Insights into the Molecular Mechanism of Triptan Transport by P-glycoprotein

Abstract

The P-glycoprotein (Pgp) transporter reduces the penetration of a chemically diverse range of neurotherapeutics at the blood-brain barrier (BBB), but the molecular features of drugs and drug-Pgp interactions that drive transport remain to be clarified. In particular, the triptan neurotherapeutics, eletriptan (ETT) and sumatriptan (STT), were identified to have a greater than 10-fold difference in transport rates despite being from the same drug class. Consistent with these transport differences, ETT activated Pgp-mediated ATP hydrolysis ~2-fold, while STT slightly inhibited Pgp-mediated ATP hydrolysis by ~10%. The interactions between them were also non-competitive, suggesting that they occupy different binding sites on the transporter. Despite these differences, protein fluorescence spectroscopy revealed that the drugs have similar affinity to the transporter. NMR with Pgp and the drugs showed that they have distinct interactions with the transporter. Tertiary conformational changes probed by acrylamide quenching of Pgp tryptophan fluorescence with the drugs and a non-hydrolyzable ATP analog implied that the STT-bound Pgp must undergo larger conformational changes to hydrolyze ATP than ETT-bound Pgp. These results and previous transport studies were used to build a conformationally driven model for triptan transport with Pgp where STT presents a higher conformational barrier for ATP hydrolysis and transport than ETT.

INTRODUCTION

P-glycoprotein (Pgp) is a member of the ATP-binding cassette superfamily and acts as the gatekeeper of the blood-brain barrier (BBB) by limiting the penetration of a chemically and structurally diverse range of molecules.^1,2 This function protects the brain from toxic insults, but also prevents the penetration of commercially available and potentially beneficial neurotherapeutics,³ which has made neurotherapeutic drug development a formidable challenge.^4,5 One strategy to overcome this barrier has been to modulate Pgp transport with inhibitors, but so far this has been unsuccessful in the clinic.^6,7 Therefore, there is a lot of interest in the pharmaceutical industry to identify molecular features of neurotherapeutics that drive Pgp transport and to determine molecular mechanisms of neurotherapeutic transport by Pgp.

Pgp is localized to the outer membrane of endothelial cells of the BBB.^8,2,9 The X-ray crystal structure of mouse Pgp revealed that the transporter is a 140 kDa asymmetric monomer comprised of twelve transmembrane helices and two nucleotide-binding domains (NBDs), which lie on the cytosolic side of the membrane and bind ATP.^10,11 The promiscuity of Pgp to neurotherapeutics is the result of a 6,000 ų cavity formed by the transmembrane helices,^10,11 which allows multiple drugs to bind to the transporter simultaneously.^12,13 The hydrolysis of ATP at the NBDs drive directional transport as a result of large conformational changes that bring the NBDs within contact, as seen in functionally related bacterial transporters.^14,15

Triptan drugs represent a class of neurotherapeutics that abort migraines by acting on the 5-hydroxytryptamine 1B and 1D (5-HT1B/1D) receptors that are distributed throughout the brain.^16–18 Several triptan drugs are substrates for Pgp and show varying degrees of brain penetration and Pgp transport,^1,19–21 but it is unclear what molecular features of triptans drive transport. At one extreme is the triptan drug sumatriptan (STT, Figure 1A). It was the first clinically available triptan drug²² and it is one of the weakest Pgp substrates.^{e.g. 19} In vitro studies with the drug and different cell types overexpressing Pgp revealed a range of efflux ratios from 1.1 to 2.9.^1,19,21 Some studies have identified STT to be a good Pgp substrate, with efflux ratios higher than other triptan drug class members,¹⁹ while other studies suggest STT is a non-transported substrate of Pgp.¹ This phenomenon has also been previously noted for Pgp and the cardiovascular drug verapamil.^23–25 In vivo, STT was shown to be weakly transported by Pgp at the BBB in mice.¹⁹ On the other extreme, eletriptan (ETT, Figure 1B) is a second-generation triptan drug²² and a very good Pgp substrate. In vitro studies of ETT identified the drug to have Pgp efflux ratios ranging from 11¹⁹ to 46.7²¹. In vivo, Pgp knockout mice showed a ~50 fold increase in ETT concentration in the brain compared to wildtype.²⁰

Figure 1.

The molecular structure of (A) STT and (B) ETT with proton assignments labeled.

The molecular mechanism of transport and the specific interactions of triptan drugs with Pgp are currently unknown. The large difference in log P values of 4.1 and 0.90 for ETT and STT, respectively, may provide an explanation for the large differences in their transport rates.²⁶ Therefore, we hypothesized that the higher transport rate of ETT by Pgp is the result of stronger interactions with Pgp because of its relatively high log P value when compared with STT. To test this hypothesis, the molecular interactions of ETT and STT with the transporter were examined. First, the effect of these drugs on Pgp-coupled ATP hydrolysis was determined. The interactions of the triptan drugs with Pgp were probed by protein fluorescence spectroscopy and the saturation transfer double difference (STDD) NMR technique. Finally, the effect of these drugs on Pgp conformation in the absence and presence of a non-hydrolyzable ATP analog was examined by acrylamide quenching of tryptophan fluorescence. These experimental results and previously published transport studies from the literature^1,19–21 were used to build a conformationally gated model describing the molecular mechanism of triptan transport by Pgp.

