In Silico Study of the Resistance to Organophosphorus Pesticides Associated with Point Mutations in Acetylcholinesterase of Lepidoptera: B. mandarina, B. mori, C. auricilius, C. suppressalis, C. pomonella, H. armígera, P. xylostella, S. frugiperda, and S. litura

An in silico analysis of the interaction between the complex-ligands of nine acetylcholinesterase (AChE) structures of Lepidopteran organisms and 43 organophosphorus (OPs) pesticides with previous resistance reports was carried out. To predict the potential resistance by structural modifications in Lepidoptera insects, due to proposed point mutations in AChE, a broad analysis was performed using computational tools, such as homology modeling and molecular docking. Two relevant findings were revealed: (1) Docking results give a configuration of the most probable spatial orientation of two interacting molecules (AChE enzyme and OP pesticide) and (2) a predicted ΔGb. The mutations evaluated in the form 1 acetylcholinesterase (AChE-1) and form 2 acetylcholinesterase (AChE-2) structures of enzymes do not affect in any way (there is no regularity of change or significant deviations) the values of the binding energy (ΔGb) recorded in the AChE–OPs complexes. However, the mutations analyzed in AChE are associated with a structural modification that causes an inadequate interaction to complete the phosphorylation of the enzyme.


Introduction
A great diversity of organisms belonging to the order Lepidoptera is of great economic interest. As a study model, nine structures of Lepidoptera acetylcholinesterase (AChE) were treated: Bombyx mandarina, Bombyx mori, Chilo auricilius, Chilo suppressalis, Cydia pomonella, Helicoverpa armígera, Plutella xylostella, Spodoptera frugiperda, and Spodoptera litura. The mulberry silkworm, B. mori, is a Lepidopteran insect of great economic importance because of its use in natural silk fiber production and because it is a valuable insect model that has greatly enhanced our understanding of the biology and altered target-site resistance (the toxin binding site becomes modified to reduce the insecticide's effects); often, more than one of these mechanisms occurs at the same time [11].
According to the mechanisms of action of pesticides, more than 25 types of resistance have been identified and at least 55 types of chemical species [12,13]. One main group is acetylcholinesterase (AChE) inhibitors, which are divided into two subgroups, cataloged as carbamates and organophosphates; both affect the nervous system. AChE is a hydrolytic enzyme that acts on acetylcholine (ACh)-its natural substrate, a neurotransmitter-generating choline and acetic acid. In the presence of organophosphorus (OPs) pesticides, AChE is phosphorylated and, as a consequence, is inhibited [14,15].
The inhibition process involves several stages, outlined in Figure 1. The first is the affinity of OPs for AChE, which determines the reversible inhibition of the enzyme (described by the affinity constant, K a = k +1 /k −1 ) [14]. The second is the phosphorylation constant (known as k +2 or k p ), which governs the rate of formation of the stable covalent bond, causing permanent inhibition of the enzyme and the release of a leaving group (BH, Figure 1) [16]. The third is a hydrolysis reaction considered in homologous kinetic systems and widely studied in lines of research concerning self-reactivating cholinesterases [14,15]. The fourth process is known as the aging of phosphorylation [17,18]. It consists of the hydrolysis of one of the alkyl residues of the phosphate group bound to the active site, giving rise to a very stable complex, characterized by an acid group in the phosphoric center. The irreversibility of the reaction is established by the magnitude of the constant k +4 [18,19]. Regarding the ideal characteristics that a