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

Authors


Address correspondence and reprint requests to Yu-Shin Ding, PhD, Chemistry and Medical Departments, Brookhaven National Laboratory, Upton, New York 11973, USA. E-mail: ding@bnl.gov

Abstract

We have synthesized and evaluated several new ligands for imaging the norepinephrine transporter (NET) system in baboons with positron emission tomography (PET). Ligands possessing high brain penetration, high affinity and selectivity, appropriate lipophilicity (log P = 1.0–3.5), high plasma free fraction and reasonable stability in plasma were selected for further studies. Based on our characterization studies in baboons, including 11C-labeled (R)-nisoxetine (Nis), oxaprotiline (Oxap), lortalamine (Lort) and new analogs of methylreboxetine (MRB), in conjunction with our earlier evaluation of 11C and 18F derivatives of reboxetine, MRB and their individual (R,R) and (S,S) enantiomers, we have identified the superiority of (S,S)-[11C]MRB and the suitability of MRB analogs [(S,S)-[11C]MRB > (S,S)-[11C]3-Cl-MRB > (S,S)-[18F]fluororeboxetine] as potential NET ligands for PET. In contrast, Nis, Oxap and Lort displayed high uptake in striatum (higher than in thalamus). The use of these ligands is further limited by high non-specific binding and relatively low specific signal, as is characteristic of many earlier NET ligands. Thus, to our knowledge (S,S)-[11C]MRB remains by far the most promising NET ligand for PET studies.

Abbreviations used
ADHD

attention deficit hyperactivity disorder

CBF

cerebral blood flow

DASB

N,N-dimethyl-2-(2-amino-cyanophenylthio) benzylamine

DAT

dopamine transporter

DMF

dimethylformate

DV

distribution volume

DVR

distribution volume ratio

EOB

end of bombardment

FRB

fluororeboxetine

Lort

lortalamine

MRB

methylreboxetine

NET

norepinephrine transporter

Nis

nisoxetine

Oxap

oxaprotiline

PET

positron emission tomography

PPB

plasma protein binding

ROI

region of interest

SERT

serotonin transporter

S/N

signal to noise

TMS

tetramethylsilane

The involvement of the norepinephrine transporter (NET) in the pathophysiology and treatment of attention deficit hyperactivity disorder (ADHD), substance abuse, neurodegenerative disorders (e.g. Alzheimer's disease and Parkinson's disease) and depression has long been recognized. However, many of these important findings have resulted from studies in vitro using post-mortem tissues (Ohama and Ikuta 1976; Carlsson 1979; Tejani-Butt et al. 1993; Jellinger 1997; Klimek et al. 1997; Eshleman et al. 1999; Macey et al. 2003). Up to now, these results have never been verified by in vivo methods because brain imaging of the NET in living systems has been hampered by the lack of suitable radioligands. The fact that all three monoamine transporters [dopamine transporter (DAT), NET and serotonin transporter (SERT)] are involved in various neurological and psychiatric diseases (Zahniser and Doolen 2001) further emphasizes the need to develop suitable NET ligands so that researchers will be able to determine the contributions of each monoamine transporter system to specific CNS disorders.

Norepinephrine transport in sympathetic nerves of peripheral tissues has been labeled with (+/–)-[11C]norepinephrine (Fowler et al. 1974; Farde et al. 1994), [18F]fluoronorepinephrine (Ding et al. 1991), [131I]MIBG(metaiodobenzylguanidine) (Wieland et al. 1981) and [11C]hydroxyephedrine (Raffel and Wieland 1999); however, this class of tracers is unsuited to the study of brain NET owing to their inability to cross the blood–brain barrier. Several potent NET reuptake inhibitors have been labeled for in vitro or in vivo mapping of brain NET, but the results were not promising because of the high non-specific binding of these agents. To name a few, [3H]desipramine (Lee and Snyder 1981, 1982; Biegon and Rainbow 1983) and [3H]mazindol (Javitch et al. 1985) showed high non-specific binding in vitro; similarly, [11C]desipramine is also not suitable as an in vivo radiotracer (Van Dort et al. 1997). [3H]Nisoxetine (Nis) is a suitable radioligand for in vitro study (Tejani-Butt 1992); unfortunately, the binding of racemic (R/S)-[11C]Nis in vivo appeared to be non-specific (Haka and Kilbourn 1989). Recently, McConathy et al. (2004)reported the preparation of [11C]talopram and [11C]talsupram. Although both compounds displayed high affinity and selectivity for the human NET in vitro (IC50 2.9 and 0.79 nm for talopram and talsupram respectively; Bogeso et al. 1985), the biodistribution studies in rats showed that the brain uptake of these two 11C-labeled tracers was low, which diminished their potential application for imaging brain NET. Although an iodinated analog of tomoxetine showed high non-specific binding in vivo (Chumpradit et al. 1992), tomoxetine itself displayed high in vitro affinity (Kd = 2 nm, Tatsumi et al. 1997). However, our intention to develop tomoxetine as a selective NET tracer for positron emission tomography (PET) studies was discouraged by our characterization studies (Ding et al., unpublished results) in baboon using [11C] DASB, a selective SERT ligand (Houle et al. 2000). In these studies, we demonstrated that tomoxetine exhibited the same blocking effect on [11C] DASB binding as fluoxetine (a selective SERT inhibitor). A thorough investigation indicated that tomoxetine is an equally potent in vivo inhibitor of both the NET and the SERT. These results were consistent with previous in vitro data showing that (R)-[3H]tomoxetine bound not only to the NET but also to the SERT (Gehlert et al. 1995). Thus, the search for useful brain-penetrating NET tracers has been unsuccessful until the recent development of 11C-labeled reboxetine derivatives that show specific localization and highly encouraging binding kinetics in rats and non-human primates with PET (Ding et al. 2003; Schou et al. 2003; Wilson et al. 2003).

