Optimizing Targeted Inhibitors of P-Glycoprotein Using Computational and Structure-Guided Approaches

John G Wise; Amila K Nanayakkara; Maha Aljowni; Gang Chen; Maisa C De Oliveira; Lauren Ammerman; Ketetha Olengue; Alexander R Lippert; Pia D Vogel

doi:10.1021/acs.jmedchem.9b00966

. Author manuscript; available in PMC: 2020 Dec 12.

Published in final edited form as: J Med Chem. 2019 Nov 26;62(23):10645–10663. doi: 10.1021/acs.jmedchem.9b00966

Optimizing Targeted Inhibitors of P-Glycoprotein Using Computational and Structure-Guided Approaches

John G Wise ^†,^‡,^§,^*, Amila K Nanayakkara ^†,^‡,^§,^#, Maha Aljowni ^‡,^‖,^#, Gang Chen ^†,^‡,^§,^#, Maisa C De Oliveira ^†,^‡,^§,^#, Lauren Ammerman ^†,^‡,^§,^#, Ketetha Olengue ^†,^‡,^§,^⊥, Alexander R Lippert ^‡,^‖, Pia D Vogel ^†,^‡,^§,^*

PMCID: PMC7031812 NIHMSID: NIHMS1068831 PMID: 31702922

Abstract

Overexpression of ABC transporters like P-glycoprotein (P-gp) has been correlated with resistances in cancer chemotherapy. Intensive efforts to identify P-gp inhibitors for use in combination therapy have not led to clinically approved inhibitors to date. Here, we describe computational approaches combined with structure-based design to improve the characteristics of a P-gp inhibitor previously identified by us. This hit compound represents a novel class of P-gp inhibitors that specifically targets and inhibits P-gp ATP hydrolysis while not being transported by the pump. We describe here a new program for virtual chemical synthesis and computational assessment, ChemGen, to produce hit compound variants with improved binding characteristics. The chemical syntheses of several variants, efficacy in reversing multidrug resistance in cell culture, and biochemical assessment of the inhibition mechanism are described. The usefulness of the computational predictions of binding characteristics of the inhibitor variants is discussed and compared to more traditional structure-based approaches.

Graphical Abstract

graphic file with name nihms-1068831-f0001.jpg

INTRODUCTION

Resistances of cancers to chemically unrelated anticancer drugs are frequently caused by the expression of members of the ABC transporter superfamily, including ABCB1 (P-glycoprotein or P-gp)^1–3 and/or ABCG2 (the breast cancer resistance protein or BCRP).^3,4 The phenomenon of multidrug resistance (MDR) remains a major obstacle in the treatment of both adult and pediatric cancers.^5–7 Despite nearly 40 years of intense research, no inhibitor of these proteins has yet been approved for clinical use.^5,8–13 A significant number of failures during the clinical trials may be attributed to the fact that many of the assessed MDR inhibitors were also good transport substrates of the pumps. This latter characteristic likely required elevated systemic inhibitor concentrations for efficacy that may have resulted in off-target toxicities. The fact that a phase III trial using the immunosuppressant, cyclosporine A, led to improved patient outcomes in poor-risk acute myeloid lymphoma patients¹⁴ suggests that these proteins are important targets for drug discovery and development, despite the limited success in finding clinically successful inhibitors of MDR pumps to date. Over the last several years, significant advances in our knowledge of the structure^15–20 and mechanism^21–25 of P-gp, BCRP, and other related pumps have emerged that may enable the design of potentent inhibitors of pumps and prove to be more successful in clinical applications.

Using the evolutionary relationship of different ABC transporters and the structural knowledge of both prokaryotic and eukaryotic ABC transporters, we created dynamic models of human P-gp²¹ and simulated a putative catalytic cycle²² that correlated well with published biochemical and biophysical studies as well as with the recently elucidated outward facing structure of the human P-gp.¹⁶ We previously used these conformationally dynamic models of human P-gp in ultrahigh throughput in silico screenings to identify and characterize inhibitors of P-glycoprotein.²⁶ One novel characteristic sought in these screens was that the potential inhibitors identified should interact strongly with the nucleotide-binding domains (NBD) of the protein but not be transport substrates of the pump.²⁶ For this reason, compounds were discarded from further evaluation if significant binding to the drug-binding sites was predicted by the in silico docking calculations.²⁶ Using this approach, we initially identified and characterized three compounds that reversed the multidrug resistance phenotype of various P-gp overexpressing cancer cell lines in both conventional and microtumor-spheroid cell cultures.^26–28 All three compounds were observed to increase cancer cell lethality to various chemotherapeutics. This increased lethality was directly correlated with increased cellular retention of the chemotherapeutics when an inhibitor was present.²⁸ Importantly, studies also showed that these three P-gp inhibitors were not significantly transported by P-gp,²⁸ supporting our predictions that the inhibitors would not bind effectively to the drug-binding domains of the pump.²⁶

One of these compounds that reversed MDR phenotypes in cancer cells was computationally predicted to be an allosteric inhibitor of P-glycoprotein (“compound 29”, 2-[(5-cyclopropyl-4H-1,2,4-triazol-3-yl)sulfanyl]-N-[2-phenyl-5-(2,4,5-trimethylphenyl)-pyrazol-3-yl, Figure 1A).²⁶ The presence of compound 29 caused increased penetration of a P-gp pump substrate into microtumors of a highly P-gp-overexpressing prostate cancer,²⁸ and coadministration of 29 with a chemotherapeutic resulted in increased cell death via apoptotic mechanisms and size reduction of tumor spheroids.²⁸ The binding pose of 29 docked at the highest affinity interaction site on P-gp as observed in the original study²⁶ is presented in Figure 1B. This site is located in the N-terminal half of the protein near the interface of the two nucleotide-binding domains and significantly outside of the ATP-binding sites. The computational prediction that 29 acted as a potential allosteric inhibitor of P-gp ATP hydrolysis as assayed using purified, recombinantly expressed P-gp was supported by the observation that 29 did not affect the binding of an ESR active analog of ATP (SL-ATP, 2′,3′-(2,2,5,5,-tetramethyl-3-pyrroline-1-oxyl-3-carboxylic acid ester ATP) to P-gp.²³ Three other compounds assessed in that same study inhibited SL-ATP binding to P-gp as well as P-gp ATP hydrolysis. In contrast to compound 29, these latter three compounds had been computationally predicted to partially overlap with the ATP-binding sites of the transporter²⁶ and were therefore anticipated to be competitive inhibitors of ATP binding. All four compounds inhibited both basal and transport-substrate stimulated ATP hydrolysis activities of purified P-gp,²⁶ suggesting direct interaction with the energy-harvesting steps of substrate pumping by P-gp.

Inspection of the computational docking of 29 at the putative allosteric site on P-gp (Figure 1C,D) indicated that considerable space was available for additional interactions between the protein and some parts of compound 29, especially around what we call the “Western” end of the inhibitor (see Figure 1A,C,D). In contrast, the “Eastern” end of 29 docked to P-gp was partly exposed to the external surface of P-gp (the phenyl group at the top of Figure 1 B), while the trimethylphenyl group seemed to be nearly optimally fit into a cavity in P-gp (Figure 1B,C,D).

A number of computational approaches exist that seek to analyze the potential chemical space of a hit compound by creating virtual libraries of variants of a given hit molecule.^29–32 We present here our initial efforts at creating variants of compound 29 with optimized binding affinity to P-glycoprotein using a novel computer-aided, structure-based approach that was applied to the “Western” end of compound 29. In this pilot study, a small number of variants of the nearly infinite chemical space around hit compound 29 were synthesized and assessed for reversal of the MDR phenotype in a multidrug-resistant prostate cancer cell line that overexpresses P-gp (DU145TXR³³). After initial evaluation of the computationally predicted inhibitor variants in P-gp inhibition in cell-based and biochemical assays, structure-based rational design and visual inspection of the putative 29 binding site on P-gp led us to synthesize and analyze a small number of 29-derivatives with different structural and physicochemical characteristics. These latter compounds were not initially computationally evaluated using the subtractive binding routines as described in ref 26 but were chosen mostly for the shape and size of the Western half of the molecule and physicochemical characteristics like polar surface areas and solubility. The work led to the discovery of several variants of the P-gp inhibitor hit compound 29 with improved efficacy in reversing MDR in P-glycoprotein-overexpressing cancer cells by inhibiting P-gp-catalyzed substrate pumping. All of the computationally evaluated variants were shown to be more efficient in increasing lethality of P-gp-overexpressing cancer cells, while none of them was observed to be transported by the pump. By contrast, only one of the traditionally rationally designed variants was not a transport substrate of the pump. This result strongly suggests that the combination of computational medicinal chemistry and computational pre-assessment of the predicted variants described here may be a promising path for drug optimization projects, especially for hard-to-target proteins where some protein interactions are not desirable.

RESULTS

Virtual Synthesis of Novel Variants of the “Western” Half of SMU29 Using the ChemGen Computational Suite.

Evaluation of the fit of compound 29 into a putative allosteric binding site on P-gp was visualized from the results of docking studies as in ref 26 and led us to hypothesize that, if variants of inhibitor 29 were made larger and more hydrophobic on one end of the molecule, they would likely fill the relatively large hydrophobic pocket in the protein where the cyclopropyl group of 29 interacts (Figure 1C,D), increasing the overall binding affinity of the variant.

In order to test this hypothesis, a central retrosynthetic disconnection of a carbon–sulfur bond in compound 29 (Figure 2A) was generalized for virtual syntheses (Figure 2B). The ChemGen program (see Experimental Section and Supporting Information, Figure S1) was used to virtually synthesize a number of variants of compound 29 that had different substituents at the “Western” thiol-derived part of the molecule. Several thousands of sulfur-containing compounds were collected from the ZINC database³⁴ by search of carbon–sulfur single bonds. Using the ChemGen precursor marking procedures (Experimental Section and Supporting Information), this set of compounds was pruned for those that could be converted to thiols. A total of 647 thiol-containing molecules remained that were then used in the final reaction step of the computational retrosynthesis shown in Figure 2B. The scaffold compound used to react with these thiols was 2-chloro-N-[1-phenyl-3-(2,4,5-trimethylphenyl)-1H-pyrazol-5-yl]acetamide, which includes the “Eastern” part of compound 29 (Figure 1A).

Virtual syntheses of the resulting 647 variants of compound 29 were performed using ChemGen applied to the single scaffold and thiol precursors as described in the Experimental Section. Final geometrical optimization of the virtually synthesized molecules was performed using the phenix.elbow³⁵ program. Some of the potential chemical spaces of these 647 ChemGen-produced variants of 29 (“group 1” compounds) are visualized in Figure S2. This figure shows the 647 virtual 29-variants aligned on the heavy atoms of the common 1-phenyl-3-(2,4,5-trimethylphenyl)-1H-pyrazol-5-yl group.

Docking to P-Glycoprotein and Chemical Synthesis of Compound 29-Variants.

The 647 group 1 molecules created by ChemGen were used in docking studies to the same structural model of P-gp that was employed in our previous work and led to the identification of compound 29 as a potential inhibitor²⁶ (see steps 8 through 10 of Figure S1, the “in silico docking routine”). Docking was performed as in ref 26 but using a target box that encompassed the putative allosteric site on P-gp (Figure 1B and Experimental Section) instead of the larger target box described in ref 26. The original unmodified compound 29 was included in these calculations so that the estimated binding affinities calculated by docking software for each of the variants could be compared to the affinities estimated for parental compound 29. Sixty-seven of these group 1 variants of compound 29 were identified in these calculations that were predicted to interact with the putative allosteric site on P-gp more strongly than parental 29. Since it was desirable to identify variants that are not pump substrates for P-gp, these 67 variants of 29 were then docked in a second step and counterscreened for low predicted interaction affinities with the drug-binding domains of P-gp (as described in ref 26). Eleven of the promising hit variants had molecular weights less than 600 Da and ratios of the estimated K_d of compound 29/estimated K_d of the 29-variant that were greater than 10 (Table 1), suggesting that these variants might have higher affinity to P-gp than that of the original 29. One additional compound (29–551), which contains a bromo substituent and has a molecular weight of 627 Da, was later added for consideration because of its otherwise favorable predicted properties. Figure S3 shows compound 29 (left panel) and an overlay of the 12 variants shown in Table 1 (right panel) in the putative allosteric site on P-gp. It can be seen from the figure that the ChemGen synthesis and subsequent docking routines generated and identified several variant compounds that were predicted to better fill the void in the putative allosteric site on P-gp than the parental compound 29. Docking calculations suggested that these variants might show improvements in binding affinities to P-gp. In this pilot study, a small number (five) of these group 1 ChemGen-generated compounds (Table 1, 29–216, −227, −231, −541, and −551) were selected for chemical synthesis and subsequent testing for potentially improved efficacy in inhibiting P-gp. At this stage of our study, no consideration was given to “drug-likeness” of the chosen compounds except for trying to keep the relative molecular weight as low as possible. Instead, the choice for synthesis of these five particular compounds was made on the basis of ease of synthesis and commercial availability of the precursor molecules.

Table 1.

ChemGen/docking Routine-Identified Variants of P-gp Inhibitor 29^a

variant name	synthesized variant	estimated ΔG_binding (kcal mol⁻¹)	estimated K_d (nM)	ratio of K_d 29/K_d variant	molecular weight (Da)	topological polar surface area (Å²)	consensus log P
opt_0009_19		−12.4	0.8	50	599	173.8	5.1
opt_0005_53	29–541	−12.0	1.6	25	522	132.5	5.4
opt_0009_29	29–216	−11.8	2.3	17	558	89.3	6.6
opt_0010_33		−11.8	2.3	17	576	151.5	6.6
opt_0005_47		−11.8	2.3	17	591	138.6	5.5
opt_0000_30		−11.7	2.7	15	541	157	4.8
opt_0002_34		−11.7	2.7	15	579	124	6.0
opt_0005_27		−11.7	2.7	15	584	131.4	4.4
opt_0010_49	29–551	−11.6	3.2	13	627	114.2	5.9
opt_0009_33	29–231	−11.5	3.8	11	535	101.3	5.7
opt_0004_12	29–227	−11.5	3.8	11	587	92.5	6.4
opt_0006_13		−11.5	3.8	11	596	130.1	6.2
ZINC08767731	29	−10.1	40	1	459	113.8	4.6

Open in a new tab

Estimated K_d values for the ligand interactions with P-gp were calculated from the lowest estimated binding ΔG values from the AutoDock calculations. The ratio of K_d values is shown as a relative value for potentially increased affinity of the variants over the parent P-gp inhibitor 29. Molecular weights, topological polar surface areas, and consensus log P values were calculated at the SwissADME website (http://www.swissadme.ch/) as described in the Experimental Section.

