Imaging the norepinephrine transporter in humans with (S,S)-[11C]O-methyl reboxetine and PET: problems and progress

Nuclear Medicine and Biology 34 (2007) 667 – 679 www.elsevier.com/locate/nucmedbio

Imaging the norepinephrine transporter in humans with (S,S)-[ 11 C]O-methyl reboxetine and PET: problems and progress Jean Logan a,⁎, Gene-jack Wang a , Frank Telang a , Joanna S. Fowler a , David Alexoff a , John Zabroski a , Millard Jayne a , Barbara Hubbard a , Payton King a , Pauline Carter a , Colleen Shea a , Youwen Xu a , Lisa Muench a , David Schlyer a , Susan Learned-Coughlin b , Valerie Cosson b , Nora D. Volkow c , Yu-shin Ding d a

Medical Department, Brookhaven National Laboratory, Upton, NY 11973, USA b GlaxoSmithKline, Research Triangle Park, NC 27709, USA c National Institute on Drug Abuse, Bethesda, MD 20892, USA d Department of Diagnostic Radiology, Yale University School of Medicine, New Haven, CT 06520-8048, USA Received 5 January 2007; received in revised form 3 March 2007; accepted 27 March 2007

Abstract Results from human studies with the PET radiotracer (S,S)-[11C]O-methyl reboxetine ([11C](S,S)-MRB), a ligand targeting the norepinephrine transporter (NET), are reported. Quantification methods were determined from test/retest studies, and sensitivity to pharmacological blockade was tested with different doses of atomoxetine (ATX), a drug that binds to the NET with high affinity (Ki=2–5 nM). Methods: Twenty-four male subjects were divided into different groups for serial 90-min PET studies with [11C](S,S)-MRB to assess reproducibility and the effect of blocking with different doses of ATX (25, 50 and 100 mg, po). Region-of-interest uptake data and arterial plasma input were analyzed for the distribution volume (DV). Images were normalized to a template, and average parametric images for each group were formed. Results: [11C](S,S)-MRB uptake was highest in the thalamus (THL) and the midbrain (MBR) [containing the locus coeruleus (LC)] and lowest for the caudate nucleus (CDT). The CDT, a region with low NET, showed the smallest change on ATX treatment and was used as a reference region for the DV ratio (DVR). The baseline average DVR was 1.48 for both the THL and MBR with lower values for other regions [cerebellum (CB), 1.09; cingulate gyrus (CNG) 1.07]. However, more accurate information about relative densities came from the blocking studies. MBR exhibited greater blocking than THL, indicating a transporter density ∼40% greater than THL. No relationship was found between DVR change and plasma ATX level. Although the higher dose tended to induce a greater decrease than the lower dose for MBR (average decrease for 25 mg=24±7%; 100 mg=31±11%), these differences were not significant. The different blocking between MBR (average decrease=28±10%) and THL (average decrease=17±10%) given the same baseline DVR indicates that the CDT is not a good measure for non-NET binding in both regions. Threshold analysis of the difference between the average baseline DV image and the average blocked image showed the expected NET distribution with the MBR (LC) and hypothalamusNTHLNCNG and CB, as well as a significant change in the supplementary motor area. DVR reproducibility for the different brain regions was ∼10%, but intersubject variability was large. Conclusions: The highest density of NETs was found in the MBR where the LC is located, followed by THL, whereas the lowest density was found in basal ganglia (lowest in CDT), consistent with the regional localization of NETs in the nonhuman primate brain. While all three doses of ATX were found to block most regions, no significant differences between doses were found for any region, although the average percent change across subjects of the MBR did correlate with ATX dose. The lack of a dose effect could reflect a low signal-to-noise ratio coupled with the possibility that a sufficient number of transporters were blocked at the lowest dose and further differences could not be detected. However, since the lowest (25 mg) dose is less than the therapeutic doses used in children for the treatment of attention-deficit/ hyperactivity disorder (∼1.0 mg/kg/day), this would suggest that there may be additional targets for ATX's therapeutic actions. Published by Elsevier Inc. Keywords: [11C](S,S)-MRB; PET; Norepinephrine transporter; Atomoxetine

⁎ Corresponding author. Chemistry Department, Upton, NY 11973, USA. Tel.: +1 631 344 4391; fax: +1 631 344 7902. E-mail address: [email protected] (J. Logan). 0969-8051/$ – see front matter. Published by Elsevier Inc. doi:10.1016/j.nucmedbio.2007.03.013

668

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

1. Introduction Norepinephrine (NE) and its reuptake system, the NE transporter (NET), are implicated in many aspects of CNS function [1]. Antidepressants such as desipramine (DMI) and venlafaxine, which act as inhibitors of the NET, have been in use for many years. Their therapeutic action is achieved by inhibition of the reuptake of NE into the nerve terminals from which it was released, thus prolonging its action in the synapse. Reduced levels of transporters in the locus coeruleus (LC) have been identified in major depression, perhaps as a compensatory mechanism [2]. The NET is also a target for drugs treating attention-deficit/hyperactivity disorder (ADHD) [3]. Loss of noradrenergic as well as dopaminergic innervation in the limbic system may be related to the anxiety and depression often associated with Parkinson's disease [4]. It is known that Parkinson's disease is associated with degeneration of noradrenergic neurons in the LC [5] and the loss of both dopamine (DA) and NE in the cortex [6,7]. A loss of the noradrenergic neurons in the LC has also been established in patients with Alzheimer's disease [8]. The LC–NE system has been shown to mediate arousal, but recently, Aston-Jones and Cohen [9] have proposed that it has a more general role in optimizing behavioral performance. The highest density of NETs is found in the LC, a dense cluster of neurons on either side of the brain stem (STM). It is the major source of NE with projections to many areas including the cerebral cortex, thalamus (THL) and cerebellum (CB); thus, the distribution of NETs in the brain is widespread [10]. This is in contrast to the localized concentration of the DA transporter (DAT), which has a high density in the striatum and a low concentration in other regions. The density of NETs is lower than that of the DAT in the striatum, and this contributes to the low signal-to-background ratio that has been found with many potential NET ligands. [3H]Nisoxetine, which binds to a single class of saturable sites identified with the NET, has been used in autoradiography [11,12], but the [11C] analog was projected to be unsuitable as a PET ligand due to the high nonspecific binding observed in mouse studies [13]. [125I]Nisoxetine, a potential SPECT ligand, as investigated by Kung et al. [14] and while useful in homogenate and autoradiography studies, displayed slow kinetics, reaching a hypothalamus-to-striatum ratio of 1.5 at 3 h postinjection. Kiyono et al. [15] tested a related compound ([125I] MIPP) in rats, finding a high accumulation of radioactivity in the LC, but other experimental data suggested binding to secondary sites in vivo. PET studies with [11C] nisoxetine in baboons revealed very little differences in regional binding. In fact, basal ganglia, a region with little or no concentration of NETs, showed higher uptake than THL, a region with a substantial NET concentration [16,17]. This may be due to the slow kinetics of [11C] nisoxetine. In any case, the low-affinity binding appears to predominate. Problems were also found with labeled forms

