Estimation of plasma protein binding of selected antipsychotics using computed molecular properties

The plasma protein binding (PPB) data of twelve antipsychotics (aripiprazole, clozapine, olanzapine, quetiapine, risperidone, sertindole, ziprasidone, chlorpromazine, flupentixol, fluphenazine, haloperidol, zuclopenthixol) were estimated using computed molecular descriptors, which included the electronic descriptor – polar surface area (PSA), the constitutional parameter – molecular weight (Mw), the geometric descriptor – volume value (Vol), the lipophilicity descriptor (logP) and aqueous solubility data (logS), and the acidity descriptor (pKa). The relationships between computed molecular properties of the selected antipsychotics and their PPB data were investigated by simple linear regression analysis. Low correlations were obtained between the PPB data of the antipsychotics and PSA, Mw, Vol, pKa, logS (R <0.30) values, while relatively higher correlations (0.35<R2<0.70) were obtained for the majority of logP values. Multiple linear regression (MLR) analysis was applied to access reliable correlations of the PPB data of the antipsychotics and the computed molecular descriptors. Relationships with acceptable probability values (P<0.05) were established for five lipophilicity descriptors (logP values) with application of the acidity descriptor (pKa) as independent variables: AlogP (R2=0.705), XlogP3 (R2=0.679), ClogP (R2=0.590), XlogP2 (R2=0.567), as well as for the experimental lipophilicity parameter, logPexp (R2=0.635). The best correlations obtained in MLR using AlogP and pKa as independent variables were checked using three additional antipsychotics: loxapine, sulpiride and amisulpride, with the PPB values of 97%, “less than” 40% and 17%, respectively. Their predicted PPB values were relatively close to the literature data. The proposed technique confirmed that lipophilicity, together with acidity significantly influences the PPB of antipsychotics. The described procedure can be regarded as an additional in vitro approach to the modeling of the investigated group of drugs.


INTRODUCTION
Psychotic illnesses can be categorized into several mental disorders such as psychoses, neuroses and mood disorders.Antipsychotic drugs, which are historically known as antischizophrenic or neuroleptic drugs, are traditionally used in schizophrenia treatment [1][2][3][4][5].Today, there are many antipsychotic drugs and new medical entities are continuously introduced into clinical practice.They can be classified into two main groups: the first group contains originally developed drugs.These are antipsychotics of the first generation.This group is known as typical antipsychotics [1][2][3][4][5].The other group represents newly developed antipsychotics, which are are known as atypical or antipsychotics of the second generation [1][2][3][4][5].
Considering their mechanism of action, antipsychotic drugs are mostly dopamine receptor antagonists.However, they can affect other targets, including cholinergic, α adrenergic, histamine or serotonin receptors, which can increase their medical efficacy.With the aim of improving the quality of life of millions of patients, changes in the modes of application as well as the introduction of newly synthesized antipsychotic drugs has significantly increased in recent years [1][2][3][4][5].
The medical success or failure of drugs, their therapeutic effect, as well as side effects, are influenced by their pharmacokinetic properties, the absorption, distribution, metabolism, route of elimination (ADME criteria).Furthermore, a drug's in vivo efficiency is significantly influenced by its plasma protein binding (PPB).Once we understand these pharmacokinetic processes and include the obtained knowledge in the design and synthesis of new drugs, we will be able to significantly increase the drugs' therapeutic success and reduce their unwanted effects [6,7].
The physicochemical properties of molecules exert a considerable influence on the ADME properties of drugs.The molecular weight and volume, lipophilicity as well as solubility, followed by polar surface area (PSA) and acidity, significantly affect the drugs' absorption, distribution and penetration into tissues, PPB and route of elimination [8][9][10][11].If more lipophilic molecules are to be compared with less lipophilic ones with similar properties, they will mostly show higher degrees of absorption and PPB, as well as better penetration into tissues and distribution.On the other hand, less lipophilic drugs are mostly eliminated in the urine, while highly lipophilic ones usually exhibit high degrees of fecal elimination.The lipophilicity effects agree with Lipinski's "rule of 5" [12].This rule predicts that low absorption or permeation of a drug is more likely when the calculated lipophilicity descriptor is found to be greater than 5, and when the molecular weight is greater than 500, as well as when there are more than 5 hydrogen-bond donors and 10 hydrogen-bond acceptors in a drug molecule [12].
Many authors have studied this group of drugs.From the early years of their discovery and development to the present day, the design and synthesis, pharmacokinetics, pharmacodynamics and efficacy of antipsychotics have been examined [4,5,[13][14][15].
In our previous research, we studied the relationships between PPB data (also including absorption and elimination) of selected antihypertensive drugs and their computed molecular descriptors, and established suitable models [16][17][18][19][20][21].The aim of the present study was to evaluate the relationships between the PPB data of twelve selected antipsychotics and their computed molecular properties.By application of MLR, molecular descriptors which are most appro-priate for estimating the antipsychotics' PPB were identified, and in the final stage of study, the best established model was checked on three additional drugs, loxapine, sulpiride and amisulpride.