Syntheses were performed using the scheme shown in Figure S4. The virtual retrosynthetic scheme shown in Figure 2A,B was slightly modified so that the conserved pyrazole “Eastern” half would be formed from a thiol precursor. The Lewis acid-catalyzed reaction of aldehyde 1 with diazoacetonitrile 2 provided the α-cyanoketone 3.³⁶ The core pyrazole motif was formed through condensation of 3 with phenylhydrazine, providing the aminopyrazole 4 in good yield.³⁷ After nucleophilic acyl substitution with chloroacetyl chloride, the thiol was formed in two steps by substitution with potassium thioacetate followed by thioester hydrolysis.³⁸ With this thiol in hand, S_N2 reactions with various alkyl chlorides provided a modular approach to diverse derivatives of compound 29. While most of the synthesized derivatives of compound 29 consisted of an amide on the “Western” half, derivative 29–216 (216) required a ketone. The target α-chloro ketone was prepared by Friedel–Crafts acylation of fluorene with chloroacteyl chloride.³⁹ The resultant chloride was then subjected to a similar sequence as described above to form the thiol after nucleophilic substitution with thioacetate. The aromatic sulfide compounds 541 and 551 were prepared by substituting the alkyl chloride 5 with the respective aromatic thiols. Details of the syntheses and product analyses are provided in the Supporting Information, “Synthetic Procedures” section.

Figure 3A shows the structures of group 1 variants 29–216 (216), 29–227 (227), 29–231 (231), 29–541 (541), and 29–551 (551) in the highest estimated affinity docking pose (shown as licorice and colored by their atoms) in the putative allosteric site of P-gp.

Figure 3B shows the chemical structures of the 29-variants below the respective docking images. The nearly identical binding of the “Eastern” portions of compound 29 and the ChemGen/docking routine-generated 29-variants is also clearly observable in Figure S3 where the right panel displays the superposition of the docking poses of all 12 of the 29-variants shown in Table 1.

Group 1 ChemGen/Docking Routine-Produced 29-Variants Resensitize a P-gp-Overexpressing Prostate Cancer Cell Line, DU145TXR, to Paclitaxel.

The mitochondrial reduction potential of cells is often used as an indicator for cell viability using MTT assays.^40–42 Using these assays, we observed that the five 29-derivatives in group 1, compounds 216, 227, 231, 541, and 551, indeed resensitized the P-gp overexpressing prostate cancer cell line, DU145TXR,³³ to the chemotherapeutic paclitaxel (Figure 4).

Analysis of the data from Figure 4 revealed that, in the presence of 3 μM 29-variant 216, the observed IC₅₀ value of paclitaxel was decreased by approximately 2.4-fold when compared to the presence of the parental compound 29 and approximately 8-fold when compared to paclitaxel alone. At higher concentrations, the efficacy of variant 216 in increasing paclitaxel toxicity decreased when compared to the parental compound 29. At the highest concentration tested (10 μM), 216 was observed to be somewhat less effective than parental compound 29 (0.4-fold compared to the 1-fold of paclitaxel + 29). Unlike 216, variants 227, 231, 541, and 551 were more effective than 29 at all concentrations tested. At 3 μM, the presence of variants 227, 231, 541, and 551 resulted in 4- to 11-fold-decreased paclitaxel IC₅₀ when compared to parental compound 29 and up to 38-fold overall sensitization to paclitaxel when compared to paclitaxel alone (compound 551). At higher concentrations (5 to 10 μM), addition of these variants resulted in increased paclitaxel toxicity and decreased paclitaxel IC₅₀ of up to 500–1000-fold at 10 μM, as compared to ~100-fold sensitization caused by the parental compound 29 at 10 μM. This data indicated that variants 227, 231, 541, and 551 were better resensitizers of the multidrug resistant cells to paclitaxel than the original compound 29 at all concentrations tested, while variant 216 appeared to be marginally better than 29 at lower concentrations. The data therefore support the hypothesis that the ChemGen-generated and docking analyses-selected group 1 variants of compound 29 had increased affinity for P-gp resulting in improved efficacy for reversing chemotherapy resistance in a P-gp overexpressing cancer cell line compared with the parental compound.

Accumulation and Cellular Retention of Calcein AM in DU145TXR Cells upon Incubation with Group 1 SMU29 Variants.

Calcein AM accumulation assays have been used by us previously to evaluate P-gp-substrate accumulation in real time in the presence or absence of P-gp inhibitors.²⁸ For these assays, P-gp-overexpressing DU145TXR cells were incubated with the respective inhibitors in the presence of the P-gp substrate, calcein AM. Inhibition of P-gp results in cellular accumulation of calcein AM and cleavage of its acetoxymethyl ester groups, generating the highly fluorescent compound, calcein. The anionic calcein is not a substrate of P-gp and remains in the cells. In these assays, the relative fluorescence of cellular calcein was measured over time and the results of these assays are shown in Figure 5 (left panel). The data indicated that when these cells were treated with any of the five group 1 29-variants, the observed cellular accumulation of fluorescent calcein was lower than that upon treatment with the parental compound 29. Only compound 551 resulted in marginally higher calcein accumulation than that of parental compound 29.

Figure 5. — Cellular accumulation of the fluorescent dye, calcein, by inclusion of group 1 variants of the P-glycoprotein inhibitor, compound 29, identified with ChemGen. The prostate cancer cell line, DU145TXR, that overexpresses P-glycoprotein was treated with calcein AM and without (left) or with (right) a 6 h preincubation with 2 μM P-gp inhibitors indicated.

To test whether the lower accumulation of calcein in the presence of the group 1 29-variants was the result of retention of the compounds in the cellular membrane due to their increased log P values relative to 29 (Table 1), similar calcein accumulation assays were performed after a 6 h preincubation with the 29-variants and parental compound 29, Figure 5 (right panel). We hypothesized that increased partitioning of variants into the hydrophobic part of the cell membrane may result in lowered cytosolic concentrations of the inhibitors and decreased binding to the putative allosteric site on P-gp, which is located in the cytoplasm, therefore potentially slowing the inhibitory effect of the compounds. The data of Figure 5, right, suggested that calcein accumulation in the presence of the variants was improved by the 6 h preincubation for all group 1 compounds compared to the “no preincubation” experiments, supporting our hypothesis that partitioning into the membrane may have been a contributing factor. Compound 541, which has the lowest log P of the group 1 variants, performed relatively similar to 29 without preincubation. All of the variants except 216, which has the highest log P, exceeded 29 in efficacy significantly upon the 6 h preincubation. Compound 216 was observed to be equivalent to 29 in efficacy even after the 6 h preincubation.

Assessing the Roles of Polarity and Size of 29-Variants in Improving Efficacy of Compound 29 Variants.

To assess the contributions of overall hydrophobicity and size of the “Western halves” of the 29-variants on efficacy in inhibiting P-gp and reversing multidrug resistance in cancer cells, five structural derivatives of 29 (“group 2 variants”, Figure 6) were synthesized, which varied in size, shape, polar surface areas, and overall hydrophobicity as judged by calculated values of molecular weight, topological polar surface areas, and log P (Table 3). The variations of structure in these molecules were again made in the “Western” half of the molecule (Figure 1) similar to group 1 variants.

Figure 6. — Rationally designed 29-variants (group 2 variants) docked at a putative allosteric inhibitory site on human P-glycoprotein. (A) The putative allosteric site on P-gp is shown in a surface representation with synthesized variants of compound 29 shown in licorice representations colored by their atoms (as indicated). Variants are shown as in Figure 1 with the surface of P-glycoprotein colored by its calculated electrostatic potential. (B) Structures of the rationally designed variant compounds.

Table 3.

29-Variants Differing in Overall Shape, Size, and Polarity that Did Not Undergo a Docking Routine^a

synthesized variant name	estimated ΔG_binding (kcal mol⁻¹)	estimated K_d (nM)	ratio of K_d 29/K_d variant	molecular weight (Da)	topological polar surface area (Å²)	consensus log P
29–238	−9.6	92	0.4	575	129.0	4.8
29–255	−11.0	9	4.4	542	142.5	5.3
29–278	−9.9	55	0.7	543	101.3	5.7
29–280	−9.6	92	0.4	575	101.3	6.0
29–286	−9.9	55	0.7	517	101.3	5.2
ZINC08767731, “29”	−10.1	40	1	459	113.8	4.6

Open in a new tab

Estimated K_d values for the ligand interactions with P-gp were calculated from the lowest estimated binding ΔG values from the AutoDock calculations. The ratio of K_d values is given as a relative value for potentially changed affinities exhibited by the respective variants over the parent P-gp inhibitor compound 29. Molecular weights, topological polar surface areas, and consensus log P values were calculated at the SwissADME website (http://www.swissadme.ch/) as described in the Experimental Section.

Group 2 variants were chosen without consideration of docking results for predicted high affinity to the putative allosteric site on P-gp; they were also not “counter-selected” against binding affinity to the drug-binding domains of P-gp using computational docking studies. Instead, the group 2 variants were rationally designed upon visual inspection of the putative binding site to provide a larger volume to fill the void visible around 29 in the putative allosteric site (Figure 1) while at the same time somewhat decreasing the hydrophobicity (compare log P and TPSA values for group 1 compounds, Table 1, with those for group 2 derivatives, Table 3). The chemical synthesis is given in the Supporting Information.

Table 3 shows that all five of these group 2 variants are somewhat larger than 29, but the calculated log P values for these group 2 compounds (Table 3) are closer to that of 29 than the log P values of the group 1 ChemGen-generated variants with the exception of 541, which has the lowest log P value of that group (Table 1). The consensus log P values^43,44 of 238 and 255 were calculated to be 4.8 and 5.3, while those of 278, 280, and 286 were calculated to be 5.7, 6.0, and 5.2, respectively. The topological polar surface areas (TPSA) of 238 and 255 were calculated to be higher than those of 278, 280, and 286. These TPSA values were also higher than those of the group 1 variants, 216, 227, and 231 (compare Tables 1 and 3). Of the five structural variants in Table 3, two had increased calculated topological polar surface areas and three had reduced calculated topological polar surface areas when compared with 29.

Docking of the group 2 variants to the putative allosteric inhibitor binding site of P-gp showed that the “Western” portions of compounds 238, 255, and 286 could penetrate deeper into the hydrophobic void than compound 29 but compounds 278 and 280 did not seem to penetrate as well as compound 29 (compare Figures 6A and 1C,D).

Effects of Group 2 Variants on the Paclitaxel Sensitivity of DU145TXR Cancer Cells.

MTT assays were again used to assess the efficacy of the new structural 29-variants on sensitizing the chemotherapy-resistant prostate cancer cell line, DU145TXR, to paclitaxel (Figure 7 and Table 4).

Figure 7. — Reversal of paclitaxel resistance by rationally designed variants of the P-glycoprotein inhibitor 29. The prostate cancer cell line, DU145TXR, that overexpresses P-glycoprotein was treated with either paclitaxel (PTX) or paclitaxel and either 3, 5, 7, or 10 μM P-gp inhibitors (upper left, upper right, lower left, and lower right panels, respectively). Metabolic activity assays using MTT were performed as described in the Experimental Section. Calculated IC₅₀ values for paclitaxel in the presence of the respective compounds calculated from this data are reported in Table 4.

Table 4.

Increased Toxicity of Paclitaxel in the Presence of Rationally Designed P-gp Inhibitors^a

inhibitor concentration	resensitization to paclitaxel with the indicated treatment and fold-increased sensitivity in the presence of inhibitors
	PTX alone	PTX + 29			PTX + 238			PTX + 255			PTX + 278			PTX + 280			PTX + 286
	IC₅₀ PTX (nM)	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29
3 μM	2120	629	3	1	185	11	3.4	172	12	3.7	231	9.2	2.7	160	13	3.9	458	4.6	1.4
5 μM	2120	194	11	1	1.9	1116	102	34	372	5.7	109	19	1.8	56	38	3.5	379	5.6	0.5
7 μM	2120	59	36	1	0.64	3312	92	2.9	731	20	21	101	2.8	16	133	3.7	192	11	0.3
10 μM	2120	21	101	1	0.07	30,286	300	0.78	2118	27	13	163	1.6	11	193	1.9	131	16	0.2

Open in a new tab

Cytotoxicity of the chemotherapeutic paclitaxel (PTX) in P-gp-overexpressing prostate cancer cells, DU145TXR, was determined in the absence and presence of the structural 29-variants, 238, 255, 276, 280, and 286. For each experimental compound, (a) IC 50 values of PTX alone or in the presence of inhibitors (in nM), (b) fold improvement of PTX sensitivity in the presence of an inhibitor, and (c) fold improvement of PTX sensitivity by variants compared to parental compound 29 are given.

The data suggest that the presence of all of the variants except for 286 increased the paclitaxel toxicities to these P-gp-overexpressing cells (Table 4). Compounds 238 and 255 increased paclitaxel toxicities to the greatest extents: At a 5 μM concentration, compound 238 decreased the paclitaxel IC₅₀ by about a thousandfold, similar to compound 255 at 7 μM. Comparison with parental compound 29 showed that 238 improved sensitization of DU145TXR to paclitaxel by 3.4-fold at 3 μM and by approximately 100-fold at 5 and 7 μM. At 10 μM, compound 238 resulted in a 300-fold decreased paclitaxel IC₅₀ of DU145TXR compared to parental compound 29 at the same concentration. Compound 255 also showed substantial improvement in sensitization of DU145TXR to paclitaxel that was comparable to 238 at 3 μM, but the effect was not as pronounced at higher concentrations. The other three compounds, 278, 280, and 286, were similar to or less effective than the group 1 variants, 216, 227, and 231.

In addition, compound 280 seemed to exhibit some toxicity to the multidrug-resistant cancer cells as judged by the lowered cell viability at very low concentrations of paclitaxel in the presence of 280 (Figure 7).

Accumulation and Cellular Retention of Calcein AM in DU145TXR Cells in the Presence of 29-Variants from Group 2.

