Journal of Pesticide Science
Online ISSN : 1349-0923
Print ISSN : 1348-589X
ISSN-L : 0385-1559
Original Articles
Substrate recognition by P-glycoprotein efflux transporters: Structure-ATPase activity relationship of diverse chemicals and agrochemicals
Satoshi Kanaoka, Yasuhisa Kimura, Masaaki Fujikawa, Yoshiaki Nakagawa, Kazumitsu Ueda, Miki Akamatsu
Author information
Keywords: P-glycoprotein, ATPase activity, structure-activity, CoMFA, agrochemical, dibenzoylhydrazine
JOURNAL FREE ACCESS FULL-TEXT HTML

2013 Volume 38 Issue 3 Pages 112-122

J-STAGE : jpestics 39 1 pp.14-21
Browse “Advance online publication” version
Details
Download PDF (3426K)
Download citation RIS

(compatible with EndNote, Reference Manager, ProCite, RefWorks)

BIB TEX

(compatible with BibDesk, LaTeX)

Text
How to download citation
Contact us

Introduction

Humans are exposed to a variety of chemicals and agrochemicals in the form of residues in food and water or in occupational use on a daily basis. Metabolism and efflux systems have evolved to protect humans from the xenobiotics absorbed in the body. P-glycoprotein (P-gp) (also known as MDR1 or ABCB1) is a member of the ATP-binding cassette (ABC) transporter family. Human P-gp is a 1280-amino acid polypeptide consisting of two transmembrane domains and two nucleotide-binding domains that actively transports a wide variety of drugs and xenobiotics out of cells and functions as an energy-dependent efflux pump.1–3) Since P-gp recognizes various compounds as substrates, it plays an important role in multidrug resistance in the treatment of cancers. In addition, from the point of view of early absorption, distribution, metabolism, and excretion (ADME) prediction of therapeutic agents, various computational prediction models have been reported for ligand interactions with P-gp.2,4–10) In general, a P-gp substrate seems to be lipophilic or amphiphilic, having a large size or molecular volume, electronegative groups and hydrogen bonding groups.11) Although X-ray structures of the complex of mouse P-gp with enantiomeric cyclic peptide inhibitors have been reported previously by Aller et al. in 2009,12) the mechanism of P-gp substrate/inhibitor recognition is complicated and still poorly understood.

Regarding the interaction of agrochemicals with P-gp, diverse agrochemicals have been evaluated for their potential to bind to human P-gp and several compounds have been shown to inhibit P-gp transport function in intact cells.13) An organochlorine insecticide, 1,1-bis(4-chlorophenyl)-2,2,2-trichloroethane (DDT), was suggested to be a weak P-gp inhibitor.14) Sreeramulu et al. also studied the effects of several commonly used insecticides on mammalian P-gp.15) However, an evaluation of diverse agrochemicals and their derivatives as human P-gp substrates has not yet been performed. Predicting the excretion levels of lipophilic or hydrophobic agrochemicals from human bodies may be useful for a risk assessment of these chemicals.

In this study, we screened diverse chemicals, especially agrochemicals by measuring the ATPase activity of human P-gp. Since the transport of chemicals by P-gp is dependent on ATP hydrolysis and most substrates transported by P-gp stimulate ATP hydrolysis, ATPase activity was used as an index to determine whether the chemical was a good substrate of P-gp. Several classes of chemicals including dibenzoylhydrazine (DBH), which is known to be an insect growth regulator and binds to the ecdysone receptor (EcR) of insects (Fig. 1),16) were found to be potential substrates of P-gp. The ATPase activity caused by DBH derivatives was quantitatively analyzed using a 3D quantitative structure-activity relationship, comparative molecular field analysis (CoMFA),17) and the favorable and unfavorable properties of ligands for the ATPase activity were clarified. A docking study of the DBH derivative having high activity, methoxyfenozide was also attempted to propose the binding site of DBHs.

Fig. 1. The common skeletal chain used for alignment drawn in bold lines with main DBH structures.

Materials and Methods

1. Materials

Chemicals, agrochemicals, and reagents of analytical grade were purchased from Nacalai Tesque (Kyoto, Japan) or Wako Pure Chemical Industries, Ltd. (Osaka, Japan). L-α-Lecithin from soybean, ATP, ADP, and AMP were purchased from Sigma-Aldrich (MO, USA).

