Exploring the P-Glycoprotein Binding Cavity with Polyoxyethylene Alkyl Ethers

Abstract

P-glycoprotein (ABCB1) moves allocrits from the cytosolic to the extracellular membrane leaflet, preventing their intrusion into the cytosol. It is generally accepted that allocrit binding from water to the cavity lined by the transmembrane domains occurs in two steps, a lipid-water partitioning step, and a cavity-binding step in the lipid membrane, whereby hydrogen-bond (i.e., weak electrostatic) interactions play a crucial role. The remaining key question was whether hydrophobic interactions also play a role for allocrit binding to the cavity. To answer this question, we chose polyoxyethylene alkyl ethers, C_mEO_n, varying in the number of methylene and ethoxyl residues as model allocrits. Using isothermal titration calorimetry, we showed that the lipid-water partitioning step was purely hydrophobic, increasing linearly with the number of methylene, and decreasing with the number of ethoxyl residues, respectively. Using, in addition, ATPase activity measurements, we demonstrated that allocrit binding to the cavity required minimally two ethoxyl residues and increased linearly with the number of ethoxyl residues. The analysis provides the first direct evidence, to our knowledge, that allocrit binding to the cavity is purely electrostatic, apparently without any hydrophobic contribution. While the polar part of allocrits forms weak electrostatic interactions with the cavity, the hydrophobic part seems to remain associated with the lipid membrane. The interplay between the two types of interactions is most likely essential for allocrit flipping.

Introduction

The ATP binding cassette (ABC) transporter P-glycoprotein (ABCB1, MDR1, P-gp) prevents intrusion of xenobiotics, including many drugs, into the cytosol by binding them in the cytosolic lipid leaflet of the plasma membrane (1–3) and flipping them back to the outer leaflet or to the extracellular aqueous phase, depending on the hydrophobicity of the compound (4,5). The binding cavity of mouse P-gp, which shows 87% identity with human P-gp, is indeed accessible from the membrane, and seems to allow simultaneous binding of two different compounds (6).

Since the discovery of P-gp in colchicine-resistant Chinese hamster ovary cells in 1978 (7), hundreds of structurally diverse compounds have been identified which bind to P-gp and get flipped. How P-gp could specifically attract so many different compounds (called “allocrits” in the following (8)) remained enigmatic for a long time. The first extensive searches for substrate recognition elements focused on chemical groups such as aromatic domains or basic nitrogens (e.g., (9)). An investigation of the structures of hundred chemically highly diverse compounds revealed that P-gp recognizes specific hydrogen-bond acceptor (HBA) patterns which may interfere with the genetic information of the cell (10), rather than specific chemical groups. The HBAs are likely to be recognized by the numerous hydrogen-donor groups in the TMDs of P-gp (11,12).

The complexity of the allocrit-P-gp interaction arises not only from the diversity of substrates but also from the fact that allocrit binding to the transporter is preceded by a lipid-water partitioning step (13–16). The free energy of allocrit binding from water to the transporter, $\Delta G_{\text{tw}}^0$, can therefore be considered as the sum of the free energy of lipid-water partitioning and the free energy of allocrit binding from the lipid membrane to the transporter cavity

$$\Delta G_{\text{tw}}^{\text{0}}=\Delta G_{\text{lw}}^{\text{0}}+\Delta G_{\text{tl}}^{\text{0}}.$$

Assessment of $\Delta G_{\text{tl}}^0$ allowed testing the hydrogen-bond acceptor hypothesis (10), and revealed that the allocrit-transporter affinity indeed increased with the number of hydrogen-bond acceptor groups (16,17).

Under the assumption that the free energy of binding of an allocrit from the lipid membrane to the transporter, $\Delta G_{\text{tl}}^0$, results exclusively from hydrogen-bond interactions, the free energy of binding per single HBA was determined as −2.5 kJ/mol for drugs exhibiting many hydrogen-bond acceptors (16). For detergents where hydrogen-bond acceptor groups are more exposed toward the aqueous phase, slightly less negative values were observed (18). Based on the concept of hydrogen-bond formation between allocrits and the transporter, a dynamic model for allocrit binding and transport was developed (19). The structures of the allocrits investigated, ranging from the large cyclosporin A to small phenothiazines or detergents, were, however, too heterogeneous to fully exclude additional hydrophobic interactions with the binding cavity.

Hydrophobic and aromatic interactions between allocrits and the binding cavity were proposed by Aller et al. (6), based on the crystal structure and by numerous classical quantitative structure-activity relationship (QSAR) studies. QSAR studies generally suggested that, in addition to hydrogen-bonding groups, hydrophobic groups play an important role for the direct allocrit-transporter interaction (e.g., (20)). In this context, diverse pharmacophores were proposed which included one up to several hydrophobes. It should be noted that classical QSAR studies are based on the assumption that allocrits access the transporter directly from the aqueous phase, not from the lipid phase (for review (21)).

The aim of this analysis was to clarify whether the hydrophobic interactions predicted by classical QSAR studies merely reflect the necessity of the compounds to partition into the lipid membrane, or whether hydrophobic interactions indeed contribute to the direct allocrit-transporter interaction in the lipid membrane. To unravel the possible role of hydrophobic allocrit-transporter interactions, we chose polyoxyethylene alkyl ethers, C_mEO_n, as allocrits for P-gp (5,18,22). Varying the number of ethoxyl groups, n, while keeping the number of methylene groups, m, constant, and varying the number of methylene groups, while keeping the number of ethoxyl groups constant, allowed a systematic investigation of the influence of methylene and ethoxyl residues for allocrit-transporter interactions.

