Sage Journals: Discover world-class research

Abstract

This article describes the kinetic modeling of [¹¹C]-harmine binding to monoamine oxidase A (MAO-A) binding sites in the human brain using positron emission tomography (PET). Positron emission tomography studies were performed in healthy volunteers at placebo conditions and after treatment with clinical doses of moclobemide. In either condition, a two-tissue compartment model (2CM) provided better fits to the data than a one-tissue model. Estimates of k₃/k₄ values from an unconstrained 2CM were highly variable. In contrast, estimates of the specifically bound radioligand distribution volume (DV_B) from an unconstrained 2CM were exceptionally stable, correlated well with the known distribution of MAO-A in the brain (cerebellum <frontal cortex ≈ putamen <temporal cortex ≈ cingulate <thalamus) and thus provided reliable indices of MAO-A density. Total distribution volume (DV) values were also highly stable and not different from those estimated with the Logan approach. Fixing the DV of free and nonspecifically bound radiotracer (DV_{F + NS}) or coupling DV_{F + NS} between brain regions enabled more stable estimates of k₃/k₄ as compared with an unconstrained 2CM. Moclobemide treatment leads to a 64% to 79% MAO-A blockade across brain regions, a result that supports the specificity of [¹¹C]-harmine binding to MAO-A. The stability and reliability of DV_B values obtained from an unconstrained 2CM, together with the computational simplicity associated with this method, support the use of DV_B as an appropriate outcome measure for [¹¹C]-harmine. These results indicate the suitability of using [¹¹C]-harmine for quantitative evaluation of MAO-A densities using PET and should enable further studies of potential MAO-A dysregulation in several psychiatric and neurologic illnesses.

Keywords

human modeling monoamine oxidase PET

Introduction

Monoamine oxidase (MAO) is a mitochondrial enzyme which catalyzes the oxidative deamination of biogenic amines. In the brain, MAO is responsible for the metabolic inactivation of dopamine, noradrenaline, and serotonin and thus plays a central role in the regulation of these neurotransmitter levels. In mammals, two isoforms of the enzyme have been identified on the basis of their biochemical properties, substrate selectivity, and gene products (Shih et al, 1999). Monoamine oxidase-A preferentially oxidizes serotonin and noradrenaline while MAO-B preferentially oxidizes the trace amine phenethylamine. Dopamine is a substrate for both isoenzymes. The physiologic roles of MAOs are related to the nature of their substrates and selective inhibition of these isoforms is used to treat a variety of pathologic conditions. Inhibition of MAO-A is useful in the treatment of depression and anxiety disorders (Livingston and Livingston, 1996) while inhibition of MAO-B is more specifically used to treat neurodegenerative disorders such as Parkinson's disease (Cohen, 1990).

The ability of MAOs to catabolize neurotransmitters has made these enzymes attractive candidates in the study of psychiatric and neurologic diseases as well as in the study of brain function. Many studies report apparent associations between abnormal platelet MAO activity and susceptibility to depression and suicidal behavior (Murphy and Weiss, 1972), schizophrenia (Lewine and Meltzer, 1984), Parkinson's disease (Jarman et al, 1993), Huntington's chorea (Mann and Chiu, 1978), alcoholism (Fowler et al, 1981), and smoking (Oreland et al, 1981). In these early studies, platelets were viewed as an accessible tissue to measure MAO activity, which was considered as an index of the enzyme activity in the brain. However, platelet MAO in human is exclusively of the MAO-B-type (Youdim, 1988) and provides no insight on the activity of the MAO-A-type.

Several lines of evidence indicate that abnormalities in brain MAO-A activity are involved in the pathogenesis of depressive illness and suicidal behavior. For instance, successful treatment of unipolar and bipolar depression is achieved using specific inhibitors of MAO-A such as clorgyline (Lipper et al, 1979) and moclobemide (Angst and Stabl, 1992). Only a few studies have aimed at measuring MAO-A activity in the brain in relation to depression and these have provided conflicting results. While a significant elevation of MAO-A activity was found in the hypothalamus of depressed suicide victims (Sherif et al, 1991), other studies showed no significant differences in MAO-A activity in any brain regions of depressed suicide victims when compared with controls (Mann and Stanley, 1984; Ordway et al, 1999). Comparison of results obtained in the post-mortem brain has however been hampered by several confounding factors such as small sample sizes, insufficient matching criteria for controls, different methodology used to measure MAO activity, comorbidity between the use of medication, alcohol, or cigarette smoking, and the disease condition.

To circumvent these problems, several radioligands have been developed for in vivo imaging of MAO-A using positron emission tomography (PET) in living humans (Ametamey et al, 1996; Bottlaender et al, 2003; Fowler et al, 1987; Ohmomo et al, 1991, 1993). Bergstrom et al (1997a) have reported on the synthesis of five different ¹¹C-labeled MAO-A inhibitors as potential radioligands for PET imaging of the enzyme. Among them, [¹¹C]-harmine is a β-carboline derivative, which is a competitive and reversible inhibitor of MAO-A with a K_i of 5 nmol/L (Kim et al, 1997). In vitro binding studies in rats have shown its binding is specific to MAO-A as it is inhibited by nanomolar concentrations of the selective MAO-A inhibitor clorgyline and only marginally affected by fairly selective MAO-B inhibitors such as deprenyl and pargyline (Bergstrom et al, 1997b). In vivo PET studies in monkeys and humans have already showed its potential for imaging MAO-A (Bergstrom et al, 1997b, c). In both species, [¹¹C]-harmine showed favorable kinetics and displayed sensitivity to pretreament with selective MAO-A inhibitors such as moclobemide and esuprone. Although these earlier studies clearly suggest that [¹¹C]-harmine may be a useful PET radioligand for the visualization and quantification of MAO-A in human, no kinetic modeling approach has ever been applied to quantify the binding of this radioligand in vivo.

The aim of the present study was to evaluate the prospect of using PET and [¹¹C]-harmine for quantification of MAO-A binding sites in the human brain. Regional brain uptake curves of [¹¹C]-harmine were quantified using kinetic modeling analyses based on either one- or two-tissue compartments, both at placebo condition and after treatment with moclobemide. Moclobemide is a short-acting, reversible, and selective MAO-A inhibitor that has proven to be efficacious in the treatment of major depression (Angst and Stabl, 1992). In addition, the performance of the Logan graphical approach (Logan et al, 1990) was also evaluated to determine the simplest kinetic method for quantification of [¹¹C]-harmine binding data in routine clinical research.

Methods

Subjects

This study was approved by the Human Subjects Review Committee of the University of Toronto. Five healthy male volunteers free of neurologic or psychiatric diseases, with no past or current history of cigarette smoking, were recruited for the study. None of the subjects took central nervous system acting medication or drugs within 3 months before enrolment in the study. Subjects were asked not to drink alcohol the day before PET scanning and not to drink beverages with caffeine on the day of a PET scan. In all cases, written informed consent was obtained before participation. The age and weight of the subjects were 34 ± 11 years (range: 20 to 49 years) and 81 ±13 kg (range: 70 to 100 kg), respectively (mean ± s.d., n = 5).

Each subject was scanned with PET and [¹¹C]-harmine on two different occasions using a double-blind, counterbalanced procedure. On one occasion, subjects were examined after 10 days treatment with oral placebo and on the other occasion, subjects were examined after 10 days treatment with moclobemide (Manerix®, Hoffmann-La Roche Limited, Mississauga, Canada). The washout period between treatment conditions (placebo versus moclobemide) was at least 8 weeks. When placed on moclobemide treatment, subjects received twice-daily dose (b.i.d.) of 150 mg moclobemide for 3 days and then 300 mg b.i.d. for 7 days. These daily doses of moclobemide are within the therapeutic dose range for moclobemide in most acute phase trials (Bonnet, 2003) and 300 mg b.i.d. is also the highest clinical dose that has a minimal risk for a hypertensive crisis. Moreover, a 10-day treatment period has been shown to be sufficient to reach steady-state plasma levels as the elimination half-life of moclomebide is approximately 2 h (Bonnet, 2003). When placed on placebo treatment, the subjects received placebo tablets twice daily for 10 consecutive days. Drug or placebo intakes took place at 8 a.m. and at 7 p.m. daily. On day 10, PET scans of [¹¹C]-harmine binding were obtained 4 h after the last morning dose of either placebo or moclobemide. Subjects then continued to take 150 mg of moclobemide or placebo 3 days after the scan date as a taper.