All DBH derivatives were synthesized and reported in previous studies.18,19)

2. P-gp ATPase activity

Human P-glycoprotein expressed by Sf9 insect cells was purified according to a previous report.20)

For lipid stocks, lecithin was solubilized in 40 mM Tris–HCl, pH 7.4, and 0.1 mM EGTA at a concentration of 20 mg/mL by sonicating in a bath sonicator (Bioruptor UCD-200 TM; TOSHO DENKI; Yamagata, Japan) until the suspension clarified. Lipid stocks were stored at 4°C. Purified P-gp 5 µg and lipid stocks were mixed at a protein/lipid ratio of 1/6 (w/w) in 800 µL of 40 mM Tris–HCl, pH 7.4, 0.1 mM EGTA, and 0.5 mM dithiothreitol at 23°C for 20–40 min and then sonicated at 4°C for 20 sec in a bath sonicator. Reconstituted protein 800 µL was mixed with 200 µL of 40 mM Tris–HCl, pH 7.4, 0.1 mM EGTA, 10 mM ATP and 10 mM MgCl2. Ten microliters of the protein solution and 10 µL of a test compound in the reaction buffer (40 mM Tris–HCl, pH 7.4 and 0.1 mM EGTA) containing less than or equal to 5% dimethylsulfoxide (final concentration ≤500 µM depending on the solubility) were mixed and the mixture was reacted at 37°C for 30 min. The concentration of dimethylsulfoxide (≤5%) did not affect the reaction. The reaction was stopped by the addition of 10 µL of 10 mM sodium ortho-vanadate solution. After the addition of 30 µL of the reaction buffer (40 mM Tris–HCl, pH 7.4 and 0.1 mM EGTA), a 10 µL sample was injected to HPLC. The concentration of ADP was measured with HPLC and ATPase activity (nmol/min/mg protein) was then calculated.

HPLC condition

Column: TSKgel ODS-100V 3 µM (4.6×150 mm)

Temperature: 40°C

Solvent: 100 mM K2HPO4–KH2PO4 (pH 5.5)

Flow rate: 1.0 mL/min (isocratic)

UV detection: 260 nm

ATPase activity was evaluated in terms of relative activity (%) to that of a potent P-gp substrate, verapamil.20)

3. Molecular modeling and CoMFA

All computations were done with the molecular modeling software package SYBYL ver. 7.3 (Sybyl Molecular Modeling Software; Tripos Associates, Inc.; MO, USA). The X-ray crystallographic structure of the unsubstituted DBH derivative21) was used as an initial structure. It was modified and the DBH-type insecticide tebufenozide (Fig. 1) was constructed, energy minimized by Tripos force fields and then used as a template. The structure of the other derivatives was constructed by modifying the structure of tebufenozide. The electrostatic potential (esp) charges of MNDO were calculated for the energy minimized structure of compounds. Compounds were automatically aligned with the common skeletal chain using the SYBYL module, Align Database. The common skeletal chain used for alignment was drawn in bold lines in Fig. 1 with the main DBH structures.

CoMFA was performed for logit-transformed activity {logit A=log [activity(%)/(100−activity(%))]}. Analyses were conducted with the “Advanced CoMFA” module of SYBYL. The procedure was similar to that described in our previous report.22) Superimposed sets of stable conformers were placed in a lattice of 23.9 Å×23.2 Å×22.8 Å (X=−13.9 to 10.0, Y=−10.3 to 12.9, Z=−8.7 to 14.1) with 2 Å spaces automatically generated by the CoMFA routine in SYBYL. The potential energy fields of each stable conformer were calculated at the lattice intersections. To calculate the Coulombic electrostatic potential at each lattice point, the charge of +1.0 as a probe and the atomic charges for each of the molecules were used. The steric interaction (Lennard–Jones) potential at the lattice points was calculated using the sp3-carbon atom as a probe. The data matrix was analyzed by the partial least squares method.23) The results of the analysis were expressed as correlation equations with the number of latent variable terms, each of which was a linear combination of original independent lattice variables. In order to show favorable and unfavorable potential regions, variables were displayed as contour diagrams of coefficients of the corresponding field descriptor terms at each lattice intersection. We initially selected the number of compounds in the set as the number of the cross-validation (the leave-one-out method) and then performed an analysis using the optimum number of latent variables deduced from the cross-validation tests without actual cross-validation.

4. Protein modeling and docking

Modeling of human P-gp was carried out using the homology modeling software PDFAMS pro ver. 2.0 (Protein Discovery Full Automatic Modeling System; In-Silico Sciences, Inc.; Tokyo, Japan), originally developed by Ogata and Umeyama24) and Sybyl. Three nucleotide-free structures of mouse P-gp were available as template proteins; a protein without a ligand (PDB code: 3G5U) and complexes with two enantiomeric cyclic peptide P-gp inhibitors (PDB code: 3G60 with QZ59-RRR and 3G61 with QZ59-SSS). Two complexes (Chain A) were selected considering the determination of the binding-site cavity for docking. The primary sequence of human P-gp and its alignment with mouse P-gp by Aller et al.12) (Supplementary Material, Fig. S1 of ref. 12) were used (sequence alignment similarity 93%) in this study. The two constructed P-gp models (model 1 from 3G60 and model 2 from 3G61) were energy minimized for 5,000 iterations of conjugated gradients using the force field and partial charges of the molecular mechanics MMFF94.25,26) The coordinates of backbone atoms were fixed during the energy minimization.