of the antidepressant DMI, which is selective for the NET [18,19]. The presence of a low-affinity binding site of DMI was noted by these authors as well as by others. Postmortem autoradiographic studies in humans with [3H] DMI found that DMI binding was generally correlated with NE innervation with the exception that caudate and putamen (CDT and PUT, respectively; regions with little NE content) had exhibited high binding [20]. [11C]DMI has been synthesized [21,22]; PET studies examining the regional distribution in the cynomolgus monkey brain found only slightly heterogeneous uptake (the ratio of THL to striatum was ∼1.15 at most), and uptake distribution was slightly affected by pretreatment with unlabeled DMI as was found in vitro [21]. More success has been achieved with the analogs of reboxetine, a potent antidepressant selective for the NET [23]. (S,S)-[11C]O-methyl reboxetine ([11C](S,S)-MRB) has been synthesized by several groups [24–26]. In vivo studies with [11C](S,S)-MRB (also referred to as [11C](S,S) MeNER) found hypothalamus-to-striatum ratios of 2.5 at 60 min in rat [25]. PET studies with the same ligand found the highest binding in THL with THL-to-striatum ratios of ∼1.5 in cynomolgus monkeys [24] and in baboons [26]. In addition, studies with [11C](S,S)-MRB in three human subjects were described in an abstract in which distribution volume ratios (DVRs) for THL were found to be in the range of 1.24 to 1.4 using the CB and the CDT as reference regions [27]. Although [11C](S,S)-MRB is an improvement over nisoxetine and DMI, this tracer has a low specific-tononspecific ratio compared to other PET tracers. Another problem encountered is the slow uptake of the tracer, necessitating scanning times greater than 1 h. This extended scanning coupled with low uptake increased the noise at the later time points due to the short half-life of carbon-11. The 18F label was introduced in place of the 11 C-methyl group to improve the signal-to-noise ratio at the latter time points [28,29]. Deuterium was introduced to slow the defluorination in order to decrease the bone radioactivity that was found to contaminate images in cortical areas. The [18F]fluoroethylated analog (S,S)-[18F] FRB showed less defluorination in vivo, and the deuteriumsubstituted (S,S)-[18F]FRB-D4 reduced the defluorination to an unobservable level [29]. Interestingly, peak radioactivity with the 18F analogs occurs early [for (S,S)-FMeNER-D2 at about 12 min] with observable washout in both high and low NET regions such as THL and CDT, respectively. This possibly indicates that this structural change altered the non-NET binding in all regions. In PET studies with cynomolgus monkeys, Seneca et al. [30] found that uptake of this ligand responded in a dose-dependent fashion to atomoxetine (ATX; a NET reuptake inhibitor used in the treatment of ADHD) in spite of the relatively low binding potentials (BPs; ∼0.3 at baseline). These results indicate that progress is being made toward the ultimate goal of designing NET radiotracers that will serve as tools for studying NET function.

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

The work reported here is based on PET studies of [11C](S,S)-MRB carried out in human subjects under conditions of test/retest and with three different oral doses of ATX in order to evaluate the sensitivity of the tracer to variations in free transporter availability as well as to address the crucial issue of reproducibility. Although the primary analysis was done by region-of-interest (ROI) analysis from individual subjects, averaged images of each group were formed from images normalized to an atlas via Statistical Parametric Mapping (SPM) software (http:// www.fil.ion.ucl.ac.uk/spm). The dynamic scans were averaged prior to the formation of parametric DV images. The purpose of doing this was to increase the signal-tonoise ratio and to allow visualization of NET distribution obtained from subtraction of the averaged blocked image from its corresponding averaged baseline image. The low signal-to-noise ratio at the voxel level of the individual images precluded the formation of parametric images for an SPM-type statistical analysis. 2. Materials and methods 2.1. Radiotracer synthesis Radiotracer synthesis and quality control are presented elsewhere [26,31]. 2.2. Subjects These studies followed the guidelines of the Institutional Review Board at Brookhaven National Laboratory, and written informed consent was obtained from each subject after the procedures had been explained. Eleven male subjects were recruited for the test/retest protocol (29.7±5.0 years), and 13 male subjects (30.8±7.0 years) were recruited for the ATX studies through newspaper advertisements and word of mouth. Exclusion criteria included current or past psychiatric or neurological disease, history of drug or alcohol abuse, history of head trauma with loss of consciousness, history of cardiovascular or endocrinological disease, current medical illness, dependence on any substance other than caffeine and positive urine screen for drugs of abuse. 2.3. PET studies in humans 2.3.1. PET camera and procedures The PET camera used was a high-resolution Siemens HR+; 63 planes; 4.5×4.5×4.5 mm, FWHM in 3D mode. An individually molded head holder was made for each subject to stabilize the head and to ensure accurate repositioning of subjects in the PET scanner for the repeated brain scans. A transmission scan was obtained with a 68Ge rotating rod source before the emission scan to correct for attenuation. Emission data were corrected for attenuation, scatter, randoms and dead time and reconstructed using filtered back projection. Catheters were placed in an antecubital

669

vein for radiotracer injection and in the radial artery for blood sampling. 2.3.2. Test/retest protocol Eleven male subjects underwent two PET scans 3 h apart on the same day. The average injected dose of [11C](S,S)MRB for Study 1 was 330±30 MBq; for Study 2, it was 311±65 MBq. The specific activity averaged 22.2±7.4 GBq/ μmol at time of injection. 2.3.3. Baseline/pretreatment protocol Each subject had a baseline scan and two studies with (different) pretreatment doses of ATX. On Day 1, each subject underwent two PET scans 3 h apart with [11C](S,S)MRB (average dose of the baseline study was 322± 18 MBq). Prior to the second scan (about 1.5 h after the first radiotracer injection), an oral dose of 25, 50 or 100 mg ATX was given. The timing of the dose was based on the pharmacokinetics of ATX. A second (pretreatment) PET study was done at least 1 week after the initial studies with a second ATX dose (different from the first pretreatment). The average injected doses of radiotracer were 316±30 MBq (25 mg ATX), 293.8±63 MBq (50 mg ATX) and 284.5±45.5 MBq (100 mg ATX). For all studies including test/retest and baseline/pretreatment, dynamic time–radioactivity–activity data were recorded for 90 min according to the following frame sequence: 10×1 min frames, 4×5 min frames, 8×7.5 min frames. Arterial blood samples were withdrawn every 5 s for the first 2 min (Ole Dich automatic blood sampler), then every minute from 2 to 6 min, then at 8, 10, 15, 20, 30, 45, 60 and 90 min. Each arterial blood sample was centrifuged, and the plasma was pipetted and counted. Plasma samples at 1, 5, 10, 20, 30, 45, 60 and 90 min were analyzed for [11C](S,S)-MRB as described previously [26]. Blood levels of ATX were measured at 1, 1.75 and 2.5 h after the oral dose was given. 2.3.4. Regions of interest For the purpose of region identification on the [11C](S, S)-MRB images, time frames from dynamic images taken from 0 to 90 min were summed and then manually resliced along the AC–PC line. Planes were summed in groups of two, placing the THL at Plane 12, to obtain 23 planes for ROI placement. An ROI template was made using an atlas for reference (Talairach and Tournoux [32]) including the following regions: CB, CDT, anterior cingulate gyrus (CNG), frontal cortex (FR), insula (INS), lateral temporal cortex (LTM), medial temporal cortex, midbrain (MBR), occipital cortex (OCC), parietal cortex (PAR), PUT, STM, THL, temporal cortex (TMP), visual cortex (VIS) and ventral striatum (VST). The template was projected onto the summed image for each subject, and the ROIs were manually adjusted to each individual. For bilateral structures, right and left were averaged. Each set of ROIs was then projected back onto the dynamic images to generate time–activity data. The template was also applied to the averaged parametric DV images.