Figure 8, left, shows that without preincubation, the presence of 5 μM compounds 238 and 255 resulted in considerably increased calcein accumulation in DU145TXR cells when compared to 29, consistent with a log P close to that of the parental compound. The effect of compound 286 was comparable to 29, while 278 and 280 caused less calcein accumulation than that of the parental compound. After a 6 h preincubation (Figure 8, right panel), the effectiveness of inhibiting P-gp-catalyzed transport of calcein AM remained strongest for the more polar 29-variants 238 and 255, while that of the more hydrophobic variants 278, 280, and 286 was comparable to 29. Even though variant 286 has a consensus log P value similar to the log P calculated for 238 and 255, it seems to group better with regard to efficacy in blocking calcein AM export with compounds that have comparable or lower calculated polar surface areas, that is, 278 and 280, see Table 3, and 216, 227, and 231 from group 1, Table 1. Group 1 compounds 216, 227, and 231, in addition, have increased consensus log P values, which might explain their somewhat reduced efficacy in blocking P-gp-catalyzed calcein AM export (Table 1 and Figure 5).

Evaluation of the Mode of Inhibition of P-Glyco-protein by Group 1 and Group 2 Variants of Compound 29.

To assess the mode of inhibition of P-gp by the novel variants of P-gp inhibitor 29, ATP hydrolysis by P-gp was evaluated in the presence or absence of the variants. Both “basal” ATP hydrolysis (assayed in the absence of an added transport substrate) and “stimulated” ATP hydrolysis (assayed in the presence of the P-gp transport substrate verapamil) were assessed as described in ref 26. We used murine P-gp (MDR3) expressed in Pichia pastoris (P. pastoris) that had all naturally occurring cysteines replaced with alanine.^45,46 It is widely assumed that the ATP hydrolytic rate of P-gp is stimulated in the presence of transport substrates when compared to ATP hydrolysis in the absence of transport substrates. Assays comparing these rates were therefore thought to be useful in identifying inhibitors of P-gp-catalyzed ATP hydrolysis and may also be a predictor for whether a compound might be a transport substrate if basal ATPase rates are stimulated by the addition of a compound.

Effects of Compound 29 Variants on Verapamil-Stimulated ATP Hydrolysis by P-gp.

The effects of compound 29-variants on P-gp ATP hydrolysis rates assayed in the presence of verapamil (a good substrate for transport by P-gp) are presented in Table 5 (“stimulated ATPase”). The respective percent ATPase activity is shown, normalized to ATP hydrolysis in the presence of a DMSO carrier and with no experimental compound added. Interestingly, the group 1 compounds differed in their effects on “stimulated” ATPase: 216 did not affect stimulated ATP hydrolysis activities, while compounds 227, 541, and 551 inhibited activity similar to parental compound 29. Compound 231 slightly stimulated ATP hydrolysis rates in the presence of verapamil. For group 2 compounds, 238 stimulated the “stimulated” ATPase rates by approximately 2-fold, while variant 280 showed only a slight stimulation of hydrolysis rates, and compounds 255 and 286 had no significant effect. Only compound 278 of the group 2 variants inhibited verapamil-stimulated ATP hydrolysis by P-gp similar to the parental compound 29.

Table 5.

Mode of Inhibition of Cysteine-less Mouse MDR3 P-Glycoprotein by Group 1 and Group 2 Compound 29 Variants^a

compound	stimulated ATPase (% DMSO, significance)		effect on stimulated ATPase	basal ATPase (% DMSO, significance)		effect on basal ATPase	cellular accumulation: ratio of plus tariquidar over no tariquidar		transport substrate for P-gp	maximum ATP binding (mol of SL-ANP bound/mol of P-gp)	SL-ANP binding to P-gp, apparent K_d (μM)	effect on SL-ANP binding
DMSO	100 ± 8			100 ± 7						1.8 ± 0.1	36.5 ± 3.6
SMU29	49 ± 2	**	inhibitor	95 ± 11	NS	none	1.0	NS	no	1.9 ± 0.1^b	71.0 ± 12.0	no
Group 1: ChemGen- and docking-selected
SMU29–216	108 ± 2	NS	none	105 ± 7	NS	none	1.1	NS	no	1.7 ± 0.1	23.1 ± 3.9	no
SMU29–227	50 ± 2	**	inhibitor	88 ± 4	NS	none	1.0	NS	no	1.8 ± 0.1	36.9 ± 4.1	no
SMU29–231	141 ± 18	*	stimulator	70 ± 0	*	inhibitor	0.9	NS	no	1.6 ± 0.1	22.2 ± 3.8	marginally
SMU29–541	68 ± 7	*	inhibitor	101 ± 12	NS	none	1.0	NS	no	1.8 ± 0.1	24.1 ± 4.0	no
SMU29–551	62 ± 5	**	inhibitor	150 ± 6	**	stimulator	1.1	*	no	1.7 ± 0.1	22.8 ± 3.6	no
Group 2: rationally designed/no docking selection
SMU29–238	194 ± 22	**	stimulator	1143 ± 46	**	stimulator	1.9	***	yes	1.2 ± 0.1	20.1 ± 4.0	yes
SMU29–255	123 ± 15	NS	none	355 ± 7	***	stimulator	1.2	*	yes	1.9 ± 0.1	25.6 ± 4.7	no
SMU29–278	41 ± 7	**	inhibitor	78 ± 4	*	inhibitor	1.0	NS	no	1.3 ± 0.1	22.9 ± 4.5	yes
SMU29–280	116 ± 4	*	stimulator	148 ± 4	*	stimulator	1.2	NS	yes	1.6 ± 0.1	21.6 ± 4.6	marginally
SMU29–286	98 ± 2	NS	none	143 ± 32	NS	none	1.3	*	yes	1.9 ± 0.1	25.7 ± 4.9	no

Open in a new tab

ATP hydrolysis assays using purified P-gp were performed without added transport substrate (“basal ATPase”) or in the presence of verapamil (“stimulated ATPase”). Results are presented compared to DMSO control ± standard deviation (three independent experiments with duplicate samples). The basal activity of P-gp was 20 to 30 nmol min⁻¹ mg⁻¹; verapamil-stimulated rates were 200 to 400 nmol min⁻¹ mg⁻¹ of P-gp. Stimulation of basal ATPase by 29-variants was used as an indicator that a compound may be a P-gp transport substrate. Effects on stimulated P-gp ATPase activity indicated whether a compound directly interfered with ATP usage by the protein (***, p < 0.001; **, p < 0.01; *, p < 0.1; NS, not significant). Quantitative cellular accumulation of 29-variants was performed using LC-MS/MS (Experimental Section) and is presented as a ratio of the cellular amounts of 29-variants in the presence of the P-gp inhibitor tariquidar, divided by amounts accumulated in its absence. A ratio of >1 indicates that the compound likely is a transport substrate of P-gp (***, very significant; *, significant; NS, not significant). The binding of an ATP analog, SL-ATP, to P-gp was used to determine whether ATP binding to P-gp was affected by the 29-variants (see Experimental Section). Values ± standard deviations are shown for at least three different P-gp preparations and three independent SL-ATP titration experiments.

The values for SL-ATP binding in the presence of 29 were taken directly from Brewer et al.^²⁶

Effects on “Basal” ATP Hydrolysis Rates of Compound 29 Variants.

Group 1 compounds 216, 227, and 541 did not significantly affect basal ATP hydrolysis by P-gp, while compound 231 was inhibitory. Only 551 of the group 1 molecules stimulated basal ATPase activities of P-gp. Of the group 2 compounds, 238 and 255 stimulated basal ATPase by ~10- and ~3-fold, respectively. Compounds 280 and 286 stimulated basal ATPase marginally or with no statistical significance. Only compound 278 inhibited basal ATPase of P-gp. The relatively strong activation of basal ATPase by compounds 238 and 255 was suggestive that these two compounds and, potentially to a lesser extent, compound 280, may be transport substrates of the pump. Compound 278 was not assumed to be a good transport substrate for P-gp since it inhibited basal ATPase by P-gp.

Intracellular Accumulation of Compound 29 Variants.

To more directly assess whether the 29 variants were indeed transport substrates for P-gp, cell accumulation assays were performed as in ref 28. These assays measured the intracellular accumulation of the experimental compounds using LC–MS/MS methods after incubation with the P-gp-overexpressing cell line, DU145TXR, in the absence and presence of the strong P-gp inhibitor, tariquidar⁴⁷ (TQR). Low levels of cellular accumulation of a compound in the absence of tariquidar accompanied by much higher levels of accumulation in the presence of tariquidar suggests that the compound in question may be a transport substrate of P-gp. If a compound is not a substrate of P-gp, no significant difference in intracellular accumulation of the compound with or without tariquidar is expected. Daunorubicin (DNR) was used as a positive control of a good transport substrate for P-gp and showed very strong cellular accumulation in these assays when P-gp was inhibited by tariquidar, but much less accumulation in the cells when P-gp was not inhibited (see Figure S5, DNR). Figure S5 “29” shows that compound 29 is not a transport substrate for P-gp as already discussed in ref 28 and no significant difference in cellular accumulation of 29 was observed with or without addition of tariquidar (TQR). Figure S5 also presents the fold accumulation of each of the experimental 29-variants in these assays, normalized to the amount of the compound found in the absence of TQR. This data is numerically presented in Table 5 as the ratio of observed accumulation in the presence of tariquidar divided by the accumulation observed in the absence of tariquidar for each of the experimental compounds. Ratios that are significantly greater than 1.0 indicate that a compound is very likely a transport substrate of P-gp.

None of the group 1 molecules resulted in intracellular accumulations that were considerably different in the absence or presence of TQR, similar to the parental compound 29 (Figure S5 and Table 5), indicating that these variants were likely not transport substrates of the pump in human cells in culture.

Of the group 2 compounds, variant 238 showed a very large and significant increase in intracellular accumulation in the presence of TQR (Figure S5 and Table 5). Compounds 255 and 286 showed more modest but statistically significant increases in intracellular accumulation when P-gp was inhibited in the presence of TQR. Compounds 278 and Compound 280 did not show significantly different intracellular accumulations in the absence or presence of TQR.

To assess whether the observed discrepancies of compounds stimulating basal P-gp ATPase activity but are not being transport substrates of the human pump in the cell culture assessments was due to the fact that, in the biochemical assays, we used a cysteine-less variant of the mouse MDR3 P-glycoprotein, we repeated the ATP hydrolysis experiments using normal human MDR1 P-gp. In order to stabilize the human protein for the activity assays, the protein was assembled into membrane nanodiscs as described in the Experimental Section. The results of the experiments are shown in Table 6. Interestingly, neither compound 29 nor any of its variants had a stimulatory effect on basal ATPase activity of the normal human protein reconstituted into membrane nanodiscs, while all of them significantly inhibited transport substrate (verapamil)-stimulated activity. The results clearly indicate that the source (human vs mouse) and potentially also the membrane environment of P-glycoprotein strongly affect the overall behavior of potential biochemical inhibitors and their impact on the catalytic mechanism of the enzyme.

Table 6.

Effects of Group 1 and Group 2 Compound 29 Variants on ATP Hydrolysis by Normal Human MDR1 P-glycoprotein^a

compound	stimulated ATPase (% DMSO, significance)		effect on stimulated ATPase	basal ATPase (% DMSO, significance)		effect on basal ATPase
DMSO	100 ± 6			100 ± 5
SMU29	62 ± 3	**	inhibitor	88 ± 9	NS	none
Group 1: ChemGen and docking selected
SMU29–216	64 ± 6	**	inhibitor	69 ± 8	**	inhibitor
SMU29–227	62 ± 5	**	inhibitor	83 ± 7	NS	none
SMU29–231	40 ± 3	****	inhibitor	70 ± 4	***	inhibitor
SMU29–541	66 ± 6	**	inhibitor	87 ± 9	NS	none
SMU29–551	27 ± 1	***	inhibitor	72 ± 9	*	inhibitor
Group 2: rationally designed/no docking selection
SMU29–238	61 ± 7	**	inhibitor	89 ± 7	NS	none
SMU29–255	59 ± 6	**	inhibitor	84 ± 9	NS	none
SMU29–278	63 ± 6	**	inhibitor	74 ± 6	**	inhibitor
SMU29–280	54 ± 7	***	inhibitor	76 ± 8	*	inhibitor
SMU29–286	59 ± 4	***	inhibitor	95 ± 3	NS	none

Open in a new tab

ATP hydrolysis assays using purified human P-glycoprotein were performed (see Experimental Section) without added transport substrate (“basal ATPase”) or in the presence of verapamil (“stimulated ATPase”). Results are presented compared to DMSO control ± standard deviation (three independent experiments with duplicate samples). The specific basal activity of normal human MDR1 P-gp was between 123 and 193 nmol min⁻¹ g⁻¹, and transport substrate (verapamil)-stimulated activity was between 193 and 263 nmol min⁻¹ mg⁻¹. Effects on stimulated P-gp ATPase activity indicated whether a compound directly interfered with ATP usage by the protein (***, p < 0.001; **, p < 0.01; *, p < 0.1; NS, not significant).

Effects of 29-Variants on Binding of an ATP Analog to Purified P-Glycoprotein.

ATP binding in the presence of the 29-variants was assessed in titration assays using a spin-labeled analog of ATP, 2′,3′-SL-ATP (2′,3′-(2,2,5,5,-tetramethyl-3-pyrroline-l-oxyl-3-carboxylic acid ester)ATP; (2′,3′ indicates a rapid equilibrium of the ester bond between the C2′ and C3′ of the ribose moiety)^48–50 and electron spin resonance spectroscopy as described in ref 26. Due to the lower stability of the human P-glycoprotein in the extended times needed for these experiments, the cysteine-less mouse protein was used here. The goal was to assess whether binding of the 29-variants to P-gp affected nucleotide binding to the protein. Results of these assays are presented in Table 5. Except for compounds 238 and 278, neither of which initially underwent the selective docking routines used for group 1 compounds, none of the novel inhibitors affected maximal binding of the ATP analog or the apparent K_d, supporting the hypothesis that the inhibitors were indeed interacting with the putative allosteric binding site on P-gp. Compounds 238 and 278 reduced SL-ATP binding to approximately 1 mol of SL-ANP (ANP, adenine nucleotide with an undefined number of phosphoryl groups) bound/mole of enzyme, suggesting that these inhibitors may also interact with the nucleotide-binding sites or may indirectly induce changes in P-gp that affect ATP binding.

DISCUSSION AND CONCLUSIONS

Using Computational Approaches To Create Novel Variants of “Hit” Molecules from Drug Discovery Programs.

