Tariquidar-induced P-glycoprotein inhibition at the rat blood-brain barrier studied with (R)-¹¹C-verapamil and PET

Abstract

The multidrug efflux transporter P-glycoprotein (P-gp) is expressed in high concentrations at the blood-brain barrier (BBB) and believed to be implicated in resistance to central nervous system drugs. We used small-animal positron emission tomography (PET) and (R)-¹¹C-verapamil together with tariquidar, a new-generation P-gp modulator, to study the functional activity of P-gp at the BBB of rats. To enable a comparison with human PET data we performed kinetic modeling to estimate the rate constants of radiotracer transport across the rat BBB.

Methods

A group of 7 Wistar Unilever rats underwent paired (R)-¹¹C-verapamil PET scans at an interval of 3 h, one baseline scan and one scan after i.v. injection of tariquidar (15 mg/kg, n=5) or vehicle (n=2).

Results

Following tariquidar administration, the distribution volume DV of (R)-¹¹C-verapamil was 12-fold higher as compared to baseline (3.68±0.81 versus 0.30±0.08; p=0.0007, paired t-test), whereas the DVs were essentially the same when only vehicle was administered. The increase in DV could mainly be attributed to an increased influx rate constant K₁ of (R)-¹¹C-verapamil into the brain, which was about 8-fold higher after tariquidar. A dose-response assessment with tariquidar provided an estimated half-maximum effect dose (ED₅₀) of 8.4±9.5 mg/kg.

Conclusion

Our data demonstrate that (R)-¹¹C-verapamil PET combined with tariquidar administration is a promising approach to measure P-gp function at the BBB.

Introduction

The ATP-binding cassette (ABC) transporter P-glycoprotein (P-gp) is expressed in different body tissues, such as liver, kidney, intestines, testes and brain. In the brain, P-gp is located at the luminal membrane of endothelial cells of blood capillaries, where it impedes the diffusion of lipophilic molecules into brain by actively effluxing them back into the vascular space (1). Apart from its well established role in multidrug resistance of cancer cells, P-gp-mediated efflux transport is also discussed to be implicated in resistance to central nervous system drugs, such as antiepileptic, anticancer, anti-HIV and antidepressant drugs (2). Moreover, changes in P-gp expression and function are discussed to be involved in the causation and pathogenesis of certain neurological disorders, such as Alzheimers’s (3) and Parkinson disease (4).

The carbon-11- (¹¹C) labeled calcium channel inhibitor verapamil has been developed as a positron emission tomography (PET) tracer to assess in vivo the function of P-gp at the blood-brain barrier (BBB) (5). Whereas the tracer was initially used as a racemic mixture (6), enantiomerically pure (R)-¹¹C-verapamil has been suggested to be preferable for kinetic modeling of PET data due to differences in metabolism and plasma protein binding between the (R)- and (S)-enantiomers (7-9).

¹¹C-verapamil is very effectively transported by P-gp at the BBB, thereby possessing low brain uptake, which in turn affords low counting statistics. For a more reliable assessment of P-gp functionality, it has been suggested to perform paired ¹¹C-verapamil PET scans, comprising one baseline scan and one scan after administration of a drug that modulates P-gp activity (10). In a proof-of-concept study conducted in healthy volunteers, Sasongko and co-workers showed significant, yet moderate (~90%), increases of ¹¹C-verapamil-derived brain activity uptake following intravenous (i.v.) infusion of the immunosuppressant cyclosporine A, which has been attributed to cyclosporine A-induced blockade of cerebral P-gp (10). However, these investigators worked with blood concentrations of cyclosporine A that were several folder higher than those achieved when the drug is used in the clinic. Unfortunately, safety concerns hamper the regular use of cyclosporine A at P-gp modulating doses in ¹¹C-verapamil-based PET protocols in human subjects, particularly in patients.