670

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679 Table 1B Average DV (n=8) and standard deviation across subjects for test/retest studies

Fig. 1. Typical uptake curves for [11C](S,S)-MRB. Uptake is expressed as percent injected dose for THL, PUT, CB and CDT.

2.3.5. Parametric images The dynamic images were normalized using SPM software to the SPM PET template (http://www.fil.ion.ucl. ac.uk/spm/). Average dynamic images weighted by the radiotracer dose were formed for each group of the 25-, 50and 100-mg ATX studies, as well as the associated baseline for each group. Average blood input functions were also formed from data of the subjects in each group. Parametric images of the DV were constructed using the method of Feng et al. [33] for a one-compartment model. A small amount of averaging to increase the signal-to-noise ratio was allowed for local pixels using the method described in Logan et al. [34]. Average number of pixels grouped was on the order of 4. The average pretreatment image was subtracted from its corresponding baseline image. In order to insure that pixels in the difference image corresponded to brain tissue, only those pixels that exceeded 50% of the maximum in the integrated radioactivity image for both were used. Thresholds of 40%, 29% and 20% of the difference between the average baseline and average pretreated DV image were

Table 1A Average K1 and standard deviation across subjects for the test/retest studies ROI

K1 #1

S.D.

K1 #2

S.D.

CB CDT CNG FR INS LTM MBR OCC PAR PUT STM THL TMP VIS VST

0.120 0.099 0.098 0.104 0.093 0.092 0.089 0.111 0.103 0.116 0.107 0.118 0.105 0.127 0.102

0.030 0.029 0.027 0.027 0.025 0.022 0.028 0.032 0.029 0.028 0.025 0.028 0.029 0.033 0.023

0.124 0.103 0.103 0.110 0.101 0.096 0.094 0.118 0.108 0.120 0.107 0.123 0.111 0.131 0.105

0.032 0.025 0.025 0.024 0.026 0.020 0.025 0.030 0.025 0.024 0.025 0.025 0.025 0.031 0.019

#1 refers to the first scan of test/retest. K1 units are in ml/min/ml.

ROI

DV #1

S.D.

DV #2

S.D.

Absolute percent change

S.D.

CB CDT CNG FR INS LTM MBR OCC PAR PUT STM THL TMP VIS VST

5.18 4.69 4.90 4.85 5.38 5.25 7.19 5.04 4.90 5.23 5.11 7.24 5.37 5.28 5.46

1.32 1.00 1.05 0.94 1.27 1.10 2.09 0.98 0.98 1.00 1.00 1.40 1.19 1.11 0.88

5.21 4.79 5.20 5.06 5.66 5.35 7.29 5.12 5.08 5.49 5.30 7.41 5.53 5.38 5.46

1.36 1.11 1.32 1.17 1.45 1.33 1.83 1.18 1.20 1.30 1.21 1.82 1.46 1.30 1.18

6.8 9.6 7.6 6.1 8.3 6.2 8.1 7.4 4.7 6.1 6.9 9.5 5.4 7.1 10.0

6.6 9.9 5.9 7.4 7.9 5.3 5.6 8.1 3.6 4.8 5.5 5.6 6.4 8.8 12.3

#1 refers to the first scan of test/retest. Units for DV are in ml/cc. Absolute percent change averaged over subjects and the corresponding standard deviations are given in the last two columns.

applied, and the corresponding voxels were displayed on the baseline-integrated image. These thresholds were chosen because they seemed to separate the high-binding regions (40%) from intermediate-binding (29%) and lower-binding (20%) regions. One additional ROI [for the supplementary motor area (SMA)] was generated from the difference images by projecting the selected area onto the dynamic images of the individual subjects to generate TACs from the normalized images for use in the DV analysis. 2.3.6. Data analysis TAC data were analyzed for the DV using a onecompartment model for estimating the DV as DV=K1/k2. Assuming two types of binding, NET and non-NET (NN),  K1
1 þ BPNET þ BPNN , where the DV is given by DV ¼ k2V k2=k2′/(1+BPNET+BPNN). Binding to the NET is characterized by BPNET, and the non-NET binding (low affinity or Table 2A Average DV, standard deviation and CV across subjects in all baseline studies (n=13) ROI

DV (base)

S.D.

CV

CB CDT CNG FR INS LTM MBR OCC PAR PUT STM THL TMP VIS VST

5.27 4.85 5.19 5.01 5.64 5.72 7.19 5.25 5.07 5.57 5.50 7.18 5.54 5.60 5.49

0.86 0.98 0.88 0.82 1.15 1.08 1.46 0.99 0.80 0.94 1.00 1.56 1.12 1.02 0.83

0.16 0.20 0.17 0.16 0.20 0.19 0.20 0.19 0.16 0.17 0.18 0.22 0.20 0.18 0.15

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679 Table 2B Average DV across subjects for blocking studies (DV 25 refers to 25 mg ATX, etc.) ROI

DV 25 (n=9)

DV 50 (n=8)

DV 100 (n=8)

CB CDT CNG FR INS LTM MBR OCC PAR PUT STM THL TMP VIS VST

4.85 4.74 4.57 4.53 4.82 4.95 5.18 4.79 4.68 5.35 4.51 5.73 5.07 5.02 5.19

4.20 4.27 4.23 4.21 4.01 4.48 4.38 4.30 4.35 4.85 4.07 5.16 4.65 4.58 4.81

4.64 4.73 4.69 4.67 4.73 5.21 4.85 4.75 4.88 5.36 4.45 5.67 5.14 4.93 5.36

specific for another receptor/transporter or nonspecific ) is described by BPNN, which is a summation over all types of (non-NET) binding contributing to the signal. Both of these terms can vary from tissue to tissue. K1 and k2′ are the tissue/plasma transport constants. For a detailed discussion of the model and effective BP, see Logan et al. [17]. DVRs were constructed using the CDT and the average of CDT and PUT as reference regions. Differences reported in test/

671

retest of DV and DVR were based on the absolute percent difference between Studies 1X and 2. A mean sum of squares 1 DVRik  DVR i Þ2 , where i within subjects [SSWS ¼ N i;k is over N subjects, k is over test/retest and DVR i is the average test/retest DVR for subject i] was used to compare DVR using CDT to the average of CDT and PUT. Differences between baseline DVR and blocked DVR for the different doses were also calculated. These differences (ΔDVR) correspond to a difference in the effective in vivo BP: DDVR ¼ DVRbase  DVRblk  f NN
NN NET NN NET ¼ REF NN fT BPbase  fT BPblk fT NN ¼ fREF DBPNET ;

ð1Þ

where NN refers to non-NET, T refers to target ROI, REF refers to the reference region and f NN =1/(1+BPNN) (see Ref. [17]). If all transporters are blocked, then the DVR difference is proportional to the baseline free transporter concentration. This gives a different value than is given by NN DVR−1 (i.e., when fREF is not equal to fTNN). (It is NN NN assumed that fREF and fT are the same in the baseline and blocked condition.) Differences between baseline and blocking studies were analyzed with one-way repeated measures analysis of variance (ANOVA) for selected regions. When significant

Fig. 2. Average percent change (decrease) in DV from baseline for the three pretreatment ATX studies.

