Conceived and designed the experiments: FK GFE PC. Performed the experiments: FK. Analyzed the data: FK GFE PC. Contributed reagents/materials/analysis tools: GFE. Wrote the paper: FK GFE. Commented on and revised the manuscript: FK PC GFE. Supervised experiments: GFE.
The authors have declared that no competing interests exist.
Overexpression of the xenotoxin transporter P-glycoprotein (P-gp) represents one major reason for the development of multidrug resistance (MDR), leading to the failure of antibiotic and cancer therapies. Inhibitors of P-gp have thus been advocated as promising candidates for overcoming the problem of MDR. However, due to lack of a high-resolution structure the concrete mode of interaction of both substrates and inhibitors is still not known. Therefore, structure-based design studies have to rely on protein homology models. In order to identify binding hypotheses for propafenone-type P-gp inhibitors, five different propafenone derivatives with known structure-activity relationship (SAR) pattern were docked into homology models of the apo and the nucleotide-bound conformation of the transporter. To circumvent the uncertainty of scoring functions, we exhaustively sampled the pose space and analyzed the poses by combining information retrieved from SAR studies with common scaffold clustering. The results suggest propafenone binding at the transmembrane helices 5, 6, 7 and 8 in both models, with the amino acid residue Y307 playing a crucial role. The identified binding site in the non-energized state is overlapping with, but not identical to, known binding areas of cyclic P-gp inhibitors and verapamil. These findings support the idea of several small binding sites forming one large binding cavity. Furthermore, the binding hypotheses for both catalytic states were analyzed and showed only small differences in their protein-ligand interaction fingerprints, which indicates only small movements of the ligand during the catalytic cycle.
A major reason for the failure of cancer, antibiotic and antiviral therapies is the development of multidrug resistance (MDR). P-glycoprotein (P-gp), an ATP-dependent transport protein located in the membrane of epithelial cells of the kidney, liver, pancreas, colon and the blood-brain barrier, has been linked to the export of a broad variety of xenotoxins. Overexpression of P-gp leads to extrusion of therapeutic drugs and therefore triggers MDR. Thus, identification of potential P-gp inhibitors represents a promising concept for treatment of multiresistant tumours. However, due to lack of high resolution structural information and the polyspecific ligand recognition pattern only very limited information is available on the molecular basis of ligand/transporter interaction. Within this study we characterized the propafenone binding site of P-gp by docking a set of derivatives with known SAR into homology models of P-gp which represent both the apo and the nucleotide-bound state. Poses retrieved are in accordance with results from previous photoaffinity labeling studies and thus pave the way for structure-based
The development of multidrug resistance (MDR) is one major impediment in cancer and antibiotic therapies
The present study aimed at elucidating the binding mode of propafenone type inhibitors of P-gp using a combined homology modeling/docking approach. Propafenones show a clear structure-activity relationship (SAR) pattern
The common scaffold represents the largest common substructure and was used for root mean square deviation (RMSD) clustering.
There are numerous studies showing that there is a basic underlying correlation between P-gp inhibitory activity and lipophilicity of the compounds. This accounts for several compound classes and has also been shown for propafenone analogues.
However, propafenones which bear a 4-hydroxy-4-phenylpiperidine moiety are generally by a factor of 10 more active than equi-lipophilic derivatives without the hydroxy-group in 4-position of the piperidine moiety (
In March 2009 Aller et al. published the crystal structure of mouse P-gp in the absence of a ligand (PDB ID: 3G5U)
As the structural difference between the apo protein and the co-crystallized structures was surprisingly low (0.61 Å of Ca atoms) the higher resolved 3G5U structure was utilized as homology modeling template (3G5U_Pgp). With the modeling program MODELLER 100 different homology models were created and refined. All models were assessed with the geometry check tool implemented in MOE, which was used as a selection criterion for the final model. As additional measure for model quality the GA341 method was used, which relies on sequence identity, compactness and the combined statistical z-score. All models obtained the highest possible GA341 value of 1. Furthermore, the final model was analyzed with the structure assessment program PROCHECK
In order to cover different catalytic states of the protein, a second homology model was generated on basis of the bacterial transporter Sav1866 in the nucleotide-bound state (PDB code: 2HYD)
For the docking process five different propafenone derivatives were selected according to their differences in lipophilic efficiency and fit quality
With the genetic algorithm based docking program GOLD
The resulting poses in both conformations were distributed largely within the TM region of P-gp (
Yellow: common scaffold cluster, grey: residual poses.
