Blocking of fatty acid amide hydrolase activity with PF-04457845 in human brain: a positron emission tomography study with the novel radioligand [¹¹C]CURB

Abstract

Positron emission tomography with [¹¹C]CURB was recently developed to quantify fatty acid amide hydrolase (FAAH), the enzyme responsible for hydrolyzing the endocannabinoid anandamide. This study investigated the test–retest reliability of [¹¹C]CURB as well as its in vivo specificity and the validity of the kinetic model by using the highly specific FAAH inhibitor, PF-04457845. Five healthy volunteers completed test–retest [¹¹C]CURB scans 1 to 2 months apart and six subjects completed baseline and blocking scans on the same day after PF-04457845 (p.o.) administration (1, 4, or 20 mg; n=2 each). The composite parameter λk₃ (an index of FAAH activity, λ=K₁/k₂) was estimated using an irreversible two-tissue compartment model with plasma input function. There were no clinically observable responses to oral PF-04457845 or [¹¹C]CURB injection. Oral administration of PF-04457845 reduced [¹¹C]CURB binding to a homogeneous level at all three doses, with λk₃ values decreased by ⩾91%. Excellent reproducibility and good reliability (test–retest variability=9% intraclass correlation coefficient=0.79) were observed across all regions of interest investigated. Our findings suggest that λk₃/[¹¹C]CURB is a reliable, highly sensitive, and selective tool to measure FAAH activity in human brain in vivo. Moreover, PF-04457845 is a highly potent FAAH inhibitor (>95% inhibition at 1 mg) in living human brain.

Introduction

The human endocannabinoid system is believed to modulate a number of human brain functions (e.g., mood, motor, appetite, sleep, pain, inflammation, and cognition) and be involved in several neuropsychiatric disorders, including depression,¹ anxiety,² schizophrenia,³ substance abuse,⁴ and neurodegeneration.⁵ Considerable efforts have been invested in the development of imaging tools to investigate the role, if any, of components of the endocannabinoid system in the pathophysiology of these disorders.^{6, 7, 8, 9, 10}

Fatty acid amide hydrolase (FAAH; EC3.5.1.99)^{11, 12} is a serine hydrolase, prominently expressed in central nervous system and in the periphery. Its main function is to degrade a wide range of fatty acid amides and esters, with preference for primary amides and arachidonoyl and oleoyl substrates, including the major endocannabinoid anandamide (N-arachidonoylethanolamide) but also more abundant oleoylethanolamide and palmitoylethanolamide. A number of preclinical^{1, 13, 14} and gene association studies of a major functional single-nucleotide polymorphism of FAAH (rs324420, C385A)^{15, 16} that leads to lower levels of the enzyme¹⁷ have linked FAAH levels to neuropsychiatric disorders including mood and anxiety.¹⁸ As such, FAAH inhibitors, which elevate levels and sustain duration of action of N-arachidonoylethanolamide by blocking its hydrolysis, have been avidly developed as potential new therapeutic tools and attractive alternatives (less psychoactive effect/safe profile) to ‘conventional' agonists/partial agonists of the endocannabinoid receptor CB1 for the treatment of a range of disorders.^{19, 20, 21, 22} In this regard, numerous FAAH inhibitors have been developed including the aryl urea-based inhibitor PF-04457845 from Pfizer. PF-04457845 is a highly potent (IC₅₀=7.2 nmol/L, with k_inact/K_i=40,300/M/s, where k_inact is the first-order rate constant of enzyme inactivation at infinite inhibitor concentration and K_i is the inhibitor dissociation constant)^{23, 24} and highly specific irreversible FAAH inhibitor in vitro and ex vivo based on the extensive preclinical characterization. In human clinical trials,²⁵ this investigational drug was shown to have good safety and tolerability profile and inhibited FAAH activity in peripheral blood leukocytes (>97% at 0.3 mg, p.o.) and increased plasma concentrations of fatty acid amides by 3.5 to 10 fold although its effectiveness and optimal dose regimen in the living human brain is unknown.

We recently reported the synthesis of [¹¹C]CURB ([¹¹C-carbonyl]-6-hydroxy-[1,10-biphenyl]-3-yl cyclohexylcarbamate (URB694)),²⁶ the first available positron emission tomography (PET) radiotracer for imaging FAAH, and showed that a two-tissue compartment model with irreversible trapping (2-TCMi) fitted the regional time-activity curves (TACs) measured in vivo in the human brain.²⁷ The composite parameter λk₃ (λ=K₁/k₂) was recommended as the preferred index for PET quantification of FAAH activity given its high identifiability and direct proportion to the irreversible trapping rate constant k₃.²⁷ However, the sensitivity of [¹¹C]CURB λk₃ to changes in FAAH activity in vivo has not been assessed and the human brain selectivity of the radiotracer and the reproducibility of the kinetic parameters (λk₃, k₃, K_i) remain to be determined.

With the availability of the Pfizer FAAH inhibitor PF-04457845, the aims of the current study were (1) to provide direct evidence that binding of [¹¹C]CURB in human brain is mediated via FAAH activity; (2) to validate the 2-TCMi, i.e., identify the sensitivity of the kinetic parameters to actual changes in brain FAAH activity; (3) to explore the in vivo occupancy of PF-04457845 at clinically relevant doses; (4) to provide test–retest quantification of parameters of interest.

Materials and methods

Subjects

All procedures were approved by the Centre for Addiction and Mental Health Research Ethics Board and complied with ethical standards of the 1975 Helsinki Declaration (5th revision, 2000). Subjects were recruited from the local community in Toronto, Canada using Internet advertisements. After provision of written informed consent, they participated in a comprehensive medical/screening interview.

