Interaction of ¹¹C-Tariquidar and ¹¹C-Elacridar with P-glycoprotein and Breast Cancer Resistance Protein at the Human Blood-Brain Barrier

Abstract

The adenosine triphosphate-binding cassette transporters P-glycoprotein (Pgp) and breast cancer resistance protein (BCRP) are 2 major gatekeepers at the blood-brain barrier (BBB) which restrict brain distribution of several clinically used drugs. In this study we investigated the suitability of the radiolabeled Pgp/BCRP inhibitors ¹¹C-tariquidar and ¹¹C-elacridar to assess Pgp density in human brain with PET.

Methods

Healthy subjects underwent a first PET scan of 120 min duration with either ¹¹C-tariquidar (n = 6) or ¹¹C-elacridar (n = 5) followed by a second PET scan of 60 min duration with (R)-¹¹C-verapamil. During scan 1 (at 60 min after radiotracer injection) unlabeled tariquidar (3 mg/kg) was intravenously administered. Data was analyzed using 1-tissue 2-rate-constant (1T2K) and 2-tissue 4-rate-constant (2T4K) compartment models using either metabolite-corrected or uncorrected arterial input functions.

Results

Following injection of ¹¹C-tariquidar or ¹¹C-elacridar, brain PET signal corrected for radioactivity in vasculature was very low (~0.1 standardized uptake value) with slow washout. In response to tariquidar injection, a moderate, but statistically significant rise in brain PET signal was observed for ¹¹C-tariquidar (+27 ± 15%, P = 0.014, paired t-test) and ¹¹C-elacridar (+21 ± 15%, P = 0.014) without changes in plasma activity concentrations. Low levels of radiolabeled metabolites (<25%) were detected in plasma at time points up to 60 min after injection of ¹¹C-tariquidar or ¹¹C-elacridar. The 2T4K model provided better data fits than the 1T2K model. Model outcome parameters were similar when metabolite-corrected or uncorrected input functions were used. There was no significant correlation between distribution volumes (V_T) of ¹¹C-tariquidar or ¹¹C-elacridar and V_Ts of (R)-¹¹C-verapamil in different brain regions.

Conclusion

The in vivo behavior of ¹¹C-tariquidar and ¹¹C-elacridar was consistent with that of dual Pgp/BCRP substrates. Both tracers were unable to visualize cerebral Pgp density, which was most likely related to insufficiently high binding affinities in relation to the very low density of Pgp in human brain (~1.3 nM). Despite their inability to visualize Pgp density, ¹¹C-tariquidar and ¹¹C-elacridar may find use as a new class of radiotracers to study the interplay of Pgp and BCRP at the human BBB in limiting brain uptake of dual substrates.

INTRODUCTION

The adenosine triphosphate (ATP)-binding cassette (ABC) transporters P-glycoprotein (Pgp) and breast cancer resistance protein (humans: BCRP; rodents: Bcrp) act together as a team of gatekeepers at the blood-brain barrier (BBB) at the level of the luminal (blood-facing) membrane of vascular endothelial cells to protect the brain from the accumulation of mostly lipophilic xeno- and endobiotics (1). Various clinically used drugs, most notably several members of the family of tyrosine kinase inhibitors for cancer treatment (e.g. gefitinib, imatinib, sorafenib), were shown to be dual substrates of Pgp and BCRP (1). It has been demonstrated that Pgp and BCRP form a cooperative drug efflux system at the BBB and that dual substrates only gain brain access when both transporters are genetically or chemically disrupted (2). Tariquidar and elacridar are potent third-generation inhibitors of Pgp, which were later discovered to also inhibit BCRP, albeit at higher concentrations than Pgp (3). Based on the assumption that tariquidar and elacridar would bind to Pgp without being transported (4), ¹¹C-tariquidar and ¹¹C-elacridar were developed as PET tracers to visualize Pgp expression levels at the BBB as opposed to substrates, like ¹¹C-verapamil or ¹¹C-N-desmethyl-loperamide, which visualize Pgp function (5). However, initial preclinical evaluation revealed that baseline brain uptake of ¹¹C-tariquidar and ¹¹C-elacridar was very low in rats or mice and significantly increased in Pgp/Bcrp combination knockout mice or after pretreatment of animals with unlabeled tariquidar or elacridar (6-9). This “substrate-like” in vivo behavior was subsequently confirmed in vitro by showing that both tariquidar and elacridar are avidly transported by Pgp and BCRP in nanomolar concentrations as used for PET (10, 11). On the other hand, PET experiments in a low and high Pgp expressing murine breast tumor graft model showed higher uptake of ¹¹C-tariquidar in Pgp overexpressing tumors, consistent with Pgp binding (12). Similarly, paired PET scans with ¹¹C-tariquidar or ¹¹C-elacridar and (R)-¹¹C-verapamil in naïve rats, with an intravenous (i.v.) administration of tariquidar (3 mg/kg) before the (R)-¹¹C-verapamil scan, revealed a significant negative correlation between regional brain distribution volumes (V_Ts) of ¹¹C-tariquidar or ¹¹C-elacridar and (R)-¹¹C-verapamil, which also pointed to Pgp-specific binding of ¹¹C-tariquidar and ¹¹C-elacridar (13). Such a paired-scan protocol could in theory be used to independently assess Pgp expression and Pgp function in one scan session.