672

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

Fig. 3. DVR for selected regions using CDT as reference averaged over subjects in the test/retest protocol. DVRs illustrate the general uniformity of all regions outside of MBR and THL.

differences were found, a multiple comparison test (Holm– Sidak) was applied to isolate differences between groups (SigmaPlot 2000 version 6.0). 3. Results 3.1. Uptake Time–activity curves shown in Fig. 1 are similar to those reported by Schou et al. (Fig. 5B in Ref. [24]) in nonhuman primates. Average uptake in THL for all test/retest (n=11 subjects tested twice) combined was 0.0033±0.0005% injected dose at 60 min postinjection. 3.2. K1 and DV for blocking and test/retest studies The average of K1 across subjects for the first and second scans of the test/retest studies are given in Table 1A. The

values for K1 are very similar to those found in the baboon studies [17] with the same radiotracer. They are on the order of 0.10 ml/min/cc with VIS and CB being the highest at 0.12. Values are much lower than typical blood flow (∼0.50 ml/min/ml). Average DVs across subjects for test and retest studies are given in Table 1B for the same regions as K1. Lowest DV values are in the CDT. Table 1B also contains the average over subjects of the absolute percent difference between DVs for test/retest. The average difference is 10% or less but with considerable variability. The baseline DV values for the blocking studies are given in Table 2A for all subjects used in the blocking studies (n=13). (When comparing baseline and blocking studies, only those baseline studies paired with the blocking studies were used in the analysis.) From Tables 1B and 2A, the DV in the untreated state (i.e., test/retest and baseline of the blocking studies) is highest in MBR and THL (∼7.2, with S.D.∼1.5) and lowest in CDT and FR. With the exception of THL and MBR, the DVs are in a narrow range on the order of 5.00±0.5. The coefficient of variation (CV) across all regions is on the order of 20%. The average DVs for the blocking studies are given in Table 2B. The standard deviation and CV (not shown) are similar to those for the baseline studies, that is, CV is on the order of 15% to 20%. The DV for 50 mg appears to be slightly less, but this is a reflection of variations in subjects. The percent change in DV for selected regions for the three ATX doses is shown in Fig. 2A, B and C. The CDT, PUT and VST showed the smallest changes at all doses, making them possible candidates for a reference region. Repeated measures ANOVA followed by the Holm–Sidak multiple comparisons test found a difference between baseline and the 50-mg dose for these regions, which may reflect some blocking of the non-NET binding. The other regions showed significant differences between baseline and two or three of the blocking doses, and FR, MBR and LTM showed a significant

Fig. 4. (A) Comparison of average absolute percent change in test/retest DVR using CDT and CDT+PUT for selected ROIs. (B) Comparison of average percent change (decrease) over subjects in baseline and blocked conditions (100 mg ATX).

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679 Table 3 Comparison of SSWS (×104) for DVR using CDT and CDT+PUT

CB CING FR INS LTMP MBR OCC STEM THL TMP VIS VSTR Mean S.D.

CDT

CDT+PUT

2.72 4.07 5.51 3.19 5.58 17.78 4.51 4.72 10.24 8.10 3.84 10.02 6.69 4.29

6.47 2.36 4.13 1.01 1.04 8.30 1.52 1.97 7.65 1.05 2.06 8.50 3.84 3.03

difference between 25 mg and one of the higher doses, although this is most likely due to the non-NET binding. 3.3. ATX in blood The blood levels of ATX (in ng/ml) averaged over subjects were 116±35 (25 mg ATX), 206±38 (50 mg ATX) and 421±114 (100 mg ATX). The average within-subject increases between doses were 109±17 (for 25 to 50 mg ATX), 140±46 (50 to 100 mg) and 280±120 (25 to 100 mg). At the 100-mg dose, six subjects experienced nausea and one had to be withdrawn from the study. 3.4. DVR test/retest Since the CDT (known to have few NETs or none) had the lowest DV as well as one of the lowest changes in the blocking studies, it would be a first choice for a reference region. Using the CDT as a reference, the DVRs [ratio of DV (ROI)/DV (CDT)] averaged over subjects for both test and

673

Table 4 Average of the difference between DVR (baseline) and DVR (blocked) for the three doses of ATX across subjects ROI

ΔDVR 25 (n=9)

ΔDVR 50 (n=8)

ΔDVR 100 (n=8)

Average CV

MBR THL STM INS CNG CB

0.38 0.28 0.19 0.16 0.09 0.08

0.40 0.22 0.10 0.09 0.05 0.05

0.45 0.27 0.21 0.15 0.08 0.10

0.40 0.62 0.85 0.77 1.1 0.90

The last column is the average of the CV for the three doses.

retest are shown in Fig. 3, illustrating the small regional differences for all regions outside of MBR and THL. The regions were arranged generally from lowest to highest. The effective BPs (assuming BP=DVR−1) are very small for most regions (∼0.10) and only 0.50 at the highest (THL and MBR). The average absolute percent difference for test/retest DVR using CDT and the average of CDT+PUT is shown in Fig. 4A. The mean and variance are generally smaller for CDT+PUT than for CDT alone. The values for SSWS for DVRs calculated for CDT and CDT+PUT are given in Table 3 for selected regions. The average CDT and CDT +PUT over these ROIs are 6.7 and 3.8 for CDT and CDT +PUT, respectively, indicating that using the average of CDT and PUT increases reproducibility somewhat. 3.5. DVR blocking studies: Although, in theory, the BP (as given by DVR−1) would be a more direct measure of variations in free transporter than the DVR, the use of the BP with CDT as reference region leads to small and negative BPs in some of the blocking cases, indicating that although the CDT has one of the lowest DVs, it most likely overestimates the non-NET binding in some regions. For this reason, results are based on the DVR.

Fig. 5. Comparison of percent change (decrease) in DVR using CDT for the three doses of ATX. (A) Averaged over ROIs for each subject. The standard deviations (not shown) are large. For MBR, the average CVover all doses is 0.37; for THL, it is 0.6, and for INS and CB, it is 0.7. For other regions, the CVs are higher. (B) ROIs from averaged images for each ATX dose.

674

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

Fig. 6. Comparison of parametric DV images of baseline and blocked (25 mg) ATX.

