (c) H437 and 8ZM interactions. The green dashes Figure 7. The binding details of (a) H363 and JUJ, (b) F379 and 9BY, and (c) H437 and 8ZM interactions. The green dashes represent the hydrogen bonding. The schemes of the atom and protein colors are the same as in Figure 5. represent the hydrogen bonding. The schemes of the atom and protein colors are the same as in Figure5. Figure 7. The binding details of (a) H363 and JUJ, (b) F379 and 9BY, and (c) H437 and 8ZM interactions. The green dashes represent the hydrogen bonding.2.5.5. Analysis The schemes of the of theSandwiched atom and colors are the same as in Figure 5. 2.5.5. Analysis of the Sandwiched Structure Studies have reported that the sandwiched structures formed by the inhibitors with Studies have reported that the sandwiched structures formed by the inhibitors with R3762.5.5. and Analysis Y436 play of the vital Sandwiched roles in cGAS Structure activi ty [10–13]. Therefore, we analyzed the bind- R376 and Y436 play vital roles in cGAS activity [10–13]. Therefore, we analyzed the binding ing modesStudies of inhibitorshave reported bound that to the the sandwiched two residues structures R376 and formed Y436 by in theFigure inhibitors 8. Generally, with modes of inhibitors bound to the two residues R376 and Y436 in Figure8. Generally, theR376 sandwiched and Y436 playstructures vital roles were in predicted cGAS activi byty the [10–13]. aromatic Therefore, rings weof the analyzed bound the inhibitors bind- the sandwiched structures were predicted by the aromatic rings of the bound inhibitors betweening modes the guanidiniumof inhibitors bound group to of the R376 two and resi thedues phenyl R376 andring Y436 of Y436. in Figure For example, 8. Generally, Figure between the guanidinium group of R376 and the phenyl ring of Y436. For example, Figure 8athe shows sandwiched the superposition structures wereof the predicted sandwiched by the structures aromatic inrings JUJ of and the JUM bound between inhibitors R376 8a shows the superposition of the sandwiched structures in JUJ and JUM between R376 and andbetween Y436. theThe guanidinium JUJ and JUM group structures of R376 do and not the overlap phenyl completely, ring of Y436. and For example,the side chainsFigure of Y436.8a shows The JUJ the and superposition JUM structures of the do sandwiched not overlap structures completely, in JUJ and and the JUM side between chains ofR376 both both R376 and Y436 formed π‒interactions with the inhibitors in different directions, R376and and Y436. Y436 The formed JUJ andπ JUM−interactions structures withdo not the overlap inhibitors completely, in different and the directions, side chains which of which was in agreement with the experimental observations [12]. Studies have also re- wasboth in agreementR376 and Y436 with formed the experimental π‒interactions observations with the inhibitors [12]. Studies in different have also directions, reported ported that the two inhibitors, JUJ and JUM, were inserted in the hydrophobic pocket thatwhich the two was inhibitors, in agreement JUJ with and JUM,the experimental were inserted observations in the hydrophobic [12]. Studies pocket have containing also re- F488containingported and that L490, F488 the whereas and two L490, inhibitors, the whereas binding JUJ theand free bindin energiesJUM, gwere free of inserted F488energies and in of L490 the F488 hydrophobic were and predictedL490 were pocket to pre- be −dicted1.13containing kJ/mol to be F488‒1.13 and and −kJ/mol1.40 L490, kJ/mol, and whereas ‒1.40 respectively thekJ/mol, bindin respectivelyg (see free Table energies S5).(see of FigureTable F488 8S5).andb shows L490Figure were that 8b KHMshowspre- wasthatdicted locatedKHM to wasbe in ‒ alocated1.13 deeper kJ/mol in pocket a anddeeper ‒ than1.40 pocket kJ/mol, the other than respectively inhibitors,the other inhibitors,(see resulting Table S5).resulting in the Figure inability in 8b the shows ofinabil- the liganditythat of theKHM to superpose. ligand was locatedto superpose. The in diazolo-/tetrazolo-pyrimidine a deeper The pocket diazolo-/ thantetrazolo-pyrimidine the other inhibitors, aromatic cores resultingaromatic of 8ZM in cores the and inabil- of KKM 8ZM wereandity KKM sandwichedof the were ligand sandwiched betweento superpose. R376 between The and diazolo-/ Y436,R376 and whiletetrazolo-pyrimidine Y436, the while side chains the side aromatic of chains R376 tended coresof R376 of to tended8ZM form hydrogento andform KKM hydrogen bonds were andsandwiched bonds cation and··· between cation·π interactions ·R376 ·π interactions and with Y436, KHM while with and the KHM side KKP and chains (see KKP Figure of R376(see5 Figure), tended which 5), waswhichto inform goodwas hydrogen in agreement good agreementbonds with and experimental cation·with experiment · ·π interactions data [11al]. data Simultaneously, with [11]. KHM Simultaneously, and one KKP 1,2,4-triazole (see one Figure 1,2,4-tri- core5), ofazolewhich ER9 core was was of sandwiched in ER9 good was agreement sandwiched between with R376 experimentbetween and Y436, Ral376 consistent data and [11]. Y436, Simultaneously, with consistent the crystal with one structure the1,2,4-tri- crystal [13]. Comparedstructureazole core [13]. to of bound ComparedER9 was 9BS sandwiched and to 9BV,bound 9BY-bound between9BS and R376R9BV,376 adoptedand9BY-bound Y436, a consistent different R376 adopted alignment with the a crystaldifferent to form sandwichedalignmentstructure to[13]. structures form Compared sandwiched in combination to bound structures 9BS with and in the combination9BV, phenyl 9BY-bound group with ofR376 the T436 phenyladopted (Figure group a8 differentc), whichof T436 was(Figurealignment consistent 8c), whichto withform was thesandwiched consistent experimental structures with data the [inexperimental10 combination]. data with [10]. the phenyl group of T436 (Figure 8c), which was consistent with the experimental data [10].