Our initial objective was to develop radiolabeled reboxetine analogs, as they have high specificity and selectivity towards the NET, and they have reasonable calculated log P values (Table 1). In addition, [3H]Nis has been used as a ‘gold standard’ for in vitro mapping of the NET and Nis has a structure quite similar to that of reboxetine (Fig. 1). Despite the biodistribution study of (R/S)-[N-11CH3]Nis in mice that showed only modest specific binding (Haka and Kilbourn 1989), we believe that 11C-labeled nisoxetine merits further investigation as a potential PET tracer. Therefore, in the present study, we labeled the more potent enantiomer of Nis in two positions (namely (R)-[O-11CH3]Nis and (R)-[N-11CH3]Nis; IC50 values for R- and S-Nis are 5.8 and 18 nm respectively; Gehlert et al. 1995) to further evaluate its potential as a PET tracer for NET using a primate model. Because the profile of labeled metabolites for the labeled N- versus O-methyl compounds may be different, we labeled Nis at different positions to probe potential differences.

Table 1.  Affinity and selectivity of several potent NET ligands
NET inhibitorDATSERTNET*DAT/NETSERT/NETCSLog P
  1. *Kd, Tatsumi et al. (1997); **IC50, Melloni et al. (1984); ***IC50, Strolin Benedetti et al. (1995). †Lort is a potent NET inhibitor with a potency higher than imipramine (13-fold) and desipramine (five-fold) (Depin et al. 1985). The affinity and selectivity for FRB will be determined if the tracer is promising. Lipophilicity was calculated as CSlog P using the ChemSilico LLC (Tewksbury, MA, USA ) family of property prediction software (CSPredict).

Nis*3601000136010001.74
MRB**> 10 0003102.540321251.13
(S,S)-MRB***  3.6   
(R,R)-MRB***  85   
FRB1.85
3-Cl-MRB**7045583.32131691.85
Oxap*434039004.98908003.50
Lort> 10 000> 100 0000.2> 10 000> 100 0001.9
Figure 1.

Structures of (S, S) and (R, R) enantiomers of [11C] MRB, [11C]3-Cl-MRB, and [18F]FRB; (R)-[O-11CH3]nisoxetine, (R)-[N-11CH3]nisoxetine, [11C]lortalamine and [11C]oxaprotiline.

Reboxetine is a specific NET inhibitor with high affinity and high selectivity, and has been approved for the treatment of depressive illness in several European countries. Based on our promising preliminary studies with 11C and 18F derivatives of reboxetine and their individual (R,R) and (S,S) enantiomers (Ding et al. 2003; Lin and Ding 2004Lin et al. 2005), we have identified the lead compound to be (S,S)-[11C]methylreboxetine (MRB). This tracer has recently been used to assess the NET occupancy by therapeutic doses of reboxetine in humans (Andree et al. 2004). Given that the initial characterization of (S,S)-[11C]MRB showed many desired in vivo properties that have not been seen with any other NET ligands, our strategy was to see whether we could optimize this ligand by, for example, decreasing the non-specific striatal uptake that characterizes in vivo binding of [11C]MRB as well as many of the NET tracers in the literature. Oxaprotiline (Oxap) and lortalamine (Lort) were chosen as candidates because they are structurally different from MRB, and they both have high affinity and selectivity towards NET (Fig. 1 and Table 1). A compound structurally similar to MRB, 3-Cl-MRB, which is more potent but less selective than reboxetine, was also chosen because it may also provide important structure–affinity relationship information in terms of binding and kinetics.