Radiochemistry

[¹¹C]-methylation of normethylharmine was performed inside an HPLC sample loop using our previously described method (Wilson et al, 2000). Briefly, precursor harmol (1-methyl-9H-pyrido[3,4-b]indol-7-ol, 1.0 mg) was dissolved in dimethylformamide (DMF) [77 μL) and treated with tetrabutylammonium hydroxide solution (1 N in MeOH, 3 μL). This solution was loaded onto an HPLC injection loop (2 mL) on a Valco HPLC injector (Houston, TX, USA) valve in the load position. [¹¹C]-iodomethane (produced as previously described; Wilson et al, 1994) was swept into the HPLC loop coated with precursor solution by a stream of N₂ gas (8 mL/min) at ambient temperature. When radioactivity peaked in the loop (3 to 4 mins), the flow of N₂ was stopped and the reaction allowed to proceed for 1 min at ambient temperature. The contents of the loop were then quantitatively injected onto the HPLC purification column by changing the position of the injection valve from the load to the inject position and the reaction mixture purified by semi-prep HPLC; Phenomenex Prodigy C₁₈ 10 μ (250 × 10 mm, 30% CH₃CN/70% H₂O + 0.1 n NH₄HCO₂, pH 4, 7 mL/min). The desired fraction was collected, evaporated to dryness at 70°C, and the residue taken up in 10 mL of sterile saline. The saline solution of [¹¹C]-harmine was passed through a sterile 0.22 μm filter into a sterile, pyrogen-free bottle containing aqueous sodium bicarbonate (1 mL, 8.4%). An aliquot (100 μL) of the formulated solution was used to establish the radiochemical purity and specific activity of the final solution by analytical HPLC; Phenomenex Prodigy C18 10 μ (250 × 4.5 mm, 30% CH₃CN:70% H₂O + 0.1 N NH₄HCO₂, pH 4, 4 mL/min).

Positron Emission Tomography System

Studies were performed on a brain PET camera system Scanditronix GEMS 2048–15B, which measured radioactivity in 15 brain sections with a thickness of 6.5 mm each (Litton et al, 1990). The intrinsic in-plane resolution of the reconstructed images is 5.5 mm full-width at half-maximum. Transmission scans were acquired with a rotating ⁶⁸Ge source and used to correct the emission scans for the attenuation of the 511 keV photons through tissue and head support. Images were reconstructed by filtered back-projection using a Hanning 5 mm filter.

Positron Emission Tomography Studies

A thermoplastic mask was made for each subject and used with a head fixation system during the PET measurements. Cannulae were inserted in the radial artery for blood sampling, and in the controlateral arm antecubital vein for radioligand injection. A saline solution of 10.9 mCi (s.d. = 1.0 mCi) of [¹¹C]-harmine at a specific radioactivity of 1,124 Ci/mmol (s.d. = 203 Ci/mmol) was injected intravenously as a bolus and immediately flushed with 10 mL saline. Radioactivity in the brain was measured in a series of sequential frames of increasing duration (from 1 to 5 mins). The total time for measurement of radioactivity in the brain was 90 mins.

An automated blood sampling system (ABSS; Scanditronix, Uppsala, Sweden) was used to measure radioactivity in arterial blood during the first 10 mins of the experiment. It consists of a 30 cm long catheter line, a scintillation crystal detector to measure the concentration of radioactivity in blood, a peristaltic pump to draw arterial blood into the catheter, and a microprocessor datalogger. The pump speed was set at 2.5 mL/min and the concentration of radioactivity in the catheter was measured and stored every 2.5 secs.

At the same time and thereafter, manual blood samples were drawn at 5, 10, 15, 20, 30, 45, 60, or 70 and 90 mins after radiotracer injection. An aliquot of each blood sample was taken to measure radioactivity concentration in total blood. The remainder of blood was centrifuged (1,500g, 5 mins) and a plasma aliquot counted together with the total blood sample using a Packard Cobra II gamma counter cross-calibrated with the PET system. The blood-to-plasma ratios determined from the manual samples were used to correct the blood radioactivity time-activity curve measured by automatic sampling and to generate the plasma radioactivity curve. The remaining volume of each manual plasma sample was used to determine unchanged radioligand and its metabolites in plasma (as described in the next section). A metabolite corrected plasma curve was generated by the product of the two curves and was used as input function (C_P(t), nCi/mL) for the kinetic and graphical analyses.

Determination of Radioactive Metabolites in Plasma

Plasma analysis of [¹¹C]-harmine was based on the column capture and switching methods developed by Luthra and Hilton (Hilton et al, 2000; Luthra et al, 1992). Arterial blood samples were withdrawn at predefined times, centrifuged, and whole unadulterated plasma (0.5 to 5 mL) injected onto a small capture column packed with the OASIS™ resin (Waters, Milford, MA, USA). Highly polar metabolites and plasma proteins were eluted with 1% CH₃CN in H₂O (2 mL/min) through a coincidence flow detector (Bioscan Flow-Count). Less polar metabolites and [¹¹C]-harmine were then back-flushed onto an HPLC column (Phenomenex AquaC₁₈, 5μ, 150 × 4.6 mm) and resolved using 30% CH₃CN/70% H₂O + 0.1 N ammonium formate (AF). pH 4 as eluent (2 mL/min).

Regions of Interest

Each subject had a brain magnetic resonance imaging (MRI) examination. T₁-weighted and proton density images were obtained on a General Electric Medical System Signa 1.5 T MRI scanner. Both T₁-weighted and proton density images were acquired with spin-echo pulse sequences. Each subject's MRI scan was coregistered to the PET scan using Rview8/mpr realignment software (Studholme et al, 1997).

Regions of interest (ROIs) for the cerebellar cortex, putamen, thalamus, prefrontal cortex, anterior temporal cortex, anterior cingulate cortex, and posterior cingulate cortex were drawn with reference to the coregistered MRI. Each ROI was drawn in two to four adjacent sections, in both hemispheres, and data from bilateral ROI were pooled to obtain the average radioactivity concentration in the volume of interest. Regional radioactivity was determined for each frame, corrected for decay, and plotted versus time.

Data Analysis

Compartmental kinetic analysis: Kinetic analysis of tissue data was performed with both a one-tissue (1CM) and a two-tissue compartment model (2CM). The 2CM includes: the radioactivity concentration of unchanged radioligand in plasma (C_P); the concentration of radioligand free and nonspecifically bound in tissue (C_{NS + F}); and the concentration of radioligand specifically bound to receptor sites (C_B). The rate constants K₁ (mL/gmin) and k₂ (min⁻¹) describe the influx and efflux rates, respectively, for radioligand diffusion through the blood-brain barrier. The rate constant k₃ and k₄ describe the radioligand transfer between the nondisplaceable compartment (i.e., C_{NS + F}) and the specific binding compartment (C_B). Model equations for the 2CM analysis of radiotracer binding to receptors are given by

\frac{d C_{N S + F} (t)}{d t} = K_{1} C_{P} (t) - (k_{2} + k_{3}) C_{N S + F} (t) + k_{4} C_{B} (t)

(1)

\frac{d C_{B} (t)}{d t} = k_{3} C_{N S + F} (t) - k_{4} C_{B} (t)

(2)

k₄ (min⁻¹) is the unimolecular receptor dissociation rate constant k_off. The rate constant k₃ (min⁻¹) is the product of the bimolecular association rate constant k_on and the concentration of available receptors. The number of available receptors is the total number of receptors B_max less the number of receptor–radioligand complexes, B_R, and the number of receptor–inhibitor complexes, B_I. Because all experiments were performed at tracer dose and high specific radioactivity, then: B_R ≪ B_max. At such conditions and when no drug is present in the brain, it follows that:

k_{3} = k_{o n} B_{max}

(3)

After treatment with an unlabeled drug I, k₃ is noted $k_{3}^{*}$ and equals:

k_{3}^{*} = k_{o n} (B_{max} - B_{I})

(4)

At high specific radioactivity, the radioligand binding potential (BP) corresponds to the ratio of the number of available receptors over K_D and is expressed as a function of rate constants as (Mintun et al, 1984):

At placebo conditions:

B P_{p l a c e b o} = \frac{k_{3}}{k_{4}} = \frac{B_{max}}{K_{D}}

(5)

After a drug treatment:

B P_{d r u g} = \frac{k_{3}^{*}}{k_{4}} = \frac{B_{max} - B_{I}}{K_{D}}

(6)

where K_D, the radioligand equilibrium dissociation constant (nmol/L), equals to (k_off/k_on).

The percentage of enzyme blockade achieved by a drug treatment can be defined as

% b l o c k a d e = \frac{B_{I}}{B_{max}} 100 %

(7)

Assuming competitive inhibition of the enzyme by the drug, the percentage of enzyme blockade can also be calculated by combining equations (5) and (6) and is defined as the percentage reduction of BP obtained after drug treatment as compared with the BP obtained at placebo conditions.

% b l o c k a d e = \frac{B P_{p l a c e b o} - B P_{d r u g}}{B P_{p l a c e b o}} 100 %

(8)

Note that calculation of % blockade using equation (8) assumes that the inhibitor-induced decrement in BP is because of a decrease in the number of available binding sites, with no change in K_B. This does not exclude a parallel change in both k_on and k_off. Invalidation of this assumption would lead to a misestimation of % blockade. For instance, a decrease in k_on with no parallel change in k_on would lead to an underestimation of the % blockade as obtained with equation (8) when compared with that obtained with equation (7).