GOLD ver. 5.1 (Cambridge Crystallographic Data Centre (CCDC) Software Ltd., Cambridge, UK) was used to dock the conformer of methoxyfenozide having the highest activity, which was constructed for CoMFA in the binding-site cavity of the human P-gp models. The original ligands, QZ59-RRR and QZ59-SSS, at the same coordinates as those of the protein–ligand complexes, Chain A, were merged with protein models 1 and 2, respectively. All protein atoms within 6.0 Å of each ligand (QZ59-RRR for model 1 and QZ59-SSS for model 2) were selected and a cavity detection algorithm was used for selected atoms to restrict the region of the binding-site cavity. Methoxyfenozide was docked into each binding site. A Genetic Algorithm (GA) was applied for GOLD docking and automatic GA settings were used for all calculations. Operator weights for the crossover, mutation, migration (95, 95, and 10, respectively), hydrogen bonding (2.5 Å), and van der Waals (4.0 Å) parameters were set as default values throughout the docking.

Results

1. ATPase activity

We screened steroids [2 sex hormones (estradiol and testosterone), 1 progestogen (progesterone), and 2 ecdysteroids which are arthropod molting hormones (ecdysone and 20-hydroxyecdysone)], bisphenol A, abamectin (insecticide: mixture of avermectin B1a (>99%) and B1b), 1 neonicotinoid (insecticide: imidacloprid), 1 triazine (herbicide: atrazine), 1 thiocarbamate (herbicide: thiobencarb), 1 organochlorine (insecticide: methoxychlor), 2 pyrethroids (insecticides: cypermethrin and fenvalerate), 11 organophosphates [9 insecticides (chlorpyrifos, diazinon, dichlorvos, fenitrothion, fenthion, methidathion, phenthoate, salithion, and trichlorfon) and 2 fungicides (iprobenfos and edifenphos)], and 4 DBHs (insecticides: chromafenozide, halofenozide, tebufenozide, and methoxyfenozide). The relative activity values of these compounds are shown in Table 1. Thirty percent relative activity to that of verapamil was temporarily set as the standard of highly active compounds.

Table 1. Relative ATPase activities caused by chemicals
Chemical class Compounds Conc. (μM) Relative activity (%) SD
Steroids estradiol 250 2.7 0.4
testosterone 250 12.2 0.7
progesterone 250 24.1 4.5
ecdysone 165 0.9 0.3
20-hydroxyecdysone 165 5.8 2.6
Pyrethroids cypermethrin 500 6.7 0.5
fenvalerate 500 6.2 0.2
Organophosphates chlorpyrifos 250 3.8 0.7
diazinon 250 29.6 11.0
dichlorvos 250 8.0 2.8
edifenphos 250 11.8 0.5
fenitrothion 250 14.1 1.0
fenthion 250 10.1 1.0
iprobenfos 250 31.4 0.3
methidathion 250 30.6 1.1
phenthonate 250 15.2 2.8
salithion 250 8.6 0.7
trichlorfon 250 7.4 0.2
DBHs chromafenozide 63 25.9 0.5
halofenozide 125 9.7 0.8
tebufenozide 125 31.9 12.0
methoxyfenozide 125 40.2 2.2
Organochlorine methoxychlor 250 1.1 1.7
Phenol bisphenol A 250 1.9 0.6
Abamectin avermectin B1a, B1b 500 5.3 0.4
Neonicotinoid imidacloprid 500 13.9 0.5
Triazine atrazine 500 15.8 7.8
Thiocarbamate thiobencarb 500 17.3 1.2

Of the steroids, progesterone derived marginal activity, 24.1% at 250 µM. No or low activities were measured by the other steroids including the ecdysteroids (≤12.2% at 165–250 µM). The activities of agrochemicals by avermectin, imidacloprid, methoxychlor, cypermethrin, fenvalerate, atrazine, and thiobencarb were not high. The highest activity of those derived by these compounds was 17.3% by thiobencarb (500 µM). Bisphenol A, which was suspected as an endocrine disruptor, also caused almost no activity. Tebufenozide and methoxyfenozide of the DBH derivatives, as well as diazinon, iprobenfos and methidathion of the organophosphates, derive relatively high ATPase activities.

Since several organophosphates derived relatively high ATPase activities, the concentration-dependent ATPase activities produced by 11 organophosphates were measured at 25, 125, and 250 µM. The results are shown in Fig. 2. ATPase activities produced by diazinon, iprobenfos, and methidathion increased with increases in concentration, showing a possibility that these compounds are P-gp substrates. Other organophosphates derived no or very low activity.

Fig. 2. Concentration-dependent ATPase activity of organophosphates.