A number of virtual chemical synthesis computer programs have been previously described: some use fragments annotated with reaction rules²⁹ or compound scaffolds with chemically reactive linkers³⁰ and again others use popular click chemistries,^31,32 just to mention a few. To make more informed choices about which of the vast numbers of possible compound variants to synthesize for subsequent testing, we have written and developed a set of computational routines (collectively called ChemGen) to in silico synthesize what could be very large numbers of variant compounds. Our methods differ from predecessor methods in that retrosynthetic approaches to the discovered hit molecule synthetic routes are mimicked in the computations. This results in advantageous translation to actual chemical syntheses of identified variants of interest and is not constrained to one or a few chemical reaction types. A disadvantage of the approach is that each reaction type must be programmed ahead of its implementation; however, the ChemGen platform can be relatively easily adapted to new chemistries.

In this presented pilot study, we created hit variants computationally and then assessed them computationally for variants that bind with increased affinity to desirable protein structures and not well to structures that were undesired for this particular protein target, that is, improved binding to regions in the nucleotide-binding domains of P-glycoprotein but low affinity to the drug-binding domains in our efforts to find inhibitors of the protein that are not transport substrates themselves. This latter part of the approach, namely, computational counter-selection for compounds with undesirable binding characteristics, was successfully used by us previously to identify molecules, including hit compound 29, that inhibited P-gp catalysis but were not transport substrates of P-gp.^26–28 When coupled with virtual synthesis of hit variants, increased efficiency and cost-effectiveness of synthesis projects are practically assured.

Compound 29 and the other hits discovered were evaluated in biochemical and biophysical studies for their mechanism of inhibition of P-gp action²⁶ as well as for their potential to reverse multidrug resistance in different cancer cell lines in culture.^27,28 Compound 29 was chosen as an initial compound for further development in this present study mostly for the fact that binding of compound 29 was predicted to be at an allosteric site, away from the nucleotide-binding sites of P-gp,²⁶ which was hypothesized to reduce potential off-target effects of the inhibitor and its derivatives. Earlier biophysical assessment using electron spin resonance spectroscopy and a spin-labeled ATP analog suggested that ATP binding was not affected in the presence of 29, while ATP hydrolysis assays showed inhibition of ATPase activity.²⁶ The putative mechanism for P-gp inhibition by 29 can be envisioned as shown in Figure 1B when comparing the position to which 29 docked with high affinity²⁶ to the recently published cryo-EM structure of the protein¹⁶ (Figure 9). Figure 9 shows a pronounced steric clash of P-gp amino acid side chains with the bound inhibitor when P-gp adopts a conformation similar to that of the published cryo-EM structure. This clash may suggest that, in the presence of bound 29, P-gp cannot undergo conformational changes that may be needed for catalytic activity. Creating variants of compound 29 with increased affinity to this particular binding site was therefore viewed as a promising strategy toward further development of specific P-gp inhibitors that would not function as a transport substrate of the protein.

Figure 9. — Putative allosteric inhibitory site on human P-glycoprotein with the ATPase inhibitor compound 29 bound. Compound 29 shown in a licorice representation with its solvent-excluded surface in a wire mesh as it was docked in the P-gp model shown in a blue cartoon. Shown in a red cartoon representation is the cryoEM-derived structure determined by Kim and Chen (2018, PDB accession 6V0C) superposed on the P-gp model used for docking. One can clearly observe a steric clash of 29 with the cryoEM structure (black arrow). Residues involved in the steric clash with 29 (atoms within 2.0 Å) are T422, E566, Q570, V584, S590, T591, and N594. Small molecules are shown in the figure with nonpolar hydrogens removed for clarity. Images were created with VMD with the MSMS surface renderer from M. Sanner and the Tachyon image renderer in VMD (see the Experimental Section for details).

Closer evaluation of the high-affinity allosteric docking site of compound 29 to P-gp revealed a relatively large hydrophobic pocket where the cyclopropyl moiety of the “Western” half of the molecule interacted with the protein (Figure 1A,C,D). In our efforts to assess putative P-gp inhibitors with increased affinity to the protein, we set out to “virtually synthesize” a number of 29-variants with larger moieties at the “Western” half of the protein. We used the synthesis scheme shown in Figure 2B and the ChemGen protocols described in the Experimental Section and Supporting Information to accomplish this goal. The resulting 647 derivatives of hit compound 29 were evaluated for binding to the allosteric site and counter-screened for low-affinity interactions to the drug-binding domains of the protein using docking methods that were similar to those described in ref 26 and that led to the discovery of the parental compound 29.²⁸

The ChemGen in silico synthesized and docking routine-selected group 1 variants of 29 were ranked by binding affinity to the allosteric site on P-gp. Compounds with molecular weights that exceeded 600 Da were originally excluded from synthesis and further evaluation. Twelve variants of 29 that were predicted to bind with relatively high affinity to the proposed allosteric site were further evaluated for some physicochemical characteristics (Table 1). All of the variants showed higher molecular weights as was to be assumed when adding larger fragments to the “Western” part of the molecule. TPSA and consensus log P values differed between the variants. Visual evaluation of the docking poses of the variants showed clear overlap of the “Eastern” parts of the molecules, highlighting the consistency of docking to this site on P-gp. The “Western” portions of the Group 1 29-variants were observed to extend further into the hydrophobic pocket of the protein compared with the original compound 29 (Figure 3A and Figure S3).

Of the twelve 29-variants given in Table 1, five were chosen for actual chemical synthesis mostly based on their perceived ease of synthesis and relatively low expense of precursor fragments (216, 227, 231, 541, and 551). Variant 551 was also chosen for synthesis even though its molecular weight of 627 Da exceeded our initial M_w cutoff mostly because of the ease of synthesis and also for its otherwise favorable characteristics. All variants added more volume to the “Western” half of the molecules, and all but 551 and to some extent 541 had lower TPSA than that of the original compound 29. All variants had somewhat higher consensus log P values than those of 29. Closer inspection of the docking poses of the five variants (Figure 3) revealed that 29-derivatives 216 and 231 both reached significantly farther into the hydrophobic pocket compared with the parental compound 29, while variant 227 seems to make significant protein interactions at the “mouth” of the hydrophobic pocket. Both 541 and 551 reached deeply into the previously observed pocket within P-gp. We hypothesized that these five variants should show improvements in reversing the MDR phenotype of cancer cells that overexpress P-gp as demonstrated for 29 in refs ²⁷ and ²⁸ and would therefore be reasonable initial choices for this proof of principle study.

Cell viability assays using a P-gp-overexpressing prostate cancer cell line indicated that all five of the 29-variants had improved characteristics over the parental 29 in causing increased cell mortality (summarized in Table 2) in the presence of the chemotherapeutic paclitaxel. This remarkable result that five out of five group 1 ChemGen-produced variants showed between 2.4-fold and 11-fold improvement over the performance of 29 in reversing the P-glycoprotein-conveyed multidrug resistance phenotype (summarized in Table 2) underscores the utility of the ChemGen virtual synthesis and docking approach for increasing variant affinity to P-gp. It is a reasonable conclusion that the larger group 1 variants were able to interact more strongly with the protein as depicted in Figures 1 and 3 than the parental compound 29 (Table 1, Figure 3, and Figure S3), adding to the overall binding affinity of the variants.

Table 2.

Increased Toxicity of Paclitaxel to DU145TXR in the Presence of P-gp Inhibitors Identified by ChemGen/Docking Routine^a

inhibitor concentration	resensitization to paclitaxel with the indicated treatment and fold-increased sensitivity in the presence of inhibitors
	PTX alone	PTX + 29			PTX + 216			PTX + 227			PTX + 231			PTX + 541			PTX + 551
	IC₅₀ PTX (nM)	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29	IC₅₀ PTX (nM)	fold vs PTX	fold vs PTX + 29
3 μM	2120	629	3	1	266	8	2.4	164	13	3.8	153	14	4.1	426	5	1.5	56	38	11
5 μM	2120	194	11	1	154	14	1.3	52	41	3.7	46	46	4.2	21	101	9.2	20	106	9.7
7 μM	2120	59	36	1	62	34	1	6	353	10	7	303	8.4	17	125	3.4	9	236	6.6
10 μM	2120	21	101	1	57	37	0.4	4	530	5	2	1060	11	17	125	3.4	6	353	3.5

Open in a new tab

Cytotoxicity of the chemotherapeutic paclitaxel (PTX) to P-gp-overexpressing prostate cancer cells, DU145TXR, was determined in the absence and presence of the ChemGen-designed 29-variants, 216, 227, 231, 541, and 551. For each experimental compound IC50 values of PTX alone or in the presence of inhibitors, fold improvement of PTX sensitivity in the presence of the inhibitor and fold improvement of PTX sensitivity by variants compared to parental compound 29 are given.

In assays designed to allow quantification of the accumulation of a P-gp transport substrate, calcein AM, in cells that overexpress P-gp, however, the larger, more hydrophobic group 1 variants did not initially perform better than 29. Comparison of the calculated consensus log P values for variants 216, 227, 231, 541, and 551 (6.6, 6.4, 5.7, 5.4, and 5.9, respectively, see Table 1) with the log P of the parental compound 29 (4.6, see Table 1) led us to hypothesize that the lack of efficacy in inhibiting P-gp-catalyzed export of calcein AM may have been due to 29-variants being too hydrophobic to efficiently transfer across the cellular membrane to the cytosol-located nucleotide-binding domains of P-gp for efficious inhibition in the nucleotide-binding domain of the protein to occur. The relatively short incubation times that are used in the calcein AM assays as shown in Figure 5, left, would hypothetically exacerbate this problem when compared to much longer exposure times in the cellular toxicity assays (Figure 4). Supporting this hypothesis was the observation of significant improvements in efficacy in the calcein AM assays when longer preincubation with the variants was performed, Figure 5, right. Except for variant 216, which has the lowest TPSA and highest consensus log P and which under these conditions performed similar to parental compound 29, all other variants performed better in retaining the P-gp substrate than 29. The variants with the most promising physicochemical characteristics (relatively high TPSA/ relatively low log P values), compounds 541 and 551, performed best in these assays. This finding supports the hypothesis that given enough time to enter the cells, the group 1 variants performed better or at least equally to the parental compound.

Since the ChemGen/docking routine computational approach for selecting variants at the putative allosteric site resulted in larger but also more hydrophobic “Western” fragments in variants 216, 227, 231, 541, and 551 than those of the parental compound 29 and increased hydrophobicity may be viewed as pharmacologically less desirable in drug candidates, we set out to rationally design another set of variants of the “Western” portion of compound 29 that might also result in improved interactions with P-gp but may in addition have also more favorable physicochemical characteristics. In this light, we synthesized five additional “Western” derivatives of 29 with varying shapes, sizes, and physicochemical properties, compounds 238, 255, 278, 280, and 286 (group 2 variants). The compounds were chosen for relative ease of synthesis as well as availability and cost of precursors. Unlike the previous set of ChemGen-generated group 1 variants, this latter group did not initially undergo the docking routines that selected for high-affinity binding to the nucleotide-binding domains and low-affinity binding to the drug-binding domains. Instead, these variants were chosen by visual inspection of the putative binding site as well as by the calculated physicochemical properties of the resulting variants.

Efficacy of Rationally Designed Group 2 Variants in Reversing MDR Caused by P-gp Overexpression.

If the hypothesis was correct that compounds 216, 227, 231, 541, and 551 were too hydrophobic for a kinetically favorable passage through the cellular membrane into the cytoplasm, thereby causing less efficient P-gp inhibition and decreased calcein AM accumulation in the P-gp overexpressing cells, then the more hydrophilic group 2 derivatives, 238, 255, as judged by their TPSA and log P values, and possibly 286 as judged by its log P value, might show improved calcein AM retention as compared to the other derivatives. Indeed, improved calcein AM retention was observed for compounds 238 and 255 already for the shorter incubation times (Figure 8, left), while 286 performed similar to 29.

The more hydrophobic compounds 278 and 280 (as judged by Table 3) performed similar to the parental compound only upon extended exposure to the cells.

All in all, TPSA values greater than ~110 Å³ and consensus log P values lower than 6 correlated best with MDR reversal efficacy in both the group 1 and group 2 variants and the most pronounced retention of the P-gp substrate, calcein AM.

Effectiveness of the Selective Docking Routines Used in Group 1 Variants To Ensure Targeting to the Nucleotide-Binding Domains and Avoiding the Drug Transport Domains of P-gp.

Since our initial premise was to generate a new class of P-gp inhibitors that are not transport substrates of the pump, it was of interest to investigate whether the subtractive docking routine performed on group 1 variants added value to the overall selection variants for synthesis. Comparison of the results from group 1 compounds to the more traditional, rationally designed group 2 variants that were chosen without additional input by docking-derived binding affinities to different substructures of the protein was therefore made. To this end, we assessed the effects of the 29-variants on basal (no transport substrate present) ATPase activity of P-gp as well as the intracellular accumulation of the hit variants in the P-gp-overexpressing cancer cells, DU145TXR, in the presence or absence of a strong inhibitor of P-gp, tariquidar. Stimulation of basal ATPase activity of P-gp is commonly believed to be a good indicator that a transport substrate is present in the assay. Interestingly, the cellular accumulation assays (Figure S5) and ATP hydrolysis assays did not correlate all too well. While all of the group 1 variants were shown not to be transport substrates in the cellular accumulation assays as in ref 28, only 231 resulted in inhibition of basal ATPase activity, while 551 stimulated the activity, and the rest had no effect.

To evaluate whether this lack of correlation may have been due to the fact that these biochemical assays were performed using a cysteine-less variant of the mouse MDR3 P-glycoprotein and not the human isoform, we repeated the experiments using normal (not cysteine-less) human P-gp that we had reconstituted into nanodiscs for enhanced stability. The results presented in Table 6 indicate that none of the 29-variants nor 29 itself stimulated basal ATP hydrolysis but that, in contrast, several of them inhibited basal ATPase activity. This finding places doubt on the general assumption in the field that transport substrates stimulate basal ATP hydrolysis. We therefore conclude that simple assessment of basal P-gp hydrolysis activities may not be a good predictor for the biochemical mechanism of an inhibitor. While not giving any information about whether or not an inhibitor is a transport substrate of P-gp, a much better prediction can be gleaned by assessing the inhibition of transport substrate stimulated ATPase activity using the normal human P-gp as shown in Table 6. All of the variants of compound 29 inhibited stimulated activity, indicating interaction with the nucleotide-binding domain as predicted from the putative binding of 29 and its variants to an allosteric site as depicted in Figure 9.

