Theoretical calculation and prediction of P-glycoprotein-interacting drugs using MolSurf parametrization and PLS statistics

Abstract

A method for the modelling and prediction of P-glycoprotein-associated ATPase activity using theoretically computed molecular descriptors and multivariate statistics has been investigated using 22 diverse drug-like compounds. The program MolSurf was used to compute theoretical molecular descriptors related to physicochemical properties such as lipophilicity, polarity, polarizability and hydrogen bonding. The multivariate partial least squares projections to latent structures (PLS) method was used to delineate the relationship between the P-glycoprotein-associated ATPase activity and the theoretically computed molecular descriptors. The PLS analysis of the entire data set, with the exclusion of tamoxifen, resulted in one significant PLS component according to cross-validation with R²=0.718, Q²=0.695, S.D.=0.475, F=48.37, RMSE^tr=0.452, p<0.001. Properties associated with the size of the molecular surface, polarizability and hydrogen bonding had the largest impact on the P-glycoprotein-associated ATPase activity. All these properties should be high to promote high ATPase activity.

Introduction

The endothelial cells, which serve as a physical lining of, for example, the intestine, the lungs and the renal tubule, constitute a complex physical and biochemical interface in the body. The endothelial cell membrane is essentially impermeable to ions and nutrients (e.g., calcium, amino acids and sugars), but permeable to lipophilic xenobiotics. A multitude of membrane-bound proteins responsible for the active transport of ions and nutrients as well as the metabolism and efflux of xenobiotics have evolved in the course of evolution (Bolhuis et al., 1997). The endothelial cells, as the first line of defence, are often affected by cancer. Chemotherapy of these cancer cells is often effective at first, although frequently a few cells survive, proliferate and develop resistance to a broad range of cytotoxic drugs. Studies have shown that these cancer cells over-express membrane-bound proteins that ‘pump’ drugs out of the cells. Two proteins, P-glycoprotein and multidrug resistance-associated protein (MRP), have been linked to multidrug resistance in cancer cells, and it has been suggested that these efflux systems are responsible for many clinical failures of cancer chemotherapy (Gottesman and Pastan, 1988, Pastan et al., 1991, Stein, 1997). Both P-glycoprotein and MRP are members of a phylogenetically conserved superfamily of ATP-dependent transporters that have many homologues in microorganisms (Bolhuis et al., 1997). P-glycoprotein (170 kDa) is composed of two homologous parts, each containing a transmembrane region involved in the efflux of xenobiotics and a cytosolic domain responsible for ATP binding and hydrolysis, separated by a phosphorylatable linker region (Gottesman and Pastan, 1988). ATP hydrolysis is required for transport, and a compound that inhibits ATPase is therefore unlikely to be transported by P-glycoprotein (Azzaria et al., 1989).

P-glycoprotein is also expressed in many organs (e.g., the kidney, lung, liver and spleen) and in hormone-producing or responsive tissues (e.g., the adrenal cortex, testis and placenta). However, its physiological role is as yet largely unknown. Physiological functions such as the transport of steroid hormones, bile acids and endogenous peptides have been proposed (Bolhuis et al., 1997). P-glycoprotein has recently been found to be of importance for the pharmacokinetics and tissue distribution of therapeutic drugs in the body.

It is intrinsically involved in the function of the blood–brain barrier, which prevents lipophilic xenobiotics and many therapeutic drugs from entering the central nervous system (Tsuji and Tamai, 1977, Schinkel et al., 1994, Schinkel et al., 1995, Schinkel et al., 1996, Jolliet-Riant and Tillement, 1999). Incomplete and/or slow intestinal drug absorption has in several cases also been explained by P-glycoprotein efflux (Terao et al., 1996, Sparrebom et al., 1997, Kim et al., 1998).

There are several in vitro methods available for the screening of drug compounds with regard to possible stimulation or inhibition of P-glycoprotein activity. A common method is to incubate the compound with cells that over-express P-glycoprotein and to measure the uptake in these cells after a defined interval. The same methodology can be used to study the inhibition of P-glycoprotein-mediated transport by measuring the uptake of known P-glycoprotein substrates (e.g., verapamil and doxorubicin) with and without the compound in the media (Bain et al., 1997). Caco-2 cell monolayers are frequently used in the pharmaceutical industry to assess membrane permeability and interactions with efflux systems (e.g., P-glycoprotein). Permeability is usually studied in both directions across the monolayers. Significantly higher permeability in the basolateral-to-apical direction compared to the apical-to-basolateral direction indicates that the compound is a substrate for P-glycoprotein or similar transporters (Karlsson et al., 1993). Another approach is to study stimulation/inhibition of the P-glycoprotein ATPase activity in membranes obtained from cells that over-express P-glycoprotein (Scarborough, 1995, Litman et al., 1997). However, from a drug development perspective, and especially in combinatorial chemistry settings, these in vitro methods are too costly and time-consuming and also require the synthesis of test compounds. Hence, there is a considerable demand for rapid and efficient computational methods to assess biopharmaceutical properties such as permeability and P-glycoprotein interactions. Such methods could be used for the refinement of drug templates and as an aid in the selection of compounds with promising properties for more elaborate in vitro and in vivo testing. The computational/statistical methods will also give valuable feedback and insight into the structural features that are of importance for the biopharmaceutical property that is being studied. We have recently reported a new method for modelling the biopharmaceutical properties of drug substances by means of MolSurf technology in conjunction with multivariate statistics (Norinder et al., 1997, Norinder et al., 1998, Norinder et al., 1999). MolSurf calculates molecular descriptors related to physicochemical properties such as lipophilicity, polarity, polarizability and hydrogen bonding (Sjöberg, 1997). In the present paper we report the use of MolSurf and PLS statistics for modelling the structure–activity relationship for the ATPase activity of structurally diverse P-glycoprotein substrates.