The ATPase activities produced by the A-ring: 3,5-dimethyl and 2-Cl derivatives and non-substituted B-ring derivatives of DBHs were further measured at 125 µM. Relative activity values are listed in Table 2. The 3-Me (compound 3), 3-OMe (5), and 3-OEt (6) compounds of the A-ring: 3,5-dimethyl derivatives as well as tebufenozide (4-Et; 19) and methoxyfenozide (2-Me, 3-OMe; 20), produced fairly high activities (more than 30%). Although chromafenozide (21) produced marginal activity (25.9%) at 63 µM, it was not tested at higher concentrations because of its low solubility. The 3-Br (35), 3-NO2 (38), 2,5-Me2 (59), 2,5-Cl2 (60) and 3,5-Me2 (66) compounds of the A-ring: 2-Cl derivatives produced activities at around 30%. Only the activity by the 2,4-Cl2 compound (104) in the non-substituted B-ring series was over 30%. In general, the activities produced by the three series were in the order of the A-ring: 3,5-dimethyl>A-ring: 2-Cl>non-substituted B-ring derivatives.

Concentration-dependent ATPase activities produced by several A-ring: 3,5-dimethyl derivatives were also evaluated. The results were drawn with that of verapamil (Km=22 µM) in Fig. 3. The activities produced by all tested compounds showed a concentration dependency; however it was difficult to estimate the kinetic properties (Km and Vmax) of the compounds because of the low solubility.