The aim of this study was to investigate the suitability of ¹¹C-tariquidar and ¹¹C-elacridar to visualize cerebral Pgp density in healthy human subjects. In analogy to a previous preclinical study (13) we performed paired PET scans with ¹¹C-tariquidar or ¹¹C-elacridar and (R)-¹¹C-verapamil including a pharmacological challenge with tariquidar (3 mg/kg).

MATERIALS AND METHODS

Subjects

The study was conducted in accordance with the Declaration of Helsinki and its amendments, registered with EUDRACT (number 2010-020759-30), and approved by the Ethics Committee of the Medical University of Vienna. The study was conducted as a prospective, single center, randomized, phase I study. All subjects provided written informed consent and were confirmed to be healthy based on medical history, physical examination, routine laboratory tests, urine drug screening, electrocardiography and vital signs. Eleven healthy men (mean age: 34 ± 8 y, mean weight: 76 ± 10 kg) were randomly assigned to receive either a ¹¹C-tariquidar (n = 6) or a ¹¹C-elacridar (n = 5) PET scan followed by a (R)-¹¹C-verapamil PET scan (Fig. 1).

FIGURE 1.

Diagram of study set-up. PET scan 1 with either ¹¹C-tariquidar or ¹¹C-elacridar over 120 min was followed by PET scan 2 with (R)-¹¹C-verapamil over 60 min with an interval of 60 min between the 2 scans. At 60 min after start of scan 1, unlabeled tariquidar was infused i.v. at a dose of 3 mg/kg over 30 min.

Radiotracer Synthesis

¹¹C-Tariquidar, ¹¹C-elacridar and (R)-¹¹C-verapamil were synthesized following previously published procedures (8, 9, 14) and formulated in 20 mL sterile 0.9% (w/v) aqueous saline solution/ethanol (9/1, v/v) (containing 1.4% (v/v) Tween-80 in the case of ¹¹C-tariquidar and ¹¹C-elacridar). Radiotracer solutions were sterilized by passage through sterile Millex-GV filters (0.22 μm) (Millipore, Bedford, MA, USA). Tween-80 was used as an excipient to prevent non-specific binding of radiotracers to filter membranes. Tween-80 has been reported to inhibit Pgp with a half-maximum inhibitory concentration of 24 μM (15). In a previous preclinical study we have administered Tween-80 at doses up to 0.3 mmol/kg and failed to observe a Pgp inhibitory effect at the BBB (16). Specific activities at the time of injection were 19 ± 7 GBq/μmol, 42 ± 20 GBq/μmol and 38 ± 18 GBq/μmol for ¹¹C-tariquidar (n = 6 batches), ¹¹C-elacridar (n = 5 batches) and (R)-¹¹C-verapamil (n = 10 batches), respectively. Radiochemical purity of all radiotracers was >98%.

PET Study Protocol

Subjects underwent a first PET scan of 120 min duration with either ¹¹C-tariquidar or ¹¹C-elacridar followed by a second PET scan of 60 min duration with (R)-¹¹C-verapamil, with an interval of 1 h between the 2 scans (see Fig. 1). During scan 1 (at 60 min after tracer injection) tariquidar (AzaTrius Pharmaceuticals Pvt Ltd, London, UK) was administered at a dose of 3 mg/kg body weight (3.6 μmol/kg) as an i.v. infusion over 30 min. For formulation of the infusion solution, a stock solution of 7.5 mg/mL of tariquidar free base in 20% ethanol/80% propylene glycol was diluted in aqueous dextrose solution (5%, w/v) to a final volume of 250 mL.