FigureFigureFigure 8.8.Superposition Superposition 8. Superposition ofof sandwichedofsandwiched sandwiched structuresstructures structures ofof of thethe the inhibitorsinhibitors inhibitors ( (a(aa))) JUM JUM andand JUJ,JUJ, ( ((bb))) KHM, KHM,KHM, 8ZM, 8ZM,8ZM, KKP, KKP,KKP, KKM, KKM,KKM, and andand ER9, ER9,ER9, and (c) 9BY, 9BS, and 9BV between the guanidinium group of R376 and the phenyl ring of Y436. Note that the C atoms of andand (c) 9BY, (c) 9BY, 9BS, 9BS, and and 9BV 9BV between between the the guanidinium guanidinium group group of of R376 R376 and and the the phenyl phenyl ringring of Y436. Note Note that that the the C C atoms atoms of of the theinhibitors inhibitors and and associated associated protei proteinsns are are shown shown in in the the same same color. color. the inhibitors and associated proteins are shown in the same color. 2.5.6.2.5.6. Binding Binding Mechanics Mechanics TheThe details details regarding regarding the the binding binding mechanics ofof the the three three categories categories of of inhibitors inhibitors with with pivotalpivotal residues residues (K362, (K362, R376, R376, Y436,Y436, and K439) areare shownshown in in Figure Figure 9 9in in terms terms of ofsingle single residueresidue binding binding free free energy energy obtainedobtained from energy decompositiondecomposition analysis. analysis. For For JUM JUM and and

Int. J. Mol. Sci. 2021, 22, 1164 11 of 16

2.5.6. Binding Mechanics The details regarding the binding mechanics of the three categories of inhibitors with Int. J. Mol. Sci. 2021, 22, x FOR PEERpivotal REVIEW residues (K362, R376, Y436, and K439) are shown in Figure9 in terms of11 single of 16