Fig. 3. Concentration-dependent ATPase activity of DBH A-ring: 3,5-dimethyl derivatives and verapamil.
Table 2. ATPase activities and log P of DBH derivatives
No. Compound Relative Activity (%)a SDb logit Ac log Pf
Obsd Calcd
Eq. 1, 2d Eq. 3e
A-ring: 3,5-dimethyl derivatives
1 B: H 24.0 0.4 −0.50 −0.86 −1.12 3.39
2 2-Me 9.6 2.2 −0.97 −1.04 −0.97 3.98
3 3-Me 46.8 1.6 −0.06 −0.43 −0.59 3.98
4 3-OH 22.8 0.8 −0.53 −0.44 −0.61 3.34
5 3-OMe 54.6 1.2 0.08 0.17 −0.19 3.75
6 3-OEt 45.3 1.5 −0.08 0.03 −0.28 4.28
7 4-Me 2.7 0.4 −1.56 −1.16 −1.26 3.95
8 4-nPr 15.3 0.4 −0.74 −0.76 −0.80 4.86
9 4-iPr 14.3 0.7 −0.78 −0.74 −0.76 4.91
10 4-nBu 10.3 0.4 −0.94 −0.75 −0.86 5.39
11 4-nPent 6.3 0.2 −1.17 −1.10 −1.15 5.92
12 4-Cl 4.2 0.2 −1.36 −1.15 −1.19 4.31
13 4-CF3 5.8 0.5 −1.21 −1.41 −1.20 4.66
14 2,3-Me2 4.9 0.2 −1.29 −0.88 −0.78 4.43
15 2-Me,3-OH 24.0 0.5 −0.50 −0.54 −0.52 3.84
16 2,3,4-Me3 5.3 0.4 −1.25 −1.04 −0.86 4.88l
17 2,3,4-F3 3.6 0.0 −1.43 −1.33 −1.25 3.64
18 2,4,5-F3 4.6 0.6 −1.32 −1.46 −1.45 3.71
19 tebufenozide(B:4-Et) 31.9 12.0 −0.33 −0.96 −0.93 4.25
20 methoxyfenozide(B: 2-Me, 3-OMe) 40.2 2.2 −0.17 −0.25 −0.31 3.70
21 chromafenozideg 25.9 0.5 −0.46 — — 2.70
A-ring: 2-Cl derivatives
22 B: H 5.1 0.2 −1.27 −1.25 −0.95 2.59
23 2-F 7.4 1.9 −1.10 −1.08 −1.11 2.63
24 2-Cl 5.2 0.3 −1.26 −1.11 −0.99 2.75
25 2-Br 8.7 0.2 −1.02 −1.10 −0.89 2.91
26 2-Ih 12.4 0.9 −0.85 — — 3.11
27 2-CF3 6.2 0.6 −1.18 −1.17 −1.22 3.02
28 2-NO2 7.7 0.2 −1.08 −1.08 −1.29 1.99
29 2-Ph 16.2 0.8 −0.71 −0.65 −0.92 3.77
30 2-OMe 17.3 0.1 −0.68 −0.69 −0.76 2.37
31 2-SMe 17.4 0.3 −0.68 −0.69 −0.76 2.84
32 3-Me 16.8 1.0 −0.69 −0.68 −0.52 3.11
33 3-F 11.9 1.3 −0.87 −0.94 −0.80 2.88
34 3-Cl 24.1 0.4 −0.50 −0.48 −0.58 3.49
35 3-Br 28.5 1.2 −0.40 −0.46 −0.53 3.62
36 3-Ih 17.0 0.7 −0.69 — — 3.87
37 3-CF3 25.0 0.5 −0.48 −0.52 −0.25 3.70
38 3-NO2 28.7 1.5 −0.40 −0.38 −0.40 2.73
39 3-CN 21.0 0.3 −0.58 −0.58 −0.47 2.48
40 3-OMe 15.6 0.2 −0.73 −0.71 −0.29 2.81
41 4-Me 10.6 1.0 −0.93 −1.07 −1.04 3.15
42 4-Et 9.5 0.5 −0.98 −0.99 −0.93 3.59
43 4-nPr 10.0 1.1 −0.95 −0.98 −0.84 4.06
44 4-iPr 8.7 0.2 −1.02 −1.00 −0.73 4.11
45 4-F 9.5 0.9 −0.98 −0.96 −0.91 2.87
46 4-Cl 6.7 0.8 −1.14 −1.12 −1.02 3.51
47 4-Br 5.2 0.4 −1.26 −1.18 −1.03 3.73
48 4-Ih 4.8 0.4 −1.30 — — 3.96
49 4-CF3 12.5 0.2 −0.85 −0.83 −0.83 3.68
50 4-CN 15.2 4.8 −0.75 −0.75 −0.97 2.44
51 4-Phi 2.1 0.4 −1.67 −1.64 — 4.49
52 4-OMe 15.8 0.5 −0.73 −0.73 −0.83 2.82
53 4-COMe 19.5 0.6 −0.62 −0.57 −0.71 2.42
54 4-SO2Me 9.5 0.1 −0.98 −0.99 −0.77 1.46
55 2,3-Me2 9.2 0.5 −0.99 −1.07 −0.62 3.40
56 2,3-Cl2 20.9 0.1 −0.58 −0.53 −0.63 3.75
57 2-Me,3-OMe 20.2 0.4 −0.60 −0.61 −0.48 3.46
58 2,4-Cl2 7.7 0.5 −1.08 −1.07 −1.14 3.71
59 2,5-Me2 32.7 0.7 −0.31 −0.33 −0.69 3.40
60 2,5-Cl2 31.1 1.5 −0.35 −0.42 −0.90 3.75
61 2,6-F2 5.2 0.1 −1.26 −1.23 −1.19 2.35
62 2,6-Cl2 5.7 0.4 −1.22 −1.22 −0.92 3.03
63 2-F,6-Cl 6.9 0.4 −1.13 −1.19 −1.16 2.67
64 3,4-Me2j 10.4 0.3 −0.94 — — 3.65
65 3,4-Cl2 26.1 0.6 −0.45 −0.51 −0.69 4.25
66 3,5-Me2 31.2 3.3 −0.34 −0.25 −0.40 3.67
67 3,5-Cl2k 15.1 1.1 −0.75 — — 4.29
Non-substituted B-ring derivatives
68 A: 2-Me 4.8 0.1 −1.30 — — 2.75
69 2-Et 9.3 0.6 −0.99 — — 2.96
70 2-F 4.0 0.2 −1.38 — — 2.38
72 2-Br 6.9 0.4 −1.13 — — 2.69
73 2-Ih 11.2 0.3 −0.90 — — 2.83
74 2-CF3 10.7 0.1 −0.92 — — 2.85
75 2-NO2 5.4 0.5 −1.24 — — 2.27
76 2-Ph 6.7 0.3 −1.14 — — 3.89
77 2-OBz 19.4 0.5 −0.62 — — 3.66
78 2-O-sBu 13.0 6.9 −0.83 — — 3.18
79 2-SMe 5.8 0.1 −1.21 — — 2.60
80 3-Me 9.4 0.5 −0.98 — — 2.79
81 3-F 5.9 0.4 −1.20 — — 2.78
82 3-Cl 3.2 0.6 −1.48 — — 3.28
83 3-Br 2.1 0.1 −1.67 — — 3.49
84 3-Ih 0.0 0.0 — — — 3.72
85 3-CF3 0.5 0.4 −2.30 — — 3.61
86 3-NO2 3.7 0.4 −1.42 — — 2.73
87 3-CN 18.3 1.1 −0.65 — — 2.34
88 3-OMe 10.1 0.7 −0.95 — — 2.56m
89 4-Me 10.8 0.6 −0.92 — — 2.99
90 4-tBu 18.3 0.3 −0.65 — — 4.11
91 4-F 4.4 0.3 −1.34 — — 2.85
92 4-Cl 2.4 0.3 −1.61 — — 3.42
93 4-Br 3.7 0.3 −1.42 — — 3.66
94 4-Ih 3.8 0.3 −1.40 — — 3.78
95 4-NO2 3.0 0.0 −1.51 — — 2.63
96 4-CN 12.7 1.8 −0.84 — — 2.50
97 4-Ph 20.8 1.0 −0.58 — — 4.24
98 4-OMe 6.4 0.1 −1.17 — — 2.56
99 4-O-(CH2)3Ph 9.8 0.0 −0.96 — — 4.50
100 2,3-Me2 15.8 0.5 −0.73 — — 3.10
101 2,3-Cl2 5.6 0.3 −1.23 — — 3.41
102 2-Me, 3-Cl 20.5 0.9 −0.59 — — 3.57
103 2,4-Me2 18.2 1.5 −0.65 — — 3.18
104 2,4-Cl2 31.9 0.7 −0.33 — — 3.55
105 2,5-Me2 10.5 3.3 −0.93 — — 3.25
106 2,5-Cl2 3.0 0.1 −1.51 — — 3.36m
107 2-OMe, 5-nPr 26.0 1.6 −0.45 — — 3.32
108 2,6-F2 3.2 1.0 −1.48 — — 2.16
109 2,6-Cl2 8.9 0.2 −1.01 — — 2.56m
110 2-F, 6-Cl 5.8 0.2 −1.21 — — 2.34
111 3,4-Me2 14.3 9.4 −0.78 — — 3.34
112 3,4-Cl2 0.7 0.2 −2.15 — — 4.25m
113 3,4-(OMe)2 6.4 0.7 −1.17 — — 2.09
115 3,5-Cl2 0.2 0.1 −2.70 — — 4.26m
116 2,5-Cl2, 3-CF3 27.9 2.8 −0.41 — — 4.68
117 2-OMe, 3,5-Me2 20.3 1.0 −0.59 — — 2.91
118 2,3,4-Cl3 12.8 1.3 −0.83 — — 4.39
119 2,3,4,5-F4 13.0 0.9 −0.83 — — 3.44
120 2,3,4,5,6-F5 4.4 0.5 −1.34 — — 3.25
121 RH5849(A: H) Others 10.8 0.4 −0.92 — — 2.45
122 halofenozide(A: H, B: 4-Cl) 9.7 0.8 −0.97 — — 3.22