PET Scans

All PET scans were performed on a GE Advance scanner run in 3-dimensional mode (GE Medical Systems). For scan 1, either ¹¹C-tariquidar (5.0 ± 0.8 MBq/kg corresponding to 0.3 ± 0.1 nmol/kg unlabeled tariquidar and 1.2 ± 0.8 μmol/kg Tween-80) or ¹¹C-elacridar (5.0 ± 0.5 MBq/kg corresponding to 0.2 ± 0.1 nmol/kg unlabeled elacridar and 0.5 ± 0.2 μmol/kg Tween-80) was i.v. injected over 30 s and dynamic emission scans were acquired for 120 min in 26 frames of increasing duration from 15 s to 10 min. For scan 2, (R)-¹¹C-verapamil (5.1 ± 0.7 MBq/kg corresponding to 0.2 ± 0.1 nmol/kg unlabeled (R)-verapamil) was injected over 30 s and dynamic emission scans were acquired for 60 min in 20 frames of increasing duration from 15 s to 10 min. Prior to each radiotracer injection a 5-min transmission scan using two ⁶⁸Ge pin sources was recorded.

Safety Monitoring

All subjects were monitored for safety parameters heart rate and blood pressure until they were discharged from the study ward the morning after the PET scan. In addition, clinical laboratory blood and urine tests were performed. Adverse events were recorded continuously and the relationship to the study drugs was assessed.

Blood Analysis

To determine arterial input function of ¹¹C-tariquidar and ¹¹C-elacridar, blood samples (2 mL each) were manually drawn from the radial artery at 7-s intervals for the first 2 min, followed by samples at 3.5, 5 and 10 min (9 mL each), 20, 30, 40 and 60 min (18 mL each) and 75, 90, 105 and 120 min (9 mL each) after radiotracer injection. Aliquots of blood and plasma were measured for radioactivity in a gamma counter which was cross-calibrated with the PET camera. Plasma samples collected at 3.5, 5, 10, 20, 30, 40 and 60 min were analyzed for radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar using a previously developed solid-phase extraction (SPE) assay (8). In brief, plasma (2-4 mL) was diluted with water (2 mL), spiked with unlabeled elacridar or tariquidar (10 μL, 20 mg/mL in dimethylsulfoxide), acidified with 5 M aqueous hydrochloric acid (40 μL) and loaded on a Sep-Pak vac tC18 cartridge (Waters Corporation, Milford, MA, USA), which had been pre-activated with methanol (3 mL) and water (5 mL). The cartridge was first washed with water (5 mL) and then eluted with methanol (2 mL) followed by aqueous ammonium acetate buffer (0.2 M, pH 5.0, 1.5 mL). Radioactivity in all 3 fractions (plasma, water, methanol/buffer) was quantified in a gamma counter. Radioactivity in the plasma and water fractions contained polar radiolabeled metabolites whereas unchanged ¹¹C-tariquidar and ¹¹C-elacridar were recovered in the methanol/buffer fraction. Recoveries of ¹¹C-tariquidar and ¹¹C-elacridar from the methanol/buffer fraction of the SPE assay ranged from 91-94%. The methanol/buffer fractions from 10, 20, 30, 40 and 60 min samples were further analyzed for radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar by reversed-phase high-performance liquid chromatography (HPLC) using a binary gradient system (see Supplemental Fig. 2 for HPLC conditions). Plasma input functions of ¹¹C-tariquidar and ¹¹C-elacridar were constructed by multiplying total activity concentrations in whole blood (from 0-60 min) with the mean ratios of plasma to whole blood activity (determined from 3.5, 5, 10, 20, 30, 40 and 60 min blood samples) with or without correcting for the fraction of polar radiolabeled metabolites, as determined by SPE, and by subsequent linear interpolation of the activity data. Blood sampling, metabolite analysis and input function generation for the (R)-¹¹C-verapamil scan were performed as described previously (14, 17). In addition, 4 venous 4-mL blood samples were collected from each subject at the end of the tariquidar infusion, at the end of the first PET scan and at the beginning and end of the second PET scan to measure tariquidar plasma concentration levels using a previously described liquid chromatography–tandem mass spectrometry (LC-MS) assay (18).

Image Analysis

T1-weighted MR images, acquired with a Philips Achieva 3.0T scanner (Philips Medical Systems, Best, The Netherlands), and the corresponding PET data were processed with Analyze 8.0 (Biomedical Imaging Resource, Mayo Foundation, MN, USA) and SPM5 (Wellcome Department of Imaging Neuroscience, UCL, UK) software as described previously (14). By using the Hammersmith n30r83 3-dimensional maximum probability atlas of the human brain (19) a template of preset volumes of interest (VOIs) was applied to the PET images to extract time-activity curves (TACs) for the following 7 gray matter regions: whole brain, hippocampus, cerebellum, caudate nucleus, putamen, thalamus and gyrus precentralis. These regions were chosen based on a previous preclinical study (20). For comparison of TACs between individual subjects, radioactivity concentrations (kBq/mL) were normalized to injected radiotracer amount and expressed as standardized uptake values (SUVs). TACs of total radioactivity in whole blood multiplied by individual fractions of blood volume in brain (V_b) derived from kinetic modeling (see next section) were used to correct brain TACs for radioactivity in the vascular compartment.