residue binding free energy obtained from energy decomposition analysis. For JUM and JUJ, which have tricyclic pyridoindole cores (category I), the free energies of K362 and R376 bindingJUJ, which with have JUJ were tricyclic−7.0 pyridoindole kJ/mol stronger cores than(category those I), of the binding free energies with JUM of K362 (Table and S5), whichR376 was binding due towith the JUJ strong were nonbonded‒7.0 kJ/mol stronger interactions than (thos∆EMMe of) andbinding weak with polar JUM solvation (Table energyS5), which between was K362 due to and the JUJ. strong This nonbonded also confirmed interactions the C− H(∆E···MMπ) andinteraction weak polar between solvation K362 andenergy JUJ, and between the absence K362 and of direct JUJ. This contacts also withconfirmed JUM (Figurethe C‒H4 and· · · Figureπ interaction S6). In between category II, whichK362 and contained JUJ, and fivethe absence inhibitors of (KHM,direct contacts KKP, KKM, with JUM 8ZM, (Figures and ER9), 4 and KHM-bound S6). In category K362 andII, R376 which showed contained higher five bindinginhibitors affinities (KHM, KKP, compared KKM, to8ZM, those and bound ER9), toKHM-bound 8ZM, KKP, K362 KKM, andand ER9 R376 (Table showed S5). higher Figure binding9 shows affinities that the compared nonbonded to those interactions bound to of 8ZM, KHM KKP, with KKM, K362 andand R376 ER9 contributed (Table S5). theFigure most 9 toshows the freethat energy the nonbonded (−25.90 andinteractions−46.42 kJ/mol, of KHM respectively). with K362 Furthermore,and R376 contributed extremely the weak most nonbonded to the free energy interaction (‒25.90 and and polar ‒46.42 solvation kJ/mol, energy respectively). of K362 andFurthermore, K439 with ER9 extremely were observed, weak nonbonded which led inte toraction weak binding and polar contributions solvation energy of −1.0 ofkJ/mol K362 (seeand Table K439 S5). with The ER9 binding were observed, free energies which of led the to fourweak residues binding contributions with 8ZM, KKP, of ‒1.0 and kJ/mol KKM were(see generally Table S5). comparable. The binding Nevertheless,free energies of the the polar four solvationresidues with energy 8ZM, between KKP, and R376 KKM and 8ZMwere contributed generally comparable. attractive energy Nevertheless, of −16.29 the kJ/mol polar solvation (Figure9 ).energy For the between remaining R376 threeand inhibitors,8ZM contributed 9BY, 9BS, attractive and 9BV, theenergy binding of ‒16.29 affinity kJ/mol between (Figure 9BY 9). and For R376 the remaining was predicted three to be inhibitors,−22.63 kJ/mol 9BY, 9BS, (Table and S5), 9BV, which the wasbinding derived affinity from between a strong 9BY nonbonded and R376 was contribution predicted of −42.06to be kJ/mol.‒22.63 kJ/mol Compared (Table to S5), 9BS which and 9BV,was derive only 9BYd from formed a strong H-bonded nonbonded contacts contribution with R376 (Figureof ‒42.065). Figure kJ/mol.6 showsCompared that to one 9BSπ and··· π9BV,stacking only 9BY was formed formed H-bonded by 9BS upon contacts binding with to Y436,R376 while (Figure multiple 5). Figureπ··· 6π showsstacking that and one O π−· H· ·π··· stackingπ interactions was formed were by formed 9BS upon by 9BV/9BYbinding to Y436, while multiple π· · ·π stacking and O‒H · · · π interactions were formed by upon binding to Y436, leading to stronger nonbonded interactions for 9BV and 9BY at 9BV/9BY upon binding to Y436, leading to stronger nonbonded interactions for 9BV and around −15.0 kJ/mol (Figure9). Despite the absence of direct connections between K439 9BY at around ‒15.0 kJ/mol (Figure 9). Despite the absence of direct connections between and all the inhibitors (Figure S6), distinct binding contributions mainly arose from the K439 and all the inhibitors (Figure S6), distinct binding contributions mainly arose from polar solvation energy with comparable attractive contributions (Figure9). Although the the polar solvation energy with comparable attractive contributions (Figure 9). Although MM/PBSA technique holds limitations in estimating the binding free energy of the protein- the MM/PBSA technique holds limitations in estimating the binding free energy of the ligand interaction [22], the above results aid in interpreting the experimental observations protein-ligand interaction [22], the above results aid in interpreting the experimental ob- andservations offer additional and offer insights additional from insights energy from decomposition energy decomposition analysis, which analysis, is not which available is not in theavailable experimental in the studies.experimental studies.

FigureFigure 9. Energy 9. Energy decomposition decomposition on theon the single single residues residues K362, K362, R376, R376, Y436, Y436, and and K439. K439. Compared Compared to to the the nonbonded nonbonded interaction interac- tion and polar solvation energy, nonpolar solvation energy are almost ignored. and polar solvation energy, nonpolar solvation energy are almost ignored. 3. Materials and Computational Methods 3. Materials3.1. Systems and Computational Methods 3.1. SystemsThe crystal structures of hcGAS inhibitors were retrieved from the RCSB Protein Data BankThe (PDB) crystal from structures the entries of hcGAS of 6MJW, inhibitors 6MJU, 6NAO, were retrieved 6NFO, 5VDU, from the 6NFG, RCSB 5VDV, Protein 5V8O, Data Bank5VDW, (PDB) and from 6LRL the [27]. entries The ofchemical 6MJW, structures 6MJU, 6NAO, of inhibitors 6NFO, bound 5VDU, to 6NFG, hcGAS 5VDV, are shown 5V8O, in Figure 1. The sequences of the inhibitor bound hcGAS proteins were almost identical, except for the mutant residues (K427E/K428E) in 6MJW and 6MJU (see Figure S7). The hcGAS-inhibitor structures were treated as one-to-one forms with an identical residue