CDT as reference for THL gave DVRs of 1.51±0.1 and 1.21±0.09 for 100 mg ATX, 1.42±0.13 and 1.20±0.17 for 50 mg ATX and 1.50±0.17 and 1.22±0.09 for 25 mg ATX (baseline and blocked, respectively). Reduction in DVR for THL was in the range of 17–19%. A comparison of the percent change in DVR between baseline and 100 mg ATX (using CDT as reference) averaged over subjects is shown in Fig. 4B for selected ROIs. VST on the left exhibits essentially no change compared to the other regions, although the DVR is similar to or greater than that of FR, CB, CNG, INS and STM (see Fig. 3). There is very little difference in results calculated with either choice (CDT or CDT+PUT) of reference region. Fig. 5A compares the percent change in the three ATX doses by ROI. MBR has the largest percent change, and only for this region does the average correlate with the dose of ATX given (24±7% at 25 mg, 27±11% at 50 mg and 31±11% at 100 mg), although the error bars are large and no statistical significance was found between the doses for any region including MBR. THL and STM are in second place (∼12% to 15%). No changes are apparent in VST. The percent change for INS was ∼10%,

with other regions averaging 10% or less. There does not appear to be a correlation with ATX dose. This was verified with the repeated measures ANOVA followed by the Holm– Sidak multiple comparisons test, which found significant differences between baseline and all of the blocked studies for MBR, THL, STM, INS, CB and CNG but no significant differences between DVRs for the three doses of ATX for these regions. The differences seen in DV for FR, MBR and LTM between 25 mg and one of the higher doses are not evident in the DVR and were probably due to the non-NET binding. Differences between baseline and at least one of the blocked studies were found for all regions except for PUT and VST (using CDT as reference). All subjects showed increased blood levels of ATX with increased dose. However, no correlation was found between changes in blood levels of ATX and the corresponding change in the DVR. The lack of sensitivity of the tracer at the doses given is suggested by the small changes seen within each subject with increasing dose of ATX. The average percent change in DVR for THL between the 25- and 50-mg ATX doses, between the 25- and 100-mg ATX doses and between the 50- and 100-mg ATX doses was −3.5±9%, −0.23±3% and −0.15±16%, respectively. The results were somewhat more variable for the MBR: 3.6±19% (25 to 50 mg ATX), 11.4±6% (25 to 100 mg ATX) and −6.2±30% (50 to 100 mg ATX). Table 4 gives the average of the difference between baseline DVR and the blocked DVR. From Eq. (1), this is proportional to the difference in BP. If the transporter were completely blocked, then this would be a measure of the “true” baseline BP. Note that the predicted BP of MBR is greater than that of THL, whereas if the BP were taken as DVRbase−1, they would be the same. 3.6. Images Sagittal and coronal views of the averaged baseline and 25-mg DV images are shown in Fig. 6. The highest binding is visible in the MBR region and is substantially blocked in the 25-mg ATX image. The CNG also appears blocked in the 25-mg images. A more quantitative view of the differences

Fig. 7. Sagittal views showing differences between baseline and blocked average DV images for different cutoff values. The voxels exceeding a set threshold are displayed in green. In Panel A, the voxels displayed in green exceed 40% of the maximum difference between baseline DV and blocked DV. In Panel B, the threshold was lowered to 29%, and in Panel C, the threshold was 20%. Panel D displays the image without the overlay.

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

between baseline and blocked images is shown in Figs. 7 and 8 for sagittal and coronal views of the average images in which changes in DV exceeding a specified threshold from baseline are shown in green superimposed on the average integrated time radioactivity image. In this case, the differences were based on the average of images from the 100-mg ATX study. In Fig. 7A, at the highest cutoff (N40% difference between baseline and blocked DV), the green area at the back of the pons/STM (designated a) would include the LC (located on the dorsal wall of the upper pons) as well as the raphe nucleus, which has a high density of NETs and borders the LC [35]. Partial volume averaging causes these regions to contribute radioactivity to neighboring tissue. The hypothalamus, an area also rich in NETs [18], located in front of the pons/STM, is the most likely source of the area labeled b. In Fig. 7B, using a lower threshold (N29%), blocking also occurs in a region (c) that appears to be the SMA. Lowering the threshold further, differences are seen in CB (d) and THL (e). In the coronal views of Fig. 8, the MBR (containing LC) contribution appears at the highest threshold (Fig. 8A), with CNG (b), THL (d) and INS (c) appearing somewhat at the intermediate threshold (Fig. 8B) but showing up more fully at the lowest (Fig. 8C). These regional changes were assessed in the ROI analysis with the exception of the SMA. Applying ROIs from the SMA region to the normalized images, TACs were generated for each subject in the baseline and blocked (100 mg) studies. The averaged DVRs were 1.07±0.08 (baseline) and 0.91±0.16 (blocked), significant at Pb.04 (paired t test) for baseline and blocked, respectively. Average ROI values from the three blocked images are shown in Fig. 5B. The images show greater changes from

675

baseline for a number of regions compared to the averaged ROI values in Fig. 5A.

4. Discussion Many difficulties have been encountered in developing radiotracers to image the NET. Figs. 3 and 5 illustrate some of these problems, which are related to background non-NET binding and sensitivity of the tracer to variations in NET density. Traditionally, the (effective) BP is used as a measure of specific binding which is related to the free transporter/ receptor concentration, Bmax′/Kd′ (here, Bmax′ and Kd′ represent the free transporter concentration and in vivo dissociation constant, respectively, which are affected by endogenous NE and nonspecific binding). Assuming that the nontransporter binding is the same in both reference and target regions, the BP can be derived from the DVR as BP=DVR−1 (see discussion in Logan et. al. [17]). However, since the background non-NET binding is most likely not constant, BPs derived in this manner will not reflect the true BP. An example of this is seen in the data for MBR and THL. Both have similar values for the DVR, which would predict similar BPs. However, MBR is significantly more blocked with all doses of ATX, indicating a greater concentration of transporters. An estimate of the relative densities of the two regions can be obtained from the change in DVR between baseline and blocking studies (Table 4). The ratio of these changes in DVR of THL to MBR is 0.60. If the assumption is made that both regions are totally blocked, then this represents the ratio of free transporter concentration for the two regions and the MBR is (at least) 40% higher than THL.

Fig. 8. Coronal views showing differences between baseline and blocked average DV images for different cutoff values. The voxels exceeding a set threshold are displayed in green. In Panel A, the voxels displayed in green exceed 40% of the maximum difference between baseline DV and blocked DV. In Panel B, the threshold was lowered to 29%, and in Panel C, the threshold was 20%. The bottom row displays integrated uptake images without the overlay.