In order to overcome P-gp-mediated drug resistance, several new-generation P-gp modulators have been developed, which have been shown to inhibit P-gp with good selectivity and high potency without causing significant side effects (11). One of the most potent P-gp modulators known to date is the anthranilic acid derivative tariquidar (XR9576) (12), which inhibits substrate transport by P-gp in vitro with a half-maximum inhibition constant (IC₅₀) of about 0.04 μM (13). Tariquidar has already advanced to clinical trials in cancer patients, where increased tumor exposure to the P-gp substrate technetium-99m-sestamibi as a surrogate marker for anticancer drug exposure was shown (14, 15).

In this study we used small-animal PET imaging in naïve rats to develop an improved PET imaging protocol, which uses enantiomerically pure (R)-¹¹C-verapamil and tariquidar to assess the functional activity of cerebral P-gp. To enable a comparison with human PET data, we analyzed our small-animal PET data by a kinetic modeling approach and estimated the rate constants of (R)-¹¹C-verapamil transport across the rat BBB. By modeling (R)-¹¹C-verapamil kinetics in rat brain with and without P-gp modulation we were able to identify those model parameters that were most sensitive to changes in cerebral P-gp function.

Materials and Methods

Animals

Adult female Wistar Unilever rats (Harlan-Winkelmann, Borchen, Germany) weighing 220-250 g were used for this study. The study was approved by the local Animal Welfare Committee and all study procedures were performed in accordance with the Austrian Animal Experiments Act. Rats had access to food and water ad libitum and were kept under a 12 h light/dark cycle.

Prior to each experiment, the animals were placed in a plexiglass container and anesthetized with 1.5% isoflurane. When unconscious, the animals were taken from the container and kept under anesthesia with 1.5% isoflurane administered via a mask during the whole experiment. Animals were warmed with a heating pad kept at 39°C. Each animal was cannulated in the carotid artery (for blood sampling) and the jugular vein (for administration of tariquidar and (R)-¹¹C-verapamil).

Chemicals

Unless otherwise stated, all chemicals were of analytical-grade and obtained from Sigma-Aldrich Chemie GmbH (Schnelldorf, Germany) or Merck (Darmstadt, Germany) and used without further purification. Female Wistar rat plasma (anticoagulant: lithium-heparin) was purchased from Lampire Biological Laboratories (Pipersville, PA, USA). Isofluran was obtained from Baxter VertriebsGmbH (Vienna, Austria). Tariquidar dimesylate was obtained from Xenova Ltd. (Slough, Berkshire, U.K.). For administration, tariquidar was dissolved freshly on each experimental day in 2.5% aqueous dextrose solution and injected at a volume of 3 mL/kg. Enantiomerically pure (R)-¹¹C-verapamil was synthesized from (R)-norverapamil (ABX advanced biochemical compounds, Radeberg, Germany) and ¹¹C-methyl triflate as described earlier (16).

PET experimental procedure

The anesthetized animals were positioned in the scanner bed and (R)-¹¹C-verapamil (84±16 MBq, 19.0±11.8 nmol, n=18) dissolved in 0.5-1 mL of phosphate-buffered saline (pH=7.4)/ethanol (9/1, v/v) was administered as an i.v. bolus via the jugular vein over approximately 40 s. At the start of radiotracer injection dynamic PET imaging was initiated using a microPET Focus220 scanner (Siemens, Medical Solutions).

During the first 3 min after radiotracer injection, 2-μL arterial blood samples were withdrawn manually with a micropipette from the carotid artery (approximately every 5 s), followed by further 2-μL samples taken at 5, 10, 20, 30, 40, 50 and 60 min. Activity in the blood samples was measured in a 1-detector Wallac gamma counter (Perkin Elmer Instruments, Wellesley, MA, USA), which was cross-calibrated with the PET camera. Moreover, one larger blood sample (0.6 mL) was collected into an heparinized vial at 10 min after tracer injection in order to determine plasma protein binding and metabolism of (R)-¹¹C-verapamil (see below). Blood activity data were corrected for radioactive decay and expressed as percent injected dose per gram (%ID/g).