Int. J. Mol. Sci. 2021, 22, 1164 12 of 16

5VDW, and 6LRL [27]. The chemical structures of inhibitors bound to hcGAS are shown in Figure1. The sequences of the inhibitor bound hcGAS proteins were almost identical, except for the mutant residues (K427E/K428E) in 6MJW and 6MJU (see Figure S7). The hcGAS-inhibitor structures were treated as one-to-one forms with an identical residue fragment of 161−522 in this study. Several missing residues of the initial crystal structures were predicted based on the different templates shown in Table S6 using the MODELLER module of the Chimera [28,29]. Water was retained in the crystal structure to avoid potential interactions between water and the hcGAS-inhibitor systems. Before performing simulations, the protonation states of the charged amino acids were estimated using the H++ web-based tool [30]. In addition to the 4 deprotonated residues H390, C396, C397, and C404, in the zinc-thumb domain of hcGAS, histidines at positions 363 and 390 in 6MJW, 6MJU, 6NFO, and 5VDW were protonated in the Nδ atom; the other charged amino acids were treated as default based on H++ analysis.

3.2. MD Simulations The 16 program was used to calculate the electrostatic potential (ESP) charges for each inhibitor at the HF/6–31G* level [31]. Restraint ESP charges were derived based on the ESP-fit charge model in AMBER20 [32]. All bonding and van der Waals parameters of the inhibitors were generated from the general amber force field [33]. Each hcGAS- inhibitor complex was placed in the center of a cubic box of 708 to 808 nm3 volume and then solvated with 23,640 to 25,860 water molecules using the transferable intermolecular potential with 3 points (TIP3P) water model [34]. Na+ and Cl− counterions were added to neutralize the system with salt concentrations consistent with the ones in experimental studies. The AMBER14SB force field was used to calculate the hcGAS-inhibitor energies for the simulations [35]. After preparing the structural parameters, energy minimizations were performed for each simulation system with a maximum of 5000 steps, followed by 2 short 1 ns simulations with position restraint for the heavy atoms and force value of 1000 kJ/(mol·nm2) in the NVT and NPT ensembles at 300 K and 1 bar. Finally, 3 separate 20 ns MD runs were performed for equilibrium with a time step of 2 fs. Temperature coupling was performed using a velocity-rescaling thermostat [36,37]. The Parrinello– Rahman approach was applied under constant pressure control [38,39]. The covalent bonds involving hydrogen were constrained using the SHAKE algorithm [40,41]. The particle mesh Ewald method was used for treating long-range electrostatic interactions [42]. The cut-off distance of van der Waals (vdW) and electrostatic forces were set to 1.2 nm. The snapshots were saved every 100 ps for further structural analysis. These simulations were performed using the GROMACS 2019 package [43].

3.3. MM/PBSA Calculations g_mmpbsa is a fast and reliable tool for estimating the binding free energy based on GROMACS and APBS programs, especially for protein-ligand interactions [23]. Here, the MM/PBSA approach, combined with 3 parallel 20 ns MD trajectories, was used to obtain the binding affinities of hcGAS-inhibitor interactions based on the g_mmpbsa program. Twenty structures for each complex were used to estimate the binding free energy in terms of every 1 ns extraction of the system coordinate. The binding free energy ∆Gbind of the hcGAS-inhibitor interaction can be evaluated using the following 3 functions:

∆Gbind = ∆EMM + ∆Gsol − T∆S, (1)

∆EMM = ∆EvdW + ∆Eele + ∆Ebonded (2)

∆Gsol = ∆Gpb/solv + ∆Gnp/solv (3)

where ∆EMM represents the sum of the bonded (∆Ebonded) and nonbonded (∆EvdW and ∆Eele) terms. In the single-trajectory setup, ∆Ebonded is always assumed to be zero owing to its accuracy and simplicity, which are close to those of the multi-trajectory approach [44]. The solvation free energy ∆Gsol includes polar solvation energy (∆Gpb/solv) and nonpo- Int. J. Mol. Sci. 2021, 22, 1164 13 of 16

lar solvation energy (∆Gnp/solv), which were estimated using the PB Equation and the solvent-accessible surface area model [45,46]. The default polar and nonpolar parameter settings were identical to those reported previously [23,47]. T and ∆S are the absolute temperature and entropy, respectively; the entropy calculations were ignored in the current calculation process, as they were time-consuming and improved the experimental data only slightly [18,23,48]. The calculations regarding per-residue binding free energy decompositions were also performed to assess the effect of each residue on the hcGAS-inhibitor interaction. The single residue-free energy was evaluated using Equation (4):

bind MM pb/solv np/solv ∆Gres = ∆Eres + ∆Gres + ∆Gres (4)