The concept of distribution volume (DV; mL g⁻¹) is also frequently used in tracer kinetics to estimate binding site density. The total DV [DV_T) has the advantage of being a more stable parameter than the individual kinetic parameters determined from compartmental analysis (Carson et al, 1993; Koeppe et al, 1991). The DV_T of a radioligand in a tissue is the volume of tissue in which the radioligand would have to distribute to reach a concentration equal to that in the plasma. It is defined as:

D_{V T} = D V_{N S + F} + D V_{B}

(9)

where DV_{NS + F} corresponds to the radioligand DV in the free and nonspecifically bound compartment and DV_B to the DV in the specifically bound compartment

DV_{NS + F} and DV_B can be expressed in terms of the kinetic rate parameters as follows:

D V_{N S + F} = \frac{K_{1}}{k_{2}}

(10)

D V_{B - p l a c e b o} = (\frac{K_{1}}{k_{2}}) (\frac{k_{3}}{k_{4}})

(11)

Note that after treatment with an unlabelled drug I, the term k₃ in equation (11) should be replaced by the term $k_{3}^{*}$ .

Assuming that DV_{NS + F} is similar at placebo and at drug treatment conditions, the percentage of drug-induced receptor blockade can be defined as the percentage reduction of DV_B obtained after drug treatment as compared with the DV_B obtained at placebo conditions.

% b l o c k a d e = \frac{{D V}_{B - p l a c e b o} - {D V B}_{B - d r u g}}{{D V}_{B - p l a c e b o}} 100 %

(12)

Note that, as in equation (8), calculation of % blockade using equation (12) assumes that the inhibitor-induced decrement in DV_B is because of a decrease in the number of available binding sites, with no change in K_D.

A simplification of the 2CM into a 1CM can be made by assuming that the concentrations C_{NS + F} and C_B equilibrate rapidly and are combined in one single compartment. The total distribution volume for the 1CM is noted as DV' and is given by:

{D V}^{'} = \frac{K_{1}}{k_{2}^{'}}

(13)

with

k_{2}^{'} = \frac{k_{2}}{(1 + k_{3} / k_{4})}

(14)

Nonlinear least-square fitting: Nonlinear least-squares fitting (NLSF) analyses using the Marquardt algorithm (Marquardt, 1963) and based on the 1CM and the 2CM were applied to the time–activity curves for regional [¹¹C]-harmine uptake. The contribution of fractional blood volume in the brain was fixed at 5% for all ROIs (Phelps et al, 1979). For each subject, the time delay between the plasma arterial input function and the brain time–activity data was estimated by fitting a whole-brain slice at the level of the striatum to an unconstrained 2CM using the 90 mins data set. However, it was difficult to characterize this delay with reasonable accuracy when using a 2CM, 5 parameters configuration. To estimate the delay, the delay value was thus increased gradually, using 1 sec increments, to minimize the residual sum of squares (RSS) of the whole-brain curve. The delay parameter value, which gave the least RSS on the fitted whole-brain curve, was used in the subsequent analyses where regional time–radioactivity curves were fitted.

Different model configurations were used for analysis of [¹¹C]-harmine (Table 1). Method A was based on the 1CM and NLSF analyses were performed to estimate K₁, k'₂, and DV'. Method B was based on the 2CM and NLSF analyses were performed to estimate the four rate constants K₁, k'₂, k₃ (or $k_{3}^{*}$ ), and k₄. To improve the stability of the fits, two other fitting strategies were tested. Both were based on the 2CM. Method C assumed the ratio K₁/K₂ to be similar between regions. Nonlinear least-squares fitting analyses were accordingly performed by fitting simultaneously data from all regions and guided by the constraint of finding a K₁/k₂ value common to all regions. A second approach, Method D, was used for the [¹¹C]-harmine placebo scans only. For each subject, this method consisted to fix the K₁/k₂ ratio in the placebo study to the value obtained in the moclobemide study using Method C. This method was based on the assumption that DV_{NS + F} is similar under both treatment conditions. Three rate constants, K, k₃ (or $k_{3}^{*}$ ), and k₄ (with k₂ = K₁/DV_{NS + F}) were estimated. Nonlinear least-squares fitting analyses were performed using the dedicated software PMOD V2.4 (PMOD Technologies Ltd, Zurich, Switzerland).

Table 1

Fitting strategies used for kinetic modeling of [¹¹C]-harmine

Method	Number of tissue compartments	Constraints
A	1	None
B	2	None
C	2	K₁/k₂ coupled
D	2	K₁/k₂ fixed

Simplified Method of Analysis

In addition to the kinetic modeling approach, [¹¹C]-harmine binding was quantified using the linear graphical approach described by Logan et al (1990) for the analysis of reversible radioligand binding. A plasma input function corrected for the presence of labeled metabolites (C_P(t)) was used. The ratio $\int_{0}^{t} R O I (t^{'}) d t^{'} / R O I (t)$ was plotted versus $\int_{0}^{t} C_{p} (t^{'}) d t^{'} / R O I (t)$ , with ROI(t) describing radioactivity in a defined ROI as a function of time. The DV_T of the radioligand in that ROI was determined from the linear portion of the plot obtained.

Statistics: Goodness-of-fit was evaluated using both the Akaike Information Criterion (AIC; Akaike, 1974), the Model Selection Criterion (MSC). Higher AIC and lower MSC values were indicative of a better fit. In addition, models of higher complexity were also successively compared using an F-test to determine whether goodness-of-fits was sufficiently improved to justify the choice of a model with a higher number of parameters (Landaw and DiStefano, 1984). Statistical significance using the igtest was assumed for P values superior or equal to 0.05.

The standard error of the parameters was given by the diagonal of the covariance matrix (Carson, 1986), expressed as percentage of the parameter value (coefficient of variation, %COV), and used to assess the parameter identifiability by the NLSF procedure. Calculation of %COV was possible only for fitted parameters. For derived parameters such as k₃/k₄ and DV_T, the s.d. of the mean (n = 5 subjects) was calculated, expressed as percentage of the mean derived parameter value, and used as an indication of the degree of intersubject variability of the parameter (%VAR = s.d. × 100%/mean).

Results

Plasma Analysis

HPLC analysis of a plasma sample could be completed in less than 8 mins allowing multiple time points to be analyzed despite the short half-life of carbon-11. [¹¹C]-harmine was rapidly metabolized with >50% of the radioactivity in the plasma attributable to polar metabolites after 20 mins. At placebo conditions, the fraction of unmetabolized [¹¹C]-harmine in the plasma decreased throughout the time course of the study and was 90% ± 5%, 72% ± 13%, 46% ± 14%, 34 %± 12%, 22 %± 7%, 16% ± 3%, and 11% ± 5% at 5, 10, 20, 30, 45, 70, and 90 mins, respectively. After moclobemide treatment, [¹¹C] -harmine metabolism was faster than that measured at placebo treatment, with 94% ± 4%, 71% ± 13%, 24% ± 8%, 18% ± 6%, 12% ± 4%, 9% ± 2%, and 11% ± 5% at 5, 10, 20, 30, 45, 70, and 90 mins, respectively. In both treatment conditions, all metabolites were significantly more polar than the parent compound and thus unlikely to pass the blood–brain barrier. Figure 1 shows representative curves obtained in the same subject for unmetabolized [¹¹C]-harmine in the plasma under placebo and moclobemide treatment conditions. As shown in Figure 1, the concentration of unmetabolized [¹¹C]-harmine in plasma peaked at higher levels after moclobemide treatment when compared with placebo conditions.

Figure 1

Arterial time–activity curves for the concentration of unmetabolized [¹¹C]-harmine in plasma after a bolus injection of [¹¹C]-harmine in a healthy human subject. Arterial measurements are shown in the same subject at placebo conditions (left panel) and during treatment with moclobemide (right panel). As shown in this figure, four of the five subjects who participated to the study showed a higher peak level of radioactivity in plasma at moclobemide treatment when compared with placebo conditions.

Kinetic Analyses

Placebo studies: Characteristic time–activity curves obtained after intravenous injection of [¹¹C]-harmine in one subject are shown in Figure 2A. At placebo conditions, the highest radioactivity concentration was observed in the thalamus, intermediate levels were observed in striatum and cortical regions, and lowest levels of radioactivity were observed in the cerebellum. Figure 2B shows regional tissue to plasma radioactivity ratio in the corresponding subject. With few exceptions, transient equilibrium between tissue and plasma levels of radioactivity was obtained after 30 mins of data acquisition in all regions and all subjects.

Figure 2

Representative [¹¹C]-harmine time–activity curves obtained at placebo conditions for (A) regional brain radioactivity and (B) tissue/plasma ratio.

As illustrated on Figure 3A, a 1CM configuration failed to describe the [¹¹C]-harmine time–activity data. Adding a second tissue compartment (Method B) significantly improved the fits in all regions and all subjects when compared with the 1CM as assessed with AIC and MSC values (Table 2) as well as F-statistics. F-statistics for each ROI time–activity curve and each subject were in the range of F = 5.7 to 200.4 and F_2,15 was significant at P ≤ 0.05 for F ≥ 3.68. Despite this, DV_T values obtained with a 1CM or a 2CM configuration were close (Figure 4A): a 1CM provided DV_T values that were, on average, approximately 5% lower than those obtained with a 2CM. This difference was observed in all brain regions examined.