The study set-up is illustrated in Fig. 1. A group of 7 animals underwent 3 consecutive PET scans (scan 1-3). First, (R)-¹¹C-verapamil was injected followed by a 60-min baseline scan (scan 1). Then, at 1 h after the end of scan 1, tariquidar (15 mg/kg, n=5) or vehicle (n=2) was administered via the jugular vein and scan 2 (90 min) was performed (scan 2 measured the remainder of circulating activity from scan 1). Finally, at 3 h after the end of scan 1, (R)-¹¹C-verapamil was injected followed by scan 3 (60 min). Four additional animals underwent only one single PET scan, which was recorded at 2 h after administration of 1, 3, 5 and 7.5 mg/kg of tariquidar, respectively.

FIGURE 1.

Diagram of study set-up. After (R)-¹¹C-verapamil PET scan 1, tariquidar (15 mg/kg, n=5) or vehicle (n=2) was administered i.v. followed by scan 2, which measured the remainder of circulating activity from scan 1. At 3 h after scan 1, (R)-¹¹C-verapamil PET scan 3 was performed.

At the end of the last PET scan, the animals were sacrificed. A terminal blood sample (5 mL) was collected and the whole brain was harvested. Plasma was obtained by centrifugation (3,000 g, 10 min). Plasma and brain samples were stored at −20°C until measurement of tariquidar concentrations (see below).

Analysis of plasma protein binding and metabolism of (R)-¹¹C-verapamil

Plasma collected at 10 min after tracer injection and from the terminal blood sample, respectively, was used to determine plasma protein binding and metabolism of (R)-¹¹C-verapamil. Protein binding was assessed by ultracentrifugation using Amicon Microcon YM-10 centrifugal filter devices (Millipore Corporation, USA). Metabolism of (R)-¹¹C-verapamil was analyzed using a previously described solid-phase extraction assay (8, 17).

Measurement of tariquidar concentrations in plasma and brain

Tariquidar concentrations in terminal plasma samples and brain tissue samples were determined with a high-performance liquid chromatography (HPLC) assay employing ultraviolet (UV) detection at a wavelength of 227 nm. All plasma and brain tissue samples were analyzed in duplicate. A Symmetry C18 HPLC column (125 × 4 mm, 5 μm) (Waters Corporation, Milford, MA, USA) was eluted isocratically with a mixture of acetonitrile, methanol and aqueous ammonium acetate buffer (0.2 M, pH=5) (35/7.8/57.2, v/v/v) containing 0.005 M 1-octane sulfonic acid (PIC B-8 Low UV reagent, Waters) at a flow rate of 1 mL/min. On this system, tariquidar eluted with a retention time of about 13-14 min.

Plasma samples (200 μL) were diluted with 1 mL of aqueous phosphate buffer (0.025 M, pH=7.0). Then, 3 mL of tert-butyl-methyl ether was added and the vials were vortexed for 5 min. After centrifugation for 5 min at 3,000 g, 2 mL of the ether layer was transferred into a new vial and concentrated to dryness. The residue was redissolved in 0.7 mL of mobile phase for HPLC, centrifuged for 5 min at 3,000 g and injected into the HPLC system, which was equipped with a 600-μL sample loop.

The frozen rat brains were weighed, brought to room temperature and homogenized with 5 mL of aqueous phosphate buffer (0.025 M, pH=7.0). 1 mL of the brain tissue homogenate was mixed with 3 mL of acetonitrile and vortexed for 5 min. After centrifugation, 3 mL of the supernatant was further processed and analyzed by HPLC as described for the plasma samples.

Calibration curves for the analysis of plasma and brain tissue samples were generated by analyzing different dilutions of a tariquidar stock solution (5.1 mg in 50 mL of water) in drug-free rat plasma or brain tissue homogenates. Prior to analysis, the calibration standards were incubated in a water bath (37°C) for 15 min. Recoveries of tariquidar from plasma and brain samples were >95% and 64%, respectively.