Kinetic Modeling

For ¹¹C-tariquidar and ¹¹C-elacridar, datasets from 0 to 60 min of scan 1 (i.e. before tariquidar administration) were used for analysis. Standard 1-tissue 2-rate-constant (1T2K) and 2-tissue 4-rate-constant (2T4K) compartment models were fitted to the time-activity data of each VOI using plasma input functions either uncorrected or corrected for radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar. Time delays of 1 s to 4 s were considered in the input functions to account for the differences in the time course of activity between the arterial catheter and the arterial capillaries in the brain. From the fits, rate constants K₁, k₂ (1T2K model) and K₁, k₂, k₃, k₄ (2T4K model), respectively, as well as blood volume fraction in brain V_b were estimated and the distribution volume (V_T) of the respective brain VOI was calculated. Fits were performed by a software developed in-house using the method of weighted nonlinear least squares as implemented in the Optimization Toolbox of MATLAB (Mathworks, Natick, MA, USA). The quality of the fits was judged by visual inspection of observed and predicted activities together with the pattern of the residuals, by the correlation between observed and predicted activities, and by estimating parameter uncertainties (standard errors) from the diagonal elements of the covariance matrix as obtained from the inverse of the corresponding Fisher information matrix. Model selection (1T2K versus 2T4K) was based on the Akaike information criterion (AIC) (21). For (R)-¹¹C-verapamil scans a 2T4K model with plasma input function corrected for polar radiolabeled metabolites was used (17). Model outcome parameters obtained with the software developed in-house were cross checked using the General Kinetic Modeling Tool in PMOD (version 2.6.1, PMOD group, Switzerland).

Analysis of ABCG2 and ABCB1 Single Nucleotide Polymorphisms

Venous blood (4 mL) was drawn during the screening examination from all study participants for assessment of common ABCG2 and ABCB1 single nucleotide polymorphisms (SNPs) using previously described procedures (22, 23). For ABCG2, the C421A variant (ABI TaqMan Genotyping Assay C__15854163_70; Applied Biosystems, Rotkreuz, Switzerland) and for ABCB1 the C1236T, G2677T and C3435T SNPs were determined.

Statistical Analysis

Statistical analysis was performed using Prism 5.0 software (GraphPad Software, La Jolla, CA, USA). A value of P less than 0.05 was considered significant.

RESULTS

Five subjects underwent ¹¹C-tariquidar and (R)-¹¹C-verapamil PET scans and 1 subject underwent only a ¹¹C-tariquidar PET scan. One subject was excluded from data analysis due to technical problems with the generation of the input function of ¹¹C-tariquidar. Five further subjects underwent ¹¹C-elacridar and (R)-¹¹C-verapamil PET scans.

There were no adverse or clinically detectable pharmacological effects related to the radiotracers injected in any of the 11 subjects. Dysgeusia and orthostatic hypotonia occurred as adverse events in 4 and 2 subjects, respectively, and were most likely related to the infusion of unlabeled tariquidar.

After the injection of ¹¹C-tariquidar or ¹¹C-elacridar, brain PET signal was very low (~0.1 SUV) and distributed uniformly throughout the brain with the exception of intense focal uptake in the venous sinus and the choroid plexus (Fig. 2). TACs of ¹¹C-tariquidar or ¹¹C-elacridar in plasma and whole brain gray matter are shown in Figure 3. One study subject who was scanned with ¹¹C-elacridar (subject 10) showed 2-3 fold higher activity concentrations in plasma and 3-4 fold higher activity concentrations in brain (Supplemental Fig. 1) as compared with the other study subjects and was therefore not included in the mean TACs shown in Fig. 3. At 60 min after injection of ¹¹C-tariquidar or ¹¹C-elacridar, unlabeled tariquidar was administered as an i.v. infusion over 30 min and PET data acquisition continued for a further 60 min (see Fig. 1). In response to tariquidar infusion, there was a small but clearly visible rise in brain PET signal both for ¹¹C-tariquidar and ¹¹C-elacridar, whereas plasma activity concentrations were essentially unchanged (Fig. 3). Brain PET signal (SUV) of ¹¹C-tariquidar was increased on average by 27 ± 15% (P = 0.014, paired t-test) at the end of scan 1 as compared with a time point before tariquidar infusion (45 min), whereas PET signal of ¹¹C-elacridar was increased by 21 ± 15% (P = 0.014). Interestingly, subject 10 showed only a very weak response to tariquidar infusion (+4% increase in brain SUV, Supplemental Fig. 1). Plasma activity concentrations before and after tariquidar infusion were not significantly different for ¹¹C-tariquidar (P = 0.368, paired t-test) and ¹¹C-elacridar (P = 0.283). In Supplemental Table 1, tariquidar plasma concentration levels at 4 different time points after tariquidar infusion are given for individual subjects.