All: 500 poses after docking, CSC: common scaffold cluster, GPV062: cluster that showed an interaction between the OH-group of GPV062 and the protein. Residues marked with an asterisk show direct interaction with docking poses.
The unprocessed complexes were energetically minimized using LigX, a minimization tool implemented in MOE for further evaluation.
The minimized poses were clustered according to the root-mean-square deviation (RMSD) of the heavy atoms of the common scaffold (
In case of 2HYD_Pgp, the RMSD clustering process resulted in 78 clusters, which were reduced to nine common scaffold clusters, containing 264 poses (
All: 500 poses after docking, CSC: common scaffold cluster. Residues marked with an asterisk show direct interaction with the docking poses.
3G5U_Pgp (non-ionized) | 3G5U_Pgp (ionized) | 2HYD_Pgp (non-ionized) | 2HYD_Pgp (ionized) | |
|
500 | 500 | 500 | 500 |
|
114 | 111 | 78 | 77 |
|
12 (184 poses) | 11 (195 poses) | 9 (264 poses) | 7 (240 poses) |
|
I, II, III | IV, V, VI | − | − |
|
I, III | IV, V, VI | − | − |
The model based on the murine 3G5U structure represents the binding competent state, whereas the model based on the nucleotide-bound 2HYD structure likely represents the off-state of P-gp ligands
In the hit-to-lead decision process as well as in lead optimization different efficiency metrics are applied to prioritise lead candidates. Briefly, in case of equi-potent compounds these parameters select for the smaller, more hydrophilic ones. As high lipophilicity correlates with promiscuity, poor solubility and poor metabolic clearance
Ligand | pIC50 | HAC | ClogP | LLE | LE | FQ |
|
6.22 | 27 | 4.38 | 1.84 | 0.23 | 0.77 |
|
6.21 | 33 | 5.15 | 1.06 | 0.19 | 0.75 |
|
7.24 | 34 | 4.15 | 3.09 | 0.21 | 0.87 |
|
6.19 | 26 | 5.54 | 0.65 | 0.24 | 0.77 |
|
5.78 | 33 | 4.94 | 0.84 | 0.18 | 0.70 |
HAC = heavy atom count, LLE = lipophilic ligand efficiency, LE = ligand efficiency, FQ = fit quality.
With 3G5U_Pgp only one quarter of all twelve common scaffold clusters showed a hydrogen bond between the hydroxyl-group of GPV062 and the protein (
CSC I (green), CSC II (yellow), CSC III (cyan). A) Top view; the three interacting amino acids are colored according to their cluster-membership. B) side view; the blue surface indicates residues that are involved in propafenone binding, determined by photoaffinity labeling
The positions of CSCs I and III are very similar, since both are forming a hydrogen bond with Y310 and a π/π-interaction with F336. In CSC II, on the contrary, a hydrogen bond interaction with A761 was observed.
For further evaluation of the poses a pharmacophore search was performed, utilizing a model published by Langer et al. that based on a set of propafenone type P-gp inhibitors
Evaluation of the docking results with 2HYD_Pgp could not be based on ligand affinity data, since this structure represents the nucleotide-bound off-state and therefore is considered as the low-affinity state for substrates. This rules out prioritization on basis of SAR-information. All common scaffold clusters of 2HYD_Pgp are in close vicinity of the 3G5U_Pgp GPV062-OH poses (
Magenta: GPV062-OH clusters of docking into 3G5U_Pgp (high affinity), green: common scaffold clusters of docking into 2HYD_Pgp (low affinity). A) overview, B) close-up view.