4. Conclusions MD simulations and MM/PBSA analyses were performed to analyze the binding of various inhibitors (JUM, JUJ, KHM, 8ZM, KKP, KKM, ER9, 9BY, 9BS, and 9BV) with hcGAS. RMSD analysis at approximately 0.25 nm indicated the convergence of all the hcGAS inhibitors. Following this, the binding free energies of hcGAS-inhibitor interactions were estimated using MM/PBSA calculations, which were generally consistent with the experimental affinities, IC50 or Kd. We determined the binding free energies at different time scales for every 2 ns, showing correlation coefficients of 0.67 (r) and 0.46 (ρ). Furthermore, energy decomposition revealed that total nonpolar energy contributed significantly to the total binding energy of each system, where the vdW interaction was stronger than that of the electrostatic interactions. To identify the residue distinctly affecting the inhibitor, we performed per-residue binding energy analysis and investigated the binding modes of the hcGAS-inhibitor complexes. In addition to K362, R376, and Y436 were the major residues that interacted with the inhibitors, which was consistent with the experimental data. New potential roles of the polar residue K439 and aromatic residues H363, F379, and H437 in various interactions, such as hydrogen bonding, salt bridge, π···π stacking, cation···π, O−H···π, and C−H···π interactions, were also revealed. In addition, we analyzed the sandwiched structures of the inhibitor bound to the guanidinium group of R376 and the phenyl ring of Y436; our results suggested that the mechanism of binding mainly involved the impacts from nonbonded interaction and polar solvation energy. Our study indicated that MM/PBSA, in combination with high-throughput virtual screening methods, could be a promising method to screen new inhibitors from the existing database of small molecules. Furthermore, the vital binding modes of hcGAS to inhibitors could be helpful in selecting inhibitors by visual inspection. However, in the current study, the entropy involved in hcGAS-inhibitor interactions was not considered; a potential improvement with the inclusion of entropy should be tested in the future. Moreover, recently, machine-learning algorithms were successfully employed together with MM/PBSA results to achieve a better prediction of the binding free energy of three kinases [49]. A specific machine learning- based predictive model may also be developed in combination with MM/PBSA to improve the estimation of the binding affinity of hcGAS inhibitors.

Supplementary Materials: The following are available online at https://www.mdpi.com/1422-006 7/22/3/1164/s1. RMSD analysis for hcGAS and inhibitors in Figures S1 and S2; correlation plot of the predicted binding free energy between multiple-trajectory and single-trajectory analysis in Figure S3; average energy component distribution as a function of time in Figure S4; binding modes between K362 and some inhibitors in Figure S5; 10 inhibitor compounds bound to K439 in Figure S6; multiple sequence alignment of hcGAS in Figure S7; collection of experimental data regarding hcGAS-inhibitor interactions in Table S1; the binding free energy estimated using MM/PBSA followed with energy decomposition components from their separate trajectories in Tables S2 to S4; the binding free energy of the selected hcGAS residues within 0.6 nm of the corresponding inhibitors in Table S5; information regarding missing residue prediction in Table S6. Int. J. Mol. Sci. 2021, 22, 1164 14 of 16