FIGURE 2.

Representative PET summation images (0-60 min) of ¹¹C-tariquidar (A) and ¹¹C-elacridar (B) in coronal (left), transaxial (middle) and sagittal views (right). Activity concentration is expressed as SUV and the radiation scale is set from 0 to 3. Anatomical structures are labeled using white arrows (1, choroid plexus; 2, venous sinus).

FIGURE 3.

Time-activity curves (mean SUV ± SD) of ¹¹C-tariquidar (open squares, n = 5) and ¹¹C-elacridar (open circles, n = 4, subject 10 not included) in arterial plasma, uncorrected for radiolabeled metabolites (A), and in whole brain gray matter, corrected for radioactivity in vasculature (B). The start of i.v. tariquidar infusion (3 mg/kg, over 30 min) is indicated by an arrow.

Radioactivity in plasma samples collected at time points between 3.5 and 60 min after injection was analyzed for radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar using a SPE/HPLC assay. At all studied time points the fraction of polar radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar in plasma was <0.25 (Fig. 4). Radio-HPLC analysis of methanol/buffer eluates from the SPE assay did not show any radioactive species other than unchanged parent compound (Supplemental Fig. 2). Plasma-to-whole blood ratios of total radioactivity over time were rather stable, i.e. 1.34 ± 0.08, 1.39 ± 0.12 and 1.36 ± 0.08 at 3.5, 30 and 60 min after injection of ¹¹C-tariquidar (n = 5) and 1.51 ± 0.10, 1.51 ± 0.12 and 1.48 ± 0.10 at the same time points after injection of ¹¹C-elacridar (n = 5).

FIGURE 4.

Fractions (mean ± SD) of unchanged ¹¹C-tariquidar (open squares, n = 5) and ¹¹C-elacridar (open circles, n= 5) in arterial plasma over time as determined by solid-phase extraction assay. The shown values are not corrected for recoveries of ¹¹C-tariquidar and ¹¹C-elacridar, which ranged from 91-94%.

Data was modeled using 1T2K and 2T4K models with plasma input functions either uncorrected (Table 1) or corrected (Supplemental Table 2) for radiolabeled metabolites. The 2T4K model provided better fits (i.e. lower AIC values) than the 1T2K model. Representative 2T4K model fits for whole brain are shown in Figure 5. K₄ values in the 2T4K model were low, pointing to partially irreversible behavior (Table 1). Outcome parameters were similar when metabolite-corrected or uncorrected input functions were used, which supported the notion that the influence of radiolabeled metabolites on radiotracer kinetics was negligible. V_T and K₁ values of subject 10 in whole brain were approximately 1.5 fold above average (2T4K: K₁: 0.007; V_T: 0.51), whereas other outcome parameters were within normal range.

TABLE 1.

Outcome Parameters for Whole Brain Gray Matter of the 1T2K and 2T4K Models ^* for ¹¹C-Tariquidar and ¹¹C-Elacridar

	1T2K		2T4K
	11C-tariquidar	11C-elacridar	11C-tariquidar	11C-elacridar
K1 (mL·mL−1·min−1)	0.003±0.001 (0.002-0.004)	0.003±0.001 (0.002-0.003)	0.004±0.001 (0.003-0.004)	0.005±0.001 (0.004-0.005)
k2 (min−1)	0.013±0.003 (0.010-0.016)	0.012±0.003 (0.009-0.014)	0.076±0.023 (0.055-0.097)	0.124±0.031 (0.094-0.155)
k3 (min−1)			0.100±0.031 (0.073-0.127)	0.108±0.032 (0.076-0.140)
k4 (min−1)			0.016±0.007 (0.010-0.022)	0.014±0.006 (0.007-0.020)
VT (mL·mL−1)	0.23±0.04 (0.20-0.27)	0.24±0.09 (0.15-0.33)	0.35±0.07 (0.29-0.42)	0.36±0.07 (0.29-0.43)
V b	0.050±0.008 (0.043-0.057)	0.046±0.003 (0.042-0.049)	0.046±0.006 (0.042-0.051)	0.042±0.003 (0.038-0.045)
AIC	−34.1±8.5	−33.8±7.7	−42.2±6.3	−49.7±1.9

^*Data from 0-60 min was used and plasma input function was not corrected for radiolabeled metabolites.