PET data analysis

PET data were sorted into frame sequences of 8 × 5 s, 3 × 10 s, 2 × 30 s, 3 × 60 s, 2 × 150 s, 2 × 300 s, 4 × 600 s for scan 1 and 3, and 5 × 60 s, 4 × 150 s, 3 × 300 s, 6 × 600 s for scan 2. PET images were reconstructed by Fourier rebinning followed by 2-dimensional filtered back projection with a ramp filter. Normalization, random and attenuation correction was applied to the data. As we were interested in global transport of (R)-¹¹C-verapamil across the BBB, whole brain was chosen as a volume of interest (VOI). Whole-brain VOIs were manually outlined on multiple planes of the PET summation images using the image analysis software AMIDE and time-activity curves (TACs), expressed in units of %ID/g, were calculated.

Kinetic modeling of (R)-¹¹C-verapamil

A standard 1-tissue 2-rate-constant (1T2K) or 2-tissue 4-rate-constant (2T4K) compartment model was fitted to the (R)-¹¹C-verapamil TACs in rat brain (7, 18-20). The input function was constructed by linear interpolation of the measured arterial blood activity data and by multiplication by the ratio of plasma to whole-blood activity, which was determined by separate measurements. A delay of 5 s was taken into account for the time-course of activity in plasma due to transport of activity in the arterial catheter.

Fits were performed by the method of weighted nonlinear least squares as implemented in the Optimization Toolbox of MATLAB (Mathworks, Natick, MA, USA). Goodness-of-fit was assessed by visual inspection of observed and predicted concentrations versus time, by the correlation between observed and predicted concentrations, by the randomness of the residuals (runs test), and by estimating parameter uncertainties (variances) from the inverse of the appropriate Fisher information matrix (21).

In order to obtain a model-independent estimate of the distribution volume DV, Logan graphical analysis (22) was applied to the PET and arterial plasma data using MATLAB. The slope DV of the linear part of the Logan plot was estimated by linear regression of the Logan variables. The linear regression was assessed by the magnitude of the squared linear correlation coefficient (r²).

A dose-response curve was fitted to the DVs of (R)-¹¹C-verapamil (effect, E) following different doses of tariquidar using SPSS (SPSS Inc., USA) according to the following equation:

$$E=E_{\max }\frac{D^n}{(E{D_{50}}^n+D^n)},$$

where E_max is the maximum effect, D the tariquidar dose (mg/kg), ED₅₀ the half-maximum effect dose, and n the Hill-factor.

Statistical analysis

For all calculated outcome parameters, differences between scan 1 (before tariquidar administration) and scan 3 (after tariquidar administration) were tested with a two-tailed paired Student’s t-test. The level of statistical significance was set to 5%.

Results

We used PET imaging to study the effect of tariquidar on brain penetration of (R)-¹¹C-verapamil in rats. We first assessed the test-retest variability of (R)-¹¹C-verapamil PET in 2 rats by performing 2 consecutive scans at an interval of 3 h (Fig. 2A). The TACs of the test-retest scans were nearly congruent. Mean DVs for the test and the retest scans were 0.23 and 0.22, respectively.

FIGURE 2.

TACs in rat brain for (R)-¹¹C-verapamil PET scans recorded (A) before (open squares, n=2, scan 1) and after (filled squares, n=2, scan 3) administration of vehicle; (B) before (open squares, n=5, scan 1) and after (filled squares, n=5, scan 3) administration of tariquidar (15 mg/kg). Activity concentration is expressed as mean %ID/g (± standard deviation, SD for n=5).

A group of 5 rats underwent paired (R)-¹¹C-verapamil PET scans (scan 1 and 3, Fig. 1) with administration of tariquidar (15 mg/kg) at 2 h before start of scan 3 (Fig. 2B). Tariquidar administration had a pronounced effect on activity uptake in brain. The mean DV of scan 3 was increased by +1,137% as compared to baseline scan 1 (p=0.0007) (table 1). In Fig. 3, PET summation images of paired scans recorded in one rat are shown. In scan 1, activity uptake in brain was considerably lower than that of surrounding tissue, whereas in scan 3 brain uptake of activity was several-fold higher.