Twelve healthy adult subjects aged 19 to 53 were recruited. All met the following criteria: (1) No past or present significant medical conditions and/or neurologic illnesses or head trauma; (2) Normal physical exam, 12-lead electrocardiogram, complete blood count/blood chemistry, electrolytes, renal and liver function tests; (3) No past or present Axis I psychiatric diagnoses as per Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders (DSM-IV); (4) No presence of metal objects in the body or implanted electronic devices that preclude safe MR scanning; (5) No claustrophobia; (6) No current pregnancy (as per serum β-HCG) or breastfeeding; (7) No current use or use during the previous month of medication that may affect the central nervous system or positive drug screening for drugs of abuse (as per dip stick urine toxicology screen (10-drug test panel from BTNX Inc., Markham, ON, Canada; at screening visit and at each visit afterwards) and broad-spectrum screen); (8) No exposure to radiation in the last 12 months exceeding permissible limit for subjects participating in research. (9) No abnormal body mass index (19 to 30). All subjects were also reminded not to consume grapefruit/pomelo containing products 7 days before the scan.

Subjects enrolled were scheduled to particiapte in either the test–retest study: two PET scans with [¹¹C]CURB performed >1 month apart (n=6); or the blocking study: two PET scans, a baseline and post-PF-04457845 scan performed on the same day, 4 hours apart (between the two tracer injections), in fixed order (n=6). Subjects were not blinded to dosing schedule. Two subjects participated in both the studies.

PF-04457845 Dosing Regimen

Clinical trials²⁵ have shown that a single 4 mg oral tablet of PF-04457845 produces a plasma maximum of the drug of 34.6±14.5 (mean±s.d.) ng/mL at 0.5  to 2 hours postdose and an inhibition of blood leukocyte FAAH activity of >97% within 2 hours. Plasma elimination half-life at this dose is around 16 to 20 hours and blood FAAH activity returns to baseline within 2 weeks. In the current study, we chose to administer 4 mg of PF-04457845 (p.o., open label) as well as two other doses within the clinical trial single-dose range of Pfizer: 20 mg and 1 mg (n=2 at each dose). The 1-mg dose was obtained by splitting the 4-mg pill. The doses were administered after completion of the baseline scan and ~2 hours before the start of the blocking scan. Subjects were reminded to have a light meal in the morning and a light lunch was permitted between the scans as food is known to have little effect on overall PF-04457845 pharmacokinetics.²⁵

Radiosynthesis of [¹¹C]CURB

The radiosynthesis of [¹¹C]CURB has been described earlier.²⁶ The method utilizes a novel one-pot automated [¹¹C]CO₂ fixation reaction to form [carbonyl-¹¹C]-O-arylcarbamates at high specific activities and in radiochemical purities exceeding 97%.²⁸ The purified radiotracer was formulated for injection as a sterile, pyrogen-free solution in saline containing 0.5% Tween-80 (Sigma-Aldrich, St Louis, MO, USA).

Image Acquisition and Reconstruction

The PET scanning was performed using a 3D HRRT brain tomograph (CPS/Siemens, Knoxville, TN, USA), which measures radioactivity in 207 slices with an interslice distance of 1.22 mm. The detectors of the HRRT are an LSO/LYSO phoswich detector, with each crystal element measuring 2 × 2 × 10 mm³.

Participants were scanned in supine position. A custom-fitted thermoplastic mask was made for each subject and used with a head fixation system (Tru Scan Imaging, Annapolis, MD, USA) to reduce movement during PET measurements. After being placed on the scanning table, a saline solution of ~370±40 MBq (~10±1 mCi) of [¹¹C]CURB was injected over a 1-minute period at a constant rate using a Harvard infusion pump (Harvard Apparatus, Holliston, MA, USA) into an intravenous line placed in an antecubital vein. Transmission scans were acquired using a single photon point source, ¹³⁷Cs (T_1/2=30.2  years, Eγ=662  keV), and used to correct the emission scans for the attenuation of 511 keV photons through tissue and head support. Radioactivity in brain was measured in a series of sequential frames of increasing duration. Scanning time was 90 minutes. The emission list mode data were rebinned into a series of 3D sinograms. The 3D sinograms were gap-filled, scatter corrected, and Fourier rebinned into 2D sonograms. The images were reconstructed from the 2D sinograms using a 2D filtered-back projection algorithm, with a HANN filter at Nyquist cutoff frequency. The reconstructed image has 256 × 256 × 207 cubic voxels measuring 1.22 × 1.22 × 1.22 mm³ and the resulting reconstructed resolution is close to isotropic 4.4 mm, full width at half maximum in plane and 4.5 mm full width at half maximum axially, averaged over measurements from the center of the transaxial field of view to 10 cm off-center in 1.0 cm increments.

Subjects underwent standard proton density weighted brain magnetic resonance imaging on a Discovery MR750 3T MRI scanner (General Electric, Mississauga, ON, Canada. Slice thickness, 2 mm, interleaved; number of slices, 84; repetition time, 6,000 ms; echo time, 8 ms; number of excitations, 2; acquisition matrix, 256 × 192; field of view, 22 × 16.5 cm) to aid region of interest (ROI) delineation of the PET images performed using ROMI (details in Rusjan et al²⁹). Regions delineated (n=12) included: prefrontal, temporal, occipital, parietal, anterior cingulate, insular and cerebellar cortices, hippocampus, amygdala, caudate, putamen, and thalamus.