Outcome parameters are given as mean ± SD (¹¹C-tariquidar: n = 5, ¹¹C-elacridar: n = 4, subject 10 excluded). In parentheses the 95% confidence intervals of the mean parameter estimates are given. K₁, k₂, k₃, k₄ : rate constants for exchange of radioactivity between plasma and brain compartments; V_T : distribution volume; V_b : blood volume fraction in brain; AIC: Akaike information criterion

FIGURE 5.

Representative fits (bold solid lines) from 2T4K model using plasma input functions uncorrected for radiolabeled metabolites (open circles) for whole brain gray matter (open squares) for ¹¹C-tariquidar (A) and ¹¹C-elacridar (B).

Regional 2T4K model-derived V_Ts of ¹¹C-tariquidar and ¹¹C-elacridar are shown in Figure 6. There was no statistically significant correlation between V_Ts of ¹¹C-tariquidar or ¹¹C-elacridar and V_T s of (R)-¹¹ C-verapamil in different brain regions (Fig. 7). All study participants were genotyped for common functional SNPs of ABCG2 and ABCB1 (Supplemental Table 3). Subject 10 was a homozygous carrier of the TTT haplotype (1236T, 2677T, 3435T) in the ABCB1 gene, but carried the wild-type ABCG2 C421A allele.

FIGURE 6.

Mean (+SD) 2T4K model-derived V_T values (using plasma input function uncorrected for radiolabeled metabolites) for ¹¹C-tariquidar (black bars, n = 5) and ¹¹C-elacridar (white bars, n = 4, subject 10 not included) in different brain regions. HI = hippocampus; CE = cerebellum; CN = caudate nucleus; PU = putamen; TH = thalamus; GP = gyrus precentralis.

FIGURE 7.

Correlation of V_Ts of ¹¹C-tariquidar (A, n = 5) and ¹¹C-elacridar (B, n = 4, subject 10 not included) (using 2T4K model and plasma input function uncorrected for radiolabeled metabolites) with V_Ts of (R)-¹¹ C-verapamil in hippocampus, cerebellum, caudate nucleus, putamen, thalamus and gyrus precentralis. Solid lines represent linear regression fits (r = Pearson correlation coefficient).

DISCUSSION

This study is to our knowledge the first to describe the brain distribution and pharmacokinetics of radiolabeled third-generation Pgp inhibitors in humans. We found that brain PET signal of ¹¹C-tariquidar and ¹¹C-elacridar, when administered in tracer doses, was very low (Fig. 2), which was consistent with previous in vitro and in vivo data suggesting that these substances are concentration-dependently transported by Pgp and BCRP at the BBB (6-11, 16).

In analogy to a recent human PET study with (R)-¹¹C-verapamil (17), we administered unlabeled tariquidar during the ¹¹C-tariquidar and ¹¹C-elacridar PET scans (Fig. 1). Administration of Pgp inhibitor during the PET scan has been shown to allow for detection of Pgp efflux transport of a radiotracer at the BBB (24). The employed tariquidar dose (3 mg/kg), when given at 1 h before start of the PET scan, corresponded to the half-maximum effect dose for inhibition of Pgp at the human BBB (18). Despite the fact that we used a higher tariquidar dose in this study than in the previous (R)-¹¹C-verapamil study (2 mg/kg), we observed a much less pronounced rise in brain PET signal of ¹¹C-tariquidar and ¹¹C-elacridar in response to tariquidar than for (R)-¹¹C-verapamil (Fig. 3). The most likely explanation for this different behavior could be that (R)-¹¹C-verapamil is only transported by Pgp at the human BBB whereas ¹¹C-tariquidar and ¹¹C-elacridar are dual Pgp/BCRP substrates. Even though tariquidar in pharmacological doses is a dual Pgp and BCRP inhibitor, preclinical data suggests that higher doses than the presently employed 3 mg/kg dose would be needed to inhibit BCRP (16). Therefore it can be assumed that while Pgp was at least half-maximally inhibited, BCRP was still fully functional during administration of 3 mg/kg tariquidar and that BCRP efflux may have effectively limited brain access of both radiotracers. This behavior is in fact completely in line with what would be expected from dual Pgp/BCRP substrates (2). It has now been documented in the literature for several dual Pgp/BCRP substrates that their brain uptake is only moderately increased in single transporter knockout mice, i.e. Mdr1a/b^(−/−) or Bcrp1^(−/−) mice, relative to wild-type mice, whereby disproportionally large increases are seen in combination knockout mice (Mdr1a/b^(−/−)Bcrp1^(−/−)) (1). This is because of a cooperative effect of both transporters in preventing brain distribution of dual substrates (1). In line with this, brain-to-blood ratios of ¹¹C-tariquidar were found to be 1.0 ± 0.1 in wild-type, 3.3 ± 0.4 in Mdr1a/b^(−/−), 1.8 ± 0.1 in Bcrp1^(−/−) and 14.4 ± 1.7 in Mdr1a/b^(−/−)Bcrp1^(−/−) mice (9).