676

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

If blocking is not complete, then the true difference between MBR and THL could be even greater. These results are similar to those reported by Seneca et al. [30], in which an intravenous infusion of 0.12 mg/kg/h of ATX blocked specific binding by ∼80% in MBR but only ∼50% in THL, although both regions had similar baseline values. The different blocking capacity between MBR and THL, given similar baseline values, is probably related to a difference in the low-affinity (non-NET) binding component. Biegon and Rainbow [18] and Lee et al. [19] report the presence of a low-affinity binding site in addition to a highaffinity binding site for the NET ligand, [3H]DMI. The highaffinity site is associated with the NET, but the low-affinity site is found in regions with low NET binding as well as in regions with the high-affinity site. Gross-Isseroff et al. [20] report high binding in the striatum in human postmortem autoradiography studies with [3H]DMI. This nonuniform, high-background binding has plagued many potential NET tracers. In fact, all of the NET ligands investigated so far seem to have some amount of this low-affinity binding, which can potentially obscure differences in the high NET binding regions. This is especially true for NIS and DMI. PET studies with [11C]nisoxetine in baboons revealed very little differences in regional binding. In fact the basal ganglia, a region with little or no concentration of NETs, showed higher uptake than THL, a region with a substantial NET concentration. The low-affinity binding appears to predominate with this tracer [17]. Apparently, the ligand [11C](S,S)-MRB has a more favorable ratio of high- to low-affinity binding than DMI or NIS, although the low-affinity component appears also to be nonuniform. This nonuniformity presents a reasonable explanation for the difference between MBR and THL. That is, they have different relative amounts of the high and low components, with MBR greater than THL in the density of high-affinity sites. The high-affinity site would be occupied by relatively low doses of ATX, whereas the lowaffinity sites would require much greater doses to block. There is some evidence that the low affinity is beginning to block at the 50-mg dose for which the DVs CDT, PUT and VST are somewhat lower than baseline (see Table 2B). The problem of the nonuniform background binding is particularly relevant to the choice of reference region. The DVRs given here were based on the CDT. We have shown that using the average of CDT+PUT increases reproducibility somewhat, although the DVRs are smaller since the DV for PUT is higher than that for CDT. In order to extract a meaningful BP, it is necessary to have a reference region without specific binding that has similar characteristics in nonspecific binding. Both PUT and VST, regions without any significant NET concentration, have DVs somewhat higher than CDT, a region also devoid of NET density (see Table 1B). The DV for CDT is among the lowest at 4.9. In the blocking studies reported here, the DVs for VST and PUT remain higher than CDT at all doses (Table 2B). CDT was also used by Seneca et al. [30] as a reference region in PET studies with (S,S)-[18F]FMeNER-D2 (FMeNER) in

nonhuman primates. Andree et al. [27] reported using the CB and the CDT as reference regions for [11C](S,S)-MRB, although, in the current study, we find that the (high-affinity) NET binding in the CB, while low, is not insignificant. It appears that although the CDT and/or the PUT are good reference regions from the point of view of low NET binding, they do not represent a good measure of non-NET binding for all regions. Due to the non-NET background, baseline images using current NET tracers do not show a true binding profile of the NET. Blocking studies are needed to reveal the greater changes in high NET density regions. The sensitivity issue is related to the ability to detect changes in the effective BP, which, in turn, is related to changes in the free transporter concentration (Bmax′). For [11C](S,S)-MRB, the effective BP in the highest regions is probably on the order of 0.40 to 0.50 (maybe higher in MBR); as a comparison, the DVR for [11C]cocaine in basal ganglia is ∼2.0 with BP∼1.0. Furthermore, the situation is worse for regions known to have low or moderate concentrations of NETs with DVRs less than 1.2. These values are similar to results previously reported in baboons with the same tracer [17]. The small BPs are not, in themselves, a problem if the reproducibility is considerably smaller than the changes expected due to drug treatment and so forth. However, due to the low signal-to-noise ratio, particularly at the later time points for the 11C tracer, the absolute percent change in DVR for test/retest (Fig. 4A), while less than 10%, has large variances. This is apparently sufficient to distinguish the baseline DVR from that with a blocking dose but not to distinguish the smaller changes in DVR between doses. For MBR, the decrease in DVR from baseline for the 25- and 100-mg ATX doses was significant; however, the difference between them (24% and 31%, respectively) was not. Furthermore, for THL, there is some indication that the 25-mg dose was close to the maximum achievable dose since there was no within-subject difference in DVR between the blocking doses. This appears to hold true for other regions with lower transporter densities (see Table 4). Seneca et al. [30] have reported dose-dependent effects in nonhuman primates with [18F]FMeNER-D2 and intravenously infused ATX. Because they measured plasma levels of N-desmethylatomoxetine, a metabolite of ATX, instead of ATX itself and used an intravenous infusion protocol for delivery of ATX, their results cannot be compared directly with the data reported here. The BP of [18F]FMeNER-D2 was at the highest in LC (0.3), slightly less than the highest value found with the 11C analog ([11C](S,S)-MRB) in baboons and the value reported here. Both the 18F and 11C compounds have been shown to have the same affinity for the NET in postmortem autoradiography in human brain [36,24]. However, the use of 18F resulted in a reduction in noise compared to the 11C tracer. Comparing uptake curves from the 11C analog (Fig. 5B in Ref. [24]) to that of the 18F (Fig. 2A in Ref. [30]) in the same nonhuman primate species shows the 18F compound to be more reversible with a peak and washout, desirable characteristics of a PET ligand.

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

Given the similarities of the BPs and affinities, the difference between these compounds may be in the effect of the structural difference on the non-NET binding. The [18F] fluorethylated analog (S,S)-[18F]FRB was also found to have relatively fast kinetics in NET-rich regions. One problem with the F-18 analogs is the potential contamination of cortical areas by uptake of 18F in bone. A dose effect on the DVR of [11C](S,S)-MRB was not seen in this work, although the plasma levels of ATX indicated that distinctly different doses should have been delivered to the brain. The lack of a dose effect in the higher-density regions such as MBR and THL raises the possibility that the lowest dose of ATX was sufficiently high to block a significant fraction of transporters. The fraction of free receptors changes more slowly with dose at higher doses. This, coupled with the noise factor, may have made it unlikely that any further changes could be detected. Whether or not a dose effect could be observed would require ATX blocking at lower doses. The apparent high occupancy observed in the low-dose (25 mg) study was unanticipated as this dose was expected to be therapeutically ineffective. The 25-mg dose corresponds to 0.36 mg/kg for a 70-kg adult, and a single 25-mg dose of ATX elicits no noradrenergic-related side effects in adults as would be expected at a high occupancy. For comparison, a study on the efficacy of three different doses of ATX in the treatment of ADHD in children found no difference between 1.8 and 1.2 mg/kg/day. Both were superior to a 0.50-mg/kg/day dose, although the lower dose did have an effect beyond placebo [37]. ATX was given in two equal portions (morning and late afternoon). At 0.25, 0.60 and 0.9 mg/kg per dose, this amount of ATX is in the range of the doses used here; therefore, it is not unexpected that some significant blocking of NETs could have occurred in our study. Given that there was no difference between the 1.8- and 1.2-mg/kg dose (or 0.9 to 0.6 mg/kg per dose) in the ADHD study suggests a plateau effect. Another result that calls into question the mechanism of action of ATX is that the half-life of ATX in plasma is, for most individuals, 4 h. Michelson et al. [38] report in another ADHD study that a single 1.0-mg/kg dose given early in the day remained effective in the evening. Given that the half-life of ATX in plasma is 4 h, the occupancy of NET would be expected to vary considerably over the time during which the drug exerted pharmacological effects. On the basis of this, the authors suggest that ATX may produce neuroregulatory changes that persist beyond the time of transporter occupancy [38]. Gastrointestinal side effects were reported by 10% to 15% of subjects in the 1.0-mg/kg study of Michelson et al. This is consistent with results reported here for the 100-mg study (corresponding to 1.4 mg/kg for a 70-kg adult) in which side effects occurred in about 50% of the subjects. Based on these results, it is not unreasonable to assume that we observed a plateau effect due to noise as well as significant transporter occupancy. In the human data reported here, there was considerable noise evident in the uptake curves particularly at the later time points (similar to those reported by Schou et al. [24] in