TABLE 1.

Outcome Parameters of the 2T4K Model

Parameter	Without tariquidar (n=9)	With tariquidar * (n=5)	Relative change (%)
K1 (mL·mL−1·min−1)	0.07±0.03 (16)	0.58±0.20 (9)	+711†
k2 (min−1)	0.34±0.06 (69)	0.27±0.07 (43)	−23
k3 (min−1)	0.29±0.22 (395)	0.37±0.17 (66)	+27
k4 (min−1)	0.73±0.29 (225)	0.66±0.49 (55)	−9
DV (mL·mL−1)	0.30±0.08 (4)	3.68±0.81 (3)	+1,137†
DV (Logan) (mL·mL−1)	0.32±0.08 (1)	3.53±0.79 (1)	+1,007†

^*Tariquidar was administered i.v. at a dose of 15 mg/kg at 2 h before start of the PET scan.

^†A statistically significant difference was observed (paired t-test, p<0.005). For statistical testing only the paired data from scan 1 and 3 were considered (n=5).

Outcome parameters are given as mean ± SD. The mean estimated COV in percent for each parameter is given in parentheses.

FIGURE 3.

Transversal and sagittal PET summation images (0-60 min) recorded before (upper row, scan 1) and after (lower row, scan 3) administration of tariquidar (15 mg/kg).

In contrast to the brain-tissue TACs (Fig. 2B), the blood TACs were only moderately increased after tariquidar administration (Fig. 4). For selected plasma samples, we determined the percentage of unchanged (R)-¹¹C-verapamil by a solid-phase extraction assay. At 10 min after radiotracer injection unchanged (R)-¹¹C-verapamil accounted for 84.1±1.5% and 86.7±4.0% of total plasma activity in PET scan 1 and 3, respectively. At 60 min after radiotracer injection (scan 3), 64.6±9.8% of total plasma activity was in the form of parent (R)-¹¹C-verapamil. At 10 min after radiotracer injection, percent non-protein bound (R)-¹¹C-verapamil was 18.7±1.3% and 18.9±3.4% for scan 1 and 3, respectively.

FIGURE 4.

Total activity concentrations in whole blood before (open squares, n=5, scan 1) and after (filled squares, n=5, scan 3) administration of tariquidar (15 mg/kg). Activity concentration is expressed as mean %ID/g (± SD).

In order to monitor the time course of P-gp inhibition by tariquidar we recorded one dynamic PET scan (scan 2) from time 0 to 90 min following tariquidar administration (Fig. 1). For this scan, peak uptake of activity was reached at 32±5 min after tariquidar injection (Fig. 5). When only vehicle was administered, no increase of activity uptake in brain over the time course of the 90-min PET scan was observed (Fig. 5).

FIGURE 5.

Brain TACs for PET scan 2 recorded from time 0-90 min following administration of vehicle (open squares, n=2) or 15 mg/kg of tariquidar (filled squares, n=5). PET scan 2 measured the remainder of circulating activity from scan 1. Activity concentration is expressed as mean %ID/g (± SD for n=5).

The 2T4K model provided better fits of the PET data than the 1T2K model. In Fig. 6, fits obtained with the 2T4K model are shown. In table 1, the outcome parameters of the 2T4K model are summarized. Compartmental-model derived DV values were in good agreement with DVs estimated by Logan graphical analysis. The influx rate constant K₁ and DV were the model parameters that were most prominently affected by tariquidar pretreatment as compared to baseline scans. K₁ and DV were significantly increased - by 8- and 12-fold respectively - following tariquidar administration (p=0.004 and 0.0007 for K₁ and DV, respectively). For k₂, the efflux rate constant from the first tissue compartment, there was a clear trend towards decreased values following tariquidar treatment, although statistical significance was not reached (p=0.057).

FIGURE 6.

