Resistance to various chemotherapeutic agents can arise via a multitude of mechanisms, but one of the common mechanisms of multidrug resistance (MDR) results from the action of ATP-dependent multidrug efflux pumps. P-glycoprotein (P-gp) is the most extensively studied multidrug efflux pump. It is overexpressed in a variety of cancers and associated with the development of MDR. Despite extensive research over almost 40 years, the molecular mechanism by which P-gp transports drugs and other endogenous molecules has not been conclusively resolved.
Over the years it has been proposed that different molecules have different binding sites in P-gp, but no specific binding sites, or any targeted inhibitors that prevent the transport of specific substrates, have been identified. In order to understand the molecular basis of substrate binding, molecular dynamics (MD) simulations have been performed to estimate the potential of mean force (PMF) for the process of substrate binding to P-gp. As a proof of concept, this approach was used to identify the binding locations of morphine and nicardipine within the transmembrane (TM) pore. It has been found that morphine and nicardipine bind at different but overlapping sites within the TM pore. The results indicate that their permeation pathways through the TM pore are not well separated. This set of PMF calculations was extended to include the canonical competitive and non-competitive P-gp substrates and inhibitors, Hoechst 33342, Rhodamine 123, paclitaxel, tariquidar and verapamil. The obtained PMF profiles show that all these molecules have an energy minimum within the TM pore. The interactions with a specific set of residues can be identified for each molecule in its minimum energy location. However, none of the molecules have a distinct, well-defined binding site. Instead, the binding locations overlap with several residues interacting with multiple P-gp substrates. This suggests that the binding locations for both competitive and non-competitive substrates are not well separated and cannot be considered as unique.
Considering the lipophilic and/or amphipathic nature of many P-gp substrates, it has been suggested that P-gp effluxes drugs directly from the membrane. In addition, it has been observed that the presence of cholesterol increases the drug-stimulated ATPase activity of P-gp. However, whether this is due to the direct effect of cholesterol on the activity of P-gp, its effect on the local concentration of substrate in the membrane, or on the rate of entry of the drug into the cell, is currently unresolved. To understand better, the role of cholesterol in drug-membrane interactions, unrestrained and umbrella-sampling simulations examining the spontaneous binding and partitioning of four P-gp substrates into a POPC bilayer in the presence or absence of 10% cholesterol has been investigated. It was found that the presence of cholesterol lowers the free energy associated with entering the middle of the bilayer in a substrate-specific manner, suggesting that P-gp substrates may preferentially accumulate in cholesterol-rich regions of the membrane.
During the investigation of the substrate binding to the mouse P-gp structure (PDBid: 3G5U), two further structures of mouse P-gp (PDBid: 4KSB and PDBid: 4M1M) became available. Both were resolved to 3.8 Å. Comparison of the structures reveals that the amino acid assignment in four (TM 3, 4, 5 and 12) of the twelve helices differs across all three structures. To identify which of these three mouse P-gp crystal structures best represents the conformation of P-gp under physiological conditions, long-timescale MD simulations of membrane-embedded P-gp structures were performed. When embedded in a cholesterol enriched POPC membrane and simulated under experimental conditions, the simulations show that P-gp is a highly flexible protein that is able to sample a multiple conformations. In each simulation, the conformation of P-gp diverged from the respective crystal structures. Furthermore, in all simulations, the widely separated NBDs moved together to form a non-specific contact interface. While the precise conformation of P-gp varied between the simulations, the overall structural conformations were broadly similar. In fact, cluster analysis revealed that 3G5U, 4M1M and 4KSB P-gp all adopted very similar conformations at some point during the simulation. Although the Cα RMSD of 4M1M P-gp exhibited the least structural fluctuations and remained closer to the crystal structures than either the 3G5U or 4KSB P-gp, it is still not possible to comment on which of the three possible structures could best represent the physiological conformation of P-gp. The choice of crystallographic starting structure can potentially have a large impact on the outcomes of an MD simulation.
The simulations presented here provide an insight into the mechanism and energetics of substrate binding in P-gp. Furthermore, they demonstrate that the results obtained for substrate binding to P-gp are independent of the choice of starting structure. These results shed light on the mechanism of substrate binding and uptake, increasing our understanding of P-gp mediated multidrug resistance, which is a significant medical problem in chemotherapeutic treatment of cancer.