Arterial Sampling of [¹¹C]CURB in Plasma

Arterial samples were taken continuously at a rate 350 mL/h for the first 7.25 minutes after [¹¹C]CURB-injection and 150 mL/h for the next 15 minutes. The continuous early arterial blood radioactivity levels were counted using an automatic blood sampling system (Model PBS-101, Veenstra Instruments, Joure, The Netherlands). In addition, 4 to 10 mL manual samples were taken 3, 7, 12, 15, 20, 30, 45, 60 and 90 minutes after injection. An aliquot of each blood sample was taken to measure radioactivity concentration in total blood. The remaining blood was centrifuged (1,500 g × 5 minutes) and a plasma aliquot counted together with the total blood sample using a Packard Cobra II γ counter (Chicago, IL, USA) cross-calibrated with the PET system. The blood-to-plasma ratios were determined from the manual samples to correct the blood radioactivity TAC measured by automatic sampling and to generate the plasma radioactivity curve. A biexponential function was used to fit the blood-to-plasma ratios. The remaining volume of each manual plasma sample (except the one at 15 minutes) was used to determine the concentration of the parent radioligand and of its metabolites in plasma. Each plasma samples (1 to 5 mL) was spiked with 20% v/v of 50% aqueous acetic acid to disrupt plasma protein binding and applied to an Oasis HLB cartridge (Waters, Mississauga, ON, Canada, 6 mL, 200 mg, 30 μ) preconditioned with ethanol (5 mL) and water (10 mL). The cartridge was then washed successively with 5 mL of 5% MeOH, and 5 mL of CH₃CN/H₂O+0.1 N ammonium formate (45/55' v/v). The radioactivity in each of the three eluted fractions as well as the cartridge (containing only unmetabolized [¹¹C]CURB) was counted in the gamma counter to determine the percentage of unmetabolized parent compound in the plasma samples. The method was validated against an HPLC method of analysis,³⁰ which did not allow as rapid a throughput of plasma samples. A Hill function was used to fit the percentage of unmetabolized tracer.

Plasma Level of PF-04457845 and Self-Reported Drug Effects

During scans that were part of the blocking study, arterial blood was taken at four time points (0 [before drug] and approximately 60 minutes, 120 minutes, and 210 minutes after dosing) for plasma measurements of PF-04457845 levels. Those were analyzed as described²⁵ by LC/MS/MS. After oral administration of PF-04457845 (and during baseline scans), subjects were also asked to answer questions on drug-related feeling on visual analog scales (measuring drug ‘liking' (from ‘dislike a lot' to ‘neutral' to ‘like a lot'), ‘good' and ‘bad' effects of the drug as well as drowsiness and calmness (from ‘not at all' to ‘extremely')) and rate side effects on the UKU side effects rating scale during 1-day and 2-week follow-up. Blood pressure and heart rate were monitored throughout.

Free Fraction

Arterial blood samples were collected from each subject before their PET scan. The whole blood was centrifuged and the plasma then stored at −80°C, and [¹¹C]CURB binding to plasma proteins was later determined by the ultrafiltration method.³¹

Kinetic Analysis

Time-activity curves acquired over 60 minutes (as per Rusjan et al²⁷) in each ROI (listed in Table 2) were extracted using ROMI (details in Rusjan et al²⁹) and analyzed using a two-tissue compartment model with irreversible binding to the second compartment (2-TCMi, k₄ ≡ 0). Briefly, the 2-TCMi has three rate parameters:

where K₁ describes the influx of [¹¹C]CURB from plasma to the free and nonspecific compartment, k₂ the efflux of the tracer from the free and nonspecific compartment to plasma, and k₃ the transfer from the free and nonspecific compartment to the specifically bound compartment. λ is the equilibrium distribution volume of the ligand in the free and nonspecifically bound compartment, F the perfusion or blood flow, PS the permeability surface area product, E the first pass extraction factor: E= 1−e^−PS/F using the Renkin-Crone model, f_ND the tissue-free fraction, k_on the [¹¹C]CURB-FAAH association rate constant, and B_avail the available density of FAAH.

The parameters of interest to quantify specific binding are the composite parameter λk₃, the rate constant k₃, and the net influx constantK_i:

Statistical Analysis

Receptor occupancy, defined as the percentage reduction in [¹¹C]CURB λk₃ from baseline scan to the PF-04457845 exposed state, was calculated for each subject using the following equation:

Differences in [¹¹C]CURB λk₃ and other parameters in ROIs as well as regional receptor occupancies were investigated with repeated-measures ANOVAs (SPSS 20.0, SPSS Inc., Chicago, IL, USA). Sphericity (for the within-subject (WS) factors: dose and ROIs) was assessed with the Mauchly test and, when indicated, correction was made with Greenhouse-Geisser adjustments. When appropriate, least significant difference t-tests, Bonferroni corrected, were applied to determine the significance of regional differences in [¹¹C]CURB binding between conditions (0 mg vs. dose). The test–retest reproducibility of λk₃, K_i, k₃, and K₁ was evaluated in three ways: (1) WS and between-subject (BS) variability were calculated and expressed as the standard deviation and coefficient of variation; (2) test–retest variability (TRV) was calculated as the mean across the subjects of the ratio of absolute difference between measurements to the average between measurements; (3) reliability (intraclass correlation coefficient (ICC)) calculated by SPSS with the one-way random effect model according to the following equation:

where BSMSS and WSMSS are the mean sum of squares BS and WS, respectively. This coefficient is an indicator of the relative contributions of the two sources of variance and while a value of −1 is associated with all variability attributable to WS variance, a value of 1 is associated with all variability associated with BS variance.

Results

Subjects

Twelve healthy volunteers were recruited and nine (5M/4F) completed the full study (two subjects did not return for follow-up scan and one scan of another subject was lost to motion artifact; therefore, the three scans were not included in the current report). Two of the nine subjects (#4 and #5 in Table 1) participated in both the blocking and test–retest study completing a total of three scans (baseline, PF-04457845 dosed and retest, 1 month later). The average age of the sample was 36.4±10.3, their body mass index was within normal range (24.5±2.4 kg/m²), all tested negative for drugs of abuse at the interview (and at all scans) and none were nicotine smokers. There were no differences in scan parameters across conditions (all P>0.5) (Table 1).

Table 1. Subject data and scan parameters.