TACs and fits obtained from the 2T4K model in the brain of one rat: (A) before (scan 1) and (B) after administration of tariquidar (15 mg/kg, scan 3). Open circles: TAC in plasma measured; open diamonds: TAC in brain VOI measured; solid squares: TAC in brain VOI model. Note that the y-axis is in a logarithmic scale.

Because baseline (R)-¹¹C-verapamil PET scans were quite similar for different animals (coefficient of variation, COV of DV values: ~25%), we abandoned the paired-scan paradigm and conducted a preliminary dose-response evaluation in a limited number of animals (n=1 per dose) by performing single PET scans after administration of 1, 3, 5 and 7.5 mg/kg of tariquidar, respectively. The TACs recorded after the different tariquidar pretreatment doses are shown in Fig 7. There was a highly significant correlation between tariquidar dose and the measured DVs of (R)-¹¹C-verapamil (r=0.94; p <0.001). A sigmoidal dose-response curve was fitted to the DVs measured after different tariquidar doses. The fitted parameters (estimate±asymptotic standard error) were as follows: ED₅₀: 8.4±9.5 mg/kg, E_max: 4.5±3.8 and n: 1.9±2.0.

FIGURE 7.

Brain TACs following administration of different tariquidar doses: 0 mg/kg (open squares, n=5), 1 mg/kg (filled squares, n=1), 3 mg/kg (open triangles, n=1), 5 mg/kg (filled triangles, n=1), 7.5 mg/kg (open circles, n=1) and 15 mg/kg (filled circles, n=5).

In table 2, the concentration levels of tariquidar in plasma and brain tissue samples collected at the end of the PET scan, that is at about 3 h after i.v. tariquidar administration, are summarized. Both plasma and brain concentrations of tariquidar were highly correlated with the DVs of (R)-¹¹C-verapamil (plasma concentration: r=0.85; p=0.007, brain concentration: r=0.95; p=0.001). There was no correlation between plasma tariquidar concentrations and the area under the blood TACs (r= −0.34; p=0.37).

TABLE 2.

Concentrations of Tariquidar in Plasma and Brain

Tariquidar dose (mg/kg)	Plasma (ng/mL)*	Brain (ng/mL)*
15 (n=5)†	1402±296	4131±1638
7.5	911	2570
5	807	1584
3	266	1449

^*Concentrations were determined at 3 h after i.v. administration of different doses of tariquidar.

^†For all other doses, except for 15 mg/kg, n=1. For the 15 mg/kg dose, tariquidar concentrations are given as mean ± SD.

Discussion

In this study, we studied tariquidar-induced modulation of P-gp at the rat BBB by means of small-animal PET imaging with the novel radiotracer (R)-¹¹C-verapamil. A series of previous studies have used PET and racemic ¹¹C-verapamil to assess P-gp modulation by older-generation P-gp inhibitors (i.e. cyclosporine A and valspodar) in rats (5, 23, 24), non-human primates (25) and in humans (10). In contrast to these previous studies, we used a P-gp inhibitor of the latest generation and analyzed our small-animal PET data by a kinetic modeling approach, which allowed us to estimate the rate constants of (R)-¹¹C-verapamil transport across the rat BBB. Thereby, a direct comparison of our rat data with previously acquired human PET data (20, 26) using the same radiotracer became possible. Moreover, by modeling (R)-¹¹C-verapamil kinetics in rat brain with and without P-gp inhibition, important data on the functional organization of P-gp-mediated efflux at the BBB could be obtained.

Following tariquidar administration, brain uptake of activity was about 12-fold higher as compared to the baseline PET scans, whereas total blood activity levels were only increased to a small extent (Fig. 2B and 4). In order to rule out the possibility that increased brain activity uptake was related to other factors than changes in verapamil transport, we assessed the influence of tariquidar treatment on radiotracer metabolism and plasma protein binding. This is important as several older-generation P-gp inhibitors, such as valspodar, have been shown to inhibit cytochrome P450 enzymes and thereby increase plasma levels of concomitantly administered therapeutic drugs (27). Our results show that both plasma protein binding and metabolism of (R)-¹¹C-verapamil were essentially left unaffected by tariquidar administration. The latter substantiates that tariquidar displays minimal cytochrome P450-mediated pharmacokinetic interactions (12).