The homology models generated in this study resemble two different states of P-gp: the open-inward or apo state and the open-outward or nucleotide-bound state. Since the publication of mouse P-gp in the absence (PDB ID: 3G5U) or in complex with ligands (PDB Ids: 3G60, 3G61) only a few homology models of the human homologue were published on the basis of these structures. Pajeva et al. presented two homology models that were based on the structure of 3G61, chain A, which is complexed with QZ59-SSS
The open-outward model relied on the structure of the bacterial homologue Sav1866 (PDB ID: 2HYD), which possesses the same domain architecture as P-gp
Although ligand docking is a commonly used tool for the identification of ligand-protein interactions, in case of P-gp it bears a lot of challenges: (i) P-gp possesses a large binding cavity that consists of several binding sites, (ii) is highly flexible, and (iii) is probably able to harbor more than one ligand simultaneously
In docking experiments, the definition of the binding site is a key parameter of the docking protocol. As only little information is available about binding of propafenones into P-gp, the whole TM region was selected as a potential interaction region. In order to avoid any bias introduced by scoring functions, a large amount of docking poses was generated. While placement algorithms of docking programs are most of the time able to find the native pose of a ligand in the binding pocket, the correct estimation of the binding energy leading to a correct ranking of the poses is still unsatisfying. To overcome this uncertainty of scoring functions, we recently implemented experimental data guided docking/scoring. In this approach prioritization of docking poses is performed on basis of mutagenesis data, biochemical data, and/or information from ligand based studies
The interaction of propafenones with P-gp follows a clear structure-activity relationship pattern (for reviews see
Although docking experiments have their limitations depending on the validity of the target structure, the results of docking into 3G5U_Pgp are very consistent. As shown in
By rotating the residue Y307 (grey:original, black: rotated) a new hydrogen bond between Y307 and the carbonyl group of the ligand was formed.
CSC II forms a weak H-bond between the hydroxyl-group of GPV062 and the backbone of A761. With respect to the ligand interaction tool in MOE the strength of this bond is only 1/10 compared to that in CSCs I and III. Applying the rotamer explorer results in either formation of a stronger hydrogen bond with the OH-group of GPV062 or formation of a new interaction with the carbonyl group (with these interactions not being coexistent). Finally, with respect to residues photoaffinity labelled by benzophenone analogous propafenones, CSCs I and III show a better match (
In consideration of these findings the pose of CSC I was preferred over the other two clusters.
It is also known that binding of propafenones to P-gp meets steric constraints in the vicinity of the nitrogen atom, because diphenyl moieties in this position lead to a log order decrease in activity
Docking into 3G5U_Pgp with ionized ligands resulted in three different CSCs that show an interaction between the OH_group of GPV062 and the protein. While one is located very central in the pore (CSC IV) forming an H-bond between GP062-OH and A727, the other two (CSC V and VI) exactly match CSC I of the docking with neutral ligands. For the latter an H-bond between the hydroxyl-group and Y310 could be observed.
As can be seen in
In contrast, group 1 and group 2 are in an up-side-down orientation when compared to each other. In this case the carbonyl group is located near Y310 and thus closer to the extracellular portion of the protein. The nitrogen atom, as well as the hydroxyl group, is oriented towards Y307 and N721, which was also observed for CSC II of the 3G5U_Pgp docking run. Therefore, group 2, comprising clusters e, f and g, corresponds to the nucleotide-bound conformations of CSC II of the apo-conformation.
CSCs h and i cannot be clearly assigned to one of these groups and have to be regarded separately. The nitrogen atom of CSC h shows a similar location as the N of group 2, however, due to a shift of the central phenyl ring downwards, H-bond interactions between the carbonyl oxygen and Y307 and the OH-group and N721 can be formed simultaneously.
CSC i shares its carbonyl group orientation with group 2, but the central phenyl ring lies in a perpendicular direction, which results in interactions between the ligand nitrogen and hydroxyl group with Q725.