a Relative ATPase activity (%) to the activity of verapamil. The activity was measured at 125 µM unless noted and the measurement was repeated at least three times. b Standard Deviation of relative activity (%). c Logit-transformed activity. d Calculated by Eq. 1 or 2. e Calculated by Eq. 3. f Referred from ref. 18 unless noted. g The activity was measured at 63 µM, and excluded from Eqs. 1, 3. h The MNDO esp charges were not calculated and excluded from equations. i Excluded from Eq. 3. j The activity was measured at 63 µM, and excluded from Eqs. 2, 3. k The activity was measured at 100 µM, and excluded from Eqs. 2, 3. l Calculated by CLOGP (MacLogP ver. 4.0). m Referred from ref. 17.

2. CoMFA of the DBH derivatives

Since the Km and Vmax values produced by DBH derivatives were not estimated, logit-transformed ATPase activity, [logit A], was used for 3D-QSAR, CoMFA. Logit-transformed data for activities presented by % have been used previously in QSAR analyses.27,28) First, the [logit A] values by the A-ring: 3,5-dimethyl and 2-Cl derivatives and non-substituted B-ring derivatives of DBHs were analyzed separately with CoMFA. Chromafenozide (21), 3,4-Me2 (64), and 3,5-Cl2 (67) compounds of the A-ring: 2-Cl derivatives were not included in the analyses because the activities produced by these compounds were measured at different concentrations from that by the other compounds. All compounds containing an iodine atom (26, 36, and 48) were also not included because their MNDO esp charges were not available. Good CoMFA equations were obtained for the other A-ring: 3,5-dimethyl and 2-Cl derivatives. However, no significant equation was obtained for non-substituted B-ring derivatives, which may have been due to the low activities produced by this set of compounds. The introduction of log P did not improve the equations, although a P-gp substrate was commonly lipophilic or amphiphilic as described previously. We also performed CoMFA for the activities produced by the combined set of the A-ring: 3,5-dimethyl and 2-Cl derivatives. An acceptable equation was derived excluding a compound (A-ring: 2-Cl, B-ring: 4-Ph). The CoMFA equations obtained are shown in Table 3. The observed and calculated [logit A] values by CoMFA equations as well as log P values are listed in Table 2.

Table 3. CoMFA equations for ATPase activity
logit A=Const. + [CoMFA field terms]
Const. CNa n b s c r 2 d Cross-validatedee RCh
s cv f q 2 g Sterici Electro.j Compounds Eq. No.
−1.30 2 20 0.269 0.757 0.432 0.373 64.8 35.2 A: 3,5-Me2 1
−1.29 8 41 0.058 0.974 0.266 0.449 62.2 37.8 A: 2-Cl 2
−1.05 2 60k 0.245 0.597 0.309 0.360 60.1 39.9 Combined 3

a Number of latent variables. b Number of compounds. c Standard error. d Correlation coefficient. e Obtained from the leave-one-out cross-validation. f Standard error. g Correlation coefficient. h Relative contribution (%). i Steric effects. j Electrostatic effects. k A:2-Cl, B:4-Ph derivative was not included.