Table 2

Statistical evaluation of fits obtained using Methods A–D as explained in

Method	AIC ± s.d.	MSC ± s.d.	A	B	C
Placebo
A	66.2 ± 10.4	2.76 ± 0.53	—
B	34.7 ± 18.2	4.42 ± 0.98	P < 0.001	—
C	36.8 ± 17.1	4.31 ± 0.93	P < 0.01	NS	—
D	39.2 ± 14.5	4.18 ± 0.78	P < 0.01	NS	NS
Moclobemide
A	98.4 ± 8.42	2.29 ± 0.42	—
B	55.6 ± 11.0	4.54 ± 0.70	P < 0.001	—
C	59.7 ± 9.84	4.32 ± 0.63	P < 0.01	NS	—

Mean ± s.d. (n = 5) values for Akaike information criterion (AIC) and model selection criterion (MSC) values are given for Methods A–D at placebo and at moclobemide treatment conditions. To the right part of the table, the levels of statistical significance for comparison of good ness-of-fits obtained between methods are given. P values were calculated using repeated measures AN OVA.

NS: nonsignificant.

Figure

3 Nonlinear least-squares fits of [¹¹C]-harmine time–activity data obtained at placebo conditions (A) and after moclobemide treatment (B) in one healthy subject. For each brain region, the symbols (triangles: thalamus; circles: putamen; diamonds; cerebellum) correspond to the experimental measured values, the dashed line corresponds to the fitted curve obtained according to a 1CM (Method A) and the solid line corresponds to the fitted curve obtained according to an unconstrained 2CM (Method B). The goodness-of-fits indicates that a 2CM was necessary to describe the in vivo kinetics of [¹¹C]-harmine at both placebo and moclobemide treatment conditions.

Figure

4 Comparison of DV_T values obtained using a 1CM (Method A), a 2CM (Methods B–D) and the Logan graphical approach at placebo (A) and at moclobemide treatment conditions (B). Group mean ± s.d. (n = 5) are shown. Note the high consistency and the similar variability between DV_T values estimated using the different approaches.

An unconstrained 2CM (Method B) provided reasonable identification of K₁ and k₄ with mean %COVs of 6.2% and 13.7%, respectively, but provided poor identification of k₂ and k₃, as exhibited by the large standard errors in parameter estimates (Table 3). As a consequence, the k₃/k₄ ratios were highly variable between subjects (Table 3) and the rank order of regional k₃/k₄ did not correlated with the known concentration of MAO-A in brain as measured by in vitro autoradiography (cerebellum < frontal cortex-≈ putamen < temporal cortex ≈ cingulate < thalamus; Saura et al, 1992, 1996a). In contrast, the rank order of DV_B values was in accordance with the known distribution of MAO-A in the brain, with mean regional values ranging from 10.5 in the cerebellum to 22.2 in the thalamus, with putamen and cerebral cortices showing intermediate values (Table 3). DV_B values were estimated with high precision, as attested by a mean %COV of 4.4% (range: 1% to 7%). DV_T values were also estimated with high precision (mean %COV = 2%; range: 1% to 5%) and the rank order of DV_T values also paralleled the rank order of MAO-A density in brain. This could be expected as a major proportion (86% ± 6% in average across regions; range: 82% to 92%) of DV_T was accounted for by specific binding.

Table 3

Average kinetic parameter estimates obtained using Method B at placebo conditions and after treatment with moclobemide

Region	K₁ ± s.d.	k₂ ± s.d.	k₃ ± s.d.	k₄ ± s.d.	k₃/k₄ ± s.d.	DV_B ± s.d.	DV_T ± s.d.
Placebo
Cerebellum	0.52 ± 0.12	0.18 ± 0.07	0.22 ± 0.12	0.06 ± 0.01	3.61 ± 1.48	10.54 ± 2.34	13.77 ± 2.52
Frontal	0.48 ± 0.11	0.24 ± 0.07	0.45 ± 0.18	0.05 ± 0.01	8.91 ± 3.36	17.18 ± 3.40	19.26 ± 3.39
Cingulate	0.44 ± 0.14	0.27 ± 0.15	0.41 ± 0.13	0.04 ± 0.01	12.33 ± 7.64	19.49 ± 3.53	21.46 ± 3.30
Putamen	0.56 ± 0.16	0.23 ± 0.11	0.40 ± 0.13	0.06 ± 0.02	7.23 ± 7.66	17.40 ± 3.48	20.04 ± 3.65
Thalamus	0.56 ± 0.13	0.21 ± 0.09	0.39 ± 0.08	0.05 ± 0.02	8.37 ± 4.26	22.15 ± 4.80	25.29 ± 5.11
Temporal	0.42 ± 0.06	0.23 ± 0.08	0.43 ± 0.15	0.04 ± 0.02	10.93 ± 5.79	19.30 ± 3.53	21.41 ± 3.60
Posterieur	0.67 ± 0.20	0.26 ± 0.13	0.39 ± 0.15	0.05 ± 0.02	8.12 ± 5.17	19.28 ± 3.86	22.17 ± 4.06
%COV ± s.d.	6.2 ± 3.2	29.3 ± 18.3	23.0 ± 16.5	13.7 ± 8.9	28.0 ± 18.1	4.4 ± 2.6	2.1 ± 1.1
%VAR ± s.d.	24.5 ± 6.0	42.4 ± 9.1	35.8 ± 10.3	28.5 ± 5.1	49.3 ± 11.1	20.0 ± 1.5	17.8 ± 1.5
Moclobemide
Cerebellum	0.37 ± 0.09	0.09 ± 0.02	0.02 ± 0.01	0.02 ± 0.01	0.98 ± 0.10	3.80 ± 0.72	7.30 ± 0.93
Frontal	0.38 ± 0.09	0.12 ± 0.02	0.04 ± 0.01	0.03 ± 0.01	1.29 ± 0.31	4.05 ± 0.71	7.26 ± 1.00
Cingulate	0.31 ± 0.08	0.11 ± 0.03	0.06 ± 0.03	0.03 ± 0.01	1.70 ± 0.35	4.61 ± 0.79	7.36 ± 0.94
Putamen	0.41 ± 0.09	0.10 ± 0.01	0.02 ± 0.01	0.03 ± 0.01	0.98 ± 0.19	4.08 ± 0.56	8.39 ± 1.33
Thalamus	0.39 ± 0.12	0.09 ± 0.02	0.03 ± 0.02	0.03 ± 0.01	0.86 ± 0.23	3.87 ± 0.96	8.44 ± 1.17
Temporal	0.32 ± 0.07	0.10 ± 0.02	0.03 ± 0.02	0.03 ± 0.01	1.27 ± 0.25	4.13 ± 0.76	7.44 ± 0.90
osterieur	0.41 ± 0.10	0.11 ± 0.04	0.03 ± 0.01	0.03 ± 0.01	1.21 ± 0.32	4.48 ± 0.85	8.32 ± 1.03
%COV ± s.d.	2.8 ± 1.1	9.6 ± 4.2	25.6 ± 8.7	22.7 ± 8.6	13.6 ± 4.9	8.9 ± 3.3	5.0 ± 2.1
%VAR ± s.d.	25.1 ± 3.3	22.1 ± 7.4	41.3 ± 11.6	31.8 ± 8.2	21.0 ± 5.5	18.4 ± 3.3	13.3 ± 1.3

Values are mean ± s.d. of 5 experiments. %C0V corresponds to the standard error of parameter estimated by the nonlinear least-squares fitting procedure, is expressed in percent of the parameter value and is given as an index of the parameter identifiability. %VAR corresponds to the inter-subject variability in the parameter estimates.

Methods C and D significantly improved the identifiability of k₃ and k₄ estimates (Table 4) when compared with Method B (Table 3) as shown by the lower %COV obtained with the two former methods. Despite this, goodness-of-fits was not statistically improved by using Method C or D when compared with Method B as assessed with AIC and MSC (Table 2). F-statistics indicated that Methods C and D slightly worsened goodness-of-fits when compared with Method B but this was not statistically significant for most fits.

Table 4

Average kinetic parameter estimates obtained using Methods C–D at placebo and moclobemide conditions