				PF-04457845	Dose inj (mCi)		Mass inj (μg)		SA (mCi/μmol)
Subject#	Gender	Age (years)	BMI	Dose (mg)	Scan 1	Scan 2	Scan 1	Scan 2	Scan 1	Scan 2
Study 1: Test–Retest
1	F	19	20	N/A	9.24	9.45	1.93	1.76	1489	1669
2	M	53	27	N/A	8.20	9.41	1.81	2.05	1409	1431
3	F	42	23	N/A	10.32	10.28	0.80	2.20	4020	1457
4	M	44	25	N/A	10.56	9.18	1.84	0.35	1788	8086
5	F	25	25	N/A	9.62	9.72	0.62	0.91	4850	3340
Mean (s.d.)		37 (14)	24 (3)		9.6 (0.9)	9.6 (0.4)	1.4 (0.6)	1.4 (0.8)	2711 (1607)	3197 (2846)
Study 2: Blocking
4	M	44	25	4	10.56	9.34	1.84	1.70	1788	1714
5	F	25	25	4	9.62	9.18	0.62	2.97	4850	963
6	M	42	23	20	9.54	9.57	0.73	0.59	4066	5091
7	F	34	26	20	10.30	9.36	0.64	0.63	4987	4628
8	M	34	27	1	10.00	10.48	1.00	0.84	3110	3896
9	M	35	23	1	8.66	9.32	1.91	0.75	1408	3872
Mean (s.d.)		36 (7)	25 (2)		9.8 (0.7)	9.5 (0.5)	1.1 (0.6)	1.3 (0.9)	3368 (1530)	3361 (1650)

BMI, body mass index; Dose inj, dose injected; Mass inj, mass injected; SA, specific activity at the time of the injection.

Safety and Self-Reported Drug Effects

Six subjects completed the PF-04457845 blocking study. The FAAH inhibitor was administered orally at the end of baseline scan at three different doses (1, 4, and 20 mg, n=2 each). There were no clinical effects of [¹¹C]CURB or PF-04457845 based on physiologic monitoring, physical examination, and self-report scales including visual analog scales of ‘drug liking' (neutral), ‘good', and ‘bad' effects of the drug, drowsiness, and calmness (visual analog scales, all P>0.05), and UKU side effects during follow-ups (no side effects reported).

[¹¹C]CURB in Plasma After Fatty Acid Amide Hydrolase Blocking

Injection of [¹¹C]CURB for the blocking scans was started 2 hours after PF-04457845 dosing and was on average 4 hours after the start of the baseline scans. Arterial blood analysis showed that metabolism of the radiotracer was not much different from the baseline scan at 1 and 4 mg doses of the FAAH blocker whereas the 20-mg dose decreased the metabolism, with %unmetabolized [¹¹C]CURB increased from 24% and 32% at baseline to 47% and 51% at the end of scan (90 minutes, Figure 1) for the two subjects, respectively. This difference suggests shared metabolizing mechanism between [¹¹C]CURB and PF-04457845, e.g., by FAAH in the periphery including liver,³² and that the 20 mg versus the 1 and 4 mg dose might also have more markedly inhibited the liver enzyme. The metabolite-corrected [¹¹C]CURB activity in plasma was slightly higher late but not early (within 20 minutes) in the scan after all three doses than that of the baseline, suggesting slightly increased supply of the radiotracer after FAAH blocking. The FAAH blocking did not affect plasma protein binding of [¹¹C]CURB, with no difference (P>0.05) in the free fraction at baseline (1.01±0.04%) and after dosing (0.98±0.07%).

Figure 1.

Plasma metabolism of [¹¹C]CURB expressed as percentage of unmetabolized radiotracer at each time points vs. injected dose at baseline (open circles, lower panels) and at different doses (1 mg, 4 mg and 20 mg) of PF-04457845 (PF, solid circles, lower panels) and the average plasma input functions (SUV, standardized uptake value, of unmetabolized [¹¹C]CURB) at baseline (solid lines, upper panels) and at the different doses of PF-04457845 (dotted lines, upper panels). The insets show details of the first 5 minutes after injection. *For illustrative purposes, we are showing data acquired over 90 minutes; note that kinetic modeling was applied to data truncated at 60 minutes as per Rusjan et al.²⁷

PF-04457845 in Blood

The dose of PF-04457845 ranged from 0.01 to 0.3 mg/kg after the 1, 4, and 20 mg doses. Plasma levels of the drug, expressed either in peak levels (C_max, 4  to 284 ng/mL) or AUC(0, t_last) (10 to 430 ng/mL per hour), were proportional (r<0.0001) to the dose and were consistent with the previous report²⁵ although the time to C_max (t_max) varied among the subjects (t_max≈1 hour, n=3; t_max≈2 hours, n=2; t_max⩾3.5 hours, n=1), in particular at the lower 1 mg and 4 mg doses (Figure 2).

Figure 2.

Plasma levels of PF-04457845 after dosing at 1, 4, and 20 mg in individual subjects. The inset shows enlarged portion of the graph at lower drug levels.

Time-Activity Curves

Consistent with our previous report,²⁷ regional TACs at baseline were heterogeneous, which plateaued after 30 minutes at different levels, reflecting differential irreversible trapping (see Figure 3A for examples of selected brain areas). After dosing with the FAAH inhibitor at 1, 4, and 20 mg, all regional TACs, normalized by injected dose/weight of the subject (standardized uptake value, SUV), despite different peak levels in early frames, decreased quickly below the baseline levels and converged after approximately 40 minutes to the same low levels (less than 10% of baseline activities at the end of the scan). This profile is consistent with a uniform and profound suppression of tracer trapping (see Supplementary Figure 1 for representative brain radioactivity images before and after PF-04457845 administration in one subject).

Figure 3.