Figure 4 shows an overlay of the structure of methoxyfenozide (20) causing the high activity with the major electrostatic and steric potential contour maps drawn according to Eq. 3 for the combined set. The red areas in Fig. 4 indicate regions where negative electrostatic interactions with the receptor binding site increase the activity, whereas the blue areas show the reverse case. The green areas in Fig. 4 indicate regions where the submolecular bulk is well accommodated with an increase in activity, whereas the yellow areas indicate regions where the submolecular bulk is unfavorable for activity.

Fig. 4. Overlay of the structure of methoxyfenozide with the major CoMFA electrostatic and steric potential contour maps drawn according to Eq. 3.

Figure 4 shows that the steric bulk in the region around the 4-position of the B-ring was unfavorable for activity while the bulky and partially electronegative 3-substituents of the B-ring were acceptable. In addition, electropositive substituents at the 2-position of the B-ring were favorable for activity. Although the CoMFA contour maps did not show clearly explainable regions for the A-ring side due to poor substituent variations on the ring (3,5-Me2 or 2-Cl), qualitative results indicated that the 3,5-Me2 groups were more favorable than the 2-Cl group as described above. A compound (A-ring: 2-Cl, B-ring: 4-Ph; 51) was excluded in Eq. 3 as it was difficult to quantitatively evaluate the longest and bulkiest 4-substituent on the B-ring (the phenyl group) in CoMFA. However, the low activity produced by the compound could be explained by the existence of the unfavorable bulky substituent at the 4-position.

3. Docking simulation

Docked poses were ranked based on the ChemPLP score, which represents the sum of receptor–ligand hydrogen bonding, van der Waals, torsional, and hydrophobic interaction energies. Higher ChemPLP scores indicate better binding interactions between the compounds and the receptors. The top 10 ranked poses obtained for each binding site (model 1 and model 2) were clustered based on the similarity of the structure and orientation of the conformers (6 and 5 clusters for models 1 and 2, respectively). The pose with the highest score in each cluster was defined as the representative. Finally 2 different representative poses with the top and second highest scores for each model (pose-1 and -2 for model 1; pose-3 and -4 for model 2) were selected. The structures of poses-1–4 were drawn in Fig. 5 with the amino acid residues of P-gp playing important roles in the protein–ligand interaction.

Fig. 5. Poses of methoxyfenozide docked into human P-gp, A: pose-1, B: pose-2, C: pose-3, D: pose-4 The backbone of the protein (3G60 PDB structure for A and B, 3G61 for C and D) is colored magenta (helices) and light blue (loops); the amino acid residues of P-gp playing important roles in the protein–ligand interaction are colored orange; the carbon, hydrogen, nitrogen, and oxygen atoms of methoxyfenozide are colored white, light blue, blue and red, respectively.

Discussion

1. Steroids and agrochemical—mammalian P-gp interactions

Estradiol, progesterone, and testosterone have been reported not to be substrates of human P-gp based on bidirectional transport experiments across MDR1-transfected MDCK(II) monolayers in previous reports.29,30) It was also shown that progesterone significantly stimulated P-gp ATPase activity.29) In our study, the activity produced by progesterone was significant while estradiol and testosterone caused almost no and low ATPase activities, respectively. Further investigations on the interaction of progesterone with P-gp are needed.

Agrochemical—mammalian P-gp interactions have been reported previously.13–15) Bain and LeBlanc evaluated diverse agrochemicals from several chemical classes, including organochlorines, carbamates, pyrethroids, organophosphates, and halophenoxy compounds, for their ability to inhibit the efflux of the P-gp substrate doxorubicin using mouse melanoma cells transfected with the human MDR1 gene.13) About 40% of tested agrochemicals inhibited P-gp, while atrazine, fenvalerate and methoxychlor which produced low ATPase activities in this study, did not inhibit the efflux of doxorubicin at 100–250 µM. These results support these compounds’ not being P-gp substrates. Sreeramulu et al. reported that methyl parathion, endosulfan, cypermethrin, and fenvalerate stimulated P-gp ATPase activity at less than 100 µM in a detergent-solubilized preparation and in reconstituted liposomes.15) The ATPase activities produced in our study by cypermethrin and fenvalerate were not high at 500 µM. We also measured the ATPase activities produced by these agrochemicals at lower concentrations (25 µM and 125 µM, data now shown); however these compounds were not effective. We used human P-gp expressed by Sf9 insect cells while they isolated P-gp from the plasma membranes of the highly multidrug-resistant Chinese hamster ovary cell line. The difference in the results obtained may be due to the different species of P-gp.

Recently, the synergistic or antagonistic activity of ecdysteroids toward doxorubicin was reported in mammalian cancer cells expressing human P-gp.31) Although ecdysone and 20-hydroxyecdysone derived very low ATPase activity (0.9 and 5.8%, respectively) in our screening, the previous report may support DBH derivatives, which are agonists of 20-hydroxyecdysone and, thus, have structure similarities to ecdysteroids, interacting with P-gp.

In this report, it was shown that several organophosphates concentration-dependently caused ATPase activity, but others did not. It was difficult to find structure similarity between active compounds compared to those with no or very low activity. More structure-activity relationship studies are necessary to clarify the binding site of organophosphates.