Method	Region	K₁	K₁/k₂	k₃	k₄	k₃/k₄	DV_B	DV_T
Placebo
C	Cerebellum	0.54 ± 0.12	2.74 ± 0.84	0.26 ± 0.10	0.06 ± 0.02	4.25 ± 1.37	10.95 ± 2.29	13.69 ± 2.49
	Frontal	0.46 ± 0.10	2.74 ± 0.84	0.36 ± ± 0.13	0.06 ± 0.02	6.45 ± 2.24	16.57 ± 3.22	19.31 ± 3.38
	Cingulate	0.40 ± 0.11	2.74 ± 0.84	0.30 ± 0.08	0.04 ± 0.01	7.44 ± 2.71	18.86 ± 3.51	21.60 ± 3.38
	Putamen	0.54 ± 0.14	2.74 ± 0.84	0.39 ± 0.13	0.06 ± 0.01	6.75 ± 2.41	17.31 ± 3.51	20.05 ± 3.64
	Thalamus	0.57 ± 0.15	2.74 ± 0.84	0.41 ± 0.08	0.05 ± 0.01	8.82 ± 3.38	22.55 ± 5.01	25.29 ± 5.06
	Temporal	0.40 ± 0.07	2.74 ± 0.84	0.34 ± 0.12	0.05 ± 0.02	7.32 ± 2.50	18.82 ± 3.45	21.54 ± 3.66
	Posterieur	0.67 ± 0.22	2.74 ± 0.84	0.39 ± 0.13	0.05 ± 0.02	7.48 ± 2.40	19.37 ± 3.68	22.11 ± 4.03
	%COV	3.0 ± 0.8	3.4 ± 1.9	10.4 ± 3.5	11.0 ± 3.7	—	—	—
	%VAR	24.9 ± 4.9	30.8	31.8 ± 6.7	29.4 ± 5.9	34.8 ± 2.2	19.8 ± 1.4	17.8 ± 1.3
D	Cerebellum	0.51 ± 0.12	3.51 ± 0.51	0.18 ± 0.08	0.06 ± 0.02	2.93 ± 0.59	10.29 ± 2.19	13.80 ± 2.49
	Frontal	0.44 ± 0.10	3.51 ± 0.51	0.25 ± 0.08	0.06 ± 0.01	4.52 ± 0.68	15.92 ± 3.10	19.43 ± 3.48
	Cingulate	0.38 ± 0.10	3.51 ± 0.51	0.24 ± 0.08	0.05 ± 0.01	5.16 ± 0.55	18.13 ± 3.05	21.64 ± 3.48
	Putamen	0.52 ± 0.13	3.51 ± 0.51	0.29 ± 0.09	0.06 ± 0.01	4.71 ± 0.66	16.60 ± 3.27	20.11 ± 3.67
	Thalamus	0.54 ± 0.14	3.51 ± 0.51	0.33 ± 0.09	0.05 ± 0.01	6.20 ± 1.04	21.86 ± 4.81	25.38 ± 5.17
	Temporal	0.39 ± 0.08	3.51 ± 0.51	0.25 ± 0.09	0.05 ± 0.01	5.17 ± 0.75	18.18 ± 3.44	21.70 ± 3.81
	Posterieur	0.65 ± 0.25	3.51 ± 0.51	0.30 ± 0.11	0.06 ± 0.02	5.32 ± 0.89	18.71 ± 3.77	22.22 ± 4.10
	%COV	2.6 ± 1.2	—	10.3 ± 5.3	11.9 ± 6.0	2.8 ± 1.5	2.8 ± 1.5	2.3 ± 1.3
	%VAR	23.7 ± 2.8	14.4	33.7 ± 5.8	25.5 ± 2.5	15.2 ± 3.2	19.7 ± 1.8	18.0 ± 1.4
Moclobemide
C	Cerebellum	0.38 ± 0.10	3.51 ± 0.51	0.03 ± 0.01	0.03 ± 0.01	1.03 ± 0.14	3.61 ±0.57	7.12 ± 0.95
	Frontal	0.37 ± 0.09	3.51 ± 0.51	0.03 ± 0.01	0.02 ± 0.01	1.21 ± 0.26	4.17 ± 0.64	7.68 ± 0.75
	Cingulate	0.29 ± 0.07	3.51 ± 0.51	0.03 ± 0.01	0.02 ± 0.01	1.36 ± 0.27	4.68 ± 0.50	8.20 ± 0.59
	Putamen	0.44 ± 0.11	3.51 ± 0.51	0.05 ± 0.03	0.04 ± 0.02	1.23 ± 0.20	4.31 ± 0.97	7.83 ± 1.35
	Thalamus	0.41 ± 0.12	3.51 ± 0.51	0.07 ± 0.02	0.05 ± 0.01	1.26 ± 0.24	4.43 ± 1.08	7.94 ± 1.46
	Temporal	0.32 ± 0.06	3.51 ± 0.51	0.03 ± 0.01	0.03 ± 0.01	1.17 ± 0.21	4.07 ± 0.68	7.58 ± 0.94
	Posterieur	0.42 ± 0.11	3.51 ± 0.51	0.04 ± 0.01	0.03 ± 0.01	1.32 ± 0.23	4.59 ± 0.78	8.11 ± 1.05
	%COV	2.4 ± 0.8	2.2 ± 1.0	12.1 ± 2.1	17.3 ± 5.2	—	—	—
	%VAR	24.8 ± 3.2	14.4	35.2 ± 8.3	30.5 ± 8.9	17.8 ± 2.6	17.5 ± 4.6	13.1 ± 3.9

Values are mean ± s.d. of five experiments. %C0V corresponds to the standard error of parameter estimated by the nonlinear least-squares fitting procedure, is expressed in percent of the parameter value and is given as an index of the parameter identifiability. %VAR corresponds to the intersubject variability in the parameter estimate.

At placebo conditions, Method C provided k₃/k₄ values in accordance with the known regional densities of MAO-A in the brain but data were affected by a large intersubject variability (mean % VAR = 34.8 ± 2.2; Table 4). DV_B and DV_T values estimated with Method C were similar and not significantly different than those estimated with Method B.

Method D provided excellent estimations of k₃/k₄ values with a %COV of 2.8 ± 1.5 and a %VAR of 15.2 ± 3.2 (Table 4). Despite this superiority in determining BPs, the mean DV_T and DV_B values obtained with Method D were very close to those estimated with Methods B and C (see Tables 3 and 4). In addition, DV_T values obtained using the three methods (B, C, and D) were not significantly different than those derived from the Logan approach (Figure 4A). DV_T values obtained with each 2CM method (irrespective of whether it was constrained on unconstrained) significantly correlated with DV values obtained with the Logan approach and this is illustrated in Figure 5 with Method B as an example (slope = 1.05; R = 0.98; Student's t-test: P < 0.001).

Figure 5

Correlation between DV_T values obtained using Method B and the Logan graphical approach. The linear regression line was calculated using individual regional values obtained among the five subjects at placebo conditions. Note that similar correlations with Logan-derived DV_T values were obtained with all 2CM configurations. R = Pearson's correlation coefficient.

A comparison between the mean regional DV_B values as measured in our experimental conditions using [¹¹C]-harmine and Method D and the mean [³H]Ro 41–1049 specific binding values reported in vitro in corresponding regions of the rat brain is shown in Figure 6. Positron emission tomography measures of [¹¹C]-harmine DV_B in the cerebellum, frontal cortex, cingulate, putamen, and thalamus were significantly correlated with in vitro measures of MAO-A binding sites densities reported in these regions by Saura et al (1992; Figure 6: R = 0.95; Student's t-test: P < 0.01). Similarly, a high correlation was observed between [¹¹C]-harmine BP values and [³H]Ro 41-1049 B_max across these brain regions (R = 0.95; Student's t-test: P < 0.05; data not shown).

Figure 6

Correlations between [¹¹C]-harmine DV_B values in various brain regions as determined using Method D and the density of [³H]Ro 41-1049 binding sites reported reported in vitro in corresponding regions of the rat brain by Saura et al (1992).

Moclobemide studies: As observed at placebo conditions, a 2CM produced better fits to the data than a 1CM configuration (Figure 3B; Table 2). F-tests also indicated that a 2CM significantly better fit the data than a 1CM for 31 of the 35 regional binding curves. DV_T values obtained with a 1CM were, in average, 19% lower than those obtained with a 2CM (Figure 4B).

A 2CM (Method B) allowed reasonable identification of K₁ and k₂ but both $k_{3}^{*}$ and k₄ values were poorly identified (Table 3). DV_B and DV_T values were estimated with good precision and were significantly reduced in all regions when compared with placebo (repeated measures ANOVA: F_1,34 = 481.6; P < 0.001 for DV_B and F_1;34 = 368.7; P < 0.001 for DV_T). Using data obtained at both placebo and moclobemide conditions, DV_T and DV_B values estimated with an unconstrained 2CM were highly correlated (Figure 7). The percentage of moclobemideinduced DV_B reduction (which is a reflection of the percentage of MAO-A blockade—see equation (12)) ranged from 63% ± 7% in the cerebellum to 82% ± 5% in the thalamus. DV_T values reduction ranged from 43% ± 8% in the cerebellum to 66% ± 6% in the thalamus.

Figure 7

Correlations between [¹¹C]-harmine DV_B and DV_T values obtained with Method B. The plot contains data obtained at placebo and after moclobemide treatment in each of the five subjects and for all ROIs. Linear regression of the data yielded a y intercept of 3.8 mL/g, a value that can be used as an overall estimate of DV_{NS + F}.

Methods C fits did not statistically modify goodness-of-fits when compared with Method B as assessed using AIC and MSC (Table 2). F-tests also indicated that Method C was not statistically different from Method B in most fits. K₁ values estimated using Method C at moclobemide conditions were significantly lower than those obtained at placebo (repeated measures ANOVA: F_1,34 = 22.9; P < 0.001; Table 4). However, DV_{NS + F} values were not significantly different between the two conditions (2.74 ± 0.84 and 3.51 ± 0.51 at placebo and moclobemide conditions, respectively; Student's t-test: P > 0.05; Table 4), suggesting that free and nonspecific binding were similar at placebo and moclobemide conditions. This was the main rationale for using K₁/k₂ values obtained in the moclobemide studies to fit the placebo data (i.e., Method D). When compared with k₃ values obtained with Method D at placebo conditions, there were significant reductions in the values of $k_{3}^{*}$ in the moclobemide studies (Table 4). Surprisingly, k₄ values were also significantly reduced by moclobemide treatment by 34% on average (repeated measures ANOVA: F_1,34 = 47.2; P < 0.001; P < 0.001; see Method C in Table 4). Fits of the moclobemide data were attempted by fixing k₄ to the values obtained in the placebo studies. This procedure worsened the data fits in all subjects (data not shown). Similarly, acceptable fits of the placebo data could not be obtained when k₄ was fixed to the value obtained in the moclobemide studies, suggesting that k₄ was truly different between the two sets of studies.