Although the general case cannot be statistically proven since the number of synthesized and tested variants was small, these experiments support the hypothesis that the ChemGen-produced variants that were selected against interactions at drug-transporting structures on P-gp were much more likely not to be transport substrates of P-gp (five of five group 1 variants were not observed to be transport substrates) than the group 2 molecules that did not undergo the docking counter-selection procedures (where four out of five molecules were likely or potentially transport substrates for P-gp). The studies strongly suggest that the docking and counter-selections procedures used to select the group 1 variants outperformed the more traditionally rationally designed molecules that were not subjected to these selections.

Mechanism of Inhibition of the 29-Variants.

To assess whether the 29-variants likely interacted with the putative allosteric site or potentially also interacted with the nucleotide-binding sites of P-glycoprotein, titration experiments using an electron spin resonance (ESR)-active ATP analog, SL-ATP, were performed and the amount of P-gp bound SL-ANP was determined in the presence of the 29-variants (Table 5). The results showed that none of the group 1 variants significantly affected ATP binding, while two of the group 2 variants reduced ATP binding from 2 to approximately 1 mol of the nucleotide analog/mol of protein. This again indicates that the selection process through ChemGen/selective docking was much more predictive of the effects the potential inhibitors had on the enzyme.

Among all compounds tested, the strongest reversal of MDR and the strongest stimulator of both “basal” and “verapamil-stimulated” ATPase activities was variant 238 of the group 2 molecules. Compound 238, while strongly reversing MDR, was also the best transport substrate of all the variants tested, a characteristic that we deemed undesirable for any clinically relevant P-gp modulator lead as discussed above. Although our curiosity on the mechanism of action of 238 is piqued by the observation that this compound also inhibited SL-ANP binding to P-gp (Table 5), we will likely not consider it for further clinical lead development.

The study presented here supports the hypothesis that the virtual synthesis of hit variants with a program suite like ChemGen combined with selection for desired characteristics predicted from docking experiments, that is, increased affinity to targeted structures and decreased affinity against substructures that should be avoided, as employed here, resulted in efficient and cost-effective identification of five out of five variants assessed that met these goals. A comparison of the synthetic capabilities of ChemGen with, for example, commercially available programs that perform synthetic chemistry reactions from ChemAxon (i.e., Reactor), shows that the potential for both packages to perform a wide variety of different chemical reactions is similar. Both programs deal with a virtual synthesis engine that can use a large library of prescreened or more randomly acquired reactant molecules to create structural variants in structure-based lead optimization studies. ChemGen also gives the option for fairly rigorous quantum mechanical optimizations of the bonding geometries of synthetic products and looks for steric interference between reactants for the identification and rejection of variants that would likely be impossible or difficult to obtain synthetically. In addition, ChemGen is set up for convenient prescreening of reactant molecules for optimized pharmacological and physiochemical fragment properties. Each program requires the expert knowledge of a synthetic chemist to specify each specific chemical reaction to be performed. While the ChemAxon Reactor has more known chemistries that have been preprogrammed to draw on due to its longer and necessarily much wider use, it should be noted that the known chemistry libraries utilized by ChemGen can be expanded on demand by any user. While both programs require either an atomic resolution structure and molecular docking of the starting “hit” ligand or ligands to start the processes, the most prominent advantage of ChemGen is that, in addition to medicinal chemistry, target selection is included in the suite. The ability to screen for preferential binding to one substructure of a target (here, a putative allosteric site in the nucleotide-binding domains) while avoiding binding to an undesirable structure (here, the drug-binding domains) was shown in the study to be highly successful. The main goal to this study was to discover derivatives of hit compound 29 that were not transport substrates of P-glycoprotein while inhibiting ATP hydrolysis and reversing drug resistance in drug-resistant cancer cells. All five variants synthesized and assayed from this group fulfilled all of these requirements. None of the variants affected nucleotide binding to the protein, suggesting that all variants were binding to the allosteric site targeted. All of the five variants rationally designed to fit into the putative allosteric binding site of hit compound 29 reversed drug resistance in cancer cells, and all inhibited transport substrate stimulated ATP hydrolysis by the human MDR1 protein, but two out of the five variants affected nucleotide binding to P-gp, indicating that these variants may also directly bind to the nucleotide binding sites that were not originally targeted. Four out of the five variants were also shown to be transport substrates of the pump, a characteristic that was not predictable from simple visual inspection of the desired binding interactions with the protein.

Even though this presented study is a pilot study in scope, the exceptionally high success rate of 100% for all desired characteristics of the ChemGen variants as compared to 60% for binding specificity to the allosteric site and only 20% for avoiding the drug-binding domains of the more traditionally rationally designed variants is a good example of how computational evaluation and selection of potential inhibitors before their actual synthesis adds to the speed and overall success rates in identifying hit-to-lead variants that possess desired characteristics. The use of ever more sophisticated presynthesis computational selections will likely yield even better variants in the future. Computational routines that employ parameters such as those used here, that is, increased affinity to targeted structures and avoidance of undesirable interactions, can in the future be easily expanded to include avoidance of targeting other proteins, optimized physicochemical properties for solubilities and membrane penetration, and even complex parameters like decreased cellular toxicity and avoidance of adverse ADME properties that can be potentially learned through deep learning derived computational studies. Such expansion of computational preselection routines would be expected to yield better molecules faster with less financial and time investment than those of current, lesser-guided rational design programs.

EXPERIMENTAL SECTION

Virtual Synthesis of Compound 29-Derivatives Using ChemGen.

Because of the nature of the putative allosteric site shown for compound 29 and in light of rational design considerations (see text), only a moderate set of variants were virtually synthesized for this study using the ChemGen programs. The ChemGen program is described in detail in the Supporting Information to this paper. In the virtual syntheses performed here, a scaffold molecule, 2-chloro-N-[1-phenyl-3-(2,4,5-trimethylphenyl)-1H-pyrazol-5-yl]acetamide, which is equivalent to the chloroacetamide that retains the “Eastern” substituent group from compound 29, was reacted with approximately 650 thiol compounds obtained by simple structural searches for thiols from the “clean drug-like” commercially available molecule set at the ZINC database.³⁴ The 2-chloro-N-[1-phenyl-3-(2,4,5-trimethylphenyl)-1H-pyrazol-5-yl]acetamide scaffold and the thiol precursor molecules were “marked” for reactions in ChemGen as described for the second reaction shown in Figure 2B (see the Supporting Information for details about the ChemGen marking and molecule building methods). The build process successfully created 647 derivatives of the “Eastern” substituents of 29, which were then geometrically optimized with the electronic Ligand Builder and Optimisation Workbench (eLBOW) – Phenix³⁵ program.

In Silico Docking of Compound 29 Variants to a Model of Human P-gp.

AutoDock Vina⁵¹ and AutoDock 4.2^52–54 were used with a model of P-glycoprotein with drug-binding domains open to the outside and nucleotide-binding sites fully formed that was extracted from targeted molecular dynamics trajectories as described in refs ²¹ and ²². This conformation of P-gp is one that is very similar to the homologous Sav1866 crystal structure reported by Dawson and Locher⁵⁵ and was the conformation with which compound 29 was originally identified.²⁶ Ligand docking was limited to a volume equivalent to 20 × 24 × 26 Å³ centered on the putative allosteric site of P-gp (see Figure 1B). One hundred twenty-eight replicates for each ligand were calculated and then ranked by the lowest estimated binding energies to the protein target calculated by the docking programs.

Calculation of Physicochemical Properties.

The SWISS-ADME server at http://www.swissadme.ch/ was used for the calculation of the physicochemical properties of the compounds as discussed in ref 44.

Imaging of P-Glycoprotein and Ligands.

The VMD (visual molecular dynamics) program suite was used extensively in this work for the analysis of structural data and for the presentation of visual images⁵⁶ and included the SURF surface representation program⁵⁷ as well as the pdb2pqr^58,59 and APBS^60,61 programs for electrostatic/solvation calculations.

Synthetic Procedures.

All synthetic procedures and analyses of products are provided in the Supporting Information, “Synthetic Procedures”. All compounds were ≥95% pure as determined by ¹H NMR and high-performance liquid chromatography (HPLC). The chromatographic separations were accomplished by using an Ascentis C18 25 cm × 2.1 mm, 5 μm column with 10 μL injection volumes. A flow rate of 0.2 mL min⁻¹ was used and eluted with a mobile phase of 100% water for 3 min, a linear gradient to 100% acetonitrile over 2 min, and 15 min of 100% acetonitrile.

Cell Lines and Cell Culture.

The chemotherapeutic sensitive DU145 human prostate cancer cells⁶² as well as the multidrug-resistant sub-line, DU145TXR³³ were generous gifts from Dr. Evan Keller (University of Michigan, Ann Arbor, MI). The multidrug-resistant DU145TXR was maintained under positive selection pressure by supplementing complete medium with 10 nM paclitaxel (Acros Organics, NJ). Both cell lines were maintained in complete media consisting of RPMI-1640 with l-glutamine, 10% fetal bovine serum (FBS; BioWest, Logan, UT), 100 U mL⁻¹ penicillin, and 100 μg mL⁻¹ streptomycin in a humidified incubator at 37 °C and 5% CO₂. The noncancerous human fetal lung cell line, HFL1,⁶³ was kindly provided by Dr. Robert Harrod (Southern Methodist University, Dallas, TX) and maintained in complete media consisting of F-12K with l-glutamine, 10% FBS (BioWest, Logan, UT), 100 U mL⁻¹ penicillin, and 100 μg mL⁻¹ streptomycin in a humidified incubator at 37 °C and 5% CO₂. To promote attachment of HFL1 cells, growth surfaces were treated with 0.1 mg mL⁻¹ rat tail collagen (BD Biosciences, Palo Alto, CA) in 0.02 N acetic acid for 10 min and rinsed with PBS prior to use. Cell culture materials were purchased from Corning Inc. (Corning, NY) unless otherwise stated.

MTT Cell Viability Assay.

Cells were trypsinized from monolayers and seeded with 3000 cells in 150 μL of complete medium in a 96-well plate. After 24 h, cells were treated for 48 h with paclitaxel (Acros Organics, NJ) and/or P-gp inhibitory compounds dissolved in DMSO or DMSO controls. All additions were diluted into complete medium. After 48 h of treatment, MTT assays were performed as described⁴¹ using 5 mg mL⁻¹ of MTT (Acros Organics, NJ) solution prepared in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.4). After 4 h of incubation with MTT, the media was removed and the formazan crystals were dissolved in 100 μL of DMSO. The absorbance at 570 nm was then measured using a BioTek Cytation 5 imaging multimode reader (Bio-Tek, Winooski, VT). Percent viability was calculated using DMSO-treated cells as representative for 100% viability. Background absorbance was determined using MTT and complete medium without cells and subtracted from all the test values.

% Viability = \frac{Absorbance at 570 nm of test well \times 100}{Absorbance at 570 nm of DMSO - treated cells}

The results were plotted as the mean with standard deviation (SD) of eight replicates per concentration from at least two independent experiments with n = 8. The graphical representations and IC₅₀ values were determined using four-parameter variable slope nonlinear regression, using the following equation: Y = bottom + (top − bottom)/(1 + 10^((log IC₅₀ − X)*hill slope) (GraphPad Prism, La Jolla California, U.S.A., version 6.05). The reported “fold sensitization” was calculated as follows.

Fold sensitization = ({IC}_{50} value of DU145TXR cells treated with chemotherapeutic only) / ({IC}_{50} value of DU145TXR cells treated with chemotherapeutic and P-gp inhibitory compound)

Calcein AM Assay.

To assess inhibition of P-gp-catalyzed transport of the P-gp pump substrate, calcein AM, DU145TXR cells were seeded in 96-well plates and allowed to grow in complete medium until confluency was reached. The medium was removed, and cells were treated without 2 μM P-gp inhibitory compounds and 1 μg mL⁻¹ calcein AM (Life Technologies, OR) and diluted into phenol red free RPMI 1640 media. To study the effect of preincubation of compounds, cells were treated with just P-gp inhibitory compounds and incubated at 37 °C for 6 h before adding calcein AM. Fluorescence excitation at 485 nm with a 20 nm gate and at emission at 535 nm with a 20 nm gate was measured using a BioTek Cytation 5 imaging multimode reader (Bio-Tek, Winooski, VT) over 60 min in 20 min intervals. DMSO was used as vehicle. Results were plotted as the mean with standard deviation (SD) of six replicates per concentration and are representative of at least two independent experiments.

Cellular Accumulation Assays for Experimental P-gp Inhibitors.

Cell use (DU145TXR), cell culturing, cell exposure to compounds, cellular handling, and extractions were performed as described in ref 28. LC–MS/MS methods were performed as described in ref 64 and as modified in ref 28.

P-Glycoprotein Purification.

Cysteine-less MDR3 and the human normal MDR1 P-glycoprotein were recombinantly expressed in the yeast P. pastoris and purified essentially as described in refs ⁴⁵,⁴⁶,⁴⁹, and ⁶⁵ and were used for assaying ATP hydrolysis and ATP binding to the protein as described in ref 26 in the presence of the 29-variants.

Nanodisc Assembly.

Human P-gp in mixed detergent micelles obtained during protein purification was reconstituted into nanodiscs similar to that in refs ⁶⁶ and ⁶⁷ with small modifications. P-gp was assembled with membrane scaffold protein MSP1E3D1 (Sigma-Aldrich) expressed in BL21 (DE3) and l-alpha-phosphatidylcholine (Sigma-Aldrich) at a ratio of 1:10:500 (P-gp:MSP:PC) in 50 mM Tris-CL (pH 8). The detergent was removed with Bio-Beads SM-2 adsorbent media (BioRad). Ni-NTA agarose (Qiagen) chromatography was used to purify P-gp reconstituted nanodiscs using six bed volumes of the start buffer (20% (v/v) glycerol, 50 mM Tris-CL pH 7.5 at 4 °C, 50 mM NaCl) and five bed volumes of the elution buffer (20% (v/v) glycerol, 50 mM Tris-Cl pH 7.5 at 4 °C, 50 mM NaCl, 300 mM imidazole).

ATPase Activity Assays.