For an analysis of the direct allocrit transporter interaction, the free energy of transporter-lipid binding, $\Delta G_{\text{tl}}^0$, of the allocrit is required. Because it cannot be measured directly, it was assessed as the difference between the free energy of transporter-water binding $\Delta G_{\text{lw}}^0$, $\Delta G_{\text{tw}}^0$, and the free energy of lipid-water partitioning of the allocrit, as shown previously for drugs (16) and detergents (18). The free energy of lipid-water partitioning of the allocrit, $\Delta G_{\text{lw}}^0$, was determined by means of isothermal titration calorimetry (ITC), using large unilamellar vesicles (LUVs) formed from POPC (1-palmitoyl-2-oleoyl-3-sn-phosphatidylcholine), which is the most abundant lipid in biological membranes. The free energy of transporter-water binding of the allocrit, $\Delta G_{\text{tw}}^0$, was calculated from transporter-water binding constants, K_tw, derived from measurements of the concentrations of half-maximum P-gp ATPase activation, K₁, of the different compounds (18). This approximation is possible because allocrit-transporter association and dissociation steps are fast compared to the catalytic step, which is ∼1–10 s⁻¹ (23).

With this analysis, we show that lipid-water partitioning of the allocrits C_mEO_n is determined by methylene and ethoxyl groups, whereas the transporter-lipid binding of allocrits (i.e., the direct interaction between allocrit and the binding cavity of P-gp in the lipid membrane) is due to ethoxyl groups only. Allocrit interactions with the binding cavity are thus exclusively due to hydrogen-bond formation or to other weak electrostatic interactions. The hydrophobic part of the allocrit seems to remain associated with the lipid membrane, and most likely plays an important role in molecular flipping.

Materials and Methods

Compounds

Polyoxyethylene alkyl ethers (C_mEO_n with m = 12 / n = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 and m = 10, 14, 16 / n = 8), and Brij 35 where C₁₂EO₂₃ is the main component, were purchased from Sigma-Aldrich (St. Louis, MO). C₁₂EO₂₅ was purchased from TCI Europe (Zwijndrecht, Belgium). We used C₁₂EO₁ and C₁₄EO₈ (≥99% purity), C₁₂EO₁₀ and C₁₂EO₂₃ (≥98% purity), and C₁₂EO₂₅ (high quality for research, purity not specified).

Isothermal titration calorimetry

Isothermal titration calorimetry (ITC) was performed with a VP ITC instrument (Microcal, Northampton, MA) at 25°C and 37°C. Buffer solutions (10 mM Tris, 100 mM NaCl) were freshly prepared and were adjusted to pH 7.4 at 25°C and pH 7.0 at 37°C. Degassed detergent solutions at a concentration of C ≤ CMC/2 were used for titrations (for CMCs, see (24,25)). For the preparation of unilamellar vesicles with a diameter of 100 nm, the buffer was added to the lipid film, dried overnight under high vacuum, and then weighed again. The lipids were resuspended in buffer, vortexed, and subjected to five freeze-thaw cycles with dry ice. Lipids were then extruded 19 times through two stacked polycarbonate membranes of 100-nm pore size (Nucleopore; Whatman, Clifton, NJ).

The molar binding enthalpy $\Delta H_{\text{D}}^0$ was measured as described in detail elsewhere (24) and the detergent/lipid molar ratio r_b (often abbreviated as X_b) was determined as

$$r_{\text{b}}^{\left(\text{i}\right)}=n_{\text{D},\text{bound}}^{\left(\text{i}\right)}/n_{\text{L}}^{\text{0}},$$ (1)

where $r_{\text{b}}^{\left(\text{i}\right)}$ is the detergent/lipid molar ratio after i injections, $n_{\text{D},\text{bound}}^{\left(\text{i}\right)}$ is the molar amount of bound detergent after i injections, and $n_{\text{L}}^0$ is the total molar amount of lipid. The partition coefficient, K_lw, describes a linear relation between the equilibrium concentration of the detergent free in solution, C_eq, and the detergent/lipid molar ratio, r_b,

$$r_{\text{b}}=K_{\text{lw}}{C}_{\text{eq}}.$$ (2)

If the detergent concentration is >10 times smaller than the lipid concentration, the detergent/lipid molar ratio, r_b, is close to the mole fraction, X,

$$r_{\text{b}}=\frac{n_{\text{D},\text{bound}}}{n_{\text{L}}}\sim X=\frac{n_{\text{D},\text{bound}}}{\left(n_{\text{L}}+n_{\text{D},\text{bound}}\right)}.$$ (3)

This approximation is no longer valid if the detergent concentration approaches the lipid concentration (see Fig. 5).

Figure 5.

Detergent mole fraction X₍₁₎ (▪) and X₍₂₎ (○) at the concentration of half-maximum activation, K₍₁₎, and inhibition, K₍₂₎, as a function of the number of ethoxyl groups, n, at 37°C.

Surface activity measurements

The air-water partition coefficient, K_aw, and the cross-sectional area, A_D, of C₁₂EO_n were determined by surface activity measurements as described previously for other compounds (26,27).

Cell line and plasma membrane preparation

MDR1-transfected mouse embryo fibroblasts (NIH-MDR1-G185) were a generous gift from Dr. M. M. Gottesman and Dr. S. V. Ambudkar (National Institutes of Health, Bethesda, MD). The crude plasma membrane vesicles were prepared as described in detail elsewhere (18,28). The lateral packing density of planar POPC bilayers was determined as π_M = 32 mN/m at 25°C (29) and may approach π_M = 30 mN/m at 37°C, corresponding to the lateral packing density of NIH-MDR1-G185 cells (16). Detergent partitioning and the concomitant changes of the phase state are therefore most likely similar for the two membranes.