We report here the syntheses and characterization of several potent NET ligands. Their log P values (calculated and measured), plasma metabolism, brain uptake and kinetics were determined and compared. Criteria for suitability as potential NET ligands for PET imaging studies are discussed. Based on these comparative studies in non-human primate brain, the ability to predict the in vivo behavior of ligands based on their log P values and affinities is also discussed.

Materials and methods

Chemistry

The procedures for the preparation of the precursors, radiosynthesis and chiral resolution leading to [11C]MRB, (S,S)-[11C]MRB and (R,R)-[11C]MRB have been described in our previous publications (Ding et al. 2003; Lin and Ding 2004). The synthetic procedures for [18F]fluororeboxetine (FRB) (S,S)-[18F]FRB, (R,R)-[18F]FRB and the tetradeuterated analog (S,S)-[18F]-FRB-D4 have been submitted as a manuscript (Lin et al. 2005). The corresponding precursors for (R)-[O-11CH3]Nis and (R)-[N-11CH3]Nis, [11C]Oxap and [11C]Lort were synthesized according to published procedures (Wilhelm et al. 1972; Olofson et al. 1977; Depin et al. 1980; Mitsunobu 1981; Srebnik et al. 1988; Liu et al. 2000). The detailed procedure for the radiosynthesis of (R)-[O-11CH3]Nis, (R)-[N-11CH3]Nis, [11C]Oxap and [11C]Lort, and the procedures for the preparation of the precursors, radiosynthesis and chiral resolution leading to [11C]3-Cl-MRB and its (S,S) and (R,R) enantiomers, will be published elsewhere.

The general procedure for the radiosynthesis of 11C-labeled ligands is described below. A V-shaped three-necked glass vessel containing approx. 1 mg of the corresponding precursor in dimethylformate (DMF) (0.25 mL) was dipped into a dry ice/acetonitrile bath 10 min before the release of [11C]CH3I from the GE PETtrace MeI Microlab (GE Medical Systems, Milwaukee, WI). [11C]CH3I was transferred into the V-shaped vessel using argon as the carrier gas. After the radioactivity trapped inside the vessel had reached its maximum, as monitored by an NaI detector, the vessel was sealed and heated in an oil bath at 100°C for 5–10 min. At the end of reaction, water (1 mL) was added to the reaction mixture, and the product was separated using a Phenomenex (Torrance, CA, USA) Luna C-18 semipreparative column (250 mm × 10 mm, 5 µm). This column was connected to a Knauer HPLC system (Sonntek Inc., Woodcliff Lake, NJ, USA) equipped with a model K-500 pump, a model 87 variable wavelength monitor (set at 254 nm), an NaI radioactivity detector and two Hewlett-Packard 3390A integrators (Palo Alto, CA, USA), and was eluted with a HPLC solvent which was developed for each individual ligand. The fraction containing the desired product was collected and evaporated. The residue was dissolved in sterile water (4 mL) and passed through a 0.22-µm Millipore filter into a sterile vial to make the final injectate for baboon studies with PET.

All other chemicals were purchased from the Aldrich Chemical Company (Milwaukee, WI, USA) and were used as received without further purification. Melting points were taken on a Fisher–Johns melting point apparatus (Fisher Scientific Co., Pittsburgh, PA, USA) and were uncorrected. All NMR spectra were recorded with samples in CDCl3 (containing internal standard tetramethylsilane (TMS)) using a Bruker AVANCE 400-MHz NMR spectrometer (400 MHz for 1H and 100 MHz for 13C) (Bruker Instruments Inc., Billerica, MA, USA), and were reported in parts per million downfield from internal tetramethylsilane. The central peak of CDCl3 signal at 77.0 ppm was used as the 13C NMR reference. Radioactivity was measured in a Capintec CRC-712MV radioisotope calibrator (Capintec Inc., Ramsey, NJ, USA). [11C]CH3I was prepared from [11C]CO2 in an automated PETtrace MeI Microlab. [11C]CO2 was produced with the BNL JSW 1710 (Japan Steel Works Ltd., Tokyo, Japan) and new EBCO (Richmond, BC, Canada) cyclotrons by the 14N(p,α)11C nuclear reaction using 100 ppm O2 in N2 as the target gas. The [11C]CO2 was first transformed into [11C]CH4 through nickel-catalyzed hydrogenation, followed by the reaction with I2 to give the no-carrier-added [11C]CH3I. No-carrier-added [18F]fluoride ion was produced by the BNL JSW 1710 cyclotron via the 18O(p,n)18F nuclear reaction on 18O-enriched water.