The percentage blockade of MAO-A calculated using DV_B and BP values were identical (calculations were performed using Methods D and C in the placebo and moclobemide studies, respectively) and ranged from 64% ± 7% in the cerebellum to 79% ± 4% in the thalamus. This could have been predicted as Methods D and C assumes that K₁/k₂ values are similar between brain regions at placebo and moclobemide conditions. Similarly, the percentage blockade of MAO-A calculated using DV_B values obtained with an unconstrained 2CM gave values that were, in average, only 2.2% lower (range: −4.1% to + 1.4%) than those estimated with BP values (Figure 8; repeated measures ANOVA: F_1,34 = 47.2; P > 0.05).

Figure 8

Percentage of monoamine oxidase (MAO)-A blockade induced by a 10 days pretreatment with clinical doses of moclobemide (see Materials and Methods) in healthy human subjects. Monoamine oxidase -A blockade was calculated in two different ways: either by using DV_B values obtained with Method B or by using BP values obtained with Methods D and C at placebo and at moclobemide conditions, respectively. The percentages of MAO-A blockade calculated using either way were not statistically different (repeated measures AN OVA; P > 0.05). Values are mean ± s.d (n = 5).

Discussion

The primary goal of this study was to evaluate methods to obtain reliable estimates of MAO-A density from the analysis of [¹¹C]-harmine brain uptake curves in humans using PET. At placebo conditions, the regional distribution of [¹¹C]-harmine uptake was concordant with the known densities of MAO-A sites in the brain. The highest levels of radioactivity were observed in the thalamus, intermediate levels were observed in the striatum and cortical regions, whereas lowest levels of radioactivity were observed in cerebellum. At moclobemide conditions, the relative regional differences in [¹¹C]-harmine uptake levels were markedly reduced, indicating a substantial blockade of the MAO-A sites. However, the peak levels of radioactivity in the brain as well as in the plasma were higher at moclobemide than at placebo conditions. Monoamine oxidase-A is abundant in peripheral organs such as the liver, kidneys, heart, lungs, and duodenum in humans (Fowler et al, 2003; Saura et al, 1996b). Moclobemide-induced inhibition at these peripheral sites likely explains the increased bioavailability of [¹¹C]-harmine in the plasma and, therefore, the increased net amount of radioactivity that enters the brain when compared with placebo conditions.

Kinetic analyses of [¹¹C]-harmine brain uptake data showed that a 2CM configuration provided significantly better fits than a 1CM configuration. This was true both at placebo and moclobemide treatment conditions and for all criteria (AIC, MSC, and F-statistics). Despite this statistical superiority of a 2CM in describing [¹¹C]-harmine time–activity data, the difference in DV_T values obtained between a 1CM and a 2CM was small, especially at placebo conditions. This likely indicated a poor separation of the free and nonspecifically bound and specifically bound compartments. This view was supported by the poor identification of k₃ and k₄ using an unconstrained 2CM. However, the statistically better fits obtained with all 2CM configurations as compared with the 1CM indicate that, although the differences were small, DV_T values should be more accurately determined with inclusion of a second tissue compartment.

A 2CM configuration provided good identification of K₁ at both treatment conditions. However, K₁ values were significantly reduced at moclobemide conditions when compared with placebo, a result suggesting a decreased cerebral blood flow (CBF) during moclobemide treatment. There is a large body of evidence showing that serotonin and norepinephrine, the two major MAO-A substrates that are known to be elevated during moclobemide treatment (Nair et al, 1993), are potent cerebral vasoconstrictors and reduce CBF (Grome and Harper, 1983; Szabo et al, 1983). Stimulations of the dorsal raphe nucleus, from which 90% of the serotonin innervation of the brain originates, have also been shown to decrease CBF in a large number of brain structures (Bonvento et al, 1989). Finally, studies in MAO-A-deficient mice, which show chronic elevation of serotonin and norepinephrine in the brain, also showed decreased blood flow in several brain regions when compared with their wild-type counterparts (Holschneider et al, 2000). In light of these data, it can be postulated that the decreased K₁ values observed during moclobemide treatment reflect vasoconstrictor-induced decreases in CBF resulting from increased brain levels of serotonin and norepinephrine.

An unconstrained 2CM yielded individual parameter values for k₂, k₃, and $k_{3}^{*}$ that were highly variable and BP values that did not correlate with the known densities of MAO-A in the brain. In contrast, DV_B and DV_T values were estimated with high precision at both placebo and moclobemide conditions. Although DV_B and DV_T includes information both on radioligand delivery and binding to specific receptor sites, at equilibrium they should be independent of tracer delivery and give an index of receptor site density (Logan et al, 1994). This view was supported by the observation that the rank of order of both parameter values correlated well with the typical rank of order of MAO densities found in both the rat and human brain post mortem (Saura et al, 1996a, 1992). It was further supported by the observation that moclobemide treatment induced 43% to 66% and 63% to 82% reductions in DV_T and DV_B values, respectively, when compared with placebo conditions, thus suggesting that variations in these parameters, and more specifically in DV_B, are more likely to reflect variations in receptor numbers than variations in nonspecific binding.

The difficulty in deriving reliable individual estimates for k₃, $k_{3}^{*}$ , and k₄ from a 2CM four parameters configuration while obtaining stable DV_T has already been reported for other radioligands (Abi-Dargham et al, 1994; Koeppe et al, 1996). By imposing constraints on the fitting procedure, more consistent estimates of receptor binding parameters were obtained. One approach that is commonly used consists of constraining the levels of free and nonspecific binding in receptor-rich regions to the level measured in a brain region, often referred as ‘reference region’, devoid of specific binding sites (Farde et al, 1989; Koeppe et al, 1991). This approach could not be used with [¹¹C]-harmine as, in accordance with in vitro measurements showing an ubiquitous distribution of MAO-A in the brain (Saura et al, 1992, 1996a), no reference region could be identified. Another approach consists of constraining parameter values in the model that are predicted not to have large variations between brain regions. Method C, which assumes that K₁/k₂ is similar in all brain structures, has already been shown to increase substantially the identifiability of kinetic parameters (Buck et al, 1996; Ginovart et al, 2001). Method C substantially reduced the COVs on k₃ and k₄ estimates when compared with Method B and gave regional BP values in accordance with the known distribution of MAO-A in the brain (Saura et al, 1996a, 1992). Method D consisted of estimating the free and nonspecific binding from the blocking studies and to subsequently use that estimate in the analysis of placebo data. This approach has been successfully used to improve the kinetic parameter estimation of radioligands such as [¹⁸F]spiperone (Logan et al, 1987) and [¹¹C]flumazenil (Price et al, 1993). Method D also yielded BP values in accordance with physiologic expectation and was the method providing the highest accuracy and lowest interindividual variability in BP estimates. For that reason, we consider it to be the method of choice for derivation of [¹¹C]-harmine BPs at placebo conditions. If no data are available at drug-treatment conditions and Method D is thus not applicable, then Method C can be used to estimate BPs.

All three 2CM-based methods (Methods B, C, and D) gave similar estimates for DV_T and DV_B and both parameters were estimated with high reliability. In addition, DV_T values estimated with the three methods were highly correlated and not significantly different from those estimated with the Logan approach. Thus, DV_T and DV_B appear to be valuable outcome measures for clinical studies as they are determined with high level of precision and are not biased by potential subject differences in blood flow or peripheral clearance. Moreover, the high correlation between DV_T and DV_B values obtained at both placebo and moclobemide conditions suggests that nonspecific binding of [¹¹C]-harmine was similar not only across brain regions but also between conditions. As MAO-A blockade increased, DV_T and DV_B decreased, increasingly approaching DV_{NS + F} values. As predicted by equation (9), which describes DV_T as a linear function of DV_B, the plot of DV_T as a function of DV_B should have a slope equal to unity and a y intercept of DV_{NS + F}. Linear regression across all subjects, all regions, and across both treatment conditions yielded values of 3.8 mL/g for DV_{NS + F}. This value is in close agreement with the value obtained at moclobemide conditions using Method C, which yielded a mean DV_{NS + F} value of 3.5 ± 0.5 mL/g (Table 4). When linear regression was performed using all subjects and all regions but only data acquired at placebo conditions, it yielded a DV_{NS + F} value of 3.1 mL/g (data not shown). Here again, this indirect way of deriving DV_{NS + F} estimate using an unconstrained 2CM (i.e., Method B) yielded a DV_{NS + F} value that was close to that determined by Method C at placebo conditions (2.74 ± 0.84; Table 4). The consistencies between the DV_{NS + F} values extrapolated from the regression data obtained with Method B and describing DV_T as a function of DV_B and those directly estimated by the model by Method C thus support the model simplifications made in this latter method and further strengthen the case for its use in clinical studies.