good OP pesticide must meet, at least four kinetic criteria are established in order to provide an optimal pesticide activity as well as safety for mammals: (1) K a (pest) < K a (mammals), (2) K +2 (pest) > K +2 (mammals), (3) K +3 (pest) < K +3 (mammals), and (4) K +4 (pest) > K +4 (mammals) [15]. Int reduce the insecticide´s effects); often, more than one of these mechanisms occurs at the same time [11]. According to the mechanisms of action of pesticides, more than 25 types of resistance have been identified and at least 55 types of chemical species [12,13]. One main group is acetylcholinesterase (AChE) inhibitors, which are divided into two subgroups, cataloged as carbamates and organophosphates; both affect the nervous system. AChE is a hydrolytic enzyme that acts on acetylcholine (ACh)-its natural substrate, a neurotransmitter-generating choline and acetic acid. In the presence of organophosphorus (OPs) pesticides, AChE is phosphorylated and, as a consequence, is inhibited [14,15].
The inhibition process involves several stages, outlined in Figure 1. The first is the affinity of OPs for AChE, which determines the reversible inhibition of the enzyme (described by the affinity constant, Ka = k+1/k−1) [14]. The second is the phosphorylation constant (known as k+2 or kp), which governs the rate of formation of the stable covalent bond, causing permanent inhibition of the enzyme and the release of a leaving group (BH, Figure 1) [16]. The third is a hydrolysis reaction considered in homologous kinetic systems and widely studied in lines of research concerning self-reactivating cholinesterases [14,15]. The fourth process is known as the aging of phosphorylation [17,18]. It consists of the hydrolysis of one of the alkyl residues of the phosphate group bound to the active site, giving rise to a very stable complex, characterized by an acid group in the phosphoric center. The irreversibility of the reaction is established by the magnitude of the constant k+4 [18,19]. Regarding the ideal characteristics that a good OP pesticide must meet, at least four kinetic criteria are established in order to provide an optimal pesticide activity as well as safety for mammals: 1) Ka (pest) < Ka (mammals), 2) K+2 (pest) > K+2 (mammals), 3) K+3 (pest) < K+3 (mammals), and 4) K+4 (pest) > K+4 (mammals) [15]. Where AChEH is the free enzyme; AB, the OP molecule; AChEA´, the dealkylated form of the phosphorylated enzyme (AChEH) [14,15].
The main interest of the research work is performed in an in silico study of the resistance to organophosphorus (OPs) pesticides associated with point mutations in AChE of Lepidoptera, using computational tools to elucidate the structural basis of the mechanism of resistance. The study considers the two forms of AChE present in Lepidoptera, known as form 1 acetylcholinesterase (AChE-1) and form 2 acetylcholinesterase (AChE-2), which present about 40% sequence identity. The OPs molecules evaluated include compounds identified in cases of resistance of Lepidoptera to OP insecticide, consulted in the Arthropod Pesticide Resistance Database (APRD) [20], as well as the database of the Insecticide Resistance Action Committee (IRAC) [21].
The study also includes two non-OPs molecules, acetylcholine (ACh)-a natural substrate of AChE-and a non-OP pesticide, psoralen. Psoralen is a naturally occurring furocoumarin, found in Psoralea, which is a genus in the legume family (Fabaceae). Psoralea corylifolia is an important medicinal plant that is used and widely studied in several traditional medicines to cure various diseases (e.g., anti-carcinogenic activity, anti-depressant activity, skin related problem/leukoderma, Alzheimer's disease) [22,23]. Where AChEH is the free enzyme; AB, the OP molecule; AChEA', the dealkylated form of the phosphorylated enzyme (AChEH) [14,15].

Computational Protocol
The main interest of the research work is performed in an in silico study of the resistance to organophosphorus (OPs) pesticides associated with point mutations in AChE of Lepidoptera, using computational tools to elucidate the structural basis of the mechanism of resistance. The study considers the two forms of AChE present in Lepidoptera, known as form 1 acetylcholinesterase (AChE-1) and form 2 acetylcholinesterase (AChE-2), which present about 40% sequence identity. The OPs molecules evaluated include compounds identified in cases of resistance of Lepidoptera to OP insecticide, consulted in the Arthropod Pesticide Resistance Database (APRD) [20], as well as the database of the Insecticide Resistance Action Committee (IRAC) [21].
The study also includes two non-OPs molecules, acetylcholine (ACh)-a natural substrate of AChE-and a non-OP pesticide, psoralen. Psoralen is a naturally occurring furocoumarin, found in Psoralea, which is a genus in the legume family (Fabaceae). Psoralea corylifolia is an important medicinal plant that is used and widely studied in several traditional medicines to cure various diseases (e.g., anti-carcinogenic activity, anti-depressant activity, skin related problem/leukoderma, Alzheimer's disease) [22,23].

Computational Protocol
The implemented computational protocol consisted of three stages: (a) Modeling by homology using the SWISS-MODEL program [24,25], (b) point mutations performed with the FoldX program [26,27], and (c) molecular docking using Autodock 4.2 [28]. The AChE molecular models, as well as the respective modified models, were constructed satisfactorily (see details about results for Homology Modeling in Table A1, Appendix A). The quality of the models built by homology modeling was analyzed using the multiple 3D alignment option provided by the PDBeFold service (summary showed in Table 1) [29]. In this 3D alignment analysis, the structures that were used as templates for constructed models were used as a reference, PDB 5YDJ (AChE of Anopheles gambiae) and 1DX4 (AChE of Drosophila melanogaster) for AChE-1 and AChE-2 enzymes, respectively. Table 1. Alignment results (performed with PDBeFOLD web server [29]). N res , number of aligned residues; % SI , % Sequence Identity; root-mean-square deviation (RMSD); quality of alignment (Q score) , with 1 being the highest score.

Construction of Molecular Models
The molecular models of the respective AChE enzymes are reliable. In the evaluation of their quality, they register acceptable scores both in their construction (Table A1, Appendix A) and in the 3D alignment, with respect to the reference crystallographic structure ( Table 1). The certainty of the structural predictions is based on the quality of the molecular models (e.g., root-mean-square deviation (RMSD) and quality of alignment (Q score )). In our study, the constructed models are reliable and reproduce the experimental reference evidence.

Docking Results of Acetylcholine Evaluations on AChE
The docking results of the AChE-ACh complex, obtained in the evaluation of AChE-1, form 1 of AChE, and AChE-2, form 2 of AChE (both wild type and modified enzymes), are highlighted below.
(A) AChE-1 and AChE-2 (both wild type enzymes), the energy threshold registered for ACh shows a difference in magnitude between ∆G b AChE-1 wild type and ∆G b AChE-2 wild type (see Figure 6), the exception is AChE of B. mandarina (Figure 2b) (the energies recorded to AChE-2-ACh complexes were equal in magnitude, ∆G b AChE-1 wild type = ∆G b AChE-2 wild type ). The result obtained in AChE of B. mandarina suggests that there is no difference in the affinity of AChEs (∆G b AChE-2 wild type and ∆G b AChE-2 wild type ) for the substrate. Whereas, in the other systems, a difference in affinity between AChE and ACh is predicted, which could be confirmed by catalytic activity studies of the enzymes. (B) AChE-1 wild type and AChE-1 modified , as seen in Figure 4, six studied systems (B. mori- Figure 4a AChE wild type and ∆G b AChE modified ; the energy thresholds of ACh in AChE-2 evaluations do not register a variation ( Figure 6).

Docking Results of Psoralen Evaluations on AChE
The docking results of the AChE-psoralen complex recorded a score −6.8 Kcal/mol < ∆G b < −6.0 Kcal/mol. This result was consistent in all the docking evaluations on AChE (AChE-1 and AChE-2, both wild type and modified with slight variations (see Figure 2a-h, Figures 4a-g and 6a-h)). The results obtained suggest that the specific modifications made to AChE do not affect the recorded energy. Results presented in Figure 5 supports this assertion; the profile of interactions recorded in the AChE-1 of P. xylostella in complex with psoralen shows the presence of hydrogen bonds (recorded in residue Y426) and hydrophobic interactions (recorded in residue W182), present in both AChE-1 wild type and AChE-1 modified . The results of the energy predictions of the enzyme AChE of Lepidoptera in the study of psoralen are promising for an in vitro study, mainly for two reasons: The first is its characteristic of being a natural molecule (no-OP) with inhibitory activity to AChE, and the second, is the scores psoralen obtained in the binding to AChE register a ∆G b < −6.0 Kcal/mol in complex with both AChE enzymes (AChE-1 and AChE-2) of the Lepidoptera organisms studies here.

Resistance-Associate Mutation on AChE
A review of the cases of registered incidences of resistance of Lepidoptera exclusively to the mechanism of action (MoA) of AChE inhibitors by OP (source: www.pesticideresistance.org, consulted in 2018-August) resulted in the top nine resistant Lepidoptera species shown in Table 2. There is a great consistency between the results presented in Figures 4 and 5 and the information contained in Table 2. Eight OPs (chlorpyrifos-methyl (26c), disulfoton (8c), dicrotophos (41c), trichlorfon (15c), mevinphos (40c), dimethoate (7c), acephate (2c), and dichlorvos (6c), in complex with AChE-1 wild type of P. xylostella, presented a hydrogen bond between S297 and O3 from ligand (OP insecticide). This interaction was very important in the effect of irreversible inhibition (Figure 1). In the respective modified enzymes, this interaction is absent-AChE-1 was modified by A201S and G227A and by A201S, G227A, and F290A. Predicted ∆G b results of AChE enzymes in complex with OPs molecules recorded an consistent energy values in the majority of the evaluations, reflected in the profile of the graphics presented in Figure 2a-h, Figure 4a-g and Figure 6a-h, according to the chemical composition of the OP molecule. However, the energy predictions between wild and modified enzymes (AChE-1 and AchE-2) do not show a relevant effect on the recorded magnitude; that is, the proposed mutations do not give rise to an enzyme insensitivity. Molecular models of modified AChE (with point mutations) provide information on predicted ∆G b , in addition to the possible conformation of the interaction mode. Docking results also showed a different location of the binding site of OPs on the enzyme, as a consequence of a change in the electronic atmosphere caused by the point mutation (e.g., acephate (2c) in complex with AChE-1 of P. xylostella, Figure 7). possible conformation of the interaction mode. Docking results also showed a different location of the binding site of OPs on the enzyme, as a consequence of a change in the electronic atmosphere caused by the point mutation (e.g., acephate (2c) in complex with AChE-1 of P. xylostella, Figure 7). The contributions of the present study about modified AChE-2 of Lepidoptera in complex with OP are mainly two: (1) Docking results give a configuration of the most probable spatial orientation of two interacting molecules (AChE enzyme and OP pesticide), and (2) a predicted ΔGb. Regarding the predicted ΔGb results obtained from the evaluations in wild type and modified AChE-2, the results are not conclusive, regarding the effect of the proposed mutations in the direction of generating insensitivity of the enzyme. Only small variations were obtained in the magnitude of the predicted energy, which could be due to the variation of the estimate. Figure 7. Graphical representation of the acephate binding on AChE-1 of P. xylostella. Spatial localization of acephate structure on the AChE-1 enzyme as a result of docking is shown in three molecular models: Conformation acquired on AChE-1 wild type, acephate is chain in red (a), conformation acquired on modified AChE-1 (A201S and G227A, and -A201S, G227A, and F290A, respectively), the acephate is chain in green and magenta (b and c, respectively). The main interactions Figure 7. Graphical representation of the acephate binding on AChE-1 of P. xylostella. Spatial localization of acephate structure on the AChE-1 enzyme as a result of docking is shown in three molecular models: Conformation acquired on AChE-1 wild type, acephate is chain in red (a), conformation acquired on modified AChE-1 (A201S and G227A, and -A201S, G227A, and F290A, respectively), the acephate is chain in green and magenta (b and c, respectively). The main interactions are shown in Figure 5. The location of SER297 is shown, which is an important residue for the effect of enzyme phosphorylation. HIS537 in AChE-1 of P. xilostella is stabilized by GLU296 and GLU423. The pKa values of ionizable residues could be predicted with PROPKA [31].
The contributions of the present study about modified AChE-2 of Lepidoptera in complex with OP are mainly two: (1) Docking results give a configuration of the most probable spatial orientation of two interacting molecules (AChE enzyme and OP pesticide), and (2) a predicted ∆G b . Regarding the predicted ∆G b results obtained from the evaluations in wild type and modified AChE-2, the results are not conclusive, regarding the effect of the proposed mutations in the direction of generating insensitivity of the enzyme. Only small variations were obtained in the magnitude of the predicted energy, which could be due to the variation of the estimate.

Homology Modeling
The AChE sequences (form 1 known as AChE-1 and form 2 known as AChE-2) of the nine lepidopterous organisms were retrieved from GenBank of the National Centre for Biotechnology Information [32]. The accession numbers of AChE-1 and AChE-2 were the following: EU262633. AQQ79918.1 and AQQ79919.1 of S. litura; and AGK44160.1 (AChE-1) of S. frugiperda. Therefore, the molecular models of AChEs were constructed by homology using the SWISS-MODEL program, operating in the search-templates mode, followed by a user-template mode [22,25]. The PDB structures of AChE that recorded the best score (Appendix A) were selected to be used as templates.