Phosphate release measurements

The P-gp-associated ATP hydrolysis was determined by quantifying the release of inorganic phosphate with a colorimetric assay, as described in detail elsewhere (28). Experiments were carried out in 96-well microtiter plates (F96 MicroWell plate, nontreated; Nalge Nunc, Rochester, NY) with reaction volumes of 60 μL containing 5 μg of protein. The release assay buffer (25 mM Tris-HCl including 50 mM KCl, 3 mM ATP, 2.5 mM MgSO₄, 3 mM DTT, 0.5 mM EGTA, 2 mM ouabain, and 3 mM sodium azide) was adjusted to pH 7.0 at 37°C. ATP hydrolysis was started by transferring the plate from ice to a water bath kept at 37°C for 1 h, and was terminated by rapidly cooling the plate on ice. Background experiments with vanadate were obtained in parallel and subtracted from measurements. For this investigation, the basal P-gp ATPase activity in the absence of drugs and detergents was 8.52 ± 1.5 nmol P_i min⁻¹ mg⁻¹. The allocrit-induced activities relative to the basal values were very constant.

Kinetic data evaluation

We fitted the bell-shaped P-gp ATPase activity curves with a model, proposed by Litman et al. (30) (see also (16)). It assumes basal activity, V₀, in the absence of allocrits, and transporter activation, if one allocrit molecule is bound to the transporter, and slowdown of transport (inhibition) if a second allocrit molecule is bound to the transporter,

$$V_s=\frac{K_1{K}_2{V}_0+K_2{V}_1{C}_{\text{s}}+V_2{C}_{\text{s}}^2}{K_1{K}_2+K_2{C}_{\text{s}}+C_{\text{s}}^2},$$ (4)

where V₁ is the maximum transporter activity (if only activation occurred), V₂ is the activity at infinite allocrit concentration, C_s is the allocrit concentration in aqueous solution, K₁ is the concentration of half-maximum activation, and K₂ is the concentration of half-maximum inhibition. The equation was fitted to the experimental data using least-squares fitting/least-squares minimization.

For most compounds it can be assumed that the aqueous concentration, C_eq, in Eq. 2. corresponds to the concentration C_s in aqueous solution in Eq. 4. The latter can then be replaced by the quotient r_b/K_lw, (Eq. 2), which yields

$$V_{\text{s}}\approx \frac{K_{\text{lw}}{K}_1{K}_2{V}_0+X_{\text{b}}{K}_2{V}_1}{K_{\text{lw}}{K}_1{K}_2+X_{\text{b}}{K}_2}.$$ (5)

If the detergent/lipid molar ratio, r_b (often denoted as X_b), corresponds to the product of the concentration of half-maximum activation, K₁, and the lipid-water partition coefficient, K_lw,

$$r_{\text{b}}=K_1{K}_{\text{lw}},$$ (6)

half-maximum rate is reached,

$$V_{\text{s}}\approx \frac{V_0+V_1}{2}.$$ (7)

The detergent/lipid molar ratio, r_b, in the cytosolic membrane leaflet at the concentration of half-maximum activation can thus be estimated according to Eqs. 5 and 7.

Results

The free energy of lipid-water partitioning as a function of the number of ethoxyl and methylene groups

Partitioning of C₁₂EO_n, (n = 3–5 and 7–10) and C_mEO₈ (m = 10, 14, 16) into large (100 nm) unilamellar vesicles, LUVs, formed from POPC was measured using ITC. Measurements were performed at T = 25°C (pH 7.4) for comparison with previous ITC measurements (24) and at T = 37°C (pH 7.0) for comparison with ATPase activity measurements (28). The small difference in pH barely affects lipid-water partition coefficients of neutral detergents. Aliquots (3–8 μL) of a suspension of LUVs (5–20 mM) were injected into the reaction cell (1.4037 mL or 1.4147 mL) containing the detergent solution at a concentration, C ≤ CMC/2. Partitioning into LUVs was endothermic for all compounds (Table 1). The reaction became more endothermic with decreasing temperature.

Table 1.

Lipid-water partition coefficients and concentrations of half-maximum activation

Compound	Klw [M−1]T = 25°C	ΔH [kJ/mol]T = 25°C	Klw [M−1]T = 37°C	ΔH [kJ/mol]T = 37°C	K1 [M]T = 37°C	K2 [M]T = 37°C	V1 [%]T = 37°C
C10EO8	450	29.41	nm	nm	1.24 × 10−5	1.16 × 10−4	152
C12EO1	nm	nm	nm	nm	No effect	No effect	No effect
C12EO2	nm	nm	nm	nm	nd	1.16 × 10−4	110
C12EO3	6.00 × 104∗	3.35†	8.00 × 104∗	−1.13∗	1.89 × 10−6	2.79 × 10−5	109
C12EO4	3.50 × 104	5.01	4.60 × 104	1.07	2.07 × 10−6	5.02 × 10−5	149
C12EO5	1.80 × 104	9.76	2.30 × 104	3.98	2.01 × 10−6	2.51 × 10−5	165
C12EO6	1.33 × 104†	nm	1.77 × 104†	nm	1.57 × 10−6	2.22 × 10−5	151
C12EO7	9.00 × 103	23.92	1.30 × 104	10.77	1.52 × 10−6	2.05 × 10−5	161
C12EO8	4.75 × 103	20.30	6.50 × 103	10.94	9.04 × 10−7	1.49 × 10−5	191
C12EO9	4.00 × 103	14.10	5.50 × 103	9.43	5.88 × 10−7	2.08 × 10−5	171
C12EO10	2.75 × 103	9.53	3.29 × 103‡	nm	4.52 × 10−7	1.63 × 10−5	177
C12EO23	nm	nm	nm	nm	6.49 × 10−7	2.87 × 10−4	143
C12EO25	nm	nm	nm	nm	1.69 × 10−7	4.90 × 10−5	140
C14EO8	5.80 × 104	2.64	nm	nm	4.72 × 10−8	2.85 × 10−7	150
C16EO8	nm	nm	nm	nm	No effect	No effect	No effect

Key: nm, not measured (values which were not measured); nd, not determined (values which were too small to be measured accurately).

^∗Measurement performed with small unilamellar vesicles.

^†Data obtained by interpolation or extrapolation.

The lipid-water partition coefficient of C₁₄EO₈ revealed a small signal/noise ratio and the compound had the tendency to self-associate under the conditions used. The value obtained has therefore to be considered as lower limit. The partition coefficient of C₁₆EO₈ could not be measured by ITC because the signal/noise ratio was too small. Data measured at 25°C are in excellent agreement with previous measurements performed under analogous conditions (24). As seen in Table 1, the lipid-water partition coefficients are ∼35% higher at 37°C. The lipid-water partition coefficients of C₁₂EO_n measured at 25°C and 37°C were plotted as a function of the number of ethoxyl groups yielding straight lines (not shown). This allowed determining the lipid-water partition coefficients of C₁₂EO₆ (at 25°C and 37°C) and of C₁₂EO₁₀ (at 37°C) by interpolation and by extrapolation, respectively.

The standard Gibbs free energy of lipid-water-partitioning, $\Delta G_{\text{lw}}^0$, of the detergents C_mEO_n was determined as

$$\Delta G_{\text{lw}}^0=-RT\ln \left(C_{\text{w}}{K}_{\text{lw}}\right),$$ (8)

where RT is the product of the gas constant and the absolute temperature and C_w is the concentration of water (C_w = 55.5 mol/L at 25°C and C_w = 55.3 mol/L at 37°C). The free energy of lipid-water partitioning of C₁₂EO_n increased (i.e., the affinity of detergents to the lipid membrane decreased) as the number of ethoxyl groups per detergent increased (Fig. 1 A). Linear regression analysis in the range of n = 4–10 yielded an incremental free energy value per ethoxyl group,

$$\Delta {\left(\Delta G_{\text{lw}\left(1\right)}^0\right)}_{\text{EO}}=1\text{.}08\pm \text{0}\text{.}10\text{kJ}/\text{mol},$$

in perfect agreement with previous values (31). The intercept a = −42.06 ± 0.72 kJ/mol reflects the free energy of lipid-water partitioning, $\Delta G_{\text{lw}}^0$, of C₁₂EO₀ (i.e., dodecanol). Data are summarized in Table 2.

Figure 1.

Free energies of binding of C₁₂EO_n (n = 3–10, 23) to membrane and transporter as a function of the number of ethoxyl groups (n) (T = 37°C). (A) Free energy of lipid-water partitioning, $\Delta G_{\text{lw}}^0$ (▴); free energy of transporter-water binding (first binding region) $\Delta G_{\text{tw}\left(1\right)}^0$ (□), free energy of transporter-water binding (second binding region) $\Delta G_{\text{tw}\left(\text{2}\right)}^0$ (○), both derived from ATPase activity measurements. (B) Free energy of transporter-lipid binding (first binding region) $\Delta G_{\text{tl}\left(1\right)}^0$ (■), (second binding region) $\Delta G_{\text{tw}\left(\text{2}\right)}^0$ (●). Linear fits were performed in the range of n = 4–10. Values for $\Delta G_{\text{tw}\left(1\right)}^0$ and $\Delta G_{\text{tw}\left(\text{2}\right)}^0$ are the averages of at least five measurements.

Table 2.

Linear regression analyses of data in Figs. 1 and 2

No.	Plot	Slope [kJ/mol per m or n]	Intercept [kJ/mol]
1	$\Delta G_{\text{lw}}^0$ vs. n (-O-)	1.08 ± 0.10	−42.06 ± 0.72
2	$\Delta G_{\text{tw}\left(\text{1}\right)}^0$ vs. n (-O-)	−0.73 ± 0.08	−40.52 ± 0.60
3	$\Delta G_{\text{tw}\left(\text{2}\right)}^0$ vs. n (-O-)	−0.26 ± 0.11	−36.11 ± 0.80
4	$\Delta G_{\text{tl}\left(\text{1}\right)}^0$ vs. n (-O-)	−1.78 ± 0.09	1.31 ± 0.60
5	$\Delta G_{\text{tl}\left(\text{2}\right)}^0$ vs. n (-O-)	−1.51 ± 0.14	6.89 ± 0.99
6	$\Delta G_{\text{lw}}^0$ vs. m (-CH2-)	−3.13 ± 0.07	5.29 ± 0.80
7	$\Delta G_{\text{tw}\left(\text{1}\right)}^0$ vs. m (-CH2-)	−3.45 ± 0.23	−5.37 ± 2.79
8	$\Delta G_{\text{tw}\left(\text{2}\right)}^0$vs. m (-CH2-)	−3.36 ± 0.30	1.29
9	$\Delta G_{\text{tl}\left(\text{1}\right)}^0$ vs. m (-CH2-)	∼0	−14.23 ± 0.89
10	$\Delta G_{\text{tl}\left(\text{2}\right)}^0$ vs. m (-CH2-)	∼0	−6.21 ± 0.03

Fig. 2 shows a plot of the free energy of lipid-water partitioning, $\Delta G_{\text{lw}}^0$, as a function of the number of methylene groups for the allocrits C_mEO₈ (m = 10, 12, 14). The number of available compounds varying in the number of methylene groups was limited. Linear regression analysis of the data in the range of m = 10–14 yielded an incremental free energy contribution per methylene group to lipid-water partitioning of

$$\Delta {\left(\Delta G_{\text{lw}\left(1\right)}^0\right)}_{{\text{CH}}_2}=-3\text{.}13\pm \text{0}\text{.}07\text{kJ}/\text{mol}\text{.}$$

This value is somewhat less negative than expected, because the lipid-water partition coefficient of C₁₄EO₈ is at a lower limit. The positive intercept (a = 5.29 ± 0.80 kJ/mol) reflects the unfavorable effect of the ethoxyl groups on lipid-water partitioning (Table 2).

Figure 2.

Free energies of binding of C_mEO₈ (m = 10, 12, 14) to the membrane and to the transporter as a function of the number of methylene groups (m). Free energy of lipid-water partitioning, $\Delta G_{\text{lw}}^0$ (▴) (measured at T = 25°C and calculated at 37°C), free energy of transporter-water binding (first binding region) $\Delta G_{\text{tw}\left(1\right)}^0$ (▪), and free energy of transporter-water binding (second binding region) $\Delta G_{\text{tw}\left(\text{2}\right)}^0$ (●) (T = 37°C).

P-gp-ATPase activity measured as a function of C₁₂EO_n concentration

The P-gp ATPase in inside-out plasma membrane vesicles of NIH-MDR1 cells was titrated with detergents of the series C₁₂EO_n differing in the number of ethoxyl groups (n = 1–10, 23, 25) (Fig. 3). Experiments were performed under steady-state conditions at pH 7.0 and T = 37°C in the concentration range C ≤ CMC/1000 to C ∼0.5 CMC (18). Phosphate release was monitored by means of a colorimetric assay. C₁₂EO₁ showed neither activation nor inhibition of the P-gp ATPase activity (i.e., no deviation from basal activity) in the whole concentration range measured. C₁₂EO₂ revealed a slight activation at low concentrations and inhibition at higher concentrations (not shown because of partial overlap with the activity profile of C₁₂EO₃). At least two hydrogen-bond acceptor groups are thus required to modulate P-gp ATPase activity. The compounds C₁₂EO₃ to C₁₂EO₁₀, as well as C₁₂EO₂₃ and C₁₂EO₂₅ induced the typical bell-shaped P-gp ATPase activity profiles with activation at low and inhibition at high concentrations, respectively (16,30).

Figure 3.

P-gp ATPase activity measured as a function of concentration of C₁₂EOn (n = 3–10, 23) in plasma membrane vesicles of NIH-MDR1-G185 cells (T = 37°C and pH 7.0). C₁₂EO₃ (▪), C₁₂EO₄ (★), C₁₂EO₅ (), C₁₂EO₆ (▴), C₁₂EO₇ (), C₁₂EO₈ (●), C₁₂EO₉ (♦), C₁₂EO₁₀ (◂), and C₁₂EO₂₃ (○). Data are expressed as the average of two measurements. (Solid lines) Fits to Eq. 4.

The activity profiles were evaluated with a two-site binding model (Eq. 4), which yielded the concentration of half-maximum activation, K₁, and inhibition, K₂, and the maximum, V₁ and minimum, V₂, activity, respectively. The kinetic parameters are summarized in Table 1. The P-gp ATPase activity, V₁, increased from C₁₂EO₃ to C₁₂EO₈ and decreased again for detergents with higher numbers of ethoxyl groups. With increasing numbers of ethoxyl groups, a slight shift of the concentration of half-maximum activity, K₁, to lower concentrations was observed. Moreover, the activity profiles broadened for compounds with higher numbers of ethoxyl groups (C₁₂EO₂₃ and C₁₂EO₂₅) compared to those with lower numbers of ethoxyl groups (C₁₂EO₂₅ is not shown because of overlap with C₁₂EO₂₃).

Allocrit affinity from water to the first and second binding region of the transporter as a function of the number of ethoxyl groups

The concentration of half-maximum activation, K₁, and half-maximum inhibition, K₂, respectively obtained from kinetic P-gp ATPase activity measurements was considered as dissociation constant from the first and the second binding region and the inverse as transporter-water binding constant to the first and the second binding region, K_tw(1) and K_tw(2), respectively (16,18). The free energy of allocrit binding to the first binding region of the transporter,$\Delta G_{\text{tw}\left(1\right)}^0$, was approximated as

$$\Delta G_{\text{tw}\left(1\right)}^0=-RT\ln \left(C_w{K}_{tw\left(1\right)}\right).$$ (9)

An analogous equation was used for the second binding region. The free energy of binding of detergents from water to the first binding region of the transporter, $\Delta G_{\text{tw}\left(1\right)}^0$, decreased (i.e., the affinity increased) with increasing number of ethoxyl groups (Fig. 1 A). Linear regression analysis in the range of n = 4–10 yielded an incremental value per ethoxyl group to C12EOn binding to the first binding region of the transporter of

$$\Delta {\left(\Delta G_{\text{tw}\left(1\right)}^0\right)}_{\text{EO}}=-0.73\pm 0.08\text{kJ}/\text{mol}$$

and an intercept of a = −40.52 ± 0.60 kJ/mol (Table 2). The free energy of binding from water to the second binding region of the transporter, $\Delta G_{\text{tw}\left(\text{2}\right)}^0$, as a function of ethoxyl groups also decreased, but to a lesser extent (Fig. 1 A). In the range of n = 4–10, linear regression analysis yielded an incremental value per ethoxyl group to binding to the second binding region of the transporter of

$$\Delta {\left(\Delta G_{\text{tw}\left(2\right)}^0\right)}_{\text{EO}}=-0.26\pm 0.11\text{kJ}/\text{mol}$$

and an intercept of a = −36.11 ± 0.80 kJ/mol (Table 2). At high numbers of ethoxyl groups, free energy of binding from water to the second binding region of the transporter becomes less negative—suggesting an overlap of more than one effect (see Discussion).

Fig. 1 B displays the free energy of binding of detergents from the lipid membrane to the first and the second binding region of the transporter, $\Delta G_{\text{tl}\left(1\right)}^0$ and $\Delta G_{\text{tl}\left(2\right)}^0$, respectively. They both decreased (i.e., the affinity increased) with increasing number of ethoxyl groups in the range of n = 3–10. Linear regression analysis (n = 4–10) yielded an incremental value per ethoxyl group to C₁₂EO_n binding to the first binding region of transporter of

$$\Delta {\left(\Delta G_{\text{tl}\left(1\right)}^0\right)}_{\text{EO}}=-1.78\pm 0.09\text{kJ}/\text{mol}$$

and an intercept of a = 1.31 ± 0.60 kJ/mol (Table 2). The slightly positive intercept reflects the free energy required for inserting the first alkyl chain (in the absence of HBAs) into the binding region of P-gp. For inserting the second molecule, the incremental value per ethoxyl group to C₁₂EO_n binding was

$$\Delta {\left(\Delta G_{\text{tl}\left(2\right)}^0\right)}_{\text{EO}}=-1.51\pm 0.14\text{kJ}/\text{mol},$$

which is somewhat less negative than for the first. The intercept was a = 6.89 ± 0.99 kJ/mol (Table 2), which suggests that insertion of a second alkyl chain into the cavity is energetically even more unfavorable. The high affinity of the alkyl chain to the lipid bilayer and the lack of an affinity with the transporter suggest that the alkyl chain remains associated with the lipid phase, which seems accessible through a cleft between the TMDs (32).

P-gp-ATPase activity measured as a function of C_mEO₈ concentration

The P-gp ATPase in inside-out plasma membrane vesicles of NIH-MDR1 cells was also stimulated with detergents of the series C_mEO₈ (m = 10, 12, 14, 16). The three compounds (m = 10, 12, 14) induced bell-shaped activity profiles as seen in Fig. 4. The highest activity in this series was observed for C₁₂EO₈. The shorter more hydrophilic compound, C₁₀EO₈, induced less activation but significant inhibition. The longer, more hydrophobic compound, C₁₄EO₈, induced only slight activation and slight inhibition, and the most hydrophobic compound, C₁₆EO₈, showed neither activation nor inhibition over the whole concentration range measured (data not shown).

Figure 4.

P-gp ATPase activity measured as a function of concentration of C_mEO₈ (m = 10, 12, 14) in plasma membrane vesicles of NIH-MDR1-G185 cells (T = 37°C and pH 7.0). C₁₀EO₈ (▪), C₁₂EO₈ (●), and C₁₄EO₈ (♦). Data are expressed as the average of two measurements. (Solid lines) Fits to Eq. 4.

Allocrit affinity from water to the first and second binding region of the transporter and its variation with the alkyl chain lengths

Fig. 2 displays the plot of the free energy of transporter-water binding to the first and the second binding region of the transporter, $\Delta G_{\text{tw}\left(1\right)}^0$ and $\Delta G_{\text{tw}\left(\text{2}\right)}^0$, for the allocrits C_mEO₈ (m = 10, 12, 14) as a function of the number of methylene groups. The slopes of the two linear fits give an incremental free energy per methylene group to transporter-water binding of allocrits (first and the second binding region) of

$$\Delta {\left(\Delta G_{\text{tw}\left(1\right)}^0\right)}_{{\text{CH}}_2}=-3\text{.}45\pm \text{0}\text{.}23\text{kJ}/\text{mol},$$

and

$$\Delta {\left(\Delta G_{\text{tw}\left(2\right)}^0\right)}_{{\text{CH}}_2}=-3\text{.}36\pm \text{0}\text{.}30\text{kJ}/\text{mol},$$

respectively. These values are in excellent agreement with previously published incremental free energy values per methylene group obtained for lipid-water partitioning of long chain alcohols (33). The intercept (Table 2, line 7) is negative (a = −5.37 ± 2.79 kJ/mol), suggesting an attractive interaction of HBAs with the hydrogen-bond donor groups in the TMDs in the absence of methylene groups. Dividing the values of the intercept by the number of ethoxyl groups (n = 8) yields an approximate free energy increment per ethoxyl group to transporter-water binding of

$$\Delta {\left(\Delta G_{\text{tw}\left(1\right)}^0\right)}_{\text{EO}}\sim -0.67\text{kJ}/\text{mol},$$

which agrees within error limits with the free energy increment per ethoxyl group obtained from the slope of the plot of $\Delta G_{\text{tw}\left(1\right)}^0$ vs. n, determined as

$$\Delta {\left(\Delta G_{\text{tw}\left(1\right)}^0\right)}_{\text{EO}}=-0.73\pm 0.08\text{kJ}/\text{mol}\text{.}$$

The free energy of transporter-water binding to the second binding region of the transporter, $\Delta G_{\text{tw}\left(\text{2}\right)}^0$, could be evaluated only for C₁₂EO₈ and C₁₀EO₈.

The detergent/lipid molar ratio in P-gp ATPase activity measurements

Detergents are known to affect the phase state and the lateral packing density of membranes at high concentrations. To investigate to what extent the detergents affect the phase state of the membrane under the present conditions, we estimated the detergent/lipid molar ratio, r_b(1) and r_b(2) (Eq. 6) and the mole fraction X₍₁₎ and X₍₂₎ (Eq. 3) at the concentration of half-maximum activation, K₁, and inhibition, K₂, respectively. The detergent/lipid molar ratios in NIH-MDR1-G185 cells can be assumed to be similar to those in POPC small unilamellar vesicles because the lateral packing densities of the two lipid membranes are similar (see above). Fig. 5 shows the detergent mole fractions X₍₁₎ and X₍₂₎ at 37°C as a function of the number of ethoxyl groups, n.

The influence of the mole fraction, X, for C₁₂EO_n, (n = 3–10) on the phase state of the POPC bilayers was investigated previously at 25°C (31). At low detergent concentrations up to a saturating mole fraction, X_sat, the sample contained mixed bilayers. Above the saturating mole fraction, X_sat, up to the concentration, X_sol, detergent-saturated lipid bilayers and lipid saturated micelles coexisted, and beyond the concentration, X_sol, only mixed micelles were observed. With decreasing number of ethoxyl units per detergent, n, and increasing temperature, an increase in the limiting detergent mole fractions X_sat and X_sol was observed.

The comparison of Fig. 5 with Fig. 7 B in Heerklotz et al. (31) shows that the mole fractions X₍₁₎ and X₍₂₎ at 37°C were below the detergent mole fractions X_sat and X_sol determined at 25°C under all conditions. Thereby it has to be noted that the lipid bilayer tolerates higher detergent mole fractions at 37°C without undergoing a phase change, as seen in Fig. 7 A in Heerklotz et al. (31). Recent NMR measurements of deuterated lipids mixed with maltosides at concentrations relevant for P-gp ATPase assays also revealed unperturbed bilayer structures (34). We thus conclude that the P-gp ATPase is surrounded by a lipid bilayer in the concentration range investigated. An indirect proof of unperturbed lipid bilayers was also obtained from P-gp ATPase measurements in living MDR1-transfected cells, which yield similar bell-shaped ATPase activity curves as a function of allocrit concentration (16). By washing cells with allocrit-free medium, ATPase inhibition at higher detergent concentrations (below the CMC) was fully reversible. Addition of detergents above the CMC led in contrast to irreversible phase changes (5). It seems thus highly unlikely that the inhibitory branch of the bell-shaped activity curves is due to phase changes (for review, see (22)).

Discussion and Conclusions

Entry of the first molecule into the binding cavity is driven by HBAs

The detergents C_mEO₈ and C₁₂EO_n are suited for a systematic investigation of the hydrophobic and polar contributions to the individual allocrit-transporter binding steps. In a first step the allocrit partitions into the lipid membrane, where the concentration achieved is generally much higher than in aqueous solution as seen from the high partition coefficients (Table 1). For the allocrits investigated, lipid-water partitioning was endothermic, which reflects an entropy-driven hydrophobic effect, due essentially to the release of ordered water molecules upon entry of the detergent into the lipid membrane (33). In a second step, the allocrit located in the cytosolic membrane leaflet binds to the transporter binding cavity. The different free energies of binding of the detergents C_mEO_n changed linearly with the number of ethoxyl, n, and methylene groups, m, respectively, at least in a certain range of m and n (Figs. 1 and 2). Because the slopes and intercepts obtained from the linear regressions revealed a strong correlation with the number of ethoxyl and methylene groups, respectively, we propose the following general equation for an estimate of the free energy of binding of the detergents C_mEO_n from water to the first binding region of the transporter as a function of the methylene and ethoxyl residues,

$$\Delta G_{\text{tw}\left(1\right)}^0=\Delta G_{1\text{w}}^0+\Delta {\text{G}}_{\text{tl}\left(\text{1}\right)}^0\approx -3\text{.}27m+1\text{.}08n-1\text{.}78n+1\text{.}31\text{.}$$ (10)

The free energy of lipid-water partitioning, $\Delta G_{\text{lw}}^0$, of the allocrit has two contributions—a negative contribution arising from methylene groups, enhancing, and a positive contribution arising from ethoxyl groups, reducing, lipid-water partitioning, respectively. The values given in Eq. 10 were taken from Table 2, except the value for the methylene contribution,

$$\Delta {\left(\Delta G_{\text{lw}}^0\right)}_{{\text{CH}}_2}=-3\text{.}27\text{kJ}/\text{mol},$$

which was taken from maltoside and glucoside measurements performed (X. Li-Blatter and A. Seelig, unpublished) because more compounds of different chain length were available than for C_mEO₈. The free energy of allocrit binding to the first binding region of the transporter in the lipid membrane, $\Delta G_{\text{tl}\left(1\right)}^0$, shows an overall attractive contribution which is due exclusively to ethoxyl groups.

Equation 10 is valid for the range of methylene groups, m = 7–14, and the range of ethoxyl groups, n = 4–10. For lower numbers of methylene groups (m < 7) molecules are not sufficiently anchored to the membrane and the contributions may be lower (X. Li-Blatter and A. Seelig, unpublished). For higher numbers of methylene groups (m > 14), no ATPase activity modulation was observed. This may be due either to a transport at basal activity or to the lack of a proper interaction with the transporter because the alkyl chain may be too long. For low numbers of ethoxyl groups (n < 4), the situation is special because of the high affinity to the lipid membrane, as will be discussed elsewhere. For large numbers of ethoxyl groups (n > 10), a lower increment per ethoxyl group can be expected—because not all HBAs may interact simultaneously with the hydrogen-bond donor groups of the TMDs of P-gp, as suggested previously, e.g., for TWEEN 80 (18) and cyclosporin A (16).

The intercepts given in Table 2, lines 1–3, represent the free energy of binding per dodecanol molecule (lacking a HBA pattern) to the lipid membrane, and to the first and the second binding sites of P-gp. The transfer of a dodecanol from water to the lipid membrane is due to hydrophobic interactions. The lipid-water partitioning reaction is energetically more favorable than binding of dodecanol from water to the first or the second P-gp binding site.

Comparison of the three intercepts (Table 2, lines 1–3) suggests that transfer of a first and a second molecule of dodecanol from the lipid membrane to the binding cavity of P-gp costs 1.54 kJ/mol and 5.95 kJ/mol, respectively, which is in reasonable agreement with the positive free energy of binding of dodecanol, from the lipid membrane to the first and second binding region of the transporter determined as

$$\Delta G_{\text{tl}\left(1\right)}^0=1.31\pm 0.60\text{kJ}/\text{mol}$$

and

$$\Delta G_{\text{tl}\left(2\right)}^0=6.89\pm 0.99\text{kJ}/\text{mol},$$

respectively (lines 4 and 5 in Table 2). This suggests that entry of the hydrophobic molecule lacking HBAs into the binding cavity is unfavorable.