Quality control of purified 11C- and 18F-labeled NET radioligands

The specific activities of radioligands were determined by the UV absorbance of the radioactive peaks compared with standard curves of unlabeled reference compounds. The radiochemical purities were determined by an analytical radio-HPLC system and a TLC system in the presence of the unlabeled reference compound as a carrier. The analytical radio-HPLC system consisted of a Phenomenex Luna C-18 analytical column (250 mm × 4.6 mm, 5 µm), a Knauer model K-1001 pump (Sonntek Inc.), a Knauer model K-1500 solvent organizer, a Knauer model K2501 UV detector (254 nm), an NaI radioactivity detector and two Hewlett-Packard 3390A integrators. The analytical column was eluted with acetonitrile/methanol/water/triethylamine/acetic acid (250 : 250 : 500 : 8 : 3.2) at 1.5 mL/min for [11C]Nis, acetonitrile/0.2 m ammonium formate (25 : 75) at 1 mL/min for [11C]Lort and acetonitrile/0.05 m ammonium formate (35 : 65) at 1.4 mL/min for [11C]Oxap. The retention times of Nis, Lort and Oxap were 9.21, 8.21 and 6.37 min respectively. For TLC analysis, Macherey-Nagel (Bodman Industries, Aston, PA, USA) polygram sil G/UV254 plastic-back TLC plate was used. The TLC plate was developed with methanol/ammonium hydroxide (10 : 0.3) for [11C]Lort, and methanol/ammonium hydroxide (10 : 0.5) for both [11C]Nis and [11C]Oxap. The developed TLC plate was scanned by using a Bioscan System 200 imaging scanner (Cambridge Scientific Products, Cambridge, MA, USA). The RF values of Nis, Lort and Oxap were 0.37, 0.34 and 0.30 respectively.

For [18F]FRB, the analytical column was eluted with a combination of acetonitrile (solvent A) and 0.1 m ammonium formate (solvent B) at a flow rate of 1.5 mL/min. The gradient HPLC system was 30 : 70 A/B to 70 : 30 A/B in 10 min. The retention time of FRB was 7.3 min. For TLC analysis, the TLC plate was developed with methanol/ammonium hydroxide (10 : 0.1) and the RF of FRB was 0.52. The enantiomeric purity of the individually labeled [18F]FRB enantiomer was determined using an analytical chiral HPLC column (Daicel Chiralpak AD, 250 mm × 4.6 mm, 10 µm; Chiral Technologies, Inc., West Chester, PA, USA) connected to the above analytical radio-HPLC system (Ding et al. 2003; Lin and Ding 2004). By eluting with hexane/2-propanol/diethylamine (92.5 : 7.5 : 0.2) at a flow rate of 1.5 mL/min, the retention times of (S,S)-FRB and (R,R)-FRB were 22.1 and 25.5 min respectively.

For 3-Cl-[11C]MRB, the analytical column was eluted with acetonitrile/0.2 m ammonium formate (34 : 66) at a flow rate of 1.5 mL/min. The retention time of 3-Cl-MRB was 7.63 min. For TLC analysis, the TLC plate was developed with methanol/ammonium hydroxide (10 : 0.2) and the RF of 3-Cl-MRB was 0.49. The enantiomeric purity of the individually labeled [11C]3-Cl-MRB enantiomers was determined using the chiral analytical radio-HPLC system described above for [18F]FRB. By eluting with hexane/ethanol/diethylamine (90 : 10 : 0.2) at a flow rate of 1.5 mL/min, the retention times of (2S,3S)-3-Cl-MRB and (2R,3R)-3-Cl-MRB were 10.56 and 13.93 min respectively.

Baboon studies

Baboon studies were carried out using the general procedures described previously (Dewey et al. 1990; Ding et al. 2003). Briefly, animals were anesthetized with ketamine (10 mg/kg) and then intubated and ventilated with a mixture of isofluorane (1–4%) and nitrous oxide (1500 mL/min) and oxygen (800 mL/min). Catheters were inserted in a popliteal artery and radial arm vein for arterial sampling and radiotracer injection respectively. There was a time interval of at least 2 h between ketamine injection and radiotracer injection to assure a minimal effect of ketamine. Animals were allowed 4 weeks between studies to recover from anesthesia and blood sampling. During the study they were monitored for heart rate, respiration rate, arterial partial pressure of oxygen and temperature using a pediatric monitor (SpaceLabs Pediatric Patient Care Monitoring System, Issaquah, WA, USA). They were transported to and from the PET facility in a heated transfer cage and a member of the staff attended them while they recovered from anesthesia. During PET, baboons were positioned on an individualized padded positioning table to eliminate motion and to minimize errors on repositioning. All animal studies were conducted in strict accordance with the NIH Guide for the Care and Use of Laboratory Animals (US Department of Health and Human Services et al. 1994) and approved by the Brookhaven National Laboratory Animal Care Committee.