Average time-activity curves (TACs) (SUV, standardized uptake value) in selected brain regions of baseline (open symbols in A) and PF-04457845 (PF) blocking (n=2 each at doses of 1, 4, and 20 mg, respectively; closed symbols in A) and test–retest (B, n=5) scans of [¹¹C]CURB. *For illustrative purposes we are showing data acquired over 90 minutes; note that kinetic modeling was applied to data truncated at 60 minutes as per Rusjan et al.²⁷ AMG, amygdala; Hippo, hippocampus; PFctx, prefrontal cortex; PUT, putamen.

Compartment Models

In the baseline condition the 2-TCMi fit significantly better the TACs than 1-TCM as we previously observed. In the blocked condition, the TACs presented a reversible appearance and it was possible to fit either with a 1-TCM or with a 2-TCMi (see Supplementary Figure 2 for representative TACs and curve fittings of a single subject). Based on the goodness of fitting (AIC (Akaike Information Criterion), or Model Selection Criterion (MSC)), while TACs of hippocampus and the cerebellar and cerebral cortices (with the exception of anterior cingulate) were fitted significantly better with the 2-TCMi vs. 1-TCM (ΔAIC=−4.39±1.53 to −11.1±4.0; MSC_(2-TCMi)=3.38±0.70 to 4.50±0.86, MSC_(1-TCM)=3.12±0.67 to 4.01±0.77; P<0.01, paired t-tests), TACs of the rest of the brain regions were fitted equally well with both models (ΔAIC=0.10±1.5 to −3.68±4.44; MSC(2-TCMi)=2.68±0.69 to 4.28±0.73, MSC_(1-TCM)=2.64±0.74 to 4.07±0.64; P⩾0.05), confirming our previous prediction that the first compartment of the 2-TCMi represents nondisplaceable binding.

Kinetic Analysis with Two-Tissue Compartment Model with Irreversible Trapping Model

Scan data truncated at 60 minutes were used for the kinetic analysis (as per Rusjan et al²⁷). Parameters estimated including K₁, k₂, k₃, K_i, and λk₃ in the current cohort of subjects (n=6) at baseline (Table 2) were within the same range as those obtained in the previous cohort (n=6; see Table 3 of Rusjan et al²⁷). After PF-04457845 dosing (1, 4, and 20 mg) λk₃ as well as k₃ and K_i (main effect of treatment irrespective of dose, F(1, 5)=425, 313, and 418, respectively; P<0.0001), but not K₁ (F(1, 5)=3.2, P=0.14) or k₂ (F(1, 5)=4.9, P=0.08), were all markedly decreased across brain regions (Table 2), reflecting occupancy of FAAH by the FAAH inhibitor and reduced trapping (k₃) of the tracer. As shown in Figure 4, overall %inhibition across all (12) regions investigated was 91±3% for λk₃ (90±2%, 93±2%, and 91±3% following the 1 mg, 4 mg, and 20 mg doses, respectively), 93±3% for k₃ (92±2%, 94±1%, and 92±4% following the 1 mg, 4 mg, and 20 mg doses, respectively), and 87±3% for K_i (84±4%, 90±4%, and 86±4% following the 1 mg, 4 mg, and 20 mg doses, respectively). Significant ROI-by-treatment interactions were observed for λk₃ (F(11, 55)=12, P<0.0001) and K_i (F(11, 55)=33, P<0.0001) but not for k₃ (F(11, 55)=0.8, P=0.6), which was explained by regional differences at baseline in λk₃ values (thalamus>other regions>amygdala) and K_i (thalamus>other regions except putamen and parietal cortex; all regions>hippocampus and amygdala).

Table 2. Kinetic rate constants estimated with a 2-TCMi at baseline and after PF-04457845 with 60 minutes of scan data.