ATP hydrolysis activity was measured using a coupled enzyme assay⁶⁸ as modified in ref 26. The specific basal activity of the mouse MDR3 cysteine-less P-gp was between 20 and 30 nmol min⁻¹ mg⁻¹, and the transport substrate (verapamil)-stimulated activity was 200–400 nmol min⁻¹ mg⁻¹. The specific basal activity of normal human MDR1 P-gp was between 123 and 193 nmol min⁻¹ mg⁻¹, and the transport substrate (verapamil)-stimulated activity was between 193 and 263 nmol min⁻¹ mg⁻¹.

ESR Measurements.

ESR measurements were as described in ref 26. The amount of the protein-bound spin-labeled (SL) adenine nucleotide was determined as the difference between the known total concentration of SL-ATP (2′,3′-(2,2,5,5,-tetramethyl-3-pyrroline-1-oxyl-3-carboxylic acid ester)–ATP⁴⁸) added and the free spin-labeled nucleotide (SL-ANP) observed in the experiment. Hyperbolic curve fitting of the results was performed using GraphPad Prism7 to determine maximum binding and apparent affinity for the spin-labeled nucleotide. The equation used for the fitting the curves was y = P₁ × x/(P₂ + x) where P₁ corresponds to the maximum binding of SL-ANP (moles of SL-ANP bound per mole P-gp) and P₂ equals the apparent dissociation constant for SL-ANP. To quantify the amount of free SL-ANP, standard curves were established where the signal amplitude of the high field signal of the ESR spectra of free SL-ANP in the absence of protein was correlated to the concentration of SL-ANP added. All ESR measurements were performed using a Bruker EMX 6/1 ESR spectrophotometer operating in the X-band mode and equipped with a high-sensitivity cavity. Spectra were acquired at a microwave frequency of 9.33 GHz, microwave power of 12.63 mW, 100 kHz modulation frequency, and a resolution of 1024 points. The centerfield of the scan was at 3325 G and an area of 100 G was scanned. The peak to peak modulation amplitude was 1 G, and the time constant was set to 10.240 ms. The conversion time was 163.84 ms, resulting in a total time sweep of 167.772 s. The signal gain was adjusted for the SL-ATP concentrations in the different experiments.

Supplementary Material

Supplemental Information

NIHMS1068831-supplement-Supplemental_Information.pdf^{(6.2MB, pdf)}

ACKNOWLEDGMENTS

The project described was supported by grant no. R15 GM094771-02 from the National Institute of General Medical Sciences (NIH/NIGMS) to J.G.W.. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health. The authors also wish to thank the SMU Hamilton Undergraduate Research Scholars Program, SMU Undergraduate Research Assistantships, SMU Undergraduate Engaged Learning Program, Communities Foundation of Texas, and Ms. Suzy Ruff of Dallas, Texas, and Ms. Myra Williams, Ph.D. of Naples, Florida, for private gifts. The authors thank Tristan Smyth for experimental assistance.

ABBREVIATIONS

P-gp: P-glycoprotein
SMU29: 2-[(5-cyclopropyl-1H-1,2,4-triazol-3-yl)sulfanyl]-N-[2-phenyl-5-(2,4,5-trimethylphenyl)-pyrazol-3-yl]acetamide (ZINC 08767731, CID 17555821)
NBD: the nucleotide-binding domains of P-gp
DBD: the transmembrane domains of P-gp that bind to transport substrates
DMSO: dimethylsulfoxide
ESR: electron spin resonance spectroscopy
SL-ATP: 2′,3′-(2,2,5,5,-tetramethyl-3-pyrroline-1-oxyl-3-carboxylic acid ester)
ATP: 2′,3′ indicates a rapid equilibrium of the ester bond between the 2′ and 3′ hydroxyl groups of ATP

Footnotes

Data Availability.

The datasets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request.

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.9b00966.

Description of ChemGen programs and workflow, figure of chemical spaces of variants produced, figure of binding interactions of variants with P-glycoprotein, chemical synthesis scheme, synthetic procedures, analytical data: HRMS and NMR, and NMR spectra and molecular strings (PDF)

SMILES, inhibition of P-gp ATPase, and inhibition of ATP binding to P-gp of given molecules (CSV)

Human P-glycoprotein (PDB)

The authors declare the following competing financial interest(s): A.R.L. discloses a financial stake in BioLum Sciences, LLC.

REFERENCES

(1).Ferreira RJ; dos Santos DJVA; Ferreira M-JU P-glycoprotein and membrane roles in multidrug resistance. Future Med. Chem 2015, 7, 929–946. [DOI] [PubMed] [Google Scholar]
(2).Sharom FJ ABC multidrug transporters: structure, function and role in chemoresistance. Pharmacogenomics 2008, 9, 105–127. [DOI] [PubMed] [Google Scholar]
(3).Jaramillo AC; Al Saig F; Cloos J; Jansen G; Peters GJ How to overcome ATP-binding cassette drug efflux transporter-mediated drug resistance? Cancer Drug Resist. 2018, 1, 6–29. [Google Scholar]
(4).Mao Q; Unadkat JD Role of the breast cancer resistance protein (BCRP/ABCG2) in drug transport–an update. AAPS J. 2015, 17, 65–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
(5).Zhao J Cancer stem cells and chemoresistance: The smartest survives the raid. Pharmacol. Ther 2016, 160, 145–158. [DOI] [PMC free article] [PubMed] [Google Scholar]
(6).Wijdeven RH; Pang B; Assaraf YG; Neefjes J Old drugs, novel ways out: Drug resistance toward cytotoxic chemotherapeutics. Drug Resist Updates 2016, 28, 65–81. [DOI] [PubMed] [Google Scholar]
(7).Fruci D; Cho WCS; Nobili V; Locatelli F; Alisi A Drug transporters and multiple drug resistance in the most common pediatric solid tumors. Curr. Drug Metab 2016, 17, 308–316. [DOI] [PubMed] [Google Scholar]
(8).Tamaki A; Ierano C; Szakacs G; Robey RW; Bates SE The controversial role of ABC transporters in clinical oncology. Essays Biochem. 2011, 50, 209–232. [DOI] [PMC free article] [PubMed] [Google Scholar]
(9).Palmeira A; Sousa E; Vasconcelos MH; Pinto MM Three decades of P-gp inhibitors: skimming through several generations and scaffolds. Curr. Med. Chem 2012, 19, 1946–2025. [DOI] [PubMed] [Google Scholar]
(10).Binkhathlan Z; Lavasanifar A P-glycoprotein inhibition as a therapeutic approach for overcoming multidrug resistance in cancer: current status and future perspectives. Curr. Cancer Drug Targets 2013, 13, 326–346. [DOI] [PubMed] [Google Scholar]
(11).Kathawala RJ; Gupta P; Ashby CR Jr.; Chen Z-S The modulation of ABC transporter-mediated multidrug resistance in cancer: A review of the past decade. Drug Resist. Updates 2015, 18, 1–17. [DOI] [PubMed] [Google Scholar]
(12).Gottesman MM; Pastan IH The Role of multidrug resistance efflux pumps in cancer: revisiting a JNCI publication exploring expression of the MDR1 (P-glycoprotein) gene. J. Natl. Cancer Inst 2015, 107, djv222. [DOI] [PMC free article] [PubMed] [Google Scholar]
(13).Marin JJG; Briz O; Rodríguez-Macias G; Díez-Martín JL; Macias RIR Role of drug transport and metabolism in the chemoresistance of acute myeloid leukemia. Blood Rev. 2016, 30, 55–64. [DOI] [PubMed] [Google Scholar]
(14).List AF; Kopecky KJ; Willman CL; Head DR; Persons DL; Slovak ML; Dorr R; Karanes C; Hynes HE; Doroshow JH; Shurafa M; Appelbaum FR Benefit of cyclosporine modulation of drug resistance in patients with poor-risk acute myeloid leukemia: a Southwest Oncology Group study. Blood 2001, 98, 3212–3220. [DOI] [PubMed] [Google Scholar]
(15).Li J; Jaimes KF; Aller SG Refined structures of mouse P-glycoprotein. Protein Sci. 2014, 23, 34–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
(16).Kim Y; Chen J Molecular structure of human P-glycoprotein in the ATP-bound, outward-facing conformation. Science 2018, 915. [DOI] [PubMed] [Google Scholar]
(17).Pan L; Aller SG Equilibrated atomic models of outward-facing P-glycoprotein and effect of ATP binding on structural dynamics. Sci. Rep 2015, 5, 7880. [DOI] [PMC free article] [PubMed] [Google Scholar]
(18).Jin MS; Oldham ML; Zhang Q; Chen J Crystal structure of the multidrug transporter P-glycoprotein from Caenorhabditis elegans. Nature 2012, 490, 566–569. [DOI] [PMC free article] [PubMed] [Google Scholar]
(19).Taylor NMI; Manolaridis I; Jackson SM; Kowal J; Stahlberg H; Locher KP Structure of the human multidrug transporter ABCG2. Nature 2017, 546, 504–509. [DOI] [PubMed] [Google Scholar]
(20).Johnson ZL; Chen J Structural basis of substrate recognition by the multidrug resistance protein MRP1. Cell 2017, 168, 1075–1085.e9. [DOI] [PubMed] [Google Scholar]
(21).Wise JG Catalytic transitions in the human MDR1 P-glycoprotein drug binding sites. Biochemistry 2012, 51, 5125–5141. [DOI] [PMC free article] [PubMed] [Google Scholar]
(22).McCormick JW; Vogel PD; Wise JG Multiple drug transport pathways through human P-glycoprotein. Biochemistry 2015, 54, 4374–4390. [DOI] [PMC free article] [PubMed] [Google Scholar]
(23).Frank GA; Shukla S; Rao P; Borgnia MJ; Bartesaghi A; Merk A; Mobin A; Esser L; Earl LA; Gottesman MM; Xia D; Ambudkar SV; Subramaniam S Cryo-EM analysis of the conformational landscape of human P-glycoprotein (ABCB1) during its catalytic cycle. Mol. Pharmacol 2016, 90, 35–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
(24).Khunweeraphong N; Stockner T; Kuchler K The structure of the human ABC transporter ABCG2 reveals a novel mechanism for drug extrusion. Sci. Rep 2017, 7, 13767. [DOI] [PMC free article] [PubMed] [Google Scholar]
(25).Zoghbi ME; Mok L; Swartz DJ; Singh A; Fendley GA; Urbatsch IL; Altenberg GA Substrate-induced conformational changes in the nucleotide-binding domains of lipid bilayer-associated P-glycoprotein during ATP hydrolysis. J. Biol. Chem 2017, 292, 20412–20424. [DOI] [PMC free article] [PubMed] [Google Scholar]
(26).Brewer FK; Follit CA; Vogel PD; Wise JG In silico screening for inhibitors of P-glycoprotein that target the nucleotide binding domains. Mol. Pharmacol 2014, 86, 716–726. [DOI] [PMC free article] [PubMed] [Google Scholar]
(27).Follit CA; Brewer FK; Wise JG; Vogel PD In silico identified targeted inhibitors of P-glycoprotein overcome multidrug resistance in human cancer cells in culture. Pharmacol. Res. Perspect 2015, 3, No. e00170. [DOI] [PMC free article] [PubMed] [Google Scholar]
(28).Nanayakkara AK; Follit CA; Chen G; Williams NS; Vogel PD; Wise JG Targeted inhibitors of P-glycoprotein increase chemotherapeutic-induced mortality of multidrug resistant tumor cells. Sci. Rep 2018, 8, 967. [DOI] [PMC free article] [PubMed] [Google Scholar]
(29).Cramer RD; Patterson DE; Clark RD; Soltanshahi F; Lawless MS Virtual compound libraries: a new approach to decision making in molecular discovery research. J. Chem. Inf. Comput. Sci 1998, 38, 1010–1023. [Google Scholar]
(30).Krier M; de Araújo-Júnior JX; Schmitt M; Duranton J; Justiano-Basaran H; Lugnier C; Bourguignon J-J; Rognan D Design of small-sized libraries by combinatorial assembly of linkers and functional groups to a given scaffold: application to the structure-based optimization of a phosphodiesterase 4 inhibitor. J. Med. Chem 2005, 48, 3816–3822. [DOI] [PubMed] [Google Scholar]
(31).Srinivasan R; Uttamchandani M; Yao SQ Rapid assembly and in situ screening of bidentate inhibitors of protein tyrosine phosphatases. Org. Lett 2006, 8, 713–716. [DOI] [PubMed] [Google Scholar]
(32).Durrant JD; McCammon JA AutoClickChem: click chemistry in silico. PLoS Comput. Biol 2012, 8, No. e1002397. [DOI] [PMC free article] [PubMed] [Google Scholar]
(33).Takeda M; Mizokami A; Mamiya K; Li YQ; Zhang J; Keller ET; Namiki M The establishment of two paclitaxel-resistant prostate cancer cell lines and the mechanisms of paclitaxel resistance with two cell lines. Prostate 2007, 67, 955–967. [DOI] [PubMed] [Google Scholar]
(34).Irwin JJ; Sterling T; Mysinger MM; Bolstad ES; Coleman RG ZINC: A free tool to discover chemistry for biology. J. Chem. Inf. Model 2012, 52, 1757–1768. [DOI] [PMC free article] [PubMed] [Google Scholar]
(35).Moriarty NW; Grosse-Kunstleve RW; Adams PD electronic Ligand Builder and Optimization Workbench (eLBOW): a tool for ligand coordinate and restraint generation. Acta Crystallogr., Sect. D: Biol. Crystallogr 2009, 65, 1074–1080. [DOI] [PMC free article] [PubMed] [Google Scholar]
(36).Yang Z; Son K-I; Li S; Zhou B; Xu J Specific 1,2-Hydride Shift in the Boron Trifluoride Catalyzed Reactions of Aromatic Aldehydes with Diazoacetonitrile: Simple Synthesis of β-Ketonitriles. Eur. J. Org. Chem 2014, 6380–6384. [Google Scholar]
(37).Su WN; Lin TP; Cheng KM; Sung KC; Lin SK; Wong FF An efficient one-pot synthesis of N-(1,3-diphenyl-1H-pyrazol-5-yl)amides. J. Heterocyclic Chem 2010, 47, 831–837. [Google Scholar]
(38).Hilton ST; Motherwell WB; Potier P; Pradet C; Selwood DL Observations on the reactivity of thiyl radicals derived from 3,6-epidithiodiketopiperazine-2,5-diones and related congeners. Bioorg. Med. Chem. Lett 2005, 15, 2239–2242. [DOI] [PubMed] [Google Scholar]
(39).Robertson DW; Krushinski JH; Beedle EE; Leander JD; Wong DT; Rathbun RC Structure-activity-relationships of (arylalkyl)imidazole anticonvulsants - comparison of the (fluorenylalkyl)imidazoles with nafimidone and denzimol. J. Med. Chem 1986, 29, 1577–1586. [DOI] [PubMed] [Google Scholar]
(40).van Meerloo J; Kaspers GJL; Cloos J Cell sensitivity assays: the MTT assay. Methods Mol. Biol 2011, 731, 237–245. [DOI] [PubMed] [Google Scholar]
(41).Berridge MV; Tan AS Characterization of the cellular reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT): subcellular localization, substrate dependence, and involvement of mitochondrial electron transport in MTT reduction. Arch. Biochem. Biophys 1993, 303, 474–482. [DOI] [PubMed] [Google Scholar]
(42).Kupcsik L Estimation of cell number based on metabolic activity: the MTT reduction assay. Methods Mol. Biol 2011, 740, 13–19. [DOI] [PubMed] [Google Scholar]
(43).Mannhold R; Poda GI; Ostermann C; Tetko IV Calculation of molecular lipophilicity: state-of-the-art and comparison of Log P methods on more than 96,000 compounds. J. Pharm. Sci 2009, 98, 861–893. [DOI] [PubMed] [Google Scholar]
(44).Daina A; Michielin O; Zoete V SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci. Rep 2017, 7, 42717. [DOI] [PMC free article] [PubMed] [Google Scholar]
(45).Lerner-Marmarosh N; Gimi K; Urbatsch IL; Gros P; Senior AE Large scale purification of detergent-soluble P-glycoprotein from Pichia pastoris cells and characterization of nucleotide binding properties of wild-type, Walker A, and Walker B mutant proteins. J. Biol. Chem 1999, 274, 34711–34718. [DOI] [PubMed] [Google Scholar]
(46).Tombline G; Urbatsch IL; Virk N; Muharemagic A; White LB; Senior AE Expression, purification, and characterization of cysteine-free mouse P-glycoprotein. Arch. Biochem. Biophys 2006, 445, 124–128. [DOI] [PubMed] [Google Scholar]
(47).Martin C; Berridge G; Mistry P; Higgins C; Charlton P; Callaghan R The molecular interaction of the high affinity reversal agent XR9576 with P-glycoprotein. Br. J. Pharmacol 1999, 128, 403–411. [DOI] [PMC free article] [PubMed] [Google Scholar]
(48).Streckenbach B; Schwarz D; Repke KRH Analysis of phosphoryl transfer mechanism and catalytic centre geometries of transport ATPase by means of spin-labelled ATP. Biochim. Biophys. Acta 1980, 601, 34–46. [DOI] [PubMed] [Google Scholar]
(49).Delannoy S; Urbatsch IL; Tombline G; Senior AE; Vogel PD Nucleotide binding to the multidrug resistance P-glycoprotein as studied by ESR spectroscopy. Biochemistry 2005, 44, 14010–14019. [DOI] [PubMed] [Google Scholar]
(50).Vogel PD; Nett JH; Sauer HE; Schmadel K; Cross RL; Trommer WE Nucleotide binding sites on mitochondrial F1-ATPase. Electron spin resonance spectroscopy and photolabeling by azido-spin-labeled adenine nucleotides support an adenylate kinase-like orientation. J. Biol. Chem 1992, 267, 11982–11986. [PubMed] [Google Scholar]
(51).Trott O; Olson AJ AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J. Comput. Chem 2010, 31, 455–461. [DOI] [PMC free article] [PubMed] [Google Scholar]
(52).Goodsell DS; Morris GM; Olson AJ Automated docking of flexible ligands: applications of AutoDock. J. Mol. Recognit 1996, 9, 1–5. [DOI] [PubMed] [Google Scholar]
(53).Morris GM; Goodsell DS; Huey R; Olson AJ Distributed automated docking of flexible ligands to proteins: parallel applications of AutoDock 2.4. J. Comput. Aided Mol. Des 1996, 10, 293–304. [DOI] [PubMed] [Google Scholar]
(54).Österberg F; Morris GM; Sanner MF; Olson AJ; Goodsell DS Automated docking to multiple target structures: incorporation of protein mobility and structural water heterogeneity in AutoDock. Proteins 2002, 46, 34–40. [DOI] [PubMed] [Google Scholar]
(55).Dawson RJP; Locher KP Structure of a bacterial multidrug ABC transporter. Nature 2006, 443, 180–185. [DOI] [PubMed] [Google Scholar]
(56).Humphrey W; Dalke A; Schulten K VMD: visual molecular dynamics. J. Mol. Graph 1996, 14, 33–38. [DOI] [PubMed] [Google Scholar]
(57).Varshney A; Brooks FP; Wright WV Computing smooth molecular surfaces. IEEE Comput. Graphics Appl 1994, 14, 19–25. [Google Scholar]
(58).Dolinsky TJ; Nielsen JE; McCammon JA; Baker NA PDB2PQR: an automated pipeline for the setup of Poisson–Boltzmann electrostatics calculations. Nucleic Acids Res. 2004, 32, W665–W667. [DOI] [PMC free article] [PubMed] [Google Scholar]
(59).Dolinsky TJ; Czodrowski P; Li H; Nielsen JE; Jensen JH; Klebe G; Baker NA PDB2PQR: expanding and upgrading automated preparation of biomolecular structures for molecular simulations. Nucleic Acids Res. 2007, 35, W522–W525. [DOI] [PMC free article] [PubMed] [Google Scholar]
(60).Baker NA; Sept D; Joseph S; Holst MJ; McCammon JA Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl. Acad. Sci 2001, 98, 10037–10041. [DOI] [PMC free article] [PubMed] [Google Scholar]
(61).Jurrus E; Engel D; Star K; Monson K; Brandi J; Felberg LE; Brookes DH; Wilson L; Chen J; Liles K; Chun M; Li P; Gohara DW; Dolinsky T; Konecny R; Koes DR; Nielsen JE; Head-Gordon T; Geng W; Kransky R; Wei G-E; Holst MJ; McCommon JA; Baker NA Improvements to the APBS biomolecular solvation software suite. Protein Sci. 2018, 27, 112–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
(62).Stone KR; Mickey DD; Wunderli H; Mickey GH; Paulson DF Isolation of a human prostate carcinoma cell line (DU 145). Int. J. Cancer 1978, 21, 274–281. [DOI] [PubMed] [Google Scholar]
(63).Breul SD; Bradley KH; Hance AJ; Schafer MP; Berg RA; Crystal RG Control of collagen production by human diploid lung fibroblasts. J. Biol. Chem 1980, 255, 5250–5260. [PubMed] [Google Scholar]
(64).Wu C-Y; Feng Y; Cardenas ER; Williams N; Floreancig PE; De Brabander JK; Roth MG Studies toward the unique pederin family member psymberin: structure–activity relationships, biochemical studies, and genetics identify the mode-of-action of psymberin. J. Am. Chem. Soc 2012, 134, 18998–19003. [DOI] [PMC free article] [PubMed] [Google Scholar]
(65).Urbatsch IL; Gimi K; Wilke-Mounts S; Lerner-Marmarosh N; Rousseau ME; Gros P; Senior AE Cysteines 431 and 1074 are responsible for inhibitory disulfide cross-linking between the two nucleotide-binding sites in human P-glycoprotein. J. Biol. Chem 2001, 276, 26980–26987. [DOI] [PubMed] [Google Scholar]
(66).Alvarez FJD; Orelle C; Davidson AL Functional reconstitution of an ABC transporter in nanodiscs for use in electron paramagnetic resonance spectroscopy. J. Am. Chem. Soc 2010, 132, 9513–9515. [DOI] [PMC free article] [PubMed] [Google Scholar]
(67).Bayburt TH; Sligar SG Membrane protein assembly into Nanodiscs. FEBS Lett. 2010, 584, 1721–1727. [DOI] [PMC free article] [PubMed] [Google Scholar]
(68).Vogel G; Steinhart R ATPase of Escherichia coli: purification, dissociation, and reconstitution of the active complex from the isolated subunits. Biochemistry 1976, 15, 208–216. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Information