677

the cynomolgus monkey). Because of the noise properties of the TAC, we used a one-compartment model to fit the data because it gave greater reproducibility on test/retest studies than the graphical method [39], which is not restricted to a one-compartment model. As a result of the one-compartment model, the DV may be underestimated in the regions of higher binding such as MBR and THL. Recently, Schou et al. [36] reported autoradiography studies in postmortem human brain using FMeNER. Binding in cortex, THL and CB was found to be about 10% of that in LC, with CDT, PUT and OCC less than 6% of LC. These are ratios based on specific binding, which are much easier to obtain in autoradiography experiments than in PET experiments. It is also very likely that the MBR/LC binding is underestimated with PET when compared to autoradiography due to partial volume averaging. The PET values reported here for THL are relatively high compared to the autoradiography results reported by Schou et al., but this may be a sampling issue since according to other researchers, this region is heterogeneous and has a mixture of high and low binding [20]. Also, in the nonhuman primate brain, the concentration in the THL between various regions was found to vary by a factor of 3, highest in the paraventricular nucleus and lowest in the paracentral/centrolateral nucleus [40]. The averaged normalized images provide a picture of NET binding with a significantly improved signal-to-noise ratio over individual images. The quality of the individual images was too poor to use for SPM analysis; however, the averaged images and differences between baseline and blocked groups illustrate the NET binding pattern. While most of the NET binding pattern was expected, the SMA appeared to be somewhat more blocked than other cortical areas, indicating a greater concentration of NETs. The change in DVR from baseline for the 100-mg ATX dose was ∼15% for the SMA. The importance of NE in this area is supported by MPTP studies in monkeys that found decreased NE concentrations as well as DA in the SMA, indicating degeneration of noradrenergic neurons [41]. The image-based analysis (Fig. 5B) shows a small dosedependent effect between the three doses for THL as well as MBR. Also, some of the lower regions (such as FR) had average values that correlated with dose of ATX. The imageaveraging technique increases the signal-to-noise ratio but does not provide a means of testing significance at this time. Work is continuing on the image-based analysis methods applied to this tracer.

5. Conclusions While all three doses of ATX were found to block most regions, no significant difference in the DVR between doses was found. From the data and the images in Fig. 6, it would appear that the lowest dose blocked a significant fraction of transporters such that the remaining differences between doses could not be detected due to noise.

678

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

CDT and a combination of CDT+PUT were tested as reference regions since they showed the least change in DV in the blocking studies and were identified by a number of other authors as regions with little NET concentration. The higher doses may have blocked a fraction of the non-NET binding, although not completely since the PUT and VST remained higher than the CDT. The combination CDT+PUT increased reproducibility in test/retest but decreased the DVRs since PUT had a relatively high DV. Although CDT had the lowest DV, it most likely overestimates the nonspecific binding in MBR and possibly other regions. This nonuniformity of background (non-NET) binding has been found with many NET ligands and obscures the specific NET binding. Based on blocking, the highest density of NETs is in the region designated MBR (which contains the LC). The next highest region was THL. Small displacements were seen in many regions such as FR, CNG and CB. A higher density was observed in INS, but this is most likely due to the fact that the INS is not as affected by partial volume averaging due to folds in the cortex. A potentially more accurate picture of NET density was found by averaging the normalized images to form parametric DV images. Thresholded differences between blocked and baseline images showed highest density in MBR/LC and hypothalamus, with THL and INS next. Interestingly, the SMA also appeared to be higher than cortical areas in general. This image-averaging technique may prove to be useful for visualizing transporter densities. Studies at lower doses of ATX might reveal a dose effect without the potential for blocking the non-NET binding, which may have occurred at the 50- and 100-mg ATX doses. The data from this study suggest that [11C](S,S)-MRB will be useful in determining whether a drug binds to the NET and perhaps, the degree of blockade. Acknowledgments This work was carried out at Brookhaven National Laboratory under contract DE-AC02-98CH10886 with the U.S. Department of Energy and supported by its Office of Biological and Environment Research. Support also came from GlaxoSmithKline, the National Institutes of Health (NIH) intramural program (NIAAA) and the NIH [National Institute for Biomedical Imaging and Bioengineering (EB002630), National Institute on Drug Abuse (DA019062) and the Office of National Drug Control Policy]. The authors thank Michael Scheuller, Donald Warner and Richard Ferrieri for cyclotron, PET and radiotracer laboratory operations and Karen Apelskog for protocol coordination. We are also especially grateful to the individuals who volunteered for these studies. References [1] Zahniser N, Doolen S. Chronic and acute regulation of Na+/Cl−dependent neurotransmitter transporters: drugs, substrates, presynaptic receptors, and signaling systems. Pharmacol Ther 2001;92:21–55.

[2] Klimek V, Stockmeier C, Overholser J, Meltzer HY, Kalka S, Dilley G, et al. Reduced levels of norepinephrine transporters in the locus coeruleus in major depression. J Neurosci 1997;17: 8451–8. [3] Bymaster F, Katner J, Nelson D, Hemrick-Luecke S, Threlkeld P, Heiligenstein J, et al. Atomoxetine increases extracellular levels of norepinephrine and dopamine in prefrontal cortex of rat: a potential mechanism for efficacy in attention deficit/hyperactivity disorder. Neuropsychopharmacology 2002;27:699–711. [4] Remy P, Doder M, Lees A, Turjanski N, Brooks D. Depression in Parkinson's disease: loss of dopamine and noradrenaline innervation in the limbic system. Brain 2005;128:1314–22. [5] Srinivasan J, Schmidt W. Potentiation of parkinsonian symptoms by depletion of locus coeruleus noradrenaline in 6-hydroxydopamineinduced partial degeneration of substantia nigra in rats. Eur J Neurosci 2003;17:2586–92. [6] Scatton B, Javoy-Agid F, Rouquier L, Dubois B, Agid Y. Reduction of cortical dopamine, noradrenaline, serotonin and their metabolites in Parkinson's disease. Brain Res 1983;275:321–8. [7] Riederer P, Birkmayer W, Seemann D, Wuketich S. Brain-noradrenaline and 3-methoxy-4-hydroxyphenylglycol in Parkinson's syndrome. J Neural Transm 1977;41:241–51. [8] Herrmann N, Lanctot KL, Khan LR. The role of norepinephrine in the behavioral and psychological symptoms of dementia. J Neuropsychiatry Clin Neurosci 2004;16:261–76. [9] Aston-Jones G, Cohen J. An integrative theory of locus coeruleus– norepinephrine function: adaptive gain and optimal performance. Annu Rev Neurosci 2005;28:403–50. [10] Aston-Jones G, Rajkowski J, Kubiak P, Alexinsky T. Locus coeruleus neurons in monkey are selectively activated by attended cues in a vigilance task. J Neurosci 1994;14:4467–80. [11] Tejani-Butt S, Brunswick D, Frazer A. [3H]Nisoxetine: a new radioligand for norepinephrine uptake sites in brain. Eur J Pharmacol 1990;191:239–43. [12] Charnay Y, Leger L, Vallet P, Hof P, Jouvet M, Bouras C. [3H] Nisoxetine binding sites in the cat brain: an autoradiographic study. Neuroscience 1995;69:259–70. [13] Haka M, Kilbourn M. Synthesis and regional mouse brain distribution of [11C]nisoxetine, a norepinephrine uptake inhibitor. Int J Rad Appl Instrum B 1989;16:771–4. [14] Kung M, Choi S, Hou C, Zhuang Z, Foulon C, Kung H. Selective binding of 2-[125I]iodo-nisoxetine to norepinephrine transporters in the brain. Nucl Med Biol 2004;31:533–41. [15] Kiyono Y, Kanegawa N, Kawashima H, Kitamura Y, Iida Y, Saji H. Evaluation of radioiodinated (R)-N-methyl-3-(2-iodophenoxy)-3-phenylpropanamine as a ligand for brain norepinephrine transporter imaging. Nucl Med Biol 2004;31:147–53. [16] Ding Y, Lin K, Logan J, Benveniste H, Carter P. Comparative evaluation of positron emission tomography radiotracers for imaging the norepinephrine transporter: (S,S) and (R,R) enantiomers of reboxetine analogs ([11C]methylreboxetine, 3-Cl-[11C]methylreboxetine and [18F]fluororeboxetine), (R)-[11C]nisoxetine, [11C]oxaprotiline and [11C]lortalamine. J Neurochem 2005;94:337–51. [17] Logan J, Ding Y, Lin K, Pareto D, Fowler J, Biegon A. Modeling and analysis of PET studies with norepinephrine transporter ligands: the search for a reference region. Nucl Med Biol 2005;32:531–42. [18] Biegon A, Rainbow T. Localization and characterization of [3H] desmethylimipramine binding sites in rat brain by quantitative autoradiography. J Neurosci 1983;3:1069–76. [19] Lee C, Javitch J, Snyder S. Characterization of [3H]desipramine binding associated with neuronal norepinephrine uptake sites in rat brain membranes. J Neurosci 1982;2:1515–25. [20] Gross-Isseroff R, Israeli M, Biegon A. Autoradiographic analysis of [3H]desmethylimipramine binding in the human brain postmortem. Brain Res 1988;456:120–6. [21] Schou M, Sovago J, Pike V, Gulyas B, Bogeso K, Farde L, et al. Synthesis and positron emission tomography evaluation of three