Author Contributions: W.L. conceived the original idea. W.L. and X.W. designed the project. X.W. performed all the simulations, analyzed the data, and wrote the original draft of the manuscript. All authors have read and approved the final version of the manuscript. Funding: This research was funded by the National Natural Science Foundation of China (grant number 31770777), the Start-up Grant for Young Scientists (860–000002110384), and the Start-up Foundation for Peacock Talents (827–000365), Shenzhen University. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data presented in this study are available in the article and supplementary material. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Zhang, B.; Davidson, M.M.; Hei, T.K. Mitochondria regulate DNA damage and genomic instability induced by high LET radiation. Life Sci. Space Res. 2014, 1, 80–88. [CrossRef][PubMed] 2. Gao, P.; Ascano, M.; Wu, Y.; Barchet, W.; Gaffney, B.L.; Zillinger, T.; Serganov, A.A.; Liu, Y.; Jones, R.A.; Hartmann, G. Cyclic [G (20, 50) pA (30, 50) p] is the metazoan second messenger produced by DNA-activated cyclic GMP-AMP synthase. Cell 2013, 153, 1094–1107. [CrossRef] 3. Kato, K.; Ishii, R.; Goto, E.; Ishitani, R.; Tokunaga, F.; Nureki, O. Structural and Functional Analyses of DNA-Sensing and Immune Activation by Human cGAS. PLoS ONE 2013, 8, e76983. [CrossRef][PubMed] 4. Li, X.; Shu, C.; Yi, G.; Chaton, C.T.; Shelton, C.L.; Diao, J.; Zuo, X.; Kao, C.C.; Herr, A.B.; Li, P. Cyclic GMP-AMP Synthase Is Activated by Double-Stranded DNA-Induced Oligomerization. Immunity 2013, 39, 1019–1031. [CrossRef][PubMed] 5. Cai, X.; Chiu, Y.-H.; Chen, Z.J. The cGAS-cGAMP-STING Pathway of Cytosolic DNA Sensing and Signaling. Mol. Cell 2014, 54, 289–296. [CrossRef][PubMed] 6. Pisetsky, D.S. Anti-DNA antibodies—quintessential biomarkers of SLE. Nat. Rev. Rheumatol. 2016, 12, 102–110. [CrossRef] [PubMed] 7. Crow, Y.J.; Hayward, B.E.; Parmar, R.; Robins, P.; Leitch, A.; Ali, M.; Black, D.N.; van Bokhoven, H.; Brunner, H.G.; Hamel, B.C.; et al. Mutations in the gene encoding the 30-50 DNA exonuclease TREX1 cause Aicardi-Goutières syndrome at the AGS1 locus. Nat. Genet. 2006, 38, 917–920. [CrossRef] 8. Gray, E.E.; Treuting, P.M.; Woodward, J.J.; Stetson, D.B. Cutting Edge: cGAS Is Required for Lethal Autoimmune Disease in the Trex1-Deficient Mouse Model of Aicardi–Goutières Syndrome. J. Immunol. 2015, 195, 1939. [CrossRef] 9. Gao, D.; Li, T.; Li, X.-D.; Chen, X.; Li, Q.-Z.; Wight-Carter, M.; Chen, Z.J. Activation of cyclic GMP-AMP synthase by self-DNA causes autoimmune diseases. Proc. Natl. Acad. Sci. USA 2015, 112, E5699. [CrossRef] 10. Hall, J.; Ralph, E.C.; Shanker, S.; Wang, H.; Byrnes, L.J.; Horst, R.; Wong, J.; Brault, A.; Dumlao, D.; Smith, J.F.; et al. The catalytic mechanism of cyclic GMP-AMP synthase (cGAS) and implications for innate immunity and inhibition. Protein Sci. 2017, 26, 2367–2380. [CrossRef] 11. Hall, J.; Brault, A.; Vincent, F.; Weng, S.; Wang, H.; Dumlao, D.; Aulabaugh, A.; Aivazian, D.; Castro, D.; Chen, M.; et al. Discovery of PF-06928215 as a high affinity inhibitor of cGAS enabled by a novel fluorescence polarization assay. PLoS ONE 2017, 12, e0184843. [CrossRef][PubMed] 12. Lama, L.; Adura, C.; Xie, W.; Tomita, D.; Kamei, T.; Kuryavyi, V.; Gogakos, T.; Steinberg, J.I.; Miller, M.; Ramos-Espiritu, L.; et al. Development of human cGAS-specific small-molecule inhibitors for repression of dsDNA-triggered interferon expression. Nat. Commun. 2019, 10, 2261. [CrossRef][PubMed] 13. Zhao, W.; Xiong, M.; Yuan, X.; Li, M.; Sun, H.; Xu, Y. In Silico Screening-Based Discovery of Novel Inhibitors of Human Cyclic GMP–AMP Synthase: A Cross-Validation Study of Molecular and Experimental Testing. J. Chem. Inf. Model. 2020, 60, 3265–3276. [CrossRef][PubMed] 14. Vincent, J.; Adura, C.; Gao, P.; Luz, A.; Lama, L.; Asano, Y.; Okamoto, R.; Imaeda, T.; Aida, J.; Rothamel, K.; et al. Small molecule inhibition of cGAS reduces interferon expression in primary macrophages from autoimmune mice. Nat. Commun. 2017, 8, 750. [CrossRef][PubMed] 15. Genheden, S.; Ryde, U. The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities. Expert Opin. Drug Discov. 2015, 10, 449–461. [CrossRef][PubMed] 16. Hou, T.; Wang, J.; Li, Y.; Wang, W. Assessing the performance of the MM/PBSA and MM/GBSA methods. 1. The accuracy of binding free energy calculations based on molecular dynamics simulations. J. Chem. Inf. Model. 2011, 51, 69–82. [CrossRef] [PubMed] 17. Wang, E.; Sun, H.; Wang, J.; Wang, Z.; Liu, H.; Zhang, J.Z.H.; Hou, T. End-Point Binding Free Energy Calculation with MM/PBSA and MM/GBSA: Strategies and Applications in Drug Design. Chem. Rev. 2019, 119, 9478–9508. [CrossRef] Int. J. Mol. Sci. 2021, 22, 1164 15 of 16