60 minutes	K1 (mL/cm3/min)	k2 (1/min)	k3 (1/min)	Ki (mL/cm3/min)	λk3 (mL/cm3/min)
Prefrontal cortex
Baseline	0.28 (0.03)	0.11 (0.02)	0.061 (0.009)	0.102 (0.010)	0.159 (0.015)
PF 1 mg	0.35 (0.02)	0.12 (0.01)	0.007 (0.002)	0.018 (0.004)	0.019 (0.005)
PF 4 mg	0.31 (0.02)	0.10 (0.01)	0.003 (0.001)	0.010 (0.002)	0.010 (0.003)
PF 20 mg	0.30 (0.06)	0.10 (0.01)	0.006 (0.001)	0.015 (0.001)	0.016 (0.001)
Temporal cortex
Baseline	0.29 (0.03)	0.11 (0.01)	0.062 (0.006)	0.106 (0.010)	0.168 (0.018)
PF 1 mg	0.34 (0.01)	0.11 (0.01)	0.007 (0.002)	0.018 (0.004)	0.020 (0.005)
PF 4 mg	0.32 (0.01)	0.10 (0.01)	0.003 (0.001)	0.011 (0.001)	0.011 (0.001)
PF 20 mg	0.31 (0.07)	0.10 (0.01)	0.005 (0.002)	0.015 (0.002)	0.016 (0.003)
Occipital cortex
Baseline	0.28 (0.04)	0.11 (0.01)	0.061 (0.009)	0.101 (0.013)	0.159 (0.021)
PF 1 mg	0.35 (0.04)	0.12 (0.02)	0.006 (0.001)	0.016 (0.003)	0.016 (0.004)
PF 4 mg	0.31 (0.03)	0.10 (0.01)	0.004 (0.001)	0.011 (0.001)	0.011 (0.001)
PF 20 mg	0.30 (0.07)	0.10 (0.01)	0.006 (0.001)	0.015 (0.001)	0.015 (0.001)
Parietal cortex
Baseline	0.31 (0.04)	0.11 (0.01)	0.063 (0.009)	0.110 (0.012)	0.172 (0.019)
PF 1 mg	0.37 (0.01)	0.12 (0.01)	0.007 (0.002)	0.020 (0.003)	0.021 (0.004)
PF 4 mg	0.33 (0.01)	0.10 (0.01)	0.005 (0.001)	0.014 (0.002)	0.015 (0.002)
PF 20 mg	0.33 (0.08)	0.10 (0.01)	0.005 (0.001)	0.016 (0.002)	0.017 (0.002)
A. cingulate
Baseline	0.26 (0.02)	0.11 (0.01)	0.068 (0.009)	0.098 (0.007)	0.159 (0.014)
PF 1 mg	0.31 (0.02)	0.11 (0.01)	0.004 (0.001)	0.010 (0.002)	0.010 (0.003)
PF 4 mg	0.29 (0.01)	0.10 (0.01)	0.004 (0.002)	0.011 (0.004)	0.012 (0.004)
PF 20 mg	0.26 (0.07)	0.09 (0.01)	0.002 (0.001)	0.006 (0.003)	0.006 (0.003)
Insular cortex
Baseline	0.27 (0.03)	0.11 (0.02)	0.066 (0.009)	0.106 (0.008)	0.173 (0.015)
PF 1 mg	0.34 (0.01)	0.12 (0.01)	0.007 (0.004)	0.019 (0.008)	0.021 (0.009)
PF 4 mg	0.30 (0.01)	0.09 (0.01)	0.004 (0.002)	0.013 (0.006)	0.013 (0.006)
PF 20 mg	0.30 (0.06)	0.10 (0.01)	0.005 (0.001)	0.015 (0.002)	0.016 (0.002)
Cerebellar cortex
Baseline	0.28 (0.02)	0.10 (0.01)	0.057 (0.007)	0.105 (0.010)	0.168 (0.017)
PF 1 mg	0.33 (0.01)	0.10 (0.01)	0.004 (0.002)	0.013 (0.003)	0.013 (0.004)
PF 4 mg	0.28 (0.02)	0.08 (0.01)	0.002 (0.001)	0.008 (0.001)	0.008 (0.001)
PF 20 mg	0.31 (0.05)	0.10 (0.01)	0.005 (0.001)	0.014 (0.001)	0.014 (0.001)
Hippocampus
Baseline	0.19 (0.03)	0.09 (0.02)	0.071 (0.018)	0.087 (0.009)	0.160 (0.016)
PF 1 mg	0.23 (0.01)	0.09 (0.02)	0.009 (0.005)	0.019 (0.008)	0.021 (0.009)
PF 4 mg	0.20 (0.01)	0.08 (0.01)	0.006 (0.001)	0.015 (0.003)	0.017 (0.004)
PF 20 mg	0.21 (0.04)	0.08 (0.01)	0.007 (0.001)	0.017 (0.003)	0.018 (0.003)
Amygdala
Baseline	0.16 (0.02)	0.08 (0.05)	0.069 (0.032)	0.073 (0.010)	0.136 (0.020)
PF 1 mg	0.19 (0.01)	0.08 (0.01)	0.008 (0.003)	0.017 (0.006)	0.019 (0.008)
PF 4 mg	0.17 (0.01)	0.07 (0.01)	0.004 (0.002)	0.010 (0.005)	0.011 (0.005)
PF 20 mg	0.16 (0.03)	0.07 (0.01)	0.007 (0.002)	0.015 (0.001)	0.016 (0.002)
Caudate
Baseline	0.29 (0.04)	0.11 (0.03)	0.057 (0.016)	0.100 (0.018)	0.155 (0.030)
PF 1 mg	0.33 (0.03)	0.10 (0.01)	0.004 (0.001)	0.014 (0.003)	0.014 (0.003)
PF 4 mg	0.31 (0.01)	0.09 (0.01)	0.001 (0.001)	0.005 (0.003)	0.005 (0.003)
PF 20 mg	0.29 (0.03)	0.09 (0.01)	0.003 (0.001)	0.010 (0.002)	0.010 (0.002)
Putamen
Baseline	0.31 (0.03)	0.10 (0.01)	0.056 (0.009)	0.110 (0.015)	0.171 (0.028)
PF 1 mg	0.36 (0.01)	0.10 (0.01)	0.004 (0.002)	0.015 (0.007)	0.015 (0.007)
PF 4 mg	0.34 (0.01)	0.09 (0.01)	0.003 (0.001)	0.010 (0.004)	0.011 (0.004)
PF 20 mg	0.32 (0.06)	0.09 (0.01)	0.003 (0.001)	0.012 (0.006)	0.012 (0.007)
Thalamus
Baseline	0.31 (0.03)	0.10 (0.01)	0.061 (0.009)	0.117 (0.009)	0.189 (0.017)
PF 1 mg	0.36 (0.02)	0.09 (0.01)	0.005 (0.001)	0.017 (0.004)	0.018 (0.005)
PF 4 mg	0.31 (0.02)	0.07 (0.01)	0.001 (0.001)	0.005 (0.003)	0.005 (0.003)
PF 20 mg	0.33 (0.04)	0.08 (0.01)	0.003 (0.001)	0.011 (0.002)	0.012 (0.003)

2-TCMi, irreversible two tissue compartment model; A. cingulate, anterior cingulate cortex; PF, PF-04457845. Data are in mean (s.d.) of n=6 for the baseline and n=2 each after PF-04457845 dosing at 1 mg, 4 mg, and 20 mg, respectively.

Table 3. Test–retest reproducibility and reliability of λk ₃ for [¹¹C]CURB determined by a 2-TCMi (n=5).