Calculation of the molecular descriptors and statistical analysis
Calculation of antipsychotics' molecular descriptors based on their molecular structures was performed using several software packages.The descriptors PSA, Mw and Vol were calculated with Molinspiration Depiction Software (www.molinspiration.com).The lipophilicity descriptors, seven different logP values (AlogPs, AClogP, milogP, AlogP, MlogP, XLOGP2, XLOGP3), and their aqueous solubility data (logS), were calculated using the software package Virtual Computational Chemistry Laboratory (www.vcclab.org).The calculation of another lipophilicity parameter, ClogP values, was performed with the Chem-Draw ultra 12.0 software package.DrugBank (www.drugbank.ca)was used for calculation of the acidity descriptors (pK a values).The PPB data, as well as values of the experimental lipophilicity parameters (log-Pexp) of the investigated drugs, were obtained using the DrugBank (www.drugbank.ca).Microsoft Excel 2003 was used for statistical analysis.

RESULTS AND DISCUSSION
For all investigated antipsychotics (Tables 1 and 2), different molecular descriptors (Table 3) were obtained.According to the available data, the selected drugs mostly have high and relatively similar values of PPB, ranging from 77% for risperidone to 99% for sertindole, ziprasidone, fluphenazine, zuclopenthixol, 100% for aripiprazole (Table 4) and 97%, <40% and 17%, for loxapine, sulpiride and amisulpride, respectively.In the preliminary investigation, the relationships between the drug PPB and all calculated molecular descriptors were investigated using simple linear regression.Low correlations with R 2 <0.30 were obtained between the PPB values and the values of PSA, Mw, Vol, pKa and logS, while relatively higher correlations (0.35<R 2 <0.70) were obtained for the majority of logP values.
In the next stage of the study, relationships between the PPB and two different molecular descriptors were investigated by MLR.When the experimental lipophilicity parameters and additional molecular descriptors, Mw, Vol and pKa as independent variables, were used for PPB estimation, a relationship with an acceptable probability value of P<0.05 and R 2 =0.635 was established only for logPexp values with the acidity descriptors (pKa) as the second variable (Eq.1).

Eq.2:
PPB pred (%)=9.904   The presented correlations can be considered as good [22].The values of the antipsychotics' PPB predicted using the presented equations are shown in Table 4.
The ADME properties [23] and their PPB influence the in vivo efficacy of drugs.The main plasmabinding proteins are albumin, α 1 -acid glycoprotein and lipoproteins.Drug molecules in vivo may be bound to proteins and lipids in the plasma, to proteins and lipids in tissues, or they can be free and diffuse in the aqueous environment of the blood and tissues [24][25][26].The degree of drug affinity for plasma proteins considerably influences their distribution in target tissue, effectiveness, duration of action, elimination, and their therapeutic and side effects.Therefore, the estimation of drug PPB is of great importance for comprehending their pharmacokinetics and pharmacodynamics [24][25][26].The collected descriptors play important roles in drug absorption, distribution, metabolism, elimination and PPB, with lipophilicity as one of the most important molecular properties that is responsible for a drug's increased absorption, penetration into tissues, higher degree of distribution and higher degree of PPB [9][10][11][12].Several lipophilicity descriptors (AlogPs, AClogP, milogP, AlogP, MlogP, XLOGP2, XLOGP3, ClogP and logPexp) were obtained for the investigated group of drugs using several software packages that include different calculation methods.The differences between these methods resulted in distinctions between absolute logP values [27].
Regarding the importance of physicochemical properties, the relationships between the PPB and all calculated molecular descriptors were investigated by simple linear regression.Low correlations (R 2 <0.30) were obtained between the PPB data and PSA, Mw, Vol, pKa and logS data, while relatively higher correlations were obtained for the logP values.The best correlations were obtained for the following parameters: AlogPs (R 2 =0.69),AlogP (R 2 =0.55), milogP (R 2 =0.49),XlogP3 (R 2 =0.47) and ClogP (R 2 =0.42).For XlogP2 and logPexp values, the coefficients R 2 were 0.36 and 0.35, respectively, while AClogP and MlogP provided correlations with R 2 <0.20.
Using MLR, all collected lipophilicity descriptors were tested as the first independent variable.Mw, Vol and pKa were chosen as the second independent variable values, while the values of PSA and logS could not be used since their relationships with logP provided correlations with R 2 >0.30.It was observed that two lipophilicity parameters, MlogP and AClogP, were exceptions since they could not be used with pKa as the second independent variable since their correlation with R 2 was about 0.50.Moreover, these two lipophilicity descriptors with additional molecular descriptors, Mw and Vol, provided low correlations with R 2 <0.40.The numbers denote the investigated drugs as indicated in Table 1.**NA (logPexp was not available).
The best established correlations were obtained with Eq.1; Eq.2 and Eq.3 by MLR analysis with log-Pexp, AlogP or XlogP3 and pKa as independent variables, and are presented in Figs. 1 and 2. The relationships between the degree of PPB obtained with the software package Drug Bank and those predicted using logPexp and pKa are presented in Fig. 1.Since for sertindole and zuclopenthixol the values of log-Pexp were not available [24], their PPB values could not be predicted using Eq.1 and consequently they are not presented on Fig 1 .The relationships between PPB obtained with the software package [24] and those predicted using AlogP or XlogP3 and pKa as independent variables are presented in Fig 2 .The best established correlation (Eq.2) obtained using AlogP and pKa as independent variables was checked using three additional antipsychotics: the typical antipsychotic loxapine and two atypical antipsychotics, sulpiride and amisulpride.Their PPB values in the literature are 97%, <40% and 17%, respectively.These values were out of the 77%-100% range where the values for PPB modelling belong.Their pKa values were 7.18, 9.12 and 9.37, respectively [24].The values of their lipophilicity parameter AlogP, which provided the best model (Eq.2), were 3.96; 0.83 and 1.13, respectively [23].The correlation presented can be considered as suitable for PPB prediction of antipsychotics since the predicted PPB values were relatively close to the literature data for loxapine (93%), sulpiride (26%) and amisulpride (27%).
The correlations between antipsychotics' PPBD and their molecular descriptors, lipophilicity parameters (logPexp, AlogP, XLOGP2, XLOGP3 and ClogP) and the acidity descriptor (pKa) as independent variables determined by MLR, confirmed that the proposed in silico technique can be considered a highthroughput screening approach for estimating PPB.Relationships between PPB data of antipsychotics collected using the software package Drug Bank (Series 1) and values predicted using logPexp and pKa (Series 1).The numbers denote the investigated antipsychotics, as indicated in Table 1.

Fig. 2.
Relationships between the PPB data data of antipsychotics collected using the software package Drug Bank (Series 1) and values predicted using AlogP and pKa (Series 2), and XlogP3 and pKa (Series 3).The numbers denote the investigated antipsychotics, as indicated in Table 1.
The important role of lipophilicity and acidity may be a consequence of drug interactions during transport to their biological targets and their interactions with their receptors.The proposed methodology, which established lipophilicity and acidity as essential for the PPB of antipsychotic drugs, can be considered as an innovative approach for investigating the degree of PPB of antipsychotic drugs.

Fig. 1 .
Fig.1.Relationships between PPB data of antipsychotics collected using the software package Drug Bank (Series 1) and values predicted using logPexp and pKa (Series 1).The numbers denote the investigated antipsychotics, as indicated in Table1.

Table 1 .
The structures of the investigated antipsychotics.