18. Spiliotopoulos, D.; Spitaleri, A.; Musco, G. Exploring PHD Fingers and H3K4me0 Interactions with Molecular Dynamics Simulations and Binding Free Energy Calculations: AIRE-PHD1, a Comparative Study. PLoS ONE 2012, 7, e46902. [CrossRef] 19. Verma, S.; Grover, S.; Tyagi, C.; Goyal, S.; Jamal, S.; Singh, A.; Grover, A. Hydrophobic Interactions Are a Key to MDM2 Inhibition by Polyphenols as Revealed by Molecular Dynamics Simulations and MM/PBSA Free Energy Calculations. PLoS ONE 2016, 11, e0149014. [CrossRef] 20. Genheden, S.; Nilsson, I.; Ryde, U. Binding Affinities of Factor Xa Inhibitors Estimated by Thermodynamic Integration and MM/GBSA. J. Chem. Inf. Model. 2011, 51, 947–958. [CrossRef] 21. Hou, T.; Wang, J.; Li, Y.; Wang, W. Assessing the performance of the molecular mechanics/Poisson Boltzmann surface area and molecular mechanics/generalized Born surface area methods. II. The accuracy of ranking poses generated from docking. J. Comput. Chem. 2011, 32, 866–877. [CrossRef][PubMed] 22. Sun, H.; Li, Y.; Tian, S.; Xu, L.; Hou, T. Assessing the performance of MM/PBSA and MM/GBSA methods. 4. Accuracies of MM/PBSA and MM/GBSA methodologies evaluated by various simulation protocols using PDBbind data set. Phys. Chem. Chem. Phys. 2014, 16, 16719–16729. [CrossRef][PubMed] 23. Kumari, R.; Kumar, R.; Lynn, A. g_mmpbsa—A GROMACS Tool for High-Throughput MM-PBSA Calculations. J. Chem. Inf. Model. 2014, 54, 1951–1962. [CrossRef][PubMed] 24. Homeyer, N.; Gohlke, H. Free Energy Calculations by the Molecular Mechanics Poisson—Boltzmann Surface Area Method. Mol. Inform. 2012, 31, 114–122. [CrossRef][PubMed] 25. Dougherty, D.A. Cation-π Interactions Involving Aromatic Amino Acids. J. Nutr. 2007, 137, 1504S–1508S. [CrossRef][PubMed] 26. Kumar, K.; Woo, S.M.; Siu, T.; Cortopassi, W.A.; Duarte, F.; Paton, R.S. Cation–π interactions in protein–ligand binding: Theory and data-mining reveal different roles for lysine and arginine. Chem. Sci. 2018, 9, 2655–2665. [CrossRef] 27. Berman, H.M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T.N.; Weissig, H.; Shindyalov, I.N.; Bourne, P.E. The Protein Data Bank. Nucleic Acids Res. 2000, 28, 235–242. [CrossRef] 28. Sali, A.; Blundell, T.L. Comparative Protein Modelling by Satisfaction of Spatial Restraints. J. Mol. Biol 1993, 234, 779–815. [CrossRef] 29. Pettersen, E.F.; Goddard, T.D.; Huang, C.C.; Couch, G.S.; Greenblatt, D.M.; Meng, E.C.; Ferrin, T.E. UCSF Chimera—A visualiza- tion system for exploratory research and analysis. J. Comput. Chem. 2004, 25, 1605–1612. [CrossRef] 30. Anandakrishnan, R.; Aguilar, B.; Onufriev, A.V. H++ 3.0: Automating pK prediction and the preparation of biomolecular structures for atomistic molecular modeling and simulations. Nucleic Acids Res. 2012, 40, W537–W541. [CrossRef] 31. Frisch, M.J.; Trucks, G.W.; Schlegel, H.B.; Scuseria, G.E.; Robb, M.A.; Cheeseman, J.R.; Scalmani, G.; Barone, V.; Petersson, G.A.; Nakatsuji, H.; et al. Gaussian 16 Rev. C.01; Gaussian, Inc.: Wallingford, CT, USA, 2016. 