NIHMS1068831-supplement-Supplemental_Information.pdf^{(6.2MB, pdf)}

[R1] (1).Ferreira RJ; dos Santos DJVA; Ferreira M-JU P-glycoprotein and membrane roles in multidrug resistance. Future Med. Chem 2015, 7, 929–946. [DOI] [PubMed] [Google Scholar]

[R2] (2).Sharom FJ ABC multidrug transporters: structure, function and role in chemoresistance. Pharmacogenomics 2008, 9, 105–127. [DOI] [PubMed] [Google Scholar]

[R3] (3).Jaramillo AC; Al Saig F; Cloos J; Jansen G; Peters GJ How to overcome ATP-binding cassette drug efflux transporter-mediated drug resistance? Cancer Drug Resist. 2018, 1, 6–29. [Google Scholar]

[R4] (4).Mao Q; Unadkat JD Role of the breast cancer resistance protein (BCRP/ABCG2) in drug transport–an update. AAPS J. 2015, 17, 65–82. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R5] (5).Zhao J Cancer stem cells and chemoresistance: The smartest survives the raid. Pharmacol. Ther 2016, 160, 145–158. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R6] (6).Wijdeven RH; Pang B; Assaraf YG; Neefjes J Old drugs, novel ways out: Drug resistance toward cytotoxic chemotherapeutics. Drug Resist Updates 2016, 28, 65–81. [DOI] [PubMed] [Google Scholar]

[R7] (7).Fruci D; Cho WCS; Nobili V; Locatelli F; Alisi A Drug transporters and multiple drug resistance in the most common pediatric solid tumors. Curr. Drug Metab 2016, 17, 308–316. [DOI] [PubMed] [Google Scholar]

[R8] (8).Tamaki A; Ierano C; Szakacs G; Robey RW; Bates SE The controversial role of ABC transporters in clinical oncology. Essays Biochem. 2011, 50, 209–232. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] (9).Palmeira A; Sousa E; Vasconcelos MH; Pinto MM Three decades of P-gp inhibitors: skimming through several generations and scaffolds. Curr. Med. Chem 2012, 19, 1946–2025. [DOI] [PubMed] [Google Scholar]

[R10] (10).Binkhathlan Z; Lavasanifar A P-glycoprotein inhibition as a therapeutic approach for overcoming multidrug resistance in cancer: current status and future perspectives. Curr. Cancer Drug Targets 2013, 13, 326–346. [DOI] [PubMed] [Google Scholar]

[R11] (11).Kathawala RJ; Gupta P; Ashby CR Jr.; Chen Z-S The modulation of ABC transporter-mediated multidrug resistance in cancer: A review of the past decade. Drug Resist. Updates 2015, 18, 1–17. [DOI] [PubMed] [Google Scholar]

[R12] (12).Gottesman MM; Pastan IH The Role of multidrug resistance efflux pumps in cancer: revisiting a JNCI publication exploring expression of the MDR1 (P-glycoprotein) gene. J. Natl. Cancer Inst 2015, 107, djv222. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R13] (13).Marin JJG; Briz O; Rodríguez-Macias G; Díez-Martín JL; Macias RIR Role of drug transport and metabolism in the chemoresistance of acute myeloid leukemia. Blood Rev. 2016, 30, 55–64. [DOI] [PubMed] [Google Scholar]

[R14] (14).List AF; Kopecky KJ; Willman CL; Head DR; Persons DL; Slovak ML; Dorr R; Karanes C; Hynes HE; Doroshow JH; Shurafa M; Appelbaum FR Benefit of cyclosporine modulation of drug resistance in patients with poor-risk acute myeloid leukemia: a Southwest Oncology Group study. Blood 2001, 98, 3212–3220. [DOI] [PubMed] [Google Scholar]

[R15] (15).Li J; Jaimes KF; Aller SG Refined structures of mouse P-glycoprotein. Protein Sci. 2014, 23, 34–46. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] (16).Kim Y; Chen J Molecular structure of human P-glycoprotein in the ATP-bound, outward-facing conformation. Science 2018, 915. [DOI] [PubMed] [Google Scholar]

[R17] (17).Pan L; Aller SG Equilibrated atomic models of outward-facing P-glycoprotein and effect of ATP binding on structural dynamics. Sci. Rep 2015, 5, 7880. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] (18).Jin MS; Oldham ML; Zhang Q; Chen J Crystal structure of the multidrug transporter P-glycoprotein from Caenorhabditis elegans. Nature 2012, 490, 566–569. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R19] (19).Taylor NMI; Manolaridis I; Jackson SM; Kowal J; Stahlberg H; Locher KP Structure of the human multidrug transporter ABCG2. Nature 2017, 546, 504–509. [DOI] [PubMed] [Google Scholar]

[R20] (20).Johnson ZL; Chen J Structural basis of substrate recognition by the multidrug resistance protein MRP1. Cell 2017, 168, 1075–1085.e9. [DOI] [PubMed] [Google Scholar]

[R21] (21).Wise JG Catalytic transitions in the human MDR1 P-glycoprotein drug binding sites. Biochemistry 2012, 51, 5125–5141. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R22] (22).McCormick JW; Vogel PD; Wise JG Multiple drug transport pathways through human P-glycoprotein. Biochemistry 2015, 54, 4374–4390. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R23] (23).Frank GA; Shukla S; Rao P; Borgnia MJ; Bartesaghi A; Merk A; Mobin A; Esser L; Earl LA; Gottesman MM; Xia D; Ambudkar SV; Subramaniam S Cryo-EM analysis of the conformational landscape of human P-glycoprotein (ABCB1) during its catalytic cycle. Mol. Pharmacol 2016, 90, 35–41. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R24] (24).Khunweeraphong N; Stockner T; Kuchler K The structure of the human ABC transporter ABCG2 reveals a novel mechanism for drug extrusion. Sci. Rep 2017, 7, 13767. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] (25).Zoghbi ME; Mok L; Swartz DJ; Singh A; Fendley GA; Urbatsch IL; Altenberg GA Substrate-induced conformational changes in the nucleotide-binding domains of lipid bilayer-associated P-glycoprotein during ATP hydrolysis. J. Biol. Chem 2017, 292, 20412–20424. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] (26).Brewer FK; Follit CA; Vogel PD; Wise JG In silico screening for inhibitors of P-glycoprotein that target the nucleotide binding domains. Mol. Pharmacol 2014, 86, 716–726. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] (27).Follit CA; Brewer FK; Wise JG; Vogel PD In silico identified targeted inhibitors of P-glycoprotein overcome multidrug resistance in human cancer cells in culture. Pharmacol. Res. Perspect 2015, 3, No. e00170. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] (28).Nanayakkara AK; Follit CA; Chen G; Williams NS; Vogel PD; Wise JG Targeted inhibitors of P-glycoprotein increase chemotherapeutic-induced mortality of multidrug resistant tumor cells. Sci. Rep 2018, 8, 967. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] (29).Cramer RD; Patterson DE; Clark RD; Soltanshahi F; Lawless MS Virtual compound libraries: a new approach to decision making in molecular discovery research. J. Chem. Inf. Comput. Sci 1998, 38, 1010–1023. [Google Scholar]