For the kinetic modeling of PET data, knowledge of the time course of radiotracer in arterial blood is required (input function). Owing to the small size of the drawn blood samples (2 μL), correction for radiolabeled metabolites of (R)-¹¹C-verapamil was not possible. Luurtsema and co-workers have previously described the metabolism of (R)-¹¹C-verapamil in the same rat strain that we used for this study (8). In rats, (R)-¹¹C-verapamil is almost exclusively metabolized by N-demethylation which generates ¹¹C-formaldehyde and other, unidentified, polar metabolites. We initially planned to apply previously reported correction factors (8) to correct total blood activity for polar radiolabeled metabolites. However, analysis of selected plasma samples showed that radiotracer metabolism was considerably slower in our study (87% and 65% versus 64% and 28% unchanged (R)-¹¹C-verapamil at 10 min and 60 min after radiotracer injection, respectively) (8). It cannot be excluded that the slower rate of radiotracer metabolism observed in our study is related to different forms of anesthesia used in the two studies. As radiotracer metabolism was very slow for all animals examined in our study, we considered it justified to employ total rather than metabolite-corrected blood activity concentrations for the kinetic modeling.

We first used a 1T2K compartment model, as described previously for the analysis of human PET data (7, 20). However, this model failed to provide good fits for the PET curves, in particular for the curves measured after tariquidar administration. These curves were characterized by a plateau of activity following peak uptake that was not accurately described by the 1T2K model. The 2T4K model provided considerably better data fits (Fig. 6). A plausible physiological correlate of the second tissue compartment could be the intracellular space of brain tissue, whereas the first tissue compartment could represent the extracellular space. ¹¹C-verapamil is a lipophilic molecule which can be expected to diffuse well across brain cell membranes into the intracellular tissue compartment. Alternatively, the first tissue compartment could represent the endothelial membrane and the second compartment could represent both the extracellular and intracellular space of brain tissue, with k₃ and k₄ describing the combined transport over the abluminal endothelial membrane and between the extra- and intracellular space (28). For some of the model parameters, in particular k₃ and k₄, the uncertainties were rather large, whereas other parameters (K₁ and DV) were fairly robust (table 1). As compared to human PET data (20, 26), the (R)-¹¹C-verapamil baseline DVs were 2- to 3-fold lower in rats. This may suggest that P-gp mediated transport of (R)-¹¹C-verapamil is more efficient at the rat BBB as compared to the human BBB thus leading to lower activity concentrations in rat brain. This is somewhat in contrast to previous findings by Hsiao et al. who reported that brain-to-blood partition ratios of racemic verapamil were similar in rats and humans (29).

Among all parameters of the 2T4K model, the influx rate constant K₁ and the distribution volume DV were most markedly affected following tariquidar administration (table 1). The about 8-fold increase of K₁ caused by P-gp inhibition is in line with the concept that P-gp acts a “gatekeeper” at the BBB and prevents substrates from diffusing across the luminal endothelial cell membrane (28). When using the 1T2K model, the influx rate constant K₁ was found to be increased to a similar extent after tariquidar treatment as for the 2T4K model (+632% for the 1T2K model versus +711% for the 2T4K model). Similar findings (i.e. that P-gp modulation leads to an increased influx rate constant K₁ of verapamil into brain) were also reported by other investigators (30, 31).

A recent study by Syvänen et al. found that the efflux rate constant k₂ of ¹¹C-verapamil was also a sensitive indicator of P-gp transport across the rat BBB (24). It is important to note that these authors used a different set-up than we did. They first gave ¹¹C-verapamil as a continuous infusion followed by administration of the P-gp inhibitor cyclosporine A at 30 min after start of radiotracer infusion. This implies that ¹¹C-verapamil had already crossed the BBB to a certain extent when the inhibitor was administered. Cyclosporine A might have thereby inhibited the active extrusion of ¹¹C-verapamil from brain parenchyma (“vacuum cleaner” function of P-gp), which could have manifested itself in a decreased k₂. In line with this assumption we also observed a trend towards k₂ decreases in our study (table 1).