We used 1T2K and 2T4K models for analysis and found that the 2T4K model provided better data fits (Fig. 5, Table 1). Importantly, we observed only low levels of radiolabeled metabolites of ¹¹C-tariquidar and ¹¹C-elacridar in plasma during the time course of the PET scan (Fig. 4 and Supplemental Fig. 2), which obviates the need to correct plasma input functions for radiolabeled metabolites. This poses a significant advantage over (R)-¹¹C-verapamil, which is extensively metabolized and gives radiolabeled metabolites, which are partly taken up into brain and may thereby confound PET measurements with this radiotracer (25). K₁ values of ¹¹C-tariquidar and ¹¹C-elacridar were up to 10-fold lower than those of (R)-¹¹C-verapamil (Table 1) (17), which suggested that ¹¹C-tariquidar and ¹¹C-elacridar were more effectively effluxed at the BBB than (R)-¹¹C-verapamil. Washout of ¹¹C-tariquidar and ¹¹C-elacridar from brain was very slow (Fig. 3). The good fits obtained with the 2T4K model (Fig. 5) suggest the presence of a second, deep brain compartment, the physiological correlate of which may be intracellular trapping of radiotracer in lysosomes, as recently reported for tariquidar (26).

Interestingly, subject 10 had approximately 1.5-fold higher V_T and K₁ values of ¹¹C-elacridar than the other subjects. Genetic analysis revealed that this subject was a homozygous carrier of the TTT haplotype in the ABCB1 gene, which has been associated in some studies with decreased Pgp expression/function and altered pharmacokinetics of Pgp substrate drugs (27). Thus, it is tempting to speculate that increased brain distribution of ¹¹C-elacridar in subject 10 may have been caused by decreased Pgp function at the BBB, although this finding needs to be confirmed in a larger cohort.

The original intention of developing ¹¹C-tariquidar and ¹¹C-elacridar was to visualize Pgp expression levels in brain. To investigate any possible Pgp binding of ¹¹C-tariquidar and ¹¹C-elacridar at the BBB, we combined ¹¹C-tariquidar and ¹¹C-elacridar PET scans with (R)-¹¹C-verapamil scans, after administration of 3 mg/kg tariquidar. We have shown previously that performing (R)-¹¹C-verapamil PET scans after half-maximal inhibition of Pgp is more sensitive to map regional differences in cerebral Pgp function than baseline scans (20). This is because brain uptake of (R)-¹¹C-verapamil is relatively low in baseline scans, which may make it difficult to detect any further reduction in brain uptake due to regionally increased Pgp activity.