EXPERIMENTAL

Materials

Cholesterol, Tris-HCl and disodium ATP (Na₂ATP) were purchased from Amresco (Solon, OH). Escherichia coli (E. Coli) total lipid extract powder was purchased from Avanti Polar Lipids Inc. (Alabaster, AL). HEPES and acrylamide were purchased from Calbiochem (San Diego, CA). Deuterium oxide (D₂O) was purchased from Cambridge Isotope Laboratories Inc. (Tewksbury, MA). Deuterated dithiothreitol (d₁₀-DTT) was purchased from CDN Isotopes (Quebec, Canada). The detergent n-dodecyl-β-D-maltoside (DDM) was purchased from EMD Millipore Corporation (San Diego, CA). Sodium orthovanadate (Na₃VO₄) was purchased from Enzo Life Sciences (Farmingdale, NY). Sumatriptan was purchased from Fagron, Inc. (St. Paul, MN). Dithiothreitol (DTT) was purchased from Gold Biotechnology (Olivette, MO). Eletriptan was given as a gift from Pfizer Inc. (New York City, NY). Ammonium chloride (NH₄Cl) and deuterated DMSO (DMSO-d₆) was purchased from Sigma Aldrich (Milwaukee, WI). The 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) detergent was purchased from VWR International (Radnor, PA). All the remaining chemicals in this study were purchased from Thermo Fisher Scientific (Waltham, MA).

Methods

Purification and reconstitution of Pgp

The his-tagged wildtype mouse Pgp (Abcb1a, MDR3) was purified from Pichia pastoris in two stages with nickel-nitrilotriacetic acid (Ni-NTA) (Thermo Fisher Scientific, Waltham, MA) and Whatman DE52 cellulose resin (Thermo Fisher Scientific, Waltham, MA) as previously described.^27,28 Detergent solubilized Pgp was directionally integrated into liposomes with the NBDs on the outside using the procedure described in.^23,24 These liposomes were composed of 80% w/v Avanti total Escherichia (E.) coli Total Lipid Extract (Avanti Polar Lipids, Alabaster, AL) and 20% w/v cholesterol with a lipid to protein ratio of 0.16 mg ml⁻¹.^23,24 The Pgp orientation in the liposomes was determined by permeabilizing the liposomes to ATP with CHAPS detergent so that any NBDs located on the inside of the liposome would hydrolyze ATP. There was no increase in the apparent ATPase activity as a result of NBDs located on the inside of the liposomes (data not shown) demonstrating that the NBDs are essentially on the external side of the liposome.²³ Proteoliposome aliquots were stored at −80°C in HE PES buffer (20 mM HEPES, 100 mM NaCl₂, 5 mM MgCl₂, 2 mM DTT pH 7.4). The concentration of protein was determined with the extinction coefficient of 1.28 mg mg⁻¹ cm⁻¹ and the DC Protein Assay Kit II (Bio-Rad, Hercules, CA).^23,24

ATPase activity measurements

The ATPase activity of Pgp was measured using the Chifflet method on a FlexStation 3 spectrometer (Molecular Devices, Sunnyvale, CA), as previously described.^29,23,24 The concentration of free inorganic phosphate (Pi) after ATP hydrolysis was estimated by the absorbance signal at 850 nm from the formation of a Pi-molybdenum complex. The ATPase activity of ETT and STT was measured in the presence of 50 nM Pgp reconstituted in liposomes in Chifflet buffer (150 mM NH₄Cl, 5 mM MgSO₄, 0.02% w/v NaN₃, 50 mM Tris-HCl, pH 7.4).^29,23,24

Simple enzyme kinetics is traditionally analyzed using linear transformations such as Lineweaver-Burk, Hans-Woolf and Eadie-Hofstee plots.^30,31 Unfortunately, these transformations suffer from lack of variable independence across the axes and they have the tendency to bias the data points.³²⁻³⁴ In contrast, non-linear regression is considerably more accurate than linear transformation, and is easily performed with modern computers.³³ Therefore, the Pgp-coupled ATP hydrolysis kinetics was fit by non-linear regression with Igor Pro 6.2 (Wavemetrics, Tigard, OR) and the following equations. Hyperbolic Pgp-coupled ATP hydrolysis kinetics of the drugs were fit to the modified Michaelis-Menten equation (Eq. 1):³⁰

$$v=\frac{V_{\max }\cdot \left[L\right]}{K_m+\left[L\right]}+v_{\mathit{basal}}$$ (1)

where V_max is the maximum ATP hydrolysis, K_m is the Michaelis-Menten constant, [L] is the concentration of ligand and v_basal is the basal ATPase activity in the absence of drugs. For drug interactions that are competitive, the ATP hydrolysis curves were fit to Eq. 1 and an apparent Michaelis-Menten constant (K_m,app) shown in Eq. 2:³⁰

$$K_{m,\mathit{app}}=K_m\cdot \left(1+\frac{\left[I\right]}{K_I}\right)$$ (2)

where [I] is the inhibitor and K_I is the inhibitory constant. For non-competitive inhibition, the apparent V_max (V_max,app) can be calculated with the following equation (Eq. 3):³⁰

$$V_{max,\mathit{app}}=\frac{V_{max}}{1+\frac{\left[I\right]}{K_I}}$$ (3)

This equation assumes that the enzyme velocity will approach 0 at saturating [I]. In cases where there is only partial inhibition of the enzyme at saturating [I], the V_max,app will approach a new V_max (V_{max,uninhibited}). In this case, Eq. 3 can be modified to Eq. 4:

$$V_{max,\mathit{app}}=\frac{V_{max,\mathit{inhibited}}}{1+\frac{\left[I\right]}{K_I}}+V_{max,\mathit{uninhibited}}$$ (4)

where V_{max,inhibited} represents the maximum inhibitable enzyme velocity at saturating [I].