[R30] (30).Krier M; de Araújo-Júnior JX; Schmitt M; Duranton J; Justiano-Basaran H; Lugnier C; Bourguignon J-J; Rognan D Design of small-sized libraries by combinatorial assembly of linkers and functional groups to a given scaffold: application to the structure-based optimization of a phosphodiesterase 4 inhibitor. J. Med. Chem 2005, 48, 3816–3822. [DOI] [PubMed] [Google Scholar]

[R31] (31).Srinivasan R; Uttamchandani M; Yao SQ Rapid assembly and in situ screening of bidentate inhibitors of protein tyrosine phosphatases. Org. Lett 2006, 8, 713–716. [DOI] [PubMed] [Google Scholar]

[R32] (32).Durrant JD; McCammon JA AutoClickChem: click chemistry in silico. PLoS Comput. Biol 2012, 8, No. e1002397. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] (33).Takeda M; Mizokami A; Mamiya K; Li YQ; Zhang J; Keller ET; Namiki M The establishment of two paclitaxel-resistant prostate cancer cell lines and the mechanisms of paclitaxel resistance with two cell lines. Prostate 2007, 67, 955–967. [DOI] [PubMed] [Google Scholar]

[R34] (34).Irwin JJ; Sterling T; Mysinger MM; Bolstad ES; Coleman RG ZINC: A free tool to discover chemistry for biology. J. Chem. Inf. Model 2012, 52, 1757–1768. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] (35).Moriarty NW; Grosse-Kunstleve RW; Adams PD electronic Ligand Builder and Optimization Workbench (eLBOW): a tool for ligand coordinate and restraint generation. Acta Crystallogr., Sect. D: Biol. Crystallogr 2009, 65, 1074–1080. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] (36).Yang Z; Son K-I; Li S; Zhou B; Xu J Specific 1,2-Hydride Shift in the Boron Trifluoride Catalyzed Reactions of Aromatic Aldehydes with Diazoacetonitrile: Simple Synthesis of β-Ketonitriles. Eur. J. Org. Chem 2014, 6380–6384. [Google Scholar]

[R37] (37).Su WN; Lin TP; Cheng KM; Sung KC; Lin SK; Wong FF An efficient one-pot synthesis of N-(1,3-diphenyl-1H-pyrazol-5-yl)amides. J. Heterocyclic Chem 2010, 47, 831–837. [Google Scholar]

[R38] (38).Hilton ST; Motherwell WB; Potier P; Pradet C; Selwood DL Observations on the reactivity of thiyl radicals derived from 3,6-epidithiodiketopiperazine-2,5-diones and related congeners. Bioorg. Med. Chem. Lett 2005, 15, 2239–2242. [DOI] [PubMed] [Google Scholar]

[R39] (39).Robertson DW; Krushinski JH; Beedle EE; Leander JD; Wong DT; Rathbun RC Structure-activity-relationships of (arylalkyl)imidazole anticonvulsants - comparison of the (fluorenylalkyl)imidazoles with nafimidone and denzimol. J. Med. Chem 1986, 29, 1577–1586. [DOI] [PubMed] [Google Scholar]

[R40] (40).van Meerloo J; Kaspers GJL; Cloos J Cell sensitivity assays: the MTT assay. Methods Mol. Biol 2011, 731, 237–245. [DOI] [PubMed] [Google Scholar]

[R41] (41).Berridge MV; Tan AS Characterization of the cellular reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT): subcellular localization, substrate dependence, and involvement of mitochondrial electron transport in MTT reduction. Arch. Biochem. Biophys 1993, 303, 474–482. [DOI] [PubMed] [Google Scholar]

[R42] (42).Kupcsik L Estimation of cell number based on metabolic activity: the MTT reduction assay. Methods Mol. Biol 2011, 740, 13–19. [DOI] [PubMed] [Google Scholar]

[R43] (43).Mannhold R; Poda GI; Ostermann C; Tetko IV Calculation of molecular lipophilicity: state-of-the-art and comparison of Log P methods on more than 96,000 compounds. J. Pharm. Sci 2009, 98, 861–893. [DOI] [PubMed] [Google Scholar]

[R44] (44).Daina A; Michielin O; Zoete V SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci. Rep 2017, 7, 42717. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] (45).Lerner-Marmarosh N; Gimi K; Urbatsch IL; Gros P; Senior AE Large scale purification of detergent-soluble P-glycoprotein from Pichia pastoris cells and characterization of nucleotide binding properties of wild-type, Walker A, and Walker B mutant proteins. J. Biol. Chem 1999, 274, 34711–34718. [DOI] [PubMed] [Google Scholar]

[R46] (46).Tombline G; Urbatsch IL; Virk N; Muharemagic A; White LB; Senior AE Expression, purification, and characterization of cysteine-free mouse P-glycoprotein. Arch. Biochem. Biophys 2006, 445, 124–128. [DOI] [PubMed] [Google Scholar]

[R47] (47).Martin C; Berridge G; Mistry P; Higgins C; Charlton P; Callaghan R The molecular interaction of the high affinity reversal agent XR9576 with P-glycoprotein. Br. J. Pharmacol 1999, 128, 403–411. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] (48).Streckenbach B; Schwarz D; Repke KRH Analysis of phosphoryl transfer mechanism and catalytic centre geometries of transport ATPase by means of spin-labelled ATP. Biochim. Biophys. Acta 1980, 601, 34–46. [DOI] [PubMed] [Google Scholar]

[R49] (49).Delannoy S; Urbatsch IL; Tombline G; Senior AE; Vogel PD Nucleotide binding to the multidrug resistance P-glycoprotein as studied by ESR spectroscopy. Biochemistry 2005, 44, 14010–14019. [DOI] [PubMed] [Google Scholar]

[R50] (50).Vogel PD; Nett JH; Sauer HE; Schmadel K; Cross RL; Trommer WE Nucleotide binding sites on mitochondrial F1-ATPase. Electron spin resonance spectroscopy and photolabeling by azido-spin-labeled adenine nucleotides support an adenylate kinase-like orientation. J. Biol. Chem 1992, 267, 11982–11986. [PubMed] [Google Scholar]

[R51] (51).Trott O; Olson AJ AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J. Comput. Chem 2010, 31, 455–461. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R52] (52).Goodsell DS; Morris GM; Olson AJ Automated docking of flexible ligands: applications of AutoDock. J. Mol. Recognit 1996, 9, 1–5. [DOI] [PubMed] [Google Scholar]

[R53] (53).Morris GM; Goodsell DS; Huey R; Olson AJ Distributed automated docking of flexible ligands to proteins: parallel applications of AutoDock 2.4. J. Comput. Aided Mol. Des 1996, 10, 293–304. [DOI] [PubMed] [Google Scholar]

[R54] (54).Österberg F; Morris GM; Sanner MF; Olson AJ; Goodsell DS Automated docking to multiple target structures: incorporation of protein mobility and structural water heterogeneity in AutoDock. Proteins 2002, 46, 34–40. [DOI] [PubMed] [Google Scholar]

[R55] (55).Dawson RJP; Locher KP Structure of a bacterial multidrug ABC transporter. Nature 2006, 443, 180–185. [DOI] [PubMed] [Google Scholar]

[R56] (56).Humphrey W; Dalke A; Schulten K VMD: visual molecular dynamics. J. Mol. Graph 1996, 14, 33–38. [DOI] [PubMed] [Google Scholar]

[R57] (57).Varshney A; Brooks FP; Wright WV Computing smooth molecular surfaces. IEEE Comput. Graphics Appl 1994, 14, 19–25. [Google Scholar]

[R58] (58).Dolinsky TJ; Nielsen JE; McCammon JA; Baker NA PDB2PQR: an automated pipeline for the setup of Poisson–Boltzmann electrostatics calculations. Nucleic Acids Res. 2004, 32, W665–W667. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R59] (59).Dolinsky TJ; Czodrowski P; Li H; Nielsen JE; Jensen JH; Klebe G; Baker NA PDB2PQR: expanding and upgrading automated preparation of biomolecular structures for molecular simulations. Nucleic Acids Res. 2007, 35, W522–W525. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R60] (60).Baker NA; Sept D; Joseph S; Holst MJ; McCammon JA Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl. Acad. Sci 2001, 98, 10037–10041. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R61] (61).Jurrus E; Engel D; Star K; Monson K; Brandi J; Felberg LE; Brookes DH; Wilson L; Chen J; Liles K; Chun M; Li P; Gohara DW; Dolinsky T; Konecny R; Koes DR; Nielsen JE; Head-Gordon T; Geng W; Kransky R; Wei G-E; Holst MJ; McCommon JA; Baker NA Improvements to the APBS biomolecular solvation software suite. Protein Sci. 2018, 27, 112–128. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R62] (62).Stone KR; Mickey DD; Wunderli H; Mickey GH; Paulson DF Isolation of a human prostate carcinoma cell line (DU 145). Int. J. Cancer 1978, 21, 274–281. [DOI] [PubMed] [Google Scholar]

[R63] (63).Breul SD; Bradley KH; Hance AJ; Schafer MP; Berg RA; Crystal RG Control of collagen production by human diploid lung fibroblasts. J. Biol. Chem 1980, 255, 5250–5260. [PubMed] [Google Scholar]

[R64] (64).Wu C-Y; Feng Y; Cardenas ER; Williams N; Floreancig PE; De Brabander JK; Roth MG Studies toward the unique pederin family member psymberin: structure–activity relationships, biochemical studies, and genetics identify the mode-of-action of psymberin. J. Am. Chem. Soc 2012, 134, 18998–19003. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R65] (65).Urbatsch IL; Gimi K; Wilke-Mounts S; Lerner-Marmarosh N; Rousseau ME; Gros P; Senior AE Cysteines 431 and 1074 are responsible for inhibitory disulfide cross-linking between the two nucleotide-binding sites in human P-glycoprotein. J. Biol. Chem 2001, 276, 26980–26987. [DOI] [PubMed] [Google Scholar]

[R66] (66).Alvarez FJD; Orelle C; Davidson AL Functional reconstitution of an ABC transporter in nanodiscs for use in electron paramagnetic resonance spectroscopy. J. Am. Chem. Soc 2010, 132, 9513–9515. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R67] (67).Bayburt TH; Sligar SG Membrane protein assembly into Nanodiscs. FEBS Lett. 2010, 584, 1721–1727. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R68] (68).Vogel G; Steinhart R ATPase of Escherichia coli: purification, dissociation, and reconstitution of the active complex from the isolated subunits. Biochemistry 1976, 15, 208–216. [DOI] [PubMed] [Google Scholar]

PERMALINK

Optimizing Targeted Inhibitors of P-Glycoprotein Using Computational and Structure-Guided Approaches

John G Wise

Amila K Nanayakkara

Maha Aljowni

Gang Chen

Maisa C De Oliveira

Lauren Ammerman

Ketetha Olengue

Alexander R Lippert

Pia D Vogel

Abstract

Graphical Abstract

INTRODUCTION

Figure 1.

RESULTS

Virtual Synthesis of Novel Variants of the “Western” Half of SMU29 Using the ChemGen Computational Suite.

Figure 2.

Docking to P-Glycoprotein and Chemical Synthesis of Compound 29-Variants.

Table 1.

Figure 3.

Group 1 ChemGen/Docking Routine-Produced 29-Variants Resensitize a P-gp-Overexpressing Prostate Cancer Cell Line, DU145TXR, to Paclitaxel.

Figure 4.

Accumulation and Cellular Retention of Calcein AM in DU145TXR Cells upon Incubation with Group 1 SMU29 Variants.

Figure 5.

Assessing the Roles of Polarity and Size of 29-Variants in Improving Efficacy of Compound 29 Variants.

Figure 6.

Table 3.

Effects of Group 2 Variants on the Paclitaxel Sensitivity of DU145TXR Cancer Cells.

Figure 7.

Table 4.

Accumulation and Cellular Retention of Calcein AM in DU145TXR Cells in the Presence of 29-Variants from Group 2.

Figure 8.

Evaluation of the Mode of Inhibition of P-Glyco-protein by Group 1 and Group 2 Variants of Compound 29.

Effects of Compound 29 Variants on Verapamil-Stimulated ATP Hydrolysis by P-gp.

Table 5.

Effects on “Basal” ATP Hydrolysis Rates of Compound 29 Variants.

Intracellular Accumulation of Compound 29 Variants.

Table 6.

Effects of 29-Variants on Binding of an ATP Analog to Purified P-Glycoprotein.

DISCUSSION AND CONCLUSIONS

Using Computational Approaches To Create Novel Variants of “Hit” Molecules from Drug Discovery Programs.

Figure 9.

Table 2.

Efficacy of Rationally Designed Group 2 Variants in Reversing MDR Caused by P-gp Overexpression.

Effectiveness of the Selective Docking Routines Used in Group 1 Variants To Ensure Targeting to the Nucleotide-Binding Domains and Avoiding the Drug Transport Domains of P-gp.

Mechanism of Inhibition of the 29-Variants.

EXPERIMENTAL SECTION

Virtual Synthesis of Compound 29-Derivatives Using ChemGen.

In Silico Docking of Compound 29 Variants to a Model of Human P-gp.

Calculation of Physicochemical Properties.

Imaging of P-Glycoprotein and Ligands.

Synthetic Procedures.

Cell Lines and Cell Culture.

MTT Cell Viability Assay.

Calcein AM Assay.

Cellular Accumulation Assays for Experimental P-gp Inhibitors.

P-Glycoprotein Purification.

Nanodisc Assembly.

ATPase Activity Assays.

ESR Measurements.

Supplementary Material

ACKNOWLEDGMENTS

ABBREVIATIONS

Footnotes

REFERENCES

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases