Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects

Chemico-Biological Interactions xxx (2012) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Chemico-Biological Interactions journal homepage: www.elsevier.com/locate/chembioint

Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects Galina F. Makhaeva a, Eugene V. Radchenko a,b, Vladimir A. Palyulin a,b, Elena V. Rudakova a, Alexey Yu. Aksinenko a, Vladimir B. Sokolov a, Nikolay S. Zefirov a,b, Rudy J. Richardson c,⇑ a b c

Institute of Physiologically Active Compounds RAS, Chernogolovka 142432, Russia Department of Chemistry, Lomonosov Moscow State University, Moscow 119991, Russia Toxicology Program, The University of Michigan, Ann Arbor, MI 48109-2029, USA

a r t i c l e

i n f o

Article history: Available online xxxx Keywords: Acetylcholinesterase (AChE) Butyrylcholinesterase (BChE) Carboxylesterase (CaE) Neuropathy target esterase (NTE) Organophosphorus compounds (OPCs) Quantitative structure–activity relationships (QSAR)

a b s t r a c t Certain organophosphorus compounds (OPCs) inhibit various serine esterases (EOHs) via phosphorylation of their active site serines. We focused on 4 EOHs of particular toxicological interest: acetylcholinesterase (AChE: acute neurotoxicity; cognition enhancement), butyrylcholinesterase (BChE: inhibition of drug metabolism and/or stoichiometric scavenging of EOH inhibitors; cognition enhancement), carboxylesterase (CaE: inhibition of drug metabolism and/or stoichiometric scavenging of EOH inhibitors), and neuropathy target esterase (NTE: delayed neurotoxicity, OPIDN). The relative degree of inhibition of these EOHs constitutes the ‘‘esterase profile’’ of an OPC and serves as a major determinant of its net physiological effects. Thus, understanding and controlling the esterase profile of OPC activity and selectivity toward these 4 target enzymes is a significant undertaking. In the present study, we analyzed the inhibitor properties of 52 OPCs against the 4 EOHs, along with pairwise and multitarget selectivities between them, using 2 QSAR approaches: Hansch modeling and Molecular Field Topology Analysis (MFTA). The general formula of the OPCs was (RO)2P(O)X, where R = alkyl, X = – SCH(Hal)COOEt (Hal = Cl, Br), –SCHCl2, –SCH2Br, –OCH(CF3)R1 (R1 = C6H5, CF3, COOEt, COOMe). The Hansch model showed that increasing neuropathic potential correlated with rising R hydrophobicity; moreover, OPC binding to scavenger EOHs (BChE and CaE) had different effects on potential acute and delayed neurotoxicity. Predicted protective roles of BChE and CaE against acute toxicity were enhanced with increasing hydrophobicity, but projected protection against OPIDN was decreased. Next, Molecular Field Topology Analysis (MFTA) models were built, considering atomic descriptors, e.g., effective charge, van der Waals radius of environment, and group lipophilicity. Activity/selectivity maps confirmed predictions from Hansch models and revealed other structural factors affecting activity and selectivity. Virtual screening based on multitarget selectivity MFTA models was used to design libraries of OPCs with favorable esterase profiles for potential application as selective inhibitors of CaE without untoward side effects. Ó 2012 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Certain organophosphorus compounds (OPCs) can inhibit various serine esterases (EOHs) via organophosphorylation of serine residues in their active sites. Varying degrees of adverse or thera-

Abbreviations: AChE, acetylcholinesterase; BChE, butyrylcholinesterase; CaE, carboxylesterase; EOH(s), serine esterase(s); MFTA, Molecular Field Topology Analysis; NTE, neuropathy target esterase; OPC(s), organophosphorus compound(s); OPIDN, organophosphorus compound-induced delayed neurotoxicity; PLS, partial least squares; QSAR, quantitative structure–activity relationships. ⇑ Corresponding author. Address: Computational Toxicology Laboratory, Toxicology Program and Risk Science Center, Department of Environmental Health Sciences, The University of Michigan, Ann Arbor, MI 48109-2029, USA. Tel.: +1 734 936 0769; fax: +1 734 763 8095. E-mail address: [email protected] (R.J. Richardson).

peutic effects arise from OPC exposure depending in part on their relative inhibitory potencies against EOHs of interest – the ‘‘esterase profile’’ [1–4]. We studied inhibitory characteristics of OPCs against a panel of 4 EOHs whose inhibition is linked to acute neurotoxicity (acetylcholinesterase, AChE, EC 3.1.1.7) [5], delayed neurotoxicity (neuropathy target esterase, NTE, 3.1.1.5) [6,7], and drug metabolism or stoichiometric scavenging of EOH inhibitors (butyrylcholinesterase, BChE, EC 3.1.1.8; and carboxylesterase, CaE, EC 3.1.1.1) [8– 10]. Inhibition of AChE and/or BChE can also exert a therapeutic effect of cognition enhancement in Alzheimer’s disease [11,12]. Clearly, the particular pattern of inhibition of these 4 targets plays an important role in shaping the pharmacodynamics and pharmacokinetics of a given OPC, thereby serving as a determinant of its overall physiological influences. Accordingly, analysis of the ester-

0009-2797/$ - see front matter Ó 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cbi.2012.10.012

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

2

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx

ase profiles of OPCs using modern QSAR methods enables us to determine the contribution of different structural elements to their potential biological effects [2–4,13,14]. The following parameters describing inhibitory activity and selectivity of the OPCs were used in our esterase profile analysis. A parameter related to inhibitory activity toward the 4 EOHs was defined as follows: AX = log ki (EX), where AX is the activity toward the enzyme EX. These activities were named AA for AChE, AB for BChE, AN for NTE and AC for CaE. Pairwise inhibitor selectivity of an OPC between two enzymes was defined with a parameter as follows: SXY = log ki (EX) log ki (EY). In this way 6 selectivity parameters were defined: SNA for NTE/AChE, SBA for BChE/AChE, SCA for CaE/ AChE, SBN for BuChE/NTE, SCN for CaE/NTE, and SCB for CaE/BChE. Thus, we had 4 indicators of activity and 6 of selectivity for a total of 10 endpoints. Each of these parameters is based on the bimolecular rate constant(s) of inhibition (ki) of a given OPC toward the enzyme(s) of interest, and each is important for certain aspects of the pharmacological and toxicological profile of the OPC [4]. For example, ki(NTE)/ki(AChE) represents the relative inhibitory potency of an OPC against targets for delayed neurotoxicity (NTE) and acute neurotoxicity (AChE). This ratio correlates with that between the LD50 and the neuropathic dose, thereby serving as an index of the neuropathic potential of an OPC that is subject to aging [6,15,16]. Likewise, ki(BChE)/ki(AChE) and ki(CaE)/ki(AChE) reflect the potential contributions of BChE and CaE to the attenuation of acute toxicity via stoichiometric scavenging. Similarly, ki(BChE)/ ki(NTE) and ki(CaE)/ki(NTE) represent the contributions of BChE and CaE to the potential mitigation of delayed neurotoxicity [2]. Finally, ki(CaE)/ki(BChE) characterizes the inter-scavenger selectivity of an OPC. In the present work, inhibitor properties of 52 OPCs – O,O-dialkylphosphates of general formula (RO)2P(O)X (Fig. 1) – against the 4 EOHs of interest, along with pairwise and multitarget selectivities between them, were analyzed using two QSAR approaches: (1) Hansch’s analysis in certain homologous series; and (2) Molecular Field Topology Analysis (MFTA). In addition, because CaE inhibition can result in reducing hydrolytic metabolism of many therapeutically important drugs [10,17,18], we applied MFTA to design a library of inhibitors as potential modulators of the pharmacokinetics of drugs containing ester or amide bonds. 2. Methods 2.1. Kinetic data on EOH inhibition The bimolecular rate constants of inhibition (ki) of the OPCs were determined using human erythrocyte (RBC) AChE, equine ser-

um BChE, and porcine liver CaE (Sigma, St. Louis, MO), as well as avian (female Gallus domesticus) brain NTE (prepared in our laboratory as described previously [19]). The use of enzymes from different tissues and species is a potential limitation of the study. However, regarding tissues, it is known that the catalytic domain of human AChE is a single gene product that is identical in RBCs and brain AChE [20], BChE is a single gene product secreted into the serum from the liver [21], hen brain is commonly used as the source for NTE [22], and CaE, considered as liver carboxylesterase 1 (CES1), is produced in the liver, which may export it to plasma in many animal species [23], but not in humans [24]. With respect to species, protein sequence identities compared to the human sequence (determined using NCBI protein BLAST, http://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Proteins) were human AChE, 100%; horse BChE, 90%; porcine CaE, 77%; and hen NTE, 63%. Despite the species differences, QSAR predictions based on these inhibition constants have been confirmed in other studies using enzymes from the same species [15,16,25–28]. Detailed descriptions of the inhibition kinetics have been presented elsewhere [29]. In brief, enzyme samples were incubated for different time intervals with an inhibitor such that [I]o >> [E]o, and the residual enzymatic activity was determined. The ki values were calculated according to [30] by linear regression using OriginPro 6.1 software and published earlier along with details of the synthesis and chemical characterization of the OPCs [2,3,31–37]. 2.2. QSAR modeling using Hansch’s approach The relationship between structure of the OPCs and their inhibitor selectivity was analyzed by polynomial regression analysis using Origin 6.1 software, OriginLab Corp., (Northampton, MA, USA). Hansch constants for hydrophobicity of substituents, p, were P used additively to yield values of p for the R-groups in the OPCs [38], and QSAR models for inhibitor selectivity of OPCs were developed. The significance of the equations obtained for N data points was estimated with values of R (correlation coefficient), S (standard deviation of the fit), and P (probability that R2 is zero). 2.3. QSAR modeling using Molecular Field Topology Analysis (MFTA) The bioactivity model in MFTA was constructed from values of local molecular descriptors (e.g., atomic properties) [39,40]. Twodimensional structures of compounds in the training set (structural formulas) were topologically superimposed to construct a molecular supergraph to provide a common frame of reference for meaningful comparison and analysis of local properties in different structures. In addition to the predictive model, constructed

I

II

III

IV

V

VI

VII

VIII

Fig. 1. General structures (I–VIII) of the OPCs in this study.

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

3

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx

by partial least squares (PLS) regression and relating these properties at all positions of the molecular supergraph to bioactivity, this method gives a graphic map of the effect of properties on activity. The predictivity of models was estimated by cross-validation with exclusion of 25% of training set compounds using the special stabilization procedure where Q2 (cross-validation parameter) values are averaged over many random reshufflings of the cross-validation subsets, thus providing a significant degree of independence of the actual data partitioning. This procedure generally gives lower Q2 values than other approaches; therefore, the usual criteria are not applicable. 3. Results and discussion 3.1. Hansch’s modeling This approach was applied to analysis of the esterase profile of two homologous series of OPCs, VI and VII (Fig. 1). Fig. 2A demonstrates a dependence on hydrophobicity for the anti-AChE activity (AA) of VI and VII and their selectivity for BChE and CaE versus AChE (SBA and SCA). Fig. 2B shows the dependence on hydrophobicity for the neuropathic potential of VI and VII (SNA) and their selectivity for BChE and CaE versus NTE (SBN and SCN). The relationships between the structures of compounds VI and VII and their inhibitor selectivity SXY was analyzed with polynomial regression analysis and the corresponding Hansch’s models

of structure-inhibitor selectivity relationships were developed. The models can be described by the general equation, SXY = P P A + B ( p) + C ( p)2. Coefficients of the best equations, equation number, and the statistical performance of the Hansch models are presented in Table 1. The dependence of the neuropathic potential, SNA, of compounds VI and VII on the hydrophobicity of alkyl radicals was found to be described by Eqs. (1) and (2), respectively (Table 1), which show that the neuropathic potential in both OPC series increases with an increase in hydrophobicity (Fig. 2B). The selectivity of these OPCs for scavenger esterases, BChE and CaE, in comparison to AChE (Fig. 2A), is described by parabolic Eqs. (3) and (4) for VI and Eqs. (5) and (6) for VII (Table 1). As can be seen from Fig. 2A and Eqs. (3) through (6), the selectivity of OPC VI and VII for scavenger esterases BChE and CaE versus the target esterase AChE was enhanced with increasing hydrophobicity. We have suggested that the ratios ki(BChE)/ki(AChE) and ki(CaE)/ki(AChE) characterize the potential contribution of BChE and CaE to the suppression of acute toxicity of an OPC by virtue of stoichiometric scavenging. Accordingly, the protective roles of BChE and CaE against acute toxicity of compounds VI and VII should be enhanced with increasing hydrophobicity. This conclusion has been validated by our results on acute toxicity of ethyl and butyl derivatives of compounds VI and VII: LD50 = 200 mg/kg for diEt-VI and >2500 mg/kg for diBu-VI [2]; LD50 = 720 mg/kg for diEt-VII and 1560 mg/kg for diBu-VII [3].

A

B

VI logkiAChE S(B/A) S(C/A)

3

4 3

SCA

2

2

SBA

SCN

1 0

SNA= RIP

-1

1

S(N/A) S(B/N) S(C/N)

SBN

2 Selectivity

Selectivity

3

AA

4

1

4

5

logki(AChE)

5

-2

0

0 0

1

2

3

4

5

6

0

7

1

2

3

4

5

6

Σπ

Σπ

VII 3

logkiAChE S(B/A) S(C/A)

SCA SBA

2 1

Selectivity

Selectivity

3

2

1

SCN

0 -1

SNA= RIP

0

0 1

2

3 Σπ

S(N/A) S(B/N) S(C/N)

2

AA

3

1

SBN

4

logki(AChE)

4

4

5

-2 1

2

3

4

5

Σπ

Fig. 2. (A) Dependence of anti-AChE activity (AA) of OPCs in classes VI and VII and their selectivities for BChE and CaE as compared to AChE (SBA and SCA) on the hydrophobicity of alkyl radicals. The dashed lines for BChE/AChE (open triangles) and CaE/AChE (stars) selectivities were plotted according to Eqs. (3–6). (B) Dependence of neuropathic potential of OPCs in classes VI and VII (SNA) and their selectivities for BChE and CaE as compared to NTE (SBN and SCN) on the hydrophobicity of alkyl radicals. The continuous lines for selectivity of NTE/AChE were obtained from Eqs. (1–2); the dashed lines for BChE/NTE (open triangles) and CaE/NTE (stars) selectivities were plotted according to Eqs. (7–10).

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

4

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx

Table 1 P P Hansch’s models which are described with a general equation SXY = A + B ( p) + C ( p)2 and their statistical performance.

a b c d e f

Endpointa/Equation #

Series

SNA/(1) SNA/(2) SBA/(3) SCA/(4) SBA/(5) SCA/(6) SBN/(7) SCN/(8) SBN/(9) SCN/(10)

VI VII VI VI VII VII VI VI VII VII

A

B 2.91 ± 0.58 2.24 ± 0.21 0.94 ± 0.63 1.26 ± 0.37 0.17 ± 0.25 1.35 ± 0.51 1.93 ± 0.33 4.15 ± 0.66 2.11 ± 0.22 0.59 ± 0.13

1.42 ± 0.35 0.58 ± 0.06 1.23 ± 0.38 0.55 ± 0.22 0.81 ± 0.18 1.05 ± 0.37 0.16 ± 0.09 0.86 ± 0.4 0.47 ± 0.16 0.71 ± 0.09

C 0.14 ± 0.05 0f 0.14 ± 0.05 0.07 ± 0.03 0.13 ± 0.03 0.08 ± 0.06 0f 0.06 ± 0.06 0.17 ± 0.03 0.12 ± 0.02

Nb

Rc

Sd

Pe

9 7 9 9 7 7 9 9 7 7

0.935 0.971 0.860 0.699 0.919 0.962 0.764 0.895 0.993 0.975

0.343 0.207 0.371 0.221 0.123 0.256 0.379 0.392 0.109 0.063

0.00196 0.00027 0.0177 0.133 0.0244 0.0056 0.0991 0.00789 0.0002 0.0025

Endpoints as defined in the introduction as corresponding selectivities. Number of compounds. Correlation coefficient. Standard deviation of the fit. Probability that R2 is zero. Fitted with a first order polynomial.

The selectivity of compounds VI and VII against scavenger esterases, BChE and CaE, in comparison to the delayed neuropathy target, NTE (Fig. 2B) is described by Eqs. (7) and (8) for VI and Eqs. (9) and (10) for VII (Table 1). With increasing hydrophobicity, the binding of OPCs VI and VII to the scavenger esterases BChE and CaE is reduced (Fig. 2B; Eqs. (7) through (10)), whereas the neuropathic potential of these compounds, SNA, is increased (Fig. 2B; Eqs. (1) and (2)). This correlates with escalated risk of delayed neurotoxicity for the longer-chain OPCs: as we showed recently, NTE was inhibited only by 10% in mouse brain by diEt-VI at a dose equal to the LD50, whereas the ratio ED50(AChE)/ED50(NTE) was equal to 4 in mouse brain after diBu-VI dosing [26]. Of course, in a biosystem the concentration, as well as the intrinsic activity, of a scavenging enzyme is undoubtedly important for the level of protection. Therefore, protection may differ in different species and the role of BChE or CaE may be predominant depending on species [41]. Indeed, a goal of current research on bioscavenger therapy is to enhance both the catalytic efficiency and concentration of such enzymes by injecting suitably engineered proteins into the bloodstream [42]. Thus, the Hansch models we developed demonstrate that increasing neuropathic potential correlates with rising hydrophobicity of substituent R. Moreover, the models show that OPC binding to scavenger EOHs (BChE and CaE) has different effects on acute and delayed neurotoxicity. Namely, the protective roles of BChE and CaE against acute toxicity were enhanced by increasing hydrophobicity, but protection against OPIDN was decreased. The results agree with data showing the low neuropathic potential of OPCs with short R-groups (Me, Et) versus the high propensity to induce OPIDN shown by OPCs with long R-groups [15,16,43–45]. 3.2. MFTA modeling MFTA models were constructed for all of the activity and selectivity endpoints. The structure of the molecular supergraph and the example superimposition for compound 10 are shown in Fig. 3.

Table 2 Statistical performance of MFTA models. Endpoint AA AB AC AN SBA SCA SNA SBN SCN SCB AABN SC|ABN

a

Nb

NFc

R2d

RMSEe

Q2f

RMSEcvg

52 52 42 27 52 42 27 27 18 42 27 18

5 2 5 6 2 6 4 2 3 1 5 2

0.91 0.87 0.96 0.96 0.57 0.83 0.94 0.88 0.94 0.60 0.91 0.78

0.41 0.54 0.44 0.31 0.43 0.46 0.31 0.35 0.35 0.80 0.35 0.43

0.82 0.80 0.91 0.88 0.35 0.62 0.83 0.73 0.83 0.47 0.71 0.42

0.57 0.66 0.68 0.56 0.52 0.70 0.52 0.51 0.57 0.92 0.56 0.68

a Endpoints as defined in the introduction AABN and SC|ABN are additional objective parameters for optimization of activity profiles as defined in Section 3.3. b Number of compounds. c Number of factors in the PLS regression model. d Squared correlation coefficient. e Root-mean-square error. f Cross-validation parameter. g Root-mean-square error for cross validation.

Optimal results were obtained when using the following local molecular descriptors [40]: the effective charge, Q, on an atom, estimated by electronegativity equalization; the effective van der Waals radius, Re, which takes into account the steric requirements of the central non-hydrogen atom and other atoms bound to it; and the group lipophilicity, Lg, which takes into account the contributions of the central non-hydrogen atom and the hydrogen atoms bound to it. The statistical parameters of the models are shown in Table 2. All of them provide good or reasonable predictivity. Fig. 4 shows the activity maps describing the effect of all local parameters on anti-AChE and anti-NTE activity as well as the neuropathicity maps reflecting the selectivity of compounds to NTE compared to AChE. Although alkoxy groups with C1–C5 chains tend

Fig. 3. Structure of the molecular supergraph and the example superimposition for OPC 10.

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

5

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx

Q

Re

Lg

AN

SNA

AA

Fig. 4. MFTA activity and selectivity maps illustrating the effect of local molecular descriptors on the ability of OPCs to inhibit AChE (AA) or NTE (AN) and the inhibitor selectivity of the OPC for NTE in comparison to AChE (SNA). In positions marked with red circles (black in the print version), an increase in the descriptor, all other things being equal, tends to increase the activity (selectivity) of OPCs. Conversely, in positions marked with blue circles (grey in the print version), an increase in the descriptor is accompanied by a probable decrease in the activity. Intensity of colors (in the online version) reflects the magnitude of the influence.

(1)

SBA

SBN

Lg

(2)

SCA

SCN

3.3. Molecular design of selective CaE inhibitors Selective CaE inhibitors suitable for use as novel modulators of pharmacokinetics of ester-containing drugs should have high activity and selectivity toward CaE, low activity toward AChE and BChE, and very low or no activity toward NTE. This multi-target/ multi-objective optimization problem is difficult to solve based on the consideration of individual activities and selectivities. We proposed that in such cases it is beneficial to introduce integral parameters of positive and negative effects that are based on domain knowledge and easily interpretable [46]. These parameters can be constructed as generalized mean values of the respective

Lg 5

Fig. 5. MFTA selectivity maps illustrating the effect of group lipophilicity (Lg) on the inhibitor selectivity of the OPC. (1) Selectivity for BChE in comparison to AChE (SBA) versus their selectivity for BChE in comparison to NTE (SBN). (2) Selectivity for CaE in comparison to AChE (SCA) versus their selectivity for CaE in comparison to NTE (SCN). For notation see Fig. 4.

to increase activity against both AChE and NTE, a more focused selectivity map demonstrates that the neuropathic potential is increased by the C2–C5 chains, especially with branching in the c-position. In addition, CF3 and phenyl substituents in the leaving group also shift the selectivity toward NTE. Fig. 5 shows the selectivity maps describing the effect of one of the local parameters, group lipophilicity (Lg), on the inhibitor selectivity of the OPC toward scavenging esterases BChE and CaE in comparison to the target enzymes AChE and NTE. One can see a positive influence of hydrophobic substituents on the OPC selectivity for BChE and CaE in comparison to AChE and a negative influence on the selectivity toward BChE and CaE in comparison to NTE. That is, OPC binding to scavenging esterases (BChE and CaE) has different effects on acute and delayed neurotoxicity. This result confirms the predictions from our Hansch models.

SC|ABN

4

3

2 1

0

-1

-2 0

2

4

6

8

10

12

14

AC Fig. 6. Distribution of selectivity and activity toward CaE in the generated structure library of OPCs. The box in the upper right-hand corner marks the desired property range.

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

6

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx CH3 F O

H3C

O

S

F F

O

H3C

O

P

H3C

O

H3C

CH3

Cl

O P

F O

O

H3C

CH3

F F

H3C

H3C CH3

H3C

O

CH3

S P

F F

O CH3

F

H3C

Fig. 7. Sample structures with high predicted selectivity and activity toward CaE.

individual activities. In this study, the power-3 (cubic) mean was used as a simple function that increases monotonically with increasing individual values while being partially dominated by larger values. Thus, we introduce the integral inhibitor activity parameter for AChE, BChE, and NTE as AABN = [1/3 (AA3 + AB3 + AN3)]1/3. The main objective parameter for optimization, selectivity toward CaE vs. AChE, BChE and NTE, is then defined as SC|ABN = AC AABN. The statistical characteristics of the MFTA models for these parameters are also listed in Table 2. The models suggest that short-chain alkoxy substituents without a- and b-branching are preferred for the optimal esterase profile, while bis-CF3 substitution in the leaving group should be avoided. However, the whole picture of influence is rather complicated (the activity and selectivity maps for AC, AABN and SC|ABN are presented in the Supplementary materials). For this reason, we have performed virtual screening [40] of potential selective CaE inhibitors in a generated structure library of OPCs. A library of 3000 structures was built from the MFTA model supergraph in stochastic mode by means of the specialized MFTA-oriented structure generator we developed [47]. After applicability domain filtering using PLS regression outlier scores, a set of 2102 structures was obtained that covered a wide range of activity and selectivity (Fig. 6). From this set, a focused library of 261 structures was selected that provided high predicted selectivity and activity toward CaE (SC|ABN > 3, AC > 9); this is represented by the box in the upper right-hand corner of Fig. 6. Some of the proposed structures from this focused library are shown in Fig. 7. 4. Conclusions The relationships between the structures of 52 dialkylphosphates and their inhibitor activity and selectivity toward 4 pharmacologically and toxicologically important EOHs were successfully analyzed by means of two approaches: (1) Hansch analysis for small homologous groups; and (2) MFTA based on local descriptors for the entire data set. Hansch models showed that increasing neuropathic potential correlated with rising hydrophobicity of substituents and that OPC binding to scavenger esterases (BChE and CaE) had different effects on acute and delayed neurotoxicity. MFTA activity/selectivity maps confirmed predictions from the Hansch models and revealed other structural factors affecting activity and selectivity. Virtual screening based on multitarget selectivity MFTA models was used to design libraries of OPCs with favorable esterase profiles for potential application as selective inhibitors of CaE that would be predicted to lack untoward side effects. Future work will include additional EOH targets of OPCs [22,48] along with a more detailed analysis of inhibition, aging, and reactivation kinetics [49]. Conflict of interest statement The authors declare that there are no conflicts of interest.

Acknowledgements The authors gratefully acknowledge the support of this work by the Russian Foundation for Basic Research (RFBR, Grant Nos. 1103-00581 and 11-03-01174), Ministry of Education and Science of Russia (Federal Program ‘‘Scientists and Educators for Innovative Russia’’ 2009–2013), Russian Academy of Sciences (RAS), NATO Science for Peace and Security Program, Grant No. 984082, and the University of Michigan Risk Science Center.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cbi.2012.10.012. References [1] G. Makhaeva, I. Filonenko, S. Fomicheva, V. Malygin, ‘‘Esterase profiles’’ of O,Odialkyl-O-dimethylchloroformimino phosphates in prediction of their toxic effects, Toxicol. Lett. 88 (Suppl. 1) (1996) 25. [2] G.F. Makhaeva, O.G. Serebryakova, N.P. Boltneva, T.G. Galenko, A.Y. Aksinenko, V.B. Sokolov, I.V. Martynov, Esterase profile and analysis of structure-inhibitor selectivity relationships for homologous phosphorylated 1hydroperfluoroisopropanols, Dokl. Biochem. Biophys. 423 (2008) 352–357. [3] E.V. Rudakova, G.F. Makhaeva, T.G. Galenko, A.Yu. Aksinenko, V.B. Sokolov, R.J. Richardson, I.V. Martynov, Esterase profile of O-phosphorylated trifluorolactates in prediction of their therapeutic and toxic effects, Dokl. Biochem. Biophys. 443 (2012) 81–85. [4] G.F. Makhaeva, E.V. Radchenko, I.I. Baskin, V.A. Palyulin, R.J. Richardson, N.S. Zefirov, Combined QSAR studies of inhibitor properties of O-phosphorylated oximes toward serine esterases involved in neurotoxicity, drug metabolism and Alzheimer’s disease, SAR QSAR Environ. Res. 23 (2012) 627–647. [5] R.J. Richardson, Anticholinesterase insecticides, in: C.A. McQueen (Ed.), Comprehensive Toxicology, vol. 13, Academic Press/Elsevier, Oxford, 2010, pp. 433–444. [6] R.J. Richardson, R.M. Worden, G.F. Makhaeva, Biomarkers and biosensors of delayed neuropathic agents, in: R.C. Gupta (Ed.), Handbook of Toxicology of Chemical Warfare Agents, Academic Press/Elsevier, Amsterdam, 2009, pp. 859–876. [7] S.J. Wijeyesakere, R.J. Richardson, Neuropathy target esterase, in: R. Krieger (Ed.), Hayes’ Handbook of Pesticide Toxicology, Academic Press/Elsevier, New York, 2010, pp. 1435–1455. [8] D.M. Maxwell, K.M. Brecht, Carboxylesterase: specificity and spontaneous reactivation of an endogenous scavenger for organophosphorus compounds, J. Appl. Toxicol. 21 (Suppl. 1) (2001) S103–S107. [9] P. Masson, O. Lockridge, Butyrylcholinesterase for protection from organophosphorus poisons: catalytic complexities and hysteretic behavior, Arch. Biochem. Biophys. 494 (2010) 107–120. [10] P.M. Potter, R.M. Wadkins, Carboxylesterases – detoxifying enzymes and targets for drug therapy, Curr. Med. Chem. 13 (2006) 1045–1054. [11] Y. Furukawa-Hibi, T. Alkam, A. Nitta, A. Matsuyama, H. Mizoguchi, K. Suzuki, S. Moussaoui, Q.S. Yu, N.H. Greig, T. Nagai, K. Yamada, Butyrylcholinesterase inhibitors ameliorate cognitive dysfunction induced by amyloid-b peptide in mice, Behav. Brain Res. 225 (2011) 222–229. [12] R.M. Lane, S.G. Potkin, A. Enz, Targeting acetylcholinesterase and butyrylcholinesterase in dementia, Int. J. Neuropsychopharmacol. 9 (2006) 101–124. [13] E.V. Radchenko, G.F. Makhaeva, V.V. Malygin, V.B. Sokolov, V.A. Palyulin, N.S. Zefirov, Modeling of the relationships between the structure of Ophosphorylated oximes and their anticholinesterase activity and selectivity using molecular field topology analysis (MFTA), Dokl. Biochem. Biophys. 418 (2008) 47–51.

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012

G.F. Makhaeva et al. / Chemico-Biological Interactions xxx (2012) xxx–xxx [14] E.V. Radchenko, G.F. Makhaeva, V.B. Sokolov, V.A. Palyulin, N.S. Zefirov, Study of the structural determinants of acute and delayed neurotoxicity of Ophosphorylated oximes by molecular field topology analysis (MFTA), Dokl. Biochem. Biophys. 429 (2009) 309–314. [15] M. Lotti, M.K. Johnson, Neurotoxicity of organophosphorus pesticides: predictions can be based on in vitro studies with hen and human enzymes, Arch. Toxicol. 41 (1978) 215–221. [16] V.V. Malygin, V.B. Sokolov, R.J. Richardson, G.F. Makhaeva, Quantitative structure–activity relationships predict the delayed neurotoxicity potential of a series of O-alkyl-O-methylchloroformino phenylphosphonates, J. Toxicol. Environ. Health, Part A 66 (2003) 611–625. [17] M.R. Redinbo, P.M. Potter, Mammalian carboxylesterases: from drug targets to protein therapeutics, Drug Discovery Today 10 (2005) 313–325. [18] M.J. Hatfield, P.M. Potter, Carboxylesterase inhibitors, Expert Opin. Ther. Pat. 21 (2011) 1159–1171. [19] G.F. Makhaeva, V.V. Malygin, A stable preparation of hen brain neuropathy target esterase for rapid biochemical assessment of neurotoxic potential of organophosphates, Chem. Biol. Interact. 119–120 (1999) 551–557. [20] J. Massoulié, N. Perrier, H. Noureddine, D. Liang, S. Bon, Old and new questions about cholinesterases, Chem. Biol. Interact. 175 (2008) 30–44. [21] P.W. Allderdice, H.A.R. Gardner, D. Galutira, O. Lockridge, B.N. LaDu, The cloned butyrylcholinesterase (BCHE) gene maps to a single chromosome site, 3q26, Genomics 11 (1991) 452–454. [22] I. Mangas, E. Vilanova, J. Estévez, NTE and non-NTE esterases in brain membrane: kinetic characterization with organophosphates, Toxicology 297 (2012) 17–25. [23] T. Satoh, M. Hosokawa, The mammalian carboxylesterases: from molecules to functions, Annu. Rev. Pharmacol. Toxicol. 38 (1998) 257–288. [24] B. Li, M. Sedlacek, I. Manoharan, R. Boopathy, E.G. Duysen, P. Masson, O. Lockridge, Butyrylcholinesterase, paraoxonase, and albumin esterase, but not carboxylesterase, are present in human plasma, Biochem. Pharmacol. 70 (2005) 1673–1684. [25] E.V. Rudakova, G.F. Makhaeva, T.G. Galenko, A.Yu. Aksinenko, V.B. Sokolov, I.V. Martynov, New selective inhibitor of mice plasma carboxylesterase, Dokl. Biochem. Biophys. (in press). [26] E.V. Rudakova, O.G. Serebryakova, N.P. Boltneva, T.G. Galenko, G.F. Makhaeva, A biochemical model in mice for assessment of neuropathic potential of organophosphorus compounds, Toxicol. Rev. (in press) (Russian). [27] G.F. Makhaeva, E.V. Rudakova, N.P. Boltneva, L.V. Sigolaeva, A.V. Eremenko, I.N. Kurochkin, R.J. Richardson, Blood esterases as a complex biomarker for exposure to organophosphorus compounds, in: C. Dishovsky, A. Pivovarov (Eds.), Counteraction to Chemical and Biological Terrorism in the East Europe Countries, Springer, 2009, pp. 177–194. [28] M.S. Gromova, N.A. Kraynova, A.A. Torgonskaya, L.V. Sigolaeva, E.V. Rudakova, G.F. Makhaeva, A.V. Eremenko, I.N. Kurochkin, Determination of blood cholinesterases using a highly sensitive amperometric biosensor based on enzyme-polyelectrolyte nanofilms, Toxicol. Rev. No. 5 (2012) 27–36. [29] G.F. Makhaeva, I.V. Filonenko, V.L. Yankovskaya, S.B. Fomicheva, V.V. Malygin, Comparative studies of O,O-dialkyl-O-chloromethylchloroformimino phosphates: interaction with neuropathy target esterase and acetylcholinesterase, Neurotoxicology 19 (1998) 623–628. [30] W.N. Aldridge, E. Reiner, Enzyme Inhibitors as Substrates, North Holland Publishing, Amsterdam, 1972. pp. 37–52. [31] G.F. Makhaeva, V.I. Fetisov, V.B. Sokolov, V.L. Yankovskaya, T.V. Goreva, V.V. Malygin, B.K. Beznosko, T.G. Galenko, A.F. Kolomiets, I.V. Martynov, Interaction of dialkyl(a-carbomethoxy-b,b,b-trifluoroethyl) phosphates with mammalian esterases, Bioorg. Khim. 13 (1987) 33–37.

7

[32] B.A. Khaskin, G.F. Makhaeva, N.A. Torgasheva, A.S. Ishmuratov, V.L. Yankovskaya, V.I. Fetisov, V.V. Malygin, I.V. Martynov, Synthesis and antiesteratic activity of O,O-dialkyl-S-carboethoxychloromethylthiol phosphates, Russ. Chem. Bull. 38 (1989) 2508–2512 (Translated from Izvestiya Akademii Nauk SSSR, Seriya Khimicheskaya, No. 12, pp. 2741– 2746, 1989). [33] B.A. Khaskin, N.A. Torgasheva, A.S. Ishmuratov, Reaction of dialkoxyphosphoryl- and thiophosphorylsulfenyl bromides with diazo compounds, J. Gen. Chem. USSR (Engl. Transl.) 61 (1991) 1090–1095 (Russian), 989–993 (Engl). [34] G.F. Makhaeva, A.Y. Aksinenko, V.B. Sokolov, O.G. Serebryakova, R.J. Richardson, Synthesis of organophosphates with fluorine-containing leaving groups as serine esterase inhibitors with potential for Alzheimer disease therapeutics, Bioorg. Med. Chem. Lett. 19 (2009) 5528–5530. [35] A.Yu. Aksinenko, V.B. Sokolov, T.V. Goreva, G.F. Makhaeva, Synthesis of Ophosphorylated 1-substituted 2,2,2-trifluoroethanols, serine hydrolase inhibitors, Russ. Chem. Bull. 59 (2010) 102–106. [36] G.F. Makhaeva, V.L. Yankovskaya, N.V. Kovaleva, V.I. Fetisov, V.V. Malygin, N.A. Torgasheva, B.A. Khaskin, O,O-Dialkyl-S-brommethylthiolphosphates – inhibitors of mammalians cholin- and carboxylesterases: structure–activity relationships, Bioorg. Chem. 25 (1999) 1–5. [37] G.F. Makhaeva, V.L. Yankovskaya, N.V. Kovaleva, V.I. Fetisov, V.V. Malygin, N.A. Torgasheva, B.A. Khaskin, Antiesterase activity and toxicity of O,O-dialkyl-Sethoxycarbonyl brommethylthiolphosphates, Bioorg. Chem. 25 (1999) 6–10. [38] C. Hansch, A. Leo, Substituent Constants for Correlation Analysis in Chemistry and Biology, Wiley, New York, 1979. [39] V.A. Palyulin, E.V. Radchenko, N.S. Zefirov, Molecular Field Topology Analysis method in QSAR studies of organic compounds, J. Chem. Inf. Comp. Sci. 40 (2000) 659–667. [40] E.V. Radchenko, V.A. Palyulin, N.S. Zefirov, Molecular Field Topology Analysis in drug design and virtual screening, in: A. Varnek, A. Tropsha (Eds.), Chemoinformatics Approaches to Virtual Screening, RSC, Cambridge, 2008, pp. 150–181. [41] E.V. Rudakova, N.P. Boltneva, G.F. Makhaeva, Comparative analysis of esterase activities of human, mouse, and rat blood, Bull. Exp. Biol. Med. 152 (2011) 73– 75. [42] P. Masson, Evolution of and perspectives on therapeutic approaches to nerve agent poisoning, Toxicol. Lett. 206 (2011) 5–13. [43] M.K. Johnson, Organophosphorus esters causing delayed neurotoxic effects: mechanism of action and structure activity studies, Arch. Toxicol. 34 (1975) 259–288. [44] G.F. Makhaeva, V.V. Malygin, I.V. Martynov, Organophosphorus pesticides induced delayed neurotoxicity (Review), Agrokhimiya No. 12 (1987) 103–123. [45] R.J. Richardson, in: P. Wexler (Ed.), Encyclopedia of Toxicology, vol. 3, 2nd ed., Elsevier, Oxford, 2005, pp. 302–306. [46] E.V. Radchenko, A.A. Melnikov, G.F. Makhaeva, V.A. Palyulin, N.S. Zefirov, Molecular design of O-phosphorylated oximes – selective butyrylcholinesterase inhibitors, Dokl. Biochem. Biophys. 443 (2012) 91–95. [47] A.A. Melnikov, V.A. Palyulin, E.V. Radchenko, N.S. Zefirov, Generation of chemical structures on the basis of QSAR Models of Molecular Field Topology Analysis, Dokl. Chem. 415 (2007) 196–199. [48] D.K. Nomura, J.E. Casida, Activity-based protein profiling of organophosphorus and thiocarbamate pesticides reveals multiple serine hydrolase targets in mouse brain, J. Agric. Food Chem. 59 (2011) 2808–2815. [49] J. Estevez, E. Vilanova, Model equations for the kinetics of covalent irreversible enzyme inhibition and spontaneous reactivation: esterases and organophosphorus compounds, Crit. Rev. Toxicol. 39 (2009) 427–448.

Please cite this article in press as: G.F. Makhaeva et al., Organophosphorus compound esterase profiles as predictors of therapeutic and toxic effects, Chemico-Biological Interactions (2012), http://dx.doi.org/10.1016/j.cbi.2012.10.012