Point Mutations on the AChE Structures
Once the 3D models of the wild type AChE enzyme were constructed, we proceeded to perform point mutations, previously described as possibly responsible for diverse resistance to carbamates and organophosphorus compounds in many insect species [33]. The AChE-2 structures of B. mandarina, B. mori, C. auricilius, C. suppressalis, C. pomonella, H. armígera, P. xylostella, and S. litura were modified in eight residues considered the most important mutations in Diptera organisms [9,34,35] (F78S, L(V)129V/T, V150L, A201S, G227A, F290Y, G328A, and G396S, A(K, S)484R corresponding AChE numbering of T. californica) with the FoldX program, using the FoldX Tool BuildModel [26,27]. In addition, molecular models of AChE-1 were generated to construct the modified enzyme by resistance-associated mutations in lepidopterous organisms [9]: A201S, G227A and L452S to B. mandarina and to B. mori; A201S to C. auricilius; A201S to C. suppressalis; F290V to C. pomonella; A201S, G227A and F290A to P. xylostella and S. litura, corresponding to the AChE numbering of T. californica; in Figure 8, a 3D representation is presented.  Figure 5. The location of SER297 is shown, which is an important residue for the effect of enzyme phosphorylation. HIS537 in AChE-1 of P. xilostella is stabilized by GLU296 and GLU423.
The pKa values of ionizable residues could be predicted with PROPKA [31].

Homology Modeling
The AChE sequences (form 1 known as AChE-1 and form 2 known as AChE-2) of the nine lepidopterous organisms were retrieved from GenBank of the National Centre for Biotechnology Information [32]. The accession numbers of AChE-1 and AChE-2 were the following: EU262633. AQQ79918.1 and AQQ79919.1 of S. litura; and AGK44160.1 (AChE-1) of S frugiperda. Therefore, the molecular models of AChEs were constructed by homology using the SWISS-MODEL program, operating in the search-templates mode, followed by a user-template mode [22,25]. The PDB structures of AChE that recorded the best score (Appendix A) were selected to be used as templates.

Point Mutations on the AChE Structures
Once the 3D models of the wild type AChE enzyme were constructed, we proceeded to perform point mutations, previously described as possibly responsible for diverse resistance to carbamates and organophosphorus compounds in many insect species [33]. The AChE-2 structures of B. mandarina, B. mori, C. auricilius, C. suppressalis, C. pomonella, H. armígera, P. xylostella, and S. litura were modified in eight residues considered the most important mutations in Diptera organisms [9,34,35] (F78S, L(V)129V/T, V150L, A201S, G227A, F290Y, G328A, and G396S, A(K, S)484R corresponding AChE numbering of T. californica) with the FoldX program, using the FoldX Tool BuildModel [26,27]. In addition, molecular models of AChE-1 were generated to construct the modified enzyme by resistance-associated mutations in lepidopterous organisms [9]: A201S, G227A and L452S to B. mandarina and to B. mori; A201S to C. auricilius; A201S to C. suppressalis; F290V to C. pomonella; A201S, G227A and F290A to P. In Figure 8, a 3D representation is presented of xylostella and S. litura, corresponding to the AChE numbering of T. californica.

Verification of Computational Methodology
The computational methodology was verified by comparing our results of the binding energy (∆G b ) predicted with the docking results reported previously (see Appendix B) by Somani et al. (2015) to AChE of humans in complex with psoralen (PDB 1EVE was used) [22] and Ranjan et al. (2018) to AChE of humans in complex with several ligands (PDB 3LII was used) [38]; phosalone (CID6199), dimefox (CID8264), dichlorvos (CID5371560), phoxim ethyl phosphonate (CID6507160), heptenophos (CID62773), and methamidophos (CID4096). These molecules were obtained from the PubChem database [36,37]. The docking evaluation was carried out as described in the molecular docking section.

Molecular Docking of Ligands Targeting AChE Enzymes
In order to perform the molecular docking, all the molecular models of the AChE (wild type and modified by point mutations) were prepared in the UCSF Chimera program using the tool dockPrep with standard protocol [39] and saved in PDBQT format, and all ligands were saved in the mol2 format, according to the standard protocol. For each protein structure (AChE-1 and AChE-2) a GridBox was built with a dimension of X = 18.8 Å, Y = 18.8 Å and Z = 18.8 Å, using the MGTool program [28]. The coordinates of the center of the GridBox used were X 1 = −59.62, Y 1 = 58.834, and Z 1 = 16.476. The docking was performed using the AutoDock4.2 program [28], which was installed on a Linux Mint 17.3 operating system implemented with a 3.4 GHz Intel Core i7 processor and 23.5 GB RAM. All molecular models were analyzed with PLIP (Protein-Ligand Interaction Profiler) [30].