Treatment with moclobemide induced substantial reductions in k₃, DV_B, and DV_T values when compared with placebo. Nonspecific binding was relatively unaffected by the pharmacological treatment as shown by the similar DV_{NS + F} values obtained using Method C at placebo and moclobemide conditions. In contrast, the dissociation rate constant k_A for [¹¹C]-harmine was substantially decreased after moclobemide treatment when compared with placebo. Such a decrease was unexpected as k₄, which equals k_off, is usually considered not to vary between experimental conditions. Investigations where a compartmental modeling approach has been applied to obtain the in vivo kinetic parameter estimates of a radioligand both at drug-free and at drug-treatment conditions are scarce. While the assumption of a constant k₄ is reasonable when comparing different brain regions in a same subject and at similar experimental conditions, k₄ may vary with pharmacological treatment. Two explanations can be proposed for a decreased k₄ under moclobemide treament. One explanation is the existence of a nonspecific binding compartment that turns over relatively slowly because of nonspecific interactions with proteins, lipids or other receptor compartment. Because a major proportion of the specific binding compartment is blocked under moclobemide conditions, this slow kinetic nonspecific binding compartment would become more prominent in the moclobemide than in the placebo studies, and this would be reflected in an apparent decrease in k₄. Another possibility is that the effect of moclobemide on k₄ is indicative of an allosteric modulation of MAO-A by the drug. Negative allosteric modulators are known not only to decrease radioligand association on the orthosteric site but also to decrease their dissociation rate (Holzgrabe and Mohr, 1998). Although the mechanism of MAO-A inhibition by moclobemide is not fully clear yet, several studies suggest that moclobemide may act through mixed competitive and noncompetitive inhibition of the enzyme (Da Prada et al, 1989; Waldmeier, 1985). The decrease in k₄ obtained after moclobemide treatment could reflect this noncompetitive interaction. Assuming that this decrease in k₄ (i.e., k_off) was accompanied by a similar decrease in k_on (and that K_n was thus unaffected; see assumptions for equations (8) and (12)), the percentage of MAO-A blockade induced by clinical doses of moclobemide was substantial, ranging from 64% in the cerebellum to 79% in the thalamus. Because our data cannot rule out the possibility of a decrease in k₄ with no parallel change in k_on, it is thus possible that the percentage of MAO-A blockade achieved by moclobemide is somewhat even higher than 64% to 79%. This substantial level of MAO-A blockade induced by moclobemide further supports the specificity and suitability of [¹¹C]-harmine for quantifying MAO-A binding sites in vivo using PET.

In summary, kinetic analyses showed that a 2CM best describes [¹¹C]-harmine kinetics, although it showed poor identifiability of BP values. However, it provided stable and reliable estimates for [¹¹C]-harmine regional DV_B and DV_T that were consistent with the known distribution of MAO-A densities in the human brain. Reduction in the complexity of the 2CM, either by fixing or by coupling DV_{NS + F} value during the fits, allowed the derivation of more stable estimates for BPs but neither improved goodness-of-fits nor improved reliability in DV_T and DV_B estimates when compared with an unconstrained 2CM. Under ideal conditions, measures of BP should be preferred as index of binding density because they vary directly with B_max. However, as these measures are estimated with uncertainty for [¹¹C]-harmine, we would recommend using DV_B as index of binding site density. Because the variability and identifiability of DV_B values estimated with all 2CM methods were similar, an unconstrained 2CM configuration should be selected for determination of this parameter as outcome measure as this method has the advantage of computational simplicity. As an alternative, DV_T could also be selected as outcome measure. Indeed, DV_T values estimated with all 2CM configurations were remarkably stable, highly correlated, and not significantly different from those estimated with the Logan method. This latter method of analysis is even simpler as it only involves a simple linear regression of the data.

The results of the present study thus support the suitability of using [¹¹C]-harmine for the quantitative evaluation of MAO-A densities in human using PET. The stability and reliability of DV_B values obtained for an unconstrained 2CM, together with the computational simplicity associated with this method, support the use of DV_B as an appropriate outcome measure for [¹¹C]-harmine.

Footnotes

Acknowledgements

The assistance of Alvina Ng during the PET experiments, Kernjit Singh for the metabolite analysis, and Armando Garcia and Alexandra Chestakova for radiochemistry support is gratefully acknowledged.

References

Abi-Dargham

Lamelle

Seibyl

Rattner

Baldwin

Zoghbi

Zea-Ponce

Bremner

Hyde

Charney

(1994) SPECT measurement of benzodiazepine receptors in human brain with iodine-123-iomazenil: kinetic and equilibrium paradigms. J Nucl Med 35:228–38

Akaike

(1974) A new look at the statistical model identification. IEEE Trans Automat Contr 19:716–23

Ametamey

Beer

Guenther

Antonini

Leenders

Waldmeier

Schubiger

(1996) Radiosynthesis of [¹¹C]brofaromine, a potential tracer for imaging monoamine oxidase A. Nucl Med Biol 23:229–34

Angst

Stabl

(1992) Efficacy of moclobemide in different patient groups: a meta-analysis of studies. Psychopharmacology (Berlin) 106(Suppl):S109–13

Bergstrom

Westerberg

Kihlberg

Langstrom

(1997a) Synthesis of some ¹¹C-labelled MAO-A inhibitors and their in vivo uptake kinetics in rhesus monkey brain. Nucl Med Biol 24:381–8

Bergstrom

Westerberg

Langstrom

(1997b) 11C-harmine as a tracer for monoamine oxidase a (MAO-A): in vitro and in vivo studies. Nucl Med Biol 24:287–93

Bergstrom

Westerberg

Nemeth

Traut

Gross

Greger

Muller-Peltzer

Safer

Eckernas

Grahner

Langstrom

(1997c) MAO-A inhibition in brain after dosing with esuprone, moclobemide and placebo in healthy volunteers: in vivo studies with positron emission tomography. Eur J Clin Pharmacol 52:121–8

Bonnet

(2003) Moclobemide: therapeutic use and clinical studies. CNS Drug Rev 9:97–140

Bonvento

Lacombe

Seylaz

(1989) Effects of electrical stimulation of the dorsal raphe nucleus on local cerebral blood flow in the rat. J Cereb Blood Flow Metab 9:251–5

10.

Bottlaender

Dolle

Guenther

Roumenov

Fuseau

Bramoulle

Curet

Jegham

Pinquier

George

Valette

(2003) Mapping the cerebral monoamine oxidase type A: positron emission tomography characterization of the reversible selective inhibitor [¹¹C]befloxatone. J Pharmacol Exp Ther 305:467–73

11.

Buck

Westera

vonSchulthess

Burger

(1996) Modeling alternatives for cerebral carbon-11-iomazenil kinetics. J Nucl Med 37:699–705

12.

Carson

(1986) Parameters estimation in positron emission tomography. In: Positron emission tomography. Principles applications for the brain and the heart ( Phelps

Mazziota

Schelbert

, eds), New York: Raven Press, 347–90

13.

Carson

Channing

Blasberg

Dunn

Cohen

Rice

Herscovitch

(1993) Comparison of bolus and infusion methods for receptor quantitation: application to [¹⁸F]cyclofoxy and positron emission tomography. J Cereb Blood Flow Metab 13:24–42

14.

Cohen

(1990) Monoamine oxidase and oxidative stress at dopaminergic synapses. J Neural Transm Suppl 32:229–38

15.

Da Prada

Kettler

Keller

Burkard

Haefely

(1989) Preclinical profiles of the novel reversible MAO-A inhibitors, moclobemide and brofaromine, in comparison with irreversible MAO inhibitors. J Neural Transm Suppl 28:5–20

16.

Farde

Eriksson

Blomquist

Halldin

(1989) Kinetic analysis of central [¹¹C]raclopride binding to D₂-dopamine receptors studied by PET—a comparison to the equilibrium analysis. J Cereb Blood Flow Metab 9:696–708

17.

Fowler

Logan

Wang

Franceschi

Volkow

Telang

Pappas

Ferrieri

Shea

Garza

King

Schlyer

Gatley

Ding

Warner

Netusil

Carter

Jayne

Alexoff

Zhu

Vaska

(2003) Monoamine oxidase a imaging in peripheral organs in healthy human subjects. Synapse 49:178–87

18.

Fowler

MacGregor

Wolf

Arnett

Dewey

Schlyer

Christman

Logan

Smith

Sachs

(1987) Mapping human brain monoamine oxidase A and B with ¹¹C-labeled suicide inactivators and PET. Science 235:481–5

19.

Fowler

Wiberg

Oreland

Danielsson

Palm

Winblad

(1981) Monoamine oxidase activity and kinetic properties in platelet-rich plasma from controls, chronic alcoholics, and patients with nonalcoholic liver disease. Biochem Med 25:356–65

20.

Ginovart

Wilson

Meyer

Hussey

Houle

(2001) Positron emission tomography quantification of [¹¹C]-DASB binding to the human serotonin transporter: modeling strategies. J Cereb Blood Flow Metab 21:1342–53

21.