J. Logan et al. / Nuclear Medicine and Biology 34 (2007) 667–679

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

norepinephrine transporter radioligands: [C-11]desipramine, [C-11] talopram and [C-11]talsupram. Mol Imaging Biol 2006;8:1–8. Van Dort M, Kim J, Tluczek L, Wieland D. Synthesis of 11C-labeled desipramine and its metabolite 2-hydroxydesipramine: potential radiotracers for PET studies of the norepinephrine transporter. Nucl Med Biol 1997;24:707–11. Wong E, Sonders M, Amara S, Tinholt P, Piercey M, Hoffmann W, et al. Reboxetine: a pharmacologically potent, selective, and specific norepinephrine reuptake inhibitor. Biol Psychiatry 2000;47:818–29. Schou M, Halldin C, Sovago J, Pike V, Gulyas B, Mozley P, et al. Specific in vivo binding to the norepinephrine transporter demonstrated with the PET radioligand, (S,S)-[11C]MeNER. Nucl Med Biol 2003;30:707–14. Wilson A, Johnson D, Mozley D, Hussey D, Ginovart N, Nobrega J, et al. Synthesis and in vivo evaluation of novel radiotracers for the in vivo imaging of the norepinephrine transporter. Nucl Med Biol 2003;30:85–92. Ding Y, Lin K, Garza V, Carter P, Alexoff D, Logan J, et al. Evaluation of a new norepinephrine transporter PET ligand in baboons, both in brain and peripheral organs. Synapse 2003;50:345–52. Andree B, Seneca N, Schou M, Mozley P, Potter W, Farde L, et al. [11C]MeNER human PET study: high occupancy of the central norepinephrine transporter induced by reboxetine at clinical doses. J Nucl Med (Suppl) 2004;45:68P. Schou M, Halldin C, Sovago J, Pike V, Hall H, Gulyas B, et al. PET evaluation of novel radiofluorinated reboxetine analogs as norepinephrine transporter probes in the monkey brain. Synapse 2004;53:57–67. Lin K, Ding Y, Kim S, Kil K. Synthesis, enantiomeric resolution, F-18 labeling and biodistribution of reboxetine analogs: promising radioligands for imaging the norepinephrine transporter with positron emission tomography. Nucl Med Biol 2005;32:415–22. Seneca N, Gulyas B, Varrone A, Schou M, Airaksinen A, Tauscher J, et al. Atomoxetine occupies the norepinephrine transporter in a dose-dependent fashion: a PET study in nonhuman primate brain using (S,S)-[18F]FMeNER-D2. Psychopharmacology (Berl) 2006; 188:119–27. Lin K, Ding Y. Synthesis, enantiomeric resolution, and selective C-11 methylation of a highly selective radioligand for imaging the

[32] [33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

679

norepinephrine transporter with positron emission tomography. Chirality 2004;16:475–81. Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain. Stuttgart: Georg Thieme Verlag; 1988. Feng D, Huang S-C, Wang Z, Ho D. An unbiased parametric imaging algorithm for nonuniformly sampled biomedical system parameter estimation. IEEE Trans Med Imaging 1996;15:512–8. Logan J, Fowler J, Ding Y, Franceschi D, Wang G, Volkow N, et al. Strategy for the formation of parametric images under conditions of low injected radioactivity applied to PET studies with the irreversible monoamine oxidase A tracers [11C]clorgyline and deuterium-substituted [11C]clorgyline. J Cereb Blood Flow Metab 2002;22: 1367–76. Ordway GA, Stockmeier CA, Cason GW, Klimek V. Pharmacology and distribution of norepinephrine transporters in the human locus coeruleus and raphe nuclei. J Neurosci 1997;17:1710–9. Schou M, Halldin C, Pike V, Mozley P, Dobson D, Innis R, et al. Postmortem human brain autoradiography of the norepinephrine transporter using (S,S)-[18F]FMeNER-D2. Eur Neuropsychopharmacol 2005; 15:517–20. Michelson D, Faries D, Wernicke J, Kelsey D, Kendrick K, Sallee FR, et al. Atomoxetine in the treatment of children and adolescents with attention-deficit/hyperactivity disorder: a randomized, placebo-controlled, dose–response study. Pediatrics 2001;108:e83. Michelson D, Allen AJ, Busner J, Casat C, Dunn D, Kratochvil C, et al. Once-daily atomoxetine treatment for children and adolescents with attention deficit hyperactivity disorder: a randomized, placebocontrolled study. Am J Psychiatry 2002;159:1896–901. Logan J, Fowler J, Volkow N, Wolf A, Dewey S, Schlyer D, et al. Graphical analysis of reversible radioligand binding from time– activity measurements applied to [N-11C-methyl]-(−)-cocaine PET studies in human subjects. J Cereb Blood Flow Metab 1990;10:740–7. Smith H, Beveridge T, Porrino L. Distribution of norepinephrine transporters in the non-human primate brain. Neuroscience 2006;138: 703–14. Elsworth J, Deutch A, Redmond D, Sladek J, Roth R. MPTP reduces dopamine and norepinephrine concentrations in the supplementary motor area and cingulate cortex of the primate. Neurosci Lett 1990;114:316–22.