Binding is achieved only if the molecule carries at least two HBAs. This is in good agreement with an early analysis of the structures of hundred compounds interacting with P-gp (10). The intercepts given in Table 2, lines 6–8, reveal the free energy contributions per detergent headgroup (n = 8) to lipid-water partitioning and transporter binding in the absence of a lipid anchor. While lipid-water partitioning of the polar headgroups is energetically unfavorable, binding of the headgroup to the transporter is favorable. This further demonstrates the interaction of HBAs with the TMDs lining the binding cavity and shows that the hydrophobic tail per se is not attracted to the binding cavity.

Entry of the second molecule into the binding cavity is also driven by HBAs but counteracted by steric effects

The affinity of the second molecule to the TMDs lining the binding cavity also increases linearly with increasing number of ethoxyl groups in the range of n = 4–10 (Fig. 1), and an equation similar to Eq. 10 can be formulated as

$$\Delta G_{\text{tw}\left(2\right)}^0=\Delta G_{\text{lw}}^0+\Delta {\text{G}}_{\text{tl}\left(\text{2}\right)}^0\approx -3\text{.}27m+1\text{.}08n-1\text{.}51n+6\text{.}89\text{.}$$ (11)

It differs in the apparent contribution per ethoxyl group to the free energy of transporter-lipid binding, $\Delta G_{\text{tl}\left(2\right)}^0$, and the intercept which reflects essentially the repulsive steric contribution. At higher numbers of ethoxyl groups (n > 10), the free energy of binding, $\Delta G_{\text{tl}\left(2\right)}^0$ vs. n, becomes less negative, which may be due either to a lack of sufficient hydrogen-bond donor groups in the TMDs or to the dominance of a steric repulsion over the hydrogen-bonding effect (see also (16) and (18)). The free energy required to accommodate a second molecule in the binding cavity can be expressed as the difference between the free energy of binding to the second, $\Delta G_{\text{tw}\left(\text{2}\right)}^0$, and the first, $\Delta G_{\text{tw}\left(1\right)}^0$ binding region. It increases with the cross-sectional area of the molecule in its membrane-bound orientation, A_D (26,35), as seen in Fig. 6. This suggests that the work required for accommodation of a second molecule in the binding cavity is proportional to the cross-sectional area of the molecule, A_D. The corresponding values for drugs measured previously (16) were included in Fig. 6. The comparison shows that the same principles are valid for drugs and the present detergents.

Figure 6.

Difference in the free energy of binding from water to the activating and inhibitory binding region of P-gp, $\Delta G_{\text{tw}\left(\text{2}\right)}^0$ – $\Delta G_{\text{tw}\left(1\right)}^0$, as a function of the cross-sectional area, A_D, of allocrits, determined by surface activity measurements. Drugs measured previously (Fig. 6 in (16)) (○). E(n) corresponds to C₁₂EO_n (n = 3, 4, 5, 8, 10, 23) (■).

Allocrit flipping is assisted by the lipid membrane

We demonstrated that the interaction of the allocrits C_mEO_n with the TMDs lining the binding cavity of P-gp is purely electrostatic, while the hydrophobic part of the allocrit most likely remains in contact with the lipid environment. Combining our results with structural knowledge allows suggesting the following flipping mechanism. To bind to P-gp, allocrits have to partition into the lipid membrane to reach the cytosolic membrane leaflet (5), where they orient such that the polar part resides between the polar headgroups and the hydrophobic part between the fatty acyl chains of the lipids. Upon encounter with the TMDs of P-gp, the hydrogen acceptor groups of the allocrit form hydrogen bonds with the hydrogen-bond donor side chains of the amino acids in the TMDs (see also (19)). Hydrogen bonds are weak electrostatic interactions that are inversely proportional to the dielectric constant of the environment and thus increase toward the center of the membrane. If allocrits contain, in addition, cationic or aromatic groups, as is the case of many drugs, then cation-π, CH-π, or π-π stacking interactions (36,37) can be formed with the numerous aromatic residues in the TMDs (11,12). It has to be noted that these interactions are also of electrostatic nature.

The polar part of the allocrit may thus move along the TMDs via transient hydrogen-bond formation, whereas the hydrophobic tail of the allocrits may remain in contact with the hydrophobic lipid environment—which seems possible due to a cleft between the TMDs that opens toward the lipid membrane (6,32,38). Because the affinity of the polar part to the TMDs is generally lower than the affinity of the hydrophobic part to the lipids (see Fig. 1, A and B), movement of the polar part may be easier than that of the more sluggish hydrophobic part (which may induce flipping).

As demonstrated previously, the presence of transmembrane stretches of proteins seems to be sufficient for facilitated flip-flop of amphiphilic compounds such as lipids (39). In the case of the P-gp ATPase, NBD dimerization upon ATP binding and concomitant narrowing of the cavity at the cytosolic side of the membrane (see (32)) imposes unidirectional flipping of the polar part of the allocrit toward the extracellular side. As soon as the polar part approaches the extracellular side of the membrane, hydrogen bonds are broken, which leads to allocrit release from the cavity. ATP is hydrolyzed, which resets the transporter to the starting position and prevents back transport of the allocrit through the transporter (see (40). A similar mechanism was suggested previously for lipopolysaccharide flipping by MsbA (41).

In conclusion, using C_mEO_n as allocrits for P-gp allowed unambiguously dissecting hydrophobic and electrostatic contributions in the process of allocrit binding from water to P-gp. The lipid-water partitioning step preceding allocrit binding to the cavity is driven by hydrophobic interactions. Allocrit binding from the lipid membrane to the cavity is driven exclusively by hydrogen bonding (or other weak electrostatic interactions in the case of more complex compounds). Quite in contrast to hydrophobic interactions which increase in the presence of water, weak electrostatic interactions decrease in the presence of water and are therefore ideal for binding in the cytosolic lipid leaflet, translocation across the membrane, and release at the extracellular lipid-water interface.