It is noteworthy that scan 2 which was recorded from time 0 to 90 min following tariquidar administration in our study (Fig. 5) was similar to the set-up described by Syvänen et al. (24). At the time when we administered tariquidar, the remainder of circulating activity from the first PET scan should have reached transient equilibrium in plasma and brain. Because our analysis had shown that metabolism of (R)-¹¹C-verapamil was slow, it can be expected that a significant fraction of circulating activity still represented unchanged parent tracer. The TACs shown in Fig. 5 apparently reflected the dynamics of the entire inhibition process, including the distribution of intravenously administered tariquidar to P-gp at the BBB. Our data suggest that tariquidar is a fast acting P-gp inhibitor with relatively low plasma clearance as reflected by the slow decline of activity following peak uptake (Fig. 5).

We performed a preliminary dose-response assessment by administering different tariquidar doses prior to PET (Fig. 7). A dose of 15 mg/kg was chosen as the maximum dose because an earlier study in a rat model of drug-resistant epilepsy had shown that higher tariquidar doses failed to produce further increases in response to antiepileptic drug treatment and also resulted in side effects of tariquidar (32). It is noteworthy that the estimated ED₅₀ of our study (8.4 mg/kg) is comparable to the ED₅₀ determined in vivo in mice, by conventional biodistribution, for another P-gp substrate (loperamide; ED₅₀: 5.6 mg/kg) (33).

HPLC measurements of tariquidar levels in plasma indicate that the highest studied dose (15 mg/kg) translates to a drug plasma concentration of about 1.7 μM, achieved at 3 h after i.v. drug administration (table 2). It is generally recognized that only the free fraction of a drug in plasma is able to exert a pharmacological effect. In humans, tariquidar is bound to 99.5% to plasma proteins (personal communication Dr. U. Elben, Aväant Pharmaceuticals). When assuming that the drug is bound to a similar extent to plasma proteins in rats, its unbound concentration attained after administration of the 15-mg/kg dose (table 2) would correspond to about 0.01 μM, which is in a similar range as the drug’s in vitro IC₅₀ (0.04 μM) (13). In humans, a single i.v. dose of 2 mg/kg of tariquidar was well tolerated and produced maximum plasma concentrations (C_max) of about 2.3 μM (12), corresponding to an unbound concentration of about 0.012 μM. Therefore, it seems reasonable to assume that clinically safe doses of tariquidar will result in a significant degree of cerebral P-gp blockade in humans.

Combined treatment with neurological drugs (e.g. antiepileptic drugs) and a P-gp modulator such as tariquidar is a promising approach in order to overcome P-gp-mediated drug resistance (2). The approach employed in this study could be translated to future clinical trials with P-gp modulators in brain disorders in order to define appropriate clinical starting doses of these drugs in humans. It could be also used to define the time window of P-gp modulation in order to optimize the time point of therapeutic drug administration. Another clinical application of (R)-¹¹C-verapamil PET could be to identify neurological patients with pronounced cerebral P-gp activity, which would most likely benefit from continued treatment with P-gp modulating drugs.

Conclusion

Our data demonstrate that (R)-¹¹C-verapamil PET combined with tariquidar administration is a useful approach for assessment of P-gp function at the BBB. It holds great promise for a future translation to animal models of drug resistance as well to studies in healthy volunteers and patients. We were able to show that tariquidar is a potent and fast acting inhibitor of cerebral P-gp that led, at the highest studied dose (15 mg/kg), to a 12-fold increase of the DV of (R)-¹¹C-verapamil in rat brain. The increased DV could be attributed to an 8-fold increased influx rate constant (K₁) of activity into the brain.