Moreover, half-maximal Pgp inhibition may increase the percentage of (R)-¹¹C-verapamil in brain relative to its radiolabeled metabolites which are not or to a lower extent transported by Pgp (25), thereby providing better sensitivity to study regional Pgp function in brain. If ¹¹C-tariquidar and ¹¹C-elacridar bind to Pgp and (R)-¹¹C-verapamil is effluxed by Pgp, brain uptake of ¹¹C-inhibitors and ¹¹C-substrates should in theory be inversely related, i.e. brain regions with high uptake of ¹¹C-inhibitor should show low uptake of ¹¹C-substrate and vice versa. Such a behavior has in fact been observed in a study in rats in which the same paired-scan protocol was employed as in the present study (13). Contrary to these preclinical results, we found no significant correlation between regional brain V_Ts of ¹¹C-tariquidar or ¹¹C-elacridar and (R)-¹¹C-verapamil (Fig. 7). This strongly suggests that other than at the rodent BBB, where ¹¹C-tariquidar and ¹¹C-elacridar appeared to be both effluxed by Pgp and also to bind to some extent to Pgp, possibly via distinct binding sites, the behavior of these probes at the human BBB is dominated by Pgp/BCRP efflux. A possible reason for this could be species-dependent differences in transporter expression levels at the human and rodent BBB. Shortly after we developed ¹¹C-elacridar and ¹¹C-tariquidar, data on absolute Pgp expression levels in isolated brain microvessels of various species, including humans, appeared in the literature (28). Using an LC/MS-based quantitative proteomics approach the expression of MDR1 in isolated human brain microvessels was found to be 2.3-fold lower than that of mdr1a in mouse brain microvessels (6.1 ± 1.7 versus 14.1 ± 2.1 fmol/μg protein) (28). A commonly used predictor for the magnitude of specific signal that can be obtained with a radioligand for imaging of a molecular target is the ratio of the concentration of target protein in whole brain (B_max) to the equilibrium dissociation constant (K_D) of the radioligand (29). B_max/K_D is also referred to as binding potential and should ideally be >5 to achieve measurable specific binding. To estimate the B_max of Pgp in whole brain one needs to know the density of brain capillary endothelial cells as a proportion of brain tissue. This is different from total brain capillary volume (~5%), which is much higher because it includes blood volume. Assuming a brain capillary endothelial cell volume of 0.2% of total brain volume (30) and a protein content of brain capillaries of 10%, the B_max of Pgp in human brain is 1.3 nM. Based on a K_D value for binding of ³H-tariquidar to Chinese hamster ovary resistant cells of 5.1 nM (4), the binding potential is 0.26, which is clearly too low to visualize Pgp in human brain with ¹¹C-tariquidar. Another factor which might contribute to the inability of ¹¹C-tariquidar and ¹¹C-elacridar to visualize Pgp density in human brain could be that BCRP expression levels and BCRP/Pgp expression ratios are higher at the human than at the rodent BBB (BCRP expression (fmol/μg protein): humans: 8.14 ± 2.26, mouse: 4.41 ± 0.69; BCRP/Pgp expression ratio: humans: 1.3, mouse: 0.3) (28). Thus, a relatively higher contribution of BCRP efflux at the human than at the rodent BBB could further reduce any possible Pgp binding of ¹¹C-tariquidar and ¹¹C-elacridar at the human BBB. In contrast to the BBB, Pgp density in the previously investigated murine tumor model may have been higher and BCRP density lower, which could explain, possibly in combination with the absence of tight junctions in tumor capillaries, why ¹¹C-tariquidar was able to visualize Pgp expression in the tumor model (12). More generally, the ability of ¹¹C-tariquidar and ¹¹C-elacridar to provide a Pgp-binding signal may be tissue-specific, depending on Pgp relative to BCRP expression levels. Following these considerations, an effective radioligand to visualize Pgp density in human brain should ideally possess an at least ten-fold higher Pgp binding affinity than tariquidar and not be transported by BCRP and Pgp.

Despite the inability of ¹¹C-tariquidar and ¹¹C-elacridar to visualize cerebral Pgp density, these probes may find use as an entirely new class of PET tracers (as opposed to Pgp-selective substrates like (R)-¹¹C-verapamil or ¹¹C-N-desmethyl-loperamide) to study the interplay of Pgp and BCRP at the human BBB in limiting brain uptake of dual substrates (16). This is important because several clinically used drugs (e.g. tyrosine kinase inhibitors) are dual substrates of Pgp and BCRP (1) and no dual Pgp/BCRP substrates are currently available for PET studies in humans. Thus ¹¹C-tariquidar or ¹¹C-elacridar could be used as generic probes to assess brain penetration of dual Pgp/BCRP substrates in healthy subjects or glioma patients under various conditions, for example after treatment with Pgp/BCRP inhibitors like elacridar or in subjects with functional SNPs in the ABCG2 and/or ABCB1 genes.

CONCLUSION

Brain PET signal of ¹¹C-tariquidar and ¹¹C-elacridar was found to be very low in humans and only moderately sensitive to a pharmacological challenge with unlabeled tariquidar, which was consistent with the behavior of dual Pgp/BCRP substrates. ¹¹C-Tariquidar and ¹¹C-elacridar were not able to visualize Pgp density in human brain, which highlights the considerable challenge in visualizing low density multidrug efflux transporters at the human BBB with PET and furthermore suggests that compounds with subnanomolar Pgp binding affinities, which are currently not available, may be needed to develop successful radioligands. Despite their inability to visualize Pgp density, ¹¹C-tariquidar and ¹¹C-elacridar may find future clinical use as an entirely new class of radiotracers to assess the interplay of Pgp and BCRP at the human BBB.