32. Case, D.A.; Belfon, K.; Ben-Shalom, I.Y.; Brozell, S.R.; Cerutti, D.S.; Cheatham III, T.E.; Cruzeiro, V.W.D.; Darden, T.A.; Duke, R.E.; Giambasu, G.; et al. AMBER 2020; University of California: San Francisco, CA, USA, 2020. 33. Wang, J.; Wolf, R.M.; Caldwell, J.W.; Kollman, P.A.; Case, D.A. Development and testing of a general amber force field. J. Comput. Chem. 2004, 25, 1157–1174. [CrossRef][PubMed] 34. Jorgensen, W.L.; Chandrasekhar, J.; Madura, J.D.; Impey, R.W.; Klein, M.L. Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 1983, 79, 926–935. [CrossRef] 35. Maier, J.A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K.E.; Simmerling, C. ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696–3713. [CrossRef][PubMed] 36. Bussi, G.; Donadio, D.; Parrinello, M. Canonical sampling through velocity rescaling. J. Chem. Phys. 2007, 126, 014101. [CrossRef] [PubMed] 37. Berendsen, H.J.C.; Postma, J.P.M.; van Gunsteren, W.F.; DiNola, A.; Haak, J.R. Molecular dynamics with coupling to an external bath. J. Chem. Phys. 1984, 81, 3684–3690. [CrossRef] 38. Parrinello, M.; Rahman, A. Polymorphic transitions in single crystals: A new molecular dynamics method. J. Appl. Phys. 1981, 52, 7182–7190. [CrossRef] 39. Nosé, S.; Klein, M.L. Constant pressure molecular dynamics for molecular systems. Mol. Phys. 1983, 50, 1055–1076. [CrossRef] 40. Miyamoto, S.; Kollman, P.A. Settle: An analytical version of the SHAKE and RATTLE algorithm for rigid water models. J. Comput. Chem. 1992, 13, 952–962. [CrossRef] 41. Andersen, H.C. Rattle: A “velocity” version of the shake algorithm for molecular dynamics calculations. J. Comput Phys. 1983, 52, 24–34. [CrossRef] 42. Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N·log(N) method for Ewald sums in large systems. J. Chem Phys. 1993, 98, 10089–10092. [CrossRef] 43. Abraham, M.J.; Murtola, T.; Schulz, R.; Páll, S.; Smith, J.C.; Hess, B.; Lindahl, E. GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 2015, 1–2, 19–25. [CrossRef] 44. Wang, C.; Greene, D.A.; Xiao, L.; Qi, R.; Luo, R. Recent Developments and Applications of the MMPBSA Method. Front. Mol. Biosci. 2018, 4.[CrossRef][PubMed] 45. Baker, N.A.; Sept, D.; Joseph, S.; Holst, M.J.; McCammon, J.A. Electrostatics of nanosystems: Application to microtubules and the ribosome. Proc. Natl. Acad. Sci. USA 2001, 98, 10037–10041. [CrossRef][PubMed] 46. Sitkoff, D.; Sharp, K.A.; Honig, B. Accurate Calculation of Hydration Free Energies Using Macroscopic Solvent Models. J. Phys. Chem. 1994, 98, 1978–1988. [CrossRef] Int. J. Mol. Sci. 2021, 22, 1164 16 of 16

47. Wang, X.; Zhang, H.; Li, W. DNA-binding Mechanisms of Human and Mouse cGAS: A Comparative MD and MM/GBSA Study. Phys. Chem. Chem. Phys. 2020, 22, 26390–26401. [CrossRef] 48. Yang, B.; Zhu, Y.; Wang, Y.; Chen, G. Interaction identification of Zif268 and TATAZF proteins with GC-/AT-rich DNA sequence: A theoretical study. J. Comput. Chem. 2011, 32, 416–428. [CrossRef] 49. Sun, H.; Pan, P.; Tian, S.; Xu, L.; Kong, X.; Li, Y.; Dan, L.; Hou, T. Constructing and Validating High-Performance MIEC-SVM Models in Virtual Screening for Kinases: A Better Way for Actives Discovery. Sci. Rep. 2016, 6, 24817. [CrossRef]