2. In silico prediction of P-gp substrates/inhibitors

Demel et al. described various approaches for the in silico prediction of the substrates of ABC transporters including P-gp based on published reports.2) There are two types of in silico models; ligand-based and structure-based models. Pharmacophore models, machine learning models, and simple rule models for the classification of P-gp substrates have been introduced for ligand-based models. The published pharmacophore models showed common features (hydrophobic and hydrogen-bond acceptor features) for P-gp substrates.10) Pharmacophore models, a variety of machine learning algorithms such as support vector machines6) have been applied for the prediction of P-gp substrates. In addition, attempts to classify P-gp substrates and non-substrates based on simple rules such as the number of hydrogen-bond acceptors and donors, molecular weight, and pKa have been made. Ishikawa et al. performed QSAR analyses for the P-gp ATPase activity of diverse compounds with chemical fragmentation codes.8) Recently, Broccatelli et al. reported a model for predicting P-gp inhibition using Molecular Interaction Fields based on the pharmacophoric features and physicochemical properties of diverse compounds.10) The combination of pharmacophore modeling, docking, and 3D-QSAR was also performed for the P-gp inhibitory effects of quinazolinones, and indolo- and pyrrolopyrimidines.9)

We constructed a CoMFA model for ATPase activity produced by DBH derivatives. The model showed sterically and electronically favorable/unfavorable regions for activity. Based on the model and qualitative SAR, the 3,5-dimethyl groups on the A-ring were more favorable than the 2-Cl group. Electropositive 2-substituents and bulkier 3-substituents on the B-ring were favorable while smaller 4-substituents were better for activity. Although we attempted to conduct classical QSAR for the same dataset, it was difficult to obtain a good equation for all compounds. This may be due to colinearity between parameters and limitation of the structure variety of substituents especially for more potent A-ring: 3,5-dimethyl derivatives. Only the length of 3-substituents on the B-ring was significant, with a positive coefficient consistent with the CoMFA result.

The X-ray structures of the complex of mouse P-gp with enantiomeric cyclic peptide inhibitors (QZ59-RRR and QZ59-SSS) revealed that each inhibitor bound to a distinct but overlapped site.12) Homology models of human P-gp for structure-based ligand design and docking studies have been reported based on the crystal structures.32) We also constructed human P-gp models based on 2 complex structures and attempted the docking of methoxyfenozide to the protein models. Four possible poses were obtained and the interaction of each pose with P-gp was investigated. Aller et al. showed that the common amino acid residues of mouse P-gp that interacted with 2 inhibitors and verapamil were F724 and V978 (F728 and V982 in human P-gp).12) All poses in Fig. 5 interacted with these 2 residues in different modes. Considering the CoMFA results, pose-3 was the most likely interaction mode of the DBH derivatives and P-gp. In pose-3, there was an enough space around the B-ring 3-substituents of methoxyfenozide while the 4-position of the B-ring was blocked by F303. In addition, the 2-methyl group of the B-ring and one of methyl groups of the A-ring seemed to have CH–π interactions with F343 and F728, respectively. The backbone NH formed a hydrogen bond with the side-chain oxygen atom of Y307. On the other hand, there was a large space around the 4-substituents of the B-ring in pose-1 and -4 as well as a small space at the B-ring 3-position in pose-2, which was inconsistent with the CoMFA results. The finding that one methyl group of the A-ring was surrounded by the hydrophobic residues L304 and L762 of the binding site in pose-3 may explain the higher activity of 3,5-dimethyl derivatives than 2-Cl derivatives. The t-butyl group also interacted with hydrophobic F337 and F978 in pose-3. Even if molecular hydrophobicity, log P, was insignificant in CoMFA, the partial hydrophobicity of DBH derivatives may have been important as P-gp substrates.

We compared the docking mode in pose-3 with the binding mode of DBH derivatives and the EcR. Based on the X-ray crystal structure of the complex of a DBH derivative (BYI06830, a chromafenozide analog) with the EcR (PDB code: 1R20),16) the t-butyl group was surrounded by the hydrophobic residues such as F336, L511 and L518. In a similar manner to methoxyfenozide and P-gp in pose-3, the CH–π interaction of one methyl group of the A-ring of BYI06830 with Y403 and a hydrogen bond of the backbone NH with Y408 were found. Although the protein structures between P-gp and EcR are very different, important interactions may be similar for the same ligand. However, it should be noted that the torsion angle of C(=O)–NH–N(t-butyl)–C(=O) differed in the conformation between methoxyfenozide in pose-3 and BYI06830 in the crystal structure. The structure-activity relationships of the B-ring substituents were also different between P-gp and EcR.19)

Although more studies are necessary to determine the binding site and mode of DBH derivatives, combined information from ligands and proteins will provide some insight into the substrate recognition mechanism of P-gp.

References
 
© 2013 Pesticide Science Society of Japan
feedback
 Top