Considering the docking run to 2HYD_Pgp with ionized ligands, group 1 could be clearly reproduced. Three out of seven CSCs form those characteristic H-bond interactions between the carbonyl oxygen and Y307 and the hydroxyl group and Y310. In contrast to the unprotonated ligands, the nitrogen atom and Y310 form a pi/cation interaction and occur at higher frequency. Overall the clusters belonging to group 1 show high homogeneity and strong interactions. In contrast to this the poses of each of the four other clusters share no consistent pattern and therefore the common binding was only reflected in geometrically similar positioning.
Interestingly, although the experimental data suggest two symmetrical binding sites, no common scaffold cluster and hardly any poses could be found at the second photoaffinity labeled site at the 2/11 interface. One possible explanation might be the asymmetry of the template crystal structure 3G5U. The region consisting of TM helices 4, 5, 7, 8, 9 and 12 in case of 3G5U_P-gp, and TM helices 3, 4, 5, 6, 7 and 8 in case of 2HYD_P-gp, in both cases showed larger sites when using the SiteFinder tool in MOE than their counterparts around the 2/11 interface. This demonstrates the limitations of docking experiments relying on one crystal structure that represents only a snapshot of a flexible protein. Thus, to rule out the possibility that every docked ligand will end up at the 5/8 interface just because of this asymmetry, a docking run with rhodamine 123 was conducted. In this case 21 of 39 clusters were found in vicinity of residues I340, L975 and V981, which are located on TM helices 6 and 12 and known to be involved in rhodamine binding
In order to gain first insights into the potential ligand translocation pathways, the compounds were docked in two different catalytic states of P-gp. Interestingly, the docking results show similar interaction patterns. In both models, ligand poses are found in close vicinity (4,5 Å) of residues Y307 and Y310 of TM helix 5, F343 of TM helix 6 and L724 of TM helix 7, which suggests involvement of both TM domains in drug binding. This is in accordance with Loo et al., who showed that both TM domains are essential for drug translocation
In
The spheres represent Cα-atoms of interacting residues of 3G5U_Pgp (panels A, B) and 2HYD_Pgp (panels C, D). Blue spheres: 2HYD_Pgp, green: 2HYD_Pgp and 3G5U_Pgp, yellow: 3G5U_Pgp. A) and b) 3G5U_Pgp in front and top view; c) and d) 2HYD_Pgp in front and top and view.
In
Furthermore, the diagram is consistent with the notion that P-gp possesses a large binding cavity, which harbors different partially overlapping drug binding sites for different ligands
Ligand docking into polyspecific antitargets such as the hERG potassium channel and the drug transporter P-glycoprotein requires thorough validation of the poses obtained. In this paper we describe the application of an SAR-guided docking protocol, which for the first time retrieves a binding hypothesis for propafenone-type inhibitors of P-gp. Although performing docking studies with homology models always bears a lot of risks the results are in agreement with experimental studies, which strengthens the applicability of the complex docking protocol we used for this study. This could pave the way for structure-based ligand design approaches.
Two homology models based on the bacterial homologue Sav1866 (PDB ID: 2HYD, resolution: 3.0 Å
For the docking study five propafenone derivatives were selected on basis of known SAR and differences in LLE and FQ. LLE was calculated by subtracting ClogP from experimentally determined IC50 values and FQ was calculated as outlined in
Minimization and protonation of the ligands was performed with MOE.
For the correct determination of ASN/GLN/HIS flips the web application MolProbity was utilized
On basis of the common scaffold an RMSD matrix of all five ligands was generated and used for clustering. The dissimilarity matrix was clustered with the program R
In case of 3G5U_Pgp those clusters were selected for final assessment that were able to form a hydrogen bond between the OH-group of GPV062 and the protein, detected by the ligand interaction tool of MOE.
(TIF)
(TIFF)
(TIFF)
(TIFF)
(PDF)
(PDF)