Fluorescence Spectroscopy

Quenching of protein fluorescence has been used to measure the dissociation constants (K_Ds) of a wide range of ligands to Pgp.^35,36,23,24 This technique was used to measure the affinity of ETT and STT to Pgp. An Olis DM 45 spectrofluorimeter (Olis Corp, Bogart, GA) was used to measure the drug-induced quenching of 1 μM Pgp reconstituted in liposomes in 100 mM potassium phosphate (KP_i) and 2 mM DTT (pH 7.4) to prevent cysteine disulfide linkage between Pgp monomers, as previously described.^23,24 Pgp protein fluorescence emission was observed between 300 and 500 nm with a fluorescence emission maximum at ~330 nm, when the protein was excited at wavelengths between 260 and 295 nm.³⁷ The ETT and STT molecules also emit fluorescence in this region, which could potentially complicate Pgp protein fluorescence measurements. Fortunately, the drug fluorescence only partially overlaps with the Pgp protein fluorescence (Figure S1), even at the highest concentrations used in the study. For drug concentrations less than 400 μM, the drug fluorescence contributions could be effectively subtracted from the overall fluorescence signal. The drug fluorescence contribution was further minimized for ETT and STT by exciting Pgp at 280 and 295 nm, respectively. The drug induced fluorescence quenching was corrected (F_corrected) for background fluorescence and inner filter effects with the following equation (Eq. 5):^37,23

$$F_{\mathit{corrected}}=\left(F-B\right)\frac{\left({\epsilon }_{ex}{b}_{ex}+{\epsilon }_{em}{b}_{em}\right)}{2}\left[Q\right]$$ (5)

where F is the measured protein fluorescence at 330 nm, B is the background, [Q] is the quenching ligand concentration, ε is the extinction coefficients for the excitation (ε_ex) and emission (ε_em) and b is the pathlength for the excitation (b_ex) and emission (b_em). The ε_ex at 280 nm and ε_em at 330 nm for ETT was 2.13 mM⁻¹ cm⁻¹ and 0.432 mM⁻¹ cm⁻¹, respectively. For STT, the ε_ex at 295 nm was determined to be 0.808 mM⁻¹ cm⁻¹ and the ε_em at 330 nm was 0.344 mM⁻¹ cm⁻¹

Protein fluorescence quenching can occur by static or dynamic mechanisms. Static quenching occurs from ligand forming a complex with the protein and is directly correlated to the ligand’s affinity.³⁷ In contrast, dynamic or collisional quenching is the result of random collisions between a ligand and protein.³⁷ To differentiate between different quenching mechanisms, protein fluorescence quenching titrations were measured at different temperatures as described.^37,23,24 The corrected fluorescence (F_corrected) of a monophasic fluorescent quenching curve was fit to the following rearrangement of the Stern-Volmer equation (Eq. 6):^37,23

$$F_{\mathit{corrected}}=\frac{F_{\mathit{corrected},0}}{1+K_{SV}\left[Q\right]}+F_{\mathit{unquenced}}$$ (6)

where F_corrected,0 is fluorescence in the absence of a quenching ligand, F_unquenched is an offset related to the unquenched fluorescence, K_SV is the Stern-Volmer constant, and [Q] is the concentration of quenching ligand. Biphasic fluorescence quenching curves were fit to the following equation (Eq. 7):³⁷

$$F_{\mathit{corrected}}=\frac{F_{L,0}}{1+K_{SV,L}\left[Q\right]}+\frac{F_{H,0}}{1+K_{SV,H}\left[Q\right]}+F_{\mathit{unquenced}}$$ (7)

where F_0,L and F_0,H are the fluorescence amplitudes at low and high quenching ligand concentrations, respectively. The K_SV,L and K_SV,H are the Stern-Volmer constants at low and high quenching ligand concentrations, respectively.

The collisional quenching of Pgp by acrylamide has been used to probe changes in exposure to solvent accessible tryptophan residues to determine conformational changes of Pgp upon ligand binding.^{e.g. 38,39,24} Acrylamide is an uncharged polar compound that has minimal penetration into lipids and the hydrophobic core of proteins.⁴⁰ Fluorescence emission of 1 μM Pgp reconstituted in liposomes was monitored at 330 nm after excitation at 295 nm in the presence of saturating concentrations of ETT and STT in 100 mM KPi and 2 mM DTT (pH 7.4). The fluorescence emission was corrected for inner filter effects and background fluorescence with Eq. 5. To compare the results with the ATP hydrolysis experiments, the fluorescence quenching was monitored in the presence of 3.2 mM of the non-hydrolyzable ATP analog adenosine 5′-(β,γ-imido)triophsophate (AMPPNP). N-acetyl-L-tryptophanamide (NATA) is an analog of solvent accessible tryptophans that was used to determine the maximum possible tryptophan quenching.³⁹ The Stern-Volmer plots were created by plotting the F_corrected,0/F_corrected as a function of acrylamide concentrations. The K_SV value is a quantitative measure of exposure to solvent accessible tryptophans, which was estimated from the slopes of the Stern-Volmer plots as described previously.^{e.g. 37,23}

Saturation transfer double difference NMR technique

All NMR experiments were performed at 25°C on a Varian INOVA 600 MHz spectrometer with a 5mm z-gradient ¹H{¹³C/¹⁵N} cryoprobe and analyzed in iNMR software (http://ww.inmr.net) and Igor Pro 6.2, as described previously.^23,24 The ¹H NMR peaks of 100 mM STT and ETT in DMSO-d₆ were assigned using standard ¹H 1D NMR techniques. The ¹H NMR peak assignments for STT and ETT are shown in Figure 1. This labeling scheme was used for the assigning ¹H NMR peaks.

The saturation transfer difference (STD) NMR technique is used to identify drug functional groups that interact with a protein receptor by selectively saturating the receptor and monitoring the saturation transfer from the receptor to the drug.⁴¹⁻⁴³ This observed degree of saturation transfer is directly correlated to the drug’s interaction with the receptor.^41,42 For membrane proteins reconstituted into a lipid system, the lipid-drug interactions can lead to nonspecific interactions and interference in the ¹H STD NMR spectrum. The saturation transfer double difference (STDD) NMR technique was developed to extract specific NMR signals corresponding to drug-protein interactions from samples with lipid membranes.^43,23,24 The STDD NMR spectrum is created by subtracting the STD NMR spectrum of the drug in the presence of lipid from the STD NMR spectrum of a drug in the presence of a protein embedded in the same lipid.^44,45,43,24 This technique was used to probe the ETT and STT binding interactions with Pgp reconstituted in liposomes. NMR samples contained 1 μM reconstituted Pgp in 80% deuterated 100 mM KP_i, pD 7.4 and saturating concentrations of triptans, 250 μM. Control liposome samples were prepared identically to reconstituted Pgp samples. A train of 50 ms gaussian shaped pulses selectively saturated reconstituted Pgp for a total saturation of 2 s. The water suppression by gradient tailored excitation (WATERGATE) pulse sequence was added to suppress the water signal.²³ The STD NMR spectra were produced by phase cycling and alternating the saturation frequency between -1.5 and 40 ppm during 512 scans.^23,24 To create the ¹H STDD NMR spectrum of the drugs with Pgp, the ¹H STD NMR spectrum of the control liposome sample with drug was subtracted from the ¹H STD NMR spectrum of the drug with reconstituted Pgp. This difference (ΔI) correlates to the molecular interactions between the drug and the transporter. This was quantitated using STDD amplification factors calculated from the following equation (Eq. 8):²³

$$\mathit{STDD}\mathit{Amplification}\mathit{Factor}=\frac{[L]}{[P]}\frac{\mathrm{\Delta }I}{I_0}$$ (8)

where [P] is the protein concentration, [L] is the concentration of ligand and I₀ is the amplitude of the ¹H NMR peaks in the spectrum without radio frequency saturation.

RESULTS

The effect of ETT and STT on Pgp ATP activation

Figure 2 shows the Pgp-coupled ATPase activity in the presence of ETT and STT. In the absence of drugs, the Pgp-coupled ATP hydrolysis activity was 560 ± 40 nmol min⁻¹ mg⁻¹ (Figure 2A), which is similar to previous estimates^23,24. The ATP hydrolysis activity was almost completely abolished in the presence of 200 μM of the Pgp inhibitor Na₃VO₄ (data not shown). In Figure 2A (closed squares), the Pgp-coupled ATP hydrolysis activity was measured with a range of ETT concentrations. The ATP hydrolysis kinetics was monophasic and reached a maximum at ~1000 nmol min⁻¹ mg⁻¹. Fitting the curve to the modified Michaelis-Menten equation (Eq. 1) generated a V_MAX and K_M value of 508 ± 27 nmol min⁻¹ mg⁻¹ and 14.8 ± 3.1 μM, respectively. Figure 2A (open circles) also shows the Pgp-coupled ATP hydrolysis kinetics in the presence of STT. In contrast to ETT, STT weakly inhibits Pgp-coupled ATP hydrolysis kinetics by < 10%. This effect is not inconsistent with STT being transported by Pgp. Several known Pgp transported substrates, colchicine, doxorubicin and chloramphenicol, have little impact on Pgp-coupled ATP hydrolysis kinetics⁴⁶⁻⁴⁸ and the fluorescent probe substrate Hoechst 33342, in fact, inhibits ATP hydrolysis by Pgp, but is still transported.⁴⁹

Figure 2. The effect of ETT and STT on Pgp-mediated ATP hydrolysis.

(A) Pgp-coupled ATPase activity as a function of basal ATPase activity (dashed line), ETT (closed squares) and STT (open circles) concentrations. (B) Pgp-coupled ATPase activity in the presence of 125 μM ETT and a range of STT concentrations. For comparison, the level for basal ATPase activity is shown as a dotted line in the panel. (C) Pgp-coupled ATPase activity in the presence of 500 μM STT and a range of ETT concentrations. The dashed line in the panel represents the maximum Pgp-coupled ATP hydrolysis of ETT. Error bars represent the standard deviation and the points represent the average of at least three independent experiments.

The fact that Pgp-mediated ATP hydrolysis kinetics in the presence of ETT and STT were distinct in Figure 2A allowed us to probe the competition of these drugs to Pgp. Classically, competitive inhibition will lead to an increase in K_m as a result of drug displacement from Pgp, while non-competitive inhibition will lead to a decrease in V_max, but will have little effect on the K_m.³⁰ Non-competitive inhibition also implies simultaneous binding of both drugs.³⁰ Figure 2B shows the effect of a range of STT concentrations on the activation of Pgp-coupled ATP hydrolysis by 125 μM ETT. Increasing STT concentrations only decreased the apparent ATP hydrolysis rate about 30%. The fact that the ATP hydrolysis is only partially inhibited implies that both drugs are bound simultaneously and that the interactions are non-competitive. Fitting the ATP hydrolysis kinetics in Figure 2B with Eqs. 1 and 4 gives a K_I, V_{max,inhibited} and V_{max,uninhibited} values of 30.2 ± 12.1 μM, 201 ± 21 nmol min⁻¹ mg⁻¹ and 187 ± 17 nmol min⁻¹ mg⁻¹, respectively. To demonstrate that ETT and STT interactions were in fact non-competitive, an ETT titration of Pgp was performed in the presence of 500 μM STT in Figure 2C. Fitting the curve to Eq. 1 reveals that the activated V_max decreases 30% to 341 ± 20 nmol min⁻¹ mg⁻¹ and the extracted K_m is 11.1 ± 2.8 μM. The decrease in V_max and essentially no change from the previous K_m value of 14.8 ± 3.1 μM (Figure 2A, closed squares) demonstrate that the interactions between ETT and STT are indeed non-competitive. The K_I for STT estimated by fitting Figure 2C with Eqs. 1 and 4 was 48.3 ± 18.0 μM, which is similar to the K_I obtained from the curve fit in Figure 2B.

Probing the interactions of ETT and STT with Pgp by intrinsic protein fluorescence

Equilibrium constants (i.e. K_m, K_I) derived from fitting Pgp-mediated ATP hydrolysis kinetics curves do not necessarily correlate directly to ETT and STT binding affinity, since the fitting equations assume specific enzyme mechanisms.³⁰ One might anticipate that there could be significant differences in the affinities of ETT and STT to Pgp because of large differences in their log P values⁵⁰ and their apparent transport rates.^e.g.20 Ligand affinities to Pgp have been directly estimated with fluorescent or radiolabeled ligands,^51,52,12 fluorescence quenching of labeled Pgp^53,54 and quenching of intrinsic Pgp tryptophan fluorescence.^35,23,24 Directly measuring drug binding with fluorescently labeled Pgp may be problematic due to interference from the label.^53,54 Therefore, the affinities of ETT and STT were estimated by measuring their impact on Pgp protein fluorescence.^35,23,24

Figure 3 shows the protein fluorescence and analysis of Pgp in the presence of ETT and STT. Figure 3A shows the Pgp fluorescence emission spectra in the presence of ETT excited at 280 nm with the background fluorescence subtracted. The protein fluorescence produces a fluorescence maximum at 330 nm (thin line) that decreases ~50% with saturating ETT (thick line). After adjustment for inner filter effects with Eq. 5, the amplitude at 330 nm was monitored as a function of ETT concentration (Figure 3B). Quenching of Pgp by ETT was biphasic with low and high concentration phases. Fitting the quenching curve to Eq. 7 gave K_SV,L and K_SV,H of 7.15 ± 1.14 μM⁻¹ and 8.69 × 10⁻³ ± 0.74 × 10⁻³ μM⁻¹, respectively. The titration was repeated at 37°C to determine the quenching mechanism for each phase as done previously^23,24. The Stern-Volmer constants, K_SV,L and K_SV,H, increased and decreased, respectively (data not shown). Therefore, the low concentration-quenching phase associated with K_SV,L quenches by a dynamic mechanism, and the high quenching concentration phase associated with K_SV,H quenches by a static mechanism. Thus, the K_A for ETT was 8.69 × 10⁻³ ± 0.74 × 10⁻³ μM⁻¹ and the corresponding dissociation constant (K_D) was 115 ± 10 μM. Figure 3C shows the effects of STT on Pgp fluorescence after excitation at 295 nm with the fluorescence background subtracted. Saturating STT quenched Pgp fluorescence about 20%. In Figure 3D, Pgp fluorescence emission corrected for inner filter effects was monitored at 330 nm as a function of STT concentration. The fluorescence quenching curve was monophasic and fitting the curve to Eq. 6 gave a K_SV value of 10.8 × 10⁻³ ± 2.9 × 10⁻³ μM⁻¹. The K_SV value increased when this titration was repeated at 10°C showing that the quenching was the result of a static mechanism (data not shown) and the K_D at 25°C was 92.6 ± 25 μM. The surprising similarities between the K_Ds between ETT and STT reveal their binding is not driven by drug hydrophobicity. Instead, the binding may be driven by other factors such as the number and strength of hydrogen bonds as suggested by a SAR study.⁵⁵

Figure 3.

The affinity of ETT and STT to Pgp deduced by protein fluorescence spectroscopy. Pgp fluorescence emission spectra in the presence of a range of (A) ETT and (C) STT concentrations. The fluorescence emission spectrum at 0 μM and 400 μM triptan drugs are shown as thin and thick black lines, respectively. Spectra at intermediate concentrations are shown in gray. The corrected fluorescence emission amplitude at 330 nm was plotted as a function of (B) ETT and (D) STT concentration. The points and error bars represent the average and standard deviations, respectively, of at least three independent experiments.

Functional groups of ETT and STT involved in molecular recognition by Pgp

Although ETT and STT affinities are similar, the effects of these drugs on Pgp-coupled ATP hydrolysis were distinct and the interactions between the two drugs on the transporter were non-competitive. Therefore, their interactions with the transporter were probed by the STDD NMR technique. The degree of drug interaction with the transporter is reflected in the amplitude of the ¹H STDD NMR peaks and quantitatively by ¹H STDD NMR amplification factors.²⁴

Figure 4 shows the STDD NMR spectra, STDD amplification factors and %ΔSTDD amplification factors of ETT and STT in the presence of Pgp. The figure follows the labeling scheme of the STT and ETT molecular structures in Figure 1. The ¹H STDD NMR spectrum of 125 μM ETT in the presence of 1 μM Pgp is illustrated in Figure 4A. ¹H STDD NMR peaks with significant saturation transfer are labeled. Protons emanating from the liposomes and exchange broadening prevented us from reliably measuring the amplitudes of several alkyl and an amine ¹H NMR peaks of ETT labeled 1, 10–11 and 21–25 (see Figure 1B). From the STDD amplitudes in Figure 4A, the STDD amplification factors are estimated in Figure 4B using Eq. 8. The average STDD amplification factor for ETT to Pgp was 15. The strongest STDD amplification factor of 20 was observed for aromatic phenyl protons labeled 15,19, while the weakest STDD amplification factor of 10 came from the methyl labeled 27 of the pyrrolidine functional group. To exemplify differences in the STDD amplification factors, they were normalized against protons of the pyrrolidine methyl in Figure 4C. The phenyl group (protons 15–19, black) of ETT had STDD amplification factors that were 72% larger than the pyrrolidine methyl. The indole ring (protons 2–7, gray) of ETT had amplification factors that were 34% larger than the pyrrolidine methyl. These results imply that the phenyl of ETT plays the largest role in molecular recognition of Pgp.

Figure 4. ETT and STT interactions with Pgp investigated by the saturation transfer double difference (STDD) NMR technique.

The ¹H STDD NMR spectra with the peaks labeled of 250 μM (A) ETT and (D) STT in the presence of 1 μM Pgp. STDD amplification factors calculated for (B) ETT and (E) STT from the ¹H STDD NMR spectra. For clarity, STDD amplification factors from the indole protons are colored gray and protons from flanking functional groups are shown in white and black. To emphasize differences in the STDD amplification factors, they were normalized against the left most (weakest) STDD amplification factor as Δ%STDD for (C) ETT and (F) STT.

Figure 4D shows the ¹H STDD NMR spectrum of 250 μM STT in the presence of 1 μM Pgp. Several ¹H STDD NMR peaks of STT were identified and labeled in the figure. Like ETT, reliable ¹H STDD NMR measurements were not possible for several alkyl protons (I and J) and amine protons (A and G) due to overlap with proton NMR signals from the liposome and exchange broadening of amines from STT (see Figure 1A). The STDD amplification factors from Figure 4D were calculated using Eq. 8 and are shown in Figure 4E. The average STDD amplification factor for STT was 6, which is less than half of ETT. This difference does not necessarily correspond to weaker interaction of STT with the transporter than ETT, since fluorescence spectroscopy with these drugs and Pgp in Figure 3 has already demonstrated that they have similar K_Ds. Instead, lower STDD amplification factors by STT can result from lower saturation transfer efficiency^41,56 because of how STT is distinctly sequestered by the transporter. For STT, the strongest STDD amplification factor of 9 was observed with proton E of the indole ring, while the weakest STDD amplification factor of 3 was observed for the methyl protons labeled K. Interestingly, the analogous methyl protons on ETT also had the weakest STDD amplification factor. To contrast the differences in STDD amplification factors of STT, they were normalized against the methyl protons labeled K in Figure 4F. The indole ring had 130% higher STDD amplification factors (gray) than the methyl protons. The N-linked methyl near the sulfone labeled H had STDD amplification factors (black) that were 75% of the methyl protons. These results imply that the indole functional group of STT plays the largest role in molecular recognition of the transporter.

Drug-induced conformational changes of Pgp in the presence of ETT and STT

Both X-ray crystal structures of analogous bacterial transporters with non-hydrolyzable ATP analogs^14,15 and studies with Pgp^{e.g. 57,58} have suggested that conformational changes are involved in Pgp-mediated ATP hydrolysis. The Pgp conformations were probed by estimating the relative exposure of solvent accessible tryptophan residues (e.g. residues 158, 799, 1104) through acrylamide quenching as we did previously.^23,24 Figure 5 shows the Stern-Volmer plots in the presence of drugs and the non-hydrolyzable ATP analog AMPPNP, where the slope is equivalent to the K_SV value. In all cases, the curves in the Stern-Volmer plots were linear with respect to acrylamide concentration, which shows that acrylamide does not directly bind to Pgp or affect Pgp’s conformation.³⁷

Figure 5. ETT and STT induced conformational changes of Pgp in the absence and presence of an ATP analog AMPPNP inferred by fluorescence spectroscopy.

(A) The Stern-Volmer plots of NATA (dashed line), Pgp in the absence of drugs (closed squares) and in the presence of AMPPNP (open circles). (B) The Stern-Volmer plot of Pgp in the presence of STT (closed squares) and STT with AMPPNP (open circles). (C) The Stern-Volmer plot of Pgp in the presence of ETT (closed squares) and ETT with AMPPNP (open circles). The points represent the average and the error bars represent the standard deviation of at least three independent experiments.

In the absence of drugs, the Stern-Volmer plot of Pgp in Figure 5A (closed squares) had a K_SV value of 1.52 ± 0.20 M⁻¹ with 34% of the total tryptophan fluorescence quenched at 1 M acrylamide. This degree of quenching correlates well to the ratio of fully solvent accessible tryptophans (3) versus the total number of tryptophan residues (11) deduced from the X-ray crystal structure of Pgp.^59,60 As a positive control, NATA was used to assess the maximum tryptophan quenching. NATA was quenched 93% at 1M acrylamide and had a K_SV value of 13.4 ± 0.30 M⁻¹ (Figure 5A, dashed line).

The addition of the ATP analog AMPPNP (open circles) to Pgp reduced the K_SV value to 0.84 ± 0.20 M⁻¹, resulting in a ΔK_SV of 0.68 ± 0.20 M⁻¹. The change in tryptophan accessibility or the ΔK_SV value is consistent with nucleotide-induced conformational shift observed in previous crystallography studies.^{e.g. 14,15}

The effect of STT on the solvent accessibility of tryptophan residues from Pgp is shown in Figure 5B (closed squares). The STT-bound complex has a significantly higher K_SV value of 4.13 ± 0.20 M⁻¹ and in the presence of AMPPNP (open circles), the K_SV value decreases to 2.69 ± 0.20 M⁻¹, resulting in a ΔK_SV value of 1.44 ± 0.20 M⁻¹. This ΔK_SV value is considerably larger than the ΔK_SV value in the absence of drugs, implying that the STT-bound complex undergoes a relatively large conformational change in the presence of the ATP analog. In the presence of ETT, the K_SV value increased modestly to 1.96 ± 0.07 M⁻¹ (Figure 5C, closed squares). Furthermore, the K_SV value of the ETT-bound complex remains essentially unchanged by the addition of AMPPNP to 2.06 ± 0.05 M⁻¹ (Figure 5C, open circles) with a ΔK_SV value of −0.10 ± 0.09 M⁻¹. These results imply that the nucleotide-induced Pgp conformational transitions are more modest in the presence of ETT than STT.

DISCUSSION

A conformationally gated model of triptan transport is presented in Figure 6 to explain the molecular mechanism driving ETT and STT transport by Pgp. The model is based on our results, the conformational changes that Pgp undergoes in the presence of nucleotides and ligands^61−64,24 and the observed differences in triptan transport.^{e.g. 1,19−21} By analogy with the bacterial transporters,¹⁵ Pgp can assume a wide range of conformations. For simplicity, we define three conformations of Pgp: “open”, “intermediate” and “closed”, as we did previously.^23,24 In the “open” conformation, the NBDs are separated with the binding cavity exposed to the cytosol. The NBDs are together in the “closed” conformation with the binding cavity exposed to the extracellular space. The “intermediate” conformation is between the “open” and “closed” Pgp conformations with equivalent exposure to the extracellular and cytosolic sides of the membrane.

Figure 6. Conformationally gated model of triptan transport by Pgp.

Pgp is depicted as a cartoon image of the “open” and “intermediate” states in the (A,B) absence of drug, the presence of (C,D) STT and (E,F) ETT. STT are shown as gray diamonds, ETT are gray circles and N is ATP. The horizontal arrows represent equilibrium between the nucleotide-bound conformational states of Pgp. The sizes of the vertical arrows represent the degree of transport. The dashed arrow reflects the fact that STT has been observed to be both transported¹⁹ and not transported by Pgp¹ depending on the cell type.

Changes to tryptophan accessibility of Pgp by ETT, STT and the non-hydrolyzable analog AMPPNP were deduced by acrylamide fluorescence quenching in Figure 5. The K_SV and ΔK_SV values were used to correlate the exposure of solvent accessible tryptophan residues to Pgp conformational changes. An increase in the K_SV value implies a shift to a more “open” conformation, while a decrease in K_SV value suggests a shift to a “closed” conformation. Moreover, the ΔK_SV value infers the degree of this conformational change. However, tryptophan accessibility and the corresponding K_SV values cannot be used to assign specific ligand-bound conformations, since it does not correlate directly to a specific Pgp conformation. We know from X-ray crystallography, cryo-em, cross-linking and mutagenesis studies with Pgp and the analogous bacterial transporters that the interaction of the NBDs is essential for ATP hydrolysis.^{65−68,57,69} Therefore, Pgp-mediated ATP hydrolysis rates, relative changes in tryptophan accessibility from Figure 5 and previous Pgp conformational assignments^23,24 were used to direct conformational assignment of Pgp in the presence of nucleotides, ETT and STT. In general, when the ligand-induced activation of Pgp-mediated ATP hydrolysis is relatively high, we assume that Pgp is shifted toward the “closed” conformation. When there is ligand-induced inhibition or no inhibition in Pgp-mediated ATP hydrolysis, we assume that Pgp is shifted toward or in an “open” conformation.

We know from X-ray crystallography of Pgp in the absence of ligands that the NBDs are relatively far apart in the “open” conformation.^{e.g. 10} Therefore, Pgp is depicted in the “open” conformation in the absence of ligands in Figure 6A. Figure 5A shows that tryptophan accessibility decreases in the presence of AMPPNP (ΔK_SV = 0.68 M⁻¹) suggesting a shift to the “closed” conformation. However, the baseline P-gp mediated ATP hydrolysis rate hovers around 600 nmol min⁻¹ mg⁻¹. Therefore, we propose that Pgp in the presence of ATP, but in the absence of ligands, is in an “intermediate” conformation in Figure 6B as we hypothesized previously²⁴.

Figure 5B shows that STT increases the K_SV value of Pgp by 2.61 M⁻¹, while STT decreases the Pgp-mediated ATP hydrolysis rate by ~10% in Figure 2. These results imply a ligand-induced separation of the NBDs and a shift to a more “open” conformation. This is consistent with X-ray crystallographic studies showing Pgp in an “open” conformation, when complexed with inhibitors.^10,70 Therefore, we propose that the STT-bound Pgp structure is in an “open” conformation and bound near the extracellular side of the transporter in Figure 6C like inhibitors in the Pgp X-ray crystal structures.^10,70 In the presence of AMPPNP, the K_SV value of the STT-bound Pgp decreases by 1.44 M⁻¹, implying a large conformational shift toward the “closed” conformation. A conformational barrier to ATP hydrolysis and transport would explain the 10% reduction in Pgp-mediated ATP hydrolysis in the presence of STT (Figure 2A) and the significantly lower Pgp-mediated transport of STT versus ETT¹⁹. Since Pgp-mediated ATP hydrolysis in the presence of STT is similar to basal activity, the STT-nucleotide-bound complex is hypothesized to be in an “intermediate” conformation in Figure 6D. In vitro transport studies with STT with Pgp have demonstrated that Pgp-mediated STT transport is cell type dependent.^1,19−21 Therefore, STT transport or lack there of is represented as a dashed arrow in Figure 6.

ETT induces ~2-fold activation of Pgp-mediated ATP hydrolysis, which is similar to digoxin-induced activation but is half of the verapamil-induced activation of Pgp.²³ Digoxin was hypothesized to shift Pgp into an “intermediate” conformation, while verapamil shifted the transporter to the “closed” conformation. Therefore, we propose that Pgp in the presence of ETT is in an “intermediate” conformation in Figure 6E like the digoxin-bound complex.²³ The Stern-Volmer plot in Figure 5C shows that the K_SV value for acrylamide quenching of Pgp in the presence of ETT changes little with saturating AMPPNP, implying that the conformation is changed little by the non-hydrolyzable ATP analog. Therefore, we propose that Pgp in the presence of ETT and ATP is also in an “intermediate” conformation in Figure 6F. With the ETT-bound Pgp being in the same conformation in the presence and absence of ATP, the conformational barrier for ATP hydrolysis is minimized when compared to STT. This reduction in the conformational barrier correlates well to the 5–20 fold higher efflux ratios observed for ETT than STT.^1,19 In addition, the interactions of STT and ETT on Pgp probed by Pgp-mediated ATP hydrolysis (Figure 2B and 2C) were non-competitive suggesting that they occupy distinct sites on the transporter. Since ETT induces significant activation of Pgp-mediated ATP hydrolysis, we place ETT far from the hypothetical STT binding site near the NBDs in Figure 6E.

Our results in this study imply that triptan transport is not merely driven by chemical properties of STT and ETT. Triptans have distinct interactions with the transporter and induce conformational changes that activate or hinder Pgp-mediated ATP hydrolysis.^1,19−21 These studies are the first to provide direct insight into the molecular mechanism of triptan transport by Pgp. Although this study is only focused on two neurotherapeutics, our simplified transport model may apply to a range of neurotherapeutics.

ABBREVIATIONS

Amp

amplitude

AMPPNP

adenosine 5′-(β,γ-imido)triophsophate

BBB

blood-brain barrier

CHAPS

3-((3-cholamidopropyl) dimethylammonio)-1-propanesulfonate

DDM

n-dodecyl-β-D-maltoside

DEAE

diethylaminoethyl cellulose

ETT

eletriptan

K_D

Dissociation constant

KPi

potassium phosphate

K_SV

Stern-Volmer quenching constant

n

number of binding sites

NATA

N-acetyl-L-tryptophanamide

NBD

nucleotide-binding domain

Ni-NTA

nickel-nitrilotriacetic acid

NMR

nuclear magnetic resonance

Pgp

P-glycoprotein

Pi

inorganic phosphate

SAR

structure-activity relationship

STDD

saturation transfer double difference

STT

sumatriptan

WATERGATE

water suppression by gradient tailored excitation