Conclusions
The Lepidoptera family has an important economic impact in the world. However, resistance to insecticides is a very common problem in different countries. This resistance is mainly associated with punctual mutations; therefore, a model that helps us know how this change affects the activity of the insecticides is necessary to obtain or apply better control strategies. Based on the results in this study, the proposed mutations are not associated with the presence of insensitivity to the enzyme. However, the mutations evaluated in the AChE-1 and AChE-2 structures of enzymes do not affect in any way (there is no regularity of change or significant deviations) the values of the binding energy (∆G b ) recorded in the AChE-OPs complexes. Therefore, it is assumed that the proposed mutations confer resistance due to an inadequate steric interaction that prevents a phosphorylation reaction of the enzyme by the OP molecule and, therefore, is an irreversible inhibition.

Results for the Homology Modeling and Multiple Alignment of AChE Enzymes
Once the sequences of AChE enzymes of Lepidoptera indicated in Section 4 were retrieved from The National Center for Biotechnology Information [32], we proceeded to construct the molecular models by homology using the SWISS-MODEL program (operating in search-templates mode, followed by user-template mode) [24,25]. The best recorded score was 5YDJ (AChE of Anopheles gambiae) and 1DX4 (AChE of Drosophila melanogaster) for AChE-1 and AChE-2, respectively. Data sheets of the SWISS-MODEL homology modelling report is presented in Table A1. The parameters recorded in the evaluation of built models refer to the structural aspects linked to the template used. The Global Model Quality Estimation (GMQE) score is expressed in values of 0 to 1. Higher numbers indicate higher reliability, reflecting the expected accuracy of a model built with that alignment and template, as well as the coverage of the target. The comprehensive scoring function for model quality assessment (QMEAN Zs ) score provides an estimate of the model on a global scale. Values recorded around zero indicate good agreement between the model structure and experimental structures of similar size, and score values of -4.0 or below indicate low quality models. The alignment of the sequences of the molecular models constructed are presented in Figure A1 (AChE-1) and Figure A2 (AChE-2). Alignments were executed with the Crustal Omega tool [40,41]. experimental structures of similar size, and score values of -4.0 or below indicate low quality models. The alignment of the sequences of the molecular models constructed are presented in Figure A1 (AChE-1) and Figure A2 (AChE-2). Alignments were executed with the Crustal Omega tool [40,41].    experimental structures of similar size, and score values of -4.0 or below indicate low quality models. The alignment of the sequences of the molecular models constructed are presented in Figure A1 (AChE-1) and Figure A2 (AChE-2). Alignments were executed with the Crustal Omega tool [40,41].

Verification of Computational Methodology
The verification of computational methodology consisted of reproducing results of ∆G b , previously reported by Ranjan et al. 2018 [38], Somani et al. 2015 [22], Chaudhry et al. 2013 [42], and Sharma et al. 2011 [43] approaches the interaction of OP with human AChE, where these authors use crystallographic structures to perform their analyses. In order to detect the variability in reproducibility in energy prediction, we evaluated 28 OPs molecules using our methodology (results not shown). We detected (performing 10 runs, in accordance with the docking protocol for the Autodoc 4.2) program variations in the determination of ∆G b of 0.5 Kcal/mol; this parameter was obtained by analyzing the variation in the results of the evaluations of the 28 OPs.
The ∆G b predictions obtained by docking with the purpose of comparing the computational protocol with other computational methodology reporters, in which six molecules on human AChE (PDB: 1EVE) were evaluated, provided results according to those previously reported [22,38] by other research groups (see Table A2). The docking study performed by Somani et al. (2015) and Ranjan et al. (2018) was carried out using the standard glide molecular docking protocol implemented within the Maestro molecular modeling suite by Schrodinger [22,38]. The results register a relative binding energy (∆∆G b ) with a value up to −1.42 Kcal/mol. Considering our variation in the determination of ∆G b (of 0.5 Kcal/mol), we suggest that the results are comparable.  [38]; ∆G b , predicted binding energy in Kcal/mol; *, previously reported; ∆∆G b = ∆G b * − ∆G b , relative binding energy.