60 minutes	Test mean (s.d.)	Retest mean (s.d.)	Mean	BSSD (%CV)	WSSD (%CV)	TRV% (SD%)	ICC
λk3
Prefrontal ctx	0.148 (0.027)	0.153 (0.023)	0.150	0.024 (16)	0.010 (6)	9 (7)	0.770
Temporal ctx	0.161 (0.028)	0.164 (0.025)	0.152	0.025 (16)	0.008 (5)	6 (5)	0.869
Occipital ctx	0.153 (0.029)	0.155 (0.022)	0.154	0.025 (16)	0.009 (6)	8 (7)	0.798
Parietal ctx	0.163 (0.028)	0.166 (0.025)	0.164	0.025 (15)	0.009 (5)	7 (7)	0.795
A. cingulate	0.158 (0.029)	0.162 (0.026)	0.160	0.025 (16)	0.011 (7)	11 (8)	0.732
Insular ctx	0.165 (0.028)	0.167 (0.029)	0.166	0.027 (16)	0.007 (4)	5 (6)	0.885
Cerebellar ctx	0.158 (0.029)	0.169 (0.030)	0.163	0.028 (17)	0.011 (7)	9 (8)	0.781
Hippocampus	0.158 (0.023)	0.161 (0.026)	0.159	0.023 (15)	0.06 (4)	5 (7)	0.853
Amygdala	0.135 (0.026)	0.138 (0.027)	0.136	0.022 (16)	0.015 (11)	15 (8)	0.546
Caudate	0.147 (0.035)	0.142 (0.022)	0.144	0.027 (19)	0.012 (8)	11 (8)	0.740
Putamen	0.163 (0.036)	0.167 (0.027)	0.165	0.030 (18)	0.009 (6)	8 (10)	0.824
Thalamus	0.181 (0.030)	0.179 (0.035)	0.180	0.032 (18)	0.008 (5)	7 (6)	0.894
All areas included						9 (7)	0.793

2-TCMi, irreversible two-tissue compartment model; A. cingulate, anterior cingulated cortex; BSSD, between-subject standard deviation; ctx, cortex; CV, %standard deviation/mean; ICC, intraclass correlation coefficient; TRV, test–retest variability—the mean across the subjects of the ratio (%) of absolute value of the difference between test-retest to the average of test–retest; WSSD, within-subject standard deviation.

Figure 4.

Changes (% of baseline, n=6) in overall kinetic parameters of [¹¹C]CURB across brain regions after administration of PF-04457845 at 1 (n=2), 4 (n=2), and 20 mg (n=2) doses, respectively. Sixty- minute scan data (mean±s.d.) are shown.

Test-Retest Reliability Analysis

Five subjects completed the test–retest study and were scanned on average 1.7 (1.1 to 2.1) months apart. Mean percentage test–retest variability (TRV%) and reliability (ICC) of regional [¹¹C]CURB λk₃ are presented in Table 3. Figure 3B illustrates the complete overlap in TACs between test and retest scans. There was no order effect between the scans (see Supplementary Figure 3 for the scatter plots). Excellent/high reproducibility was observed within the prefrontal, temporal, parietal, occipital, insular, cerebellar cortices, hippocampus, thalamus, and putamen (<10% variability) where WS variability was between 4% and 7%. Reliability in these regions ranged from 0.77 to 0.89 (ICC). In the amygdala, dorsal caudate nucleus and anterior cingulate cortex absolute % variability was good (<15%) with a stable WS covariance ranging from 7% to 11%. Reliability in these regions ranged from 0.55 to 0.74 (ICC).

Test–retest data for other kinetic parameters including K₁, k₃, and K_i are given in Supplementary Table 1. Both K₁ and K_i had excellent reproducibility across the brain regions (⩽10% variability) and good reliability in most of the brain regions (K₁: ICC=0.95 (0.78 to 0.98); K_i: ICC=0.90 (0.77 to 0.95)). In contrast, k₃ had relatively poor reproducibility in some brain regions, in particular in dorsal caudate, thalamus, and amygdala (16% to 41% variability) and poor and unstable reliability across the brain regions (ICC=0.522 (0.11 to 0.90)). Similarly, k₂ (TRV%: 15% (8% to 42%); ICC: 0.67 (0.48 to 0.94)) also showed relatively poor reproducibility and reliability among the brain regions examined.

Discussion

This is the first study imaging FAAH inhibition in the living human brain. We showed that [¹¹C]CURB is a reliable, highly sensitive and selective PET radioligand for imaging of brain FAAH activity. We reproduced our previous results²⁷ showing that a 2-TCMi is the optimal model for regional quantification of [¹¹C]CURB binding. In addition, we showed that the model behaves as predicted and thus we were able to identify the first compartment as nondisplaceable binding and the second as the specific irreversible binding to FAAH. Moreover, this study also showed that the Pfizer investigational drug PF-04457845 has good brain penetration and is highly potent in suppressing brain FAAH activity, with >95% inhibition (λk₃) achieved at a 1-mg dose.

[¹¹C]CURB was the first and is still the only PET tracer of FAAH that has been applied to human clinical research. Our previous modeling analysis (see Rusjan et al²⁷) showed excellent kinetic characteristics of this radioligand, including high brain uptake, good identifiability of the parameters, and independence of the known blood flow limitations associated with irreversible radioligands. With the availability of a highly specific FAAH inhibitor PF-04457845,^{23, 24, 25} the remaining questions of human in vivo selectivity and sensitivity to FAAH activity changes, characterization of nonspecific binding, and test–retest reliability were addressed in the current report.