Grome

Harper

(1983) The effects of serotonin on local cerebral blood flow. J Cereb Blood Flow Metab 3:71–7

22.

Hilton

Yokoi

Dannals

Ravert

Szabo

Wong

(2000) Column-switching HPLC for the analysis of plasma in PET imaging studies. Nucl Med Biol 27:627–30

23.

Holschneider

Scremin

Huynh

Chen

Seif

Shih

(2000) Regional cerebral cortical activation in monoamine oxidase A-deficient mice: differential effects of chronic versus acute elevations in serotonin and norepinephrine. Neuroscience 101:869–77

24.

Holzgrabe

Mohr

(1998) Allosteric modulators of ligand binding to muscarinic acetylcholine receptors. Drug Discovery Today 3:214–22

25.

Jarman

Glover

Sandler

Turjanski

Stern

(1993) Platelet monoamine oxidase B activity in Parkinson's disease: a re-evaluation. J Neural Transm Park Dis Dement Sect 5:1–4

26.

Kim

Sablin

Ramsay

(1997) Inhibition of monoamine oxidase A by beta-carboline derivatives. Arch Biochem Biophys 337:137–42

27.

Koeppe

Frey

Vander Borght

Karlamangla

Jewett

Lee

Kilbourn

Kuhl

(1996) Kinetic evaluation of [¹¹C]dihydrotetrabenazine by dynamic PET: measurement of vesicular monoamine transporter. J Cereb Blood Flow Metab 16: 1288–1299

28.

Koeppe

Holthoff

Frey

Kilbourn

Kuhl

(1991) Compartmental analysis of [¹¹C]flumazenil kinetics for the estimation of ligand transport rate and receptor distribution using positron emission tomography. J Cereb Blood Flow Metab 11:735–44

29.

Landaw

DiStefano

III (1984) Multiexponential, multicompartmental, and noncompartmental modeling. II. Data analysis and statistical considerations. Am J Physiol 246:R665–677

30.

Lewine

Meltzer

(1984) Negative symptoms and platelet monoamine oxidase activity in male schizophrenic patients. Psychiatry Res 12:99–109

31.

Lipper

Murphy

Slater

Buchsbaum

(1979) Comparative behavioral effects of clorgyline and pargyline in man: a preliminary evaluation. Psychopharmacology (Berlin) 62:123–8

32.

Litton

Holte

Eriksson

(1990) Evaluation of the Karolinska new positron camera: the Scanditronix PC2048–15B. IEEE Trans Nucl Sci NS 37:743–8

33.

Livingston

(1996) Monoamine oxidase inhibitors. An update on drug interactions. Drug Saf 14:219–27

34.

Logan

Fowler

Volkow

Wolf

Dewey

Schlyer

MacGregor

Hitzemann

Bendriem

Gatley

(1990) Graphical analysis of reversible radioligand binding from time–activitymeasurements applied to [N-¹¹C-methyl]-(–)-cocaine PET studies in human subjects. J Cereb Blood Flow Metab 10:740–7

35.

Logan

Volkow

Fowler

Wang

Dewey

MacGregor

Schlyer

Gatley

Pappas

King

(1994) Effects of blood flow on [¹¹C]raclopride binding in the brain: model simulations and kinetic analysis of PET data. J Cereb Blood Flow Metab 14: 995–1010

36.

Logan

Wolf

Shiue

Fowler

(1987) Kinetic modeling of receptor-ligand binding applied to positron emission tomographic studies with neuroleptic tracers. J Neurochem 48:73–83

37.

Luthra

Osman

Turton

Vaja

Dowsett

Brady

(1992) An automated system based on solid phase extraction and HPLC for the routine determinination in plasma of unchanged [¹¹C]-L-deprenyl, [¹¹C]diprenorphine, [¹¹C]flumazenil, [¹¹C]raclopridee and [¹¹C]Scherring 23390. IXth International Symposium on Badiopharmaceutical Chemistry, Paris, K15

38.

Mann

Chiu

(1978) Platelet monoamine oxidase activity in Huntington's chorea. J Neurol Neurosurg Psychiatry 41:809–12

39.

Mann

Stanley

(1984) Postmortem monoamine oxidase enzyme kinetics in the frontal cortex of suicide victims and controls. Acta Psychiatr Scand 69:135–9

40.

Marquardt

(1963) An algorithm for least-squares estimation on non-linear parameters. J Soc Ind Appl Math 11:431–41

41.

Mintun

Raichle

Kilbourn

Wooten

Welch

(1984) A quantitative model for the in vivo assessment of drug binding sites with positron emission tomography. Ann Neurol 15:217–27

42.

Murphy

Weiss

(1972) Reduced monoamine oxidase activity in blood platelets from bipolar depressed patients. Am J Psychiatry 128:1351–7

43.

Nair

Ahmed

Kin

(1993) Biochemistry and pharmacology of reversible inhibitors of MAO-A agents: focus on moclobemide. J Psychiatry Neurosci 18:214–25

44.

Ohmomo

Hirata

Murakami

Magata

Tanaka

Yokoyama

(1991) Synthesis of fluorine and iodine analogues of clorgyline and selective inhibition of monoamine oxidase A. Chem Pharm Bull (Tokyo) 39:1038–40

45.

Ohmomo

Hirata

Murakami

Magata

Tanaka

Yokoyama

(1993) Synthesis and characterization of ¹¹C-labeled fluoroclorgyline: A monoamine oxidase A specific inhibitor for positron emission tomography. Chem Pharm Bull (Tokyo) 41:1994–7

46.

Ordway

Farley

Dilley

Overholser

Meltzer

Balraj

Stockmeier

Klimek

(1999) Quantitative distribution of monoamine oxidase A in brainstem monoamine nuclei is normal in major depression. Brain Res 847:71–9

47.

Oreland

Fowler

Schalling

(1981) Low platelet monoamine oxidase activity in cigarette smokers. Life Sci 29:2511–8

48.

Phelps

Huang

Hoffman

Kuhl

(1979) Validation of tomographic measurement of cerebral blood volume with C-11-labeled carboxyhemoglobin. J Nucl Med 20:328–34

49.

Price

Mayberg

Dannals

Wilson

Ravert

Sadzot

Rattner

Kimball

Feldman

Frost

(1993) Measurement of benzodiazepine receptor number and affinity in humans using tracer kinetic modeling, positron emission tomography, and [¹¹C]flumazenil. J Cereb Blood Flow Metab 13: 656–67

50.

Saura

Bleuel

Ulrich

Mendelowitsch

Chen

Shih

Malherbe

Da Prada

Richards

(1996a) Molecular neuroanatomy of human monoamine oxidases A and B revealed by quantitative enzyme radioautography and in situ hybridization histochemistry. Neuroscience 70:755–74

51.

Saura

Kettler

Da Prada

Richards

(1992) Quantitative enzyme radioautography with ³H-Ro 41-1049 and ³H-Ro 19-6327 in vitro: localization and abundance of MAO-A and MAO-B in rat CNS, peripheral organs, and human brain. J Neurosci 12:1977–99

52.

Saura

Nadal

van den Berg

Vila

Bombi

Mahy

(1996b) Localization of monoamine oxidases in human peripheral tissues. Life Sci 59:1341–9

53.

Sherif

Marcusson

Oreland

(1991) Brain gammaaminobutyrate transaminase and monoamine oxidase activities in suicide victims. Eur Arch Psychiatry Clin Neurosci 241:139–44

54.

Shih

Chen

Ridd

(1999) Monoamine oxidase: from genes to behavior. Annu Rev Neurosci 22: 197–217

55.

Studholme

Hill

DLG

Hawkes

(1997) Automated three-dimensional registration of magnetic resonance and positron emission tomography brain images by multiresolution optimisation of voxel similarity measures. Med Phys 24:25–35

56.

Szabo

Kovach

Babosa

Greenberg

Revich

(1983) Effect of sustained norepinephrine infusion on local cerebral blood flow in the rat. Circ Shock 10:101–117

57.

Waldmeier

(1985) On the reversibility of reversible MAO inhibitors. Naunyn Schmiedebergs Arch Pharmacol 329:305–10

58.

Wilson

DaSilva

Houle

(1994) Facile radiolabelling and purification of 2β-[O-¹¹CH₃]-carbomethoxy-3β-aryltropanes: radiotracers for the dopamine transporter. J Labelled Compd Radiopharm 34:759–65

59.

Wilson

Garcia

Jin

Houle

(2000) Radiotracer synthesis from [¹¹C]-iodomethane: a remarkably simple captive solvent method. Nucl Med Biol 27:529–32

60.

Youdim

(1988) Platelet monoamine oxidase B: use and misuse. Experientia 44:137–41

Positron Emission Tomography Quantification of [ 11 C]-Harmine Binding to Monoamine Oxidase-A in the Human Brain

Abstract

Keywords

Introduction

Methods

Subjects

Radiochemistry

Positron Emission Tomography System

Positron Emission Tomography Studies

Determination of Radioactive Metabolites in Plasma

Regions of Interest

Data Analysis

Simplified Method of Analysis

Results

Plasma Analysis

Kinetic Analyses

Discussion

Footnotes

Acknowledgements

References