[¹¹C]CURB λk ₃ : Specificity for Fatty Acid Amide Hydrolase Activity

[¹¹C]CURB is a carbamate irreversible inhibitor of FAAH,²⁶ an analog of the prototypical inhibitor URB597. Since URB597 is known to have some off targets in the periphery,^{20, 33, 34, 35} the question of [¹¹C]CURB's in vivo specificity was addressed here. PF-04457845, which has been extensively characterized and is one of the most specific inhibitors of FAAH,^{23, 24} belongs to the aryl urea class of FAAH inhibitor but also inhibits FAAH irreversibly with the similar mechanism as the carbamates.³⁶ Our finding that doses of PF-0447845 ranging 20 fold apart inhibited [¹¹C]CURB binding to the same maximal extent (possibly with the exception of one subject at the 1-mg dose: subject #8 who had the lowest plasma drug levels (Figure 2) had also the lowest %inhibition of λk₃ across brain regions (−88±3%)) strongly suggests that in vivo binding of [¹¹C]CURB in human brain is mediated via FAAH. Binding obtained under the blocking conditions thus defines the nonspecific fraction of [¹¹C]CURB, which is <10% for the parameter λk₃.

One caveat of our dose occupancy study is the uncertainty in parameter estimation under the nearly full blocking conditions. As indicated in our previously published simulation analysis (see Rusjan et al²⁷), the identifiability of the parameters (λk₃, k₃, and K_i) would be markedly lost with very low FAAH activity, with COV% increased from <10% to 20% to 25% when FAAH is inhibited by 90%, which is consistent with observed COV% in this study. We were not able to explore the full range of occupancy because drug product was not available in doses lower than 1 mg. However, even allowing a 50% uncertainty in λk₃ estimation under the blocking conditions, the proportion of nonspecific component was still on average only 13% for λk₃ determined.

Under the blocking conditions where almost complete occupancy of FAAH was reached, the TACs could be well fitted with a 1-TCM although the 2-TCMi was slightly better in the cortical areas. Therefore, identification of the first compartment as free and nonspecific, the second as irreversible specific, and the meaning of the rate constants were close to ideal but not perfect. Results of the test–retest study suggested that k₃ and k₂ separately had poor reproducibility and reliability at baseline. Therefore, the overall assessment supports that a composite parameter that cancels out the correlation between k₂ and k₃ (Rusjan et al²⁷ and Fowler et al³⁷) as λk₃ is the preferred parameter for FAAH activity quantification by [¹¹C]CURB. In the absence of a reference region, it is challenging, irrespective of the reversibility of the radioligand, to discriminate the relative contribution of free, specific, and nonspecific binding to the quantification parameters. For [¹¹C]CURB, the ability of PF-04457845 to acutely and completely shut down FAAH without producing any clinical side effects, which appears to be rare in radiotracer development for PET imaging, helps to define the contribution of the nonspecific components. The negative side of the use of λk₃ as the quantification parameter is that it could be sensitive to changes in nonspecific binding. However, BP_P=λk₃/k₄ in 2-TCM reversible radioligands³⁸ would be susceptible to the same issue and the stability of using V_T to quantify occupancy is always supported by assuming V_ND does not change between conditions.

PF-04457845 Occupancy

The use of PF-04457845 allowed us to establish [¹¹C]CURB as a very specific FAAH tracer in vivo. Furthermore, this is the first study to characterize the brain occupancy of FAAH by PF-04457845 at clinical doses despite ample preclinical^{23, 24} and human peripheral blood evidence²⁵ on the specificity and potency of the investigational drug. This brain imaging study provided direct evidence to firmly establish that PF-04457845 is a highly potent inhibitor of FAAH in living human brain.

It was suggested based on preclinical evidence that the pharmacological effects of FAAH inhibitors depend on the extent of FAAH activity inhibition, with near fully blockade required to elevate endogenous levels of the endocannabinoid anandamide and presumably to produce a functional effect, e.g., for pain suppression^{19, 24, 39, 40} although whether this ‘rule' applies to other N-arachidonoylethanolamide-related functions or conditions of partial but chronic FAAH inhibition is an open question, e.g., for the genetic variant C385A.¹⁸ Preclinical data in rats suggested that a dose of at least 0.1 mg/kg (p.o.) is required to maximally inhibit brain FAAH,²³ whereas clinical trials in the human showed that a single oral dose of 0.3 mg was able to almost completely inhibit FAAH in peripheral blood lymphocytes for 24 hours.²⁵ The development of [¹¹C]CURB now allows for direct dose finding in the brain by PET. The Pfizer recommended clinical trial dose of PF-04457845 is 4 mg (p.o.) based on a conservative estimation with the peripheral inhibition data. It is unfortunate that the current study could not examine lower doses due to the unavailability of drug product dosage less than 1 mg. However, the fact that the three tested doses of 1, 4, and 20 mg produced the same extent of maximal brain FAAH inhibition (with the exception of one subject (#8) at a 1-mg dose) strongly suggests that FAAH inhibition reached the ceiling at the 4-mg dose. Given a nonspecific fraction of 9%, the inhibition of the specific component, i.e., FAAH, by a 1-mg dose in subject #8 would still be >95%.

The major limitations of the present study include the small sample size and the use of the only available clinical doses, leading to a ceiling effect and precluding a full report of dose dependency of [¹¹C]CURB binding to FAAH inhibition. Lower blocking doses may be required to establish the sensitivity of the parameters to a partial FAAH inhibition, e.g., 20% to 30%. Further brain imaging studies are also required to firmly establish PF-04457845 dose regimen in clinical trials given BS variations. In addition, [¹¹C]CURB brain imaging could also help establish the chronic effects of FAAH inhibition and other manipulation of the endocannabinoid system, e.g., possible adaptation in expression of the enzyme after long-term treatment, and the recovery of the enzyme after cessation.^{39, 40}

Conclusion

PF-04457845 blocked strongly the activity of [¹¹C]CURB in vivo in the human brain. No further changes in the [¹¹C]CURB TACs appeared with the increase in the dose higher than 1 mg; therefore, the free and nonspecific binding of the radioligand was unequivocally characterized. It can be modeled very closely by a 1-TCM, which validated the use of λk₃ from a 2-TCMi to quantify FAAH activity in the human brain. In addition, the reliability of λk₃was assessed by test–retest experiments.

The authors declare no conflict of interest.