PET imaging

We acquired dynamic PET scans using the Siemens HR + (Siemen's high-resolution, whole-body PET scanner with 4.5 × 4.5 × 4.8 mm at center of field of view, Malvern, PA, USA) in three-dimensional mode for a period of 90 min combined with pharmacologic intervention to evaluate new radioligands, as described in our recent publication for [11C]MRB (Ding et al. 2003). For all scans, a transmission scan was obtained with a 68Ge rotating rod source before the emission scan to correct for attenuation before each radiotracer injection. Typically, 2–4 mCi radioactivity was injected for each baboon study (specific activity 0.5–1.5 Ci/mol at time of injection with our JSW cyclotron; however, with our new EBCO cyclotron the specific activity has been increased by three-fold). An intrasubject (two baboons) assessment of (1) reproducibility (test/retest) and (2) specificity and selectivity (pretreatment with selective NET inhibitor Nis 1 mg/kg i.v.) was conducted for all the newly developed tracers. Further selectivity studies were performed on the more promising ligands, based upon initial results. These entailed pretreatment with GBR12909 (a selective DAT inhibitor; 1.5 mg/kg, 30 min before tracer injection) and citalopram (a selective SERT inhibitor; 2 mg/kg, 30 min before tracer injection) to confirm the lack of binding of the new tracer to these sites in vivo, similar to our previous studies demonstrating ligand selectivity (Ding et al. 1994, 1997, 2003). During all drug interventions, we continuously monitored heart rate, blood pressure and respiration rate to ensure the stability and safety of the animal.

Plasma radioactivity and metabolite analysis

Arterial blood was sampled every 2.5 s (OleDich blood sampling machine, Hvidovre, Denmark) for the first 2 min, and then at 5, 10, 20, 30 and 60 min and at the end of the study. All samples were centrifuged to obtain plasma, which was counted in a well counter (Picker, Cleveland, OH, USA) that had been cross-calibrated to the PET with a 68Ge/68Ga source. Selected plasma samples were assayed for the fraction of the parent radiotracer either by HPLC (method A) or by solid-phase extraction using a laboratory robot (method B) (Ding et al. 2003). The total radioactivity in plasma for each tracer for each scan was corrected for the presence of labeled metabolites to obtain the input function that was used in kinetic modeling.

Method A

Baboon plasma (0.05–0.4 mL), sampled at 1, 5, 10, 30 and 60 min and at the end of study, was added to 0.3 ml acetonitrile containing a standard of unlabeled parent compound (20 µL of 1 mg/mL solution) and the mixture was sonicated and centrifuged. The supernatant was analyzed by HPLC with UV detection (254 nm) and radioactivity assay of fractions. For MRB, a Phenomenex, Luna C-18 column (4.6 × 250 mm, 5 μm) was used, eluted with CH3CN: 0.2 m ammonium formate (AF) (30 : 70) at a flow rate of 1.6mL/min. A Waters (Milford, MA, USA) µbondapak C18 column (3.9 × 300 mm) was used; CH3CN: 0.05 m AF (35 : 65) was used for elution of FRB, Lort, Nis and Oxap, and CH3CN: 0.1 m AF (70 : 30) for elution of 3-Cl-MRB. The fraction of radiotracer in each sample was taken as the amount of radioactivity co-eluting with the unlabeled parent compound relative to the total amount injected into the HPLC.

Method B

Baboon plasma (0.2 mL), sampled at 1, 5, 10, 20, 30 and 60 min and the end of study, was added to 3 mL water and stored at room temperature (25°C) before being applied to activated Varian (Palo Alto, CA, USA) BondElut C18 cartridges (500 mg, preconditioned with 5 mL methanol followed by 5 mL phosphate buffer, pH 7.4) preloaded with 2 mL deionized water. This was followed by a series of three solvent rinses to remove the metabolite fractions (twice with 5 mL deionized water then twice with 5 mL 60% methanol in water). The radioactivity remaining on the cartridge was measured as the percentage of parent radiotracer.

Evaluation of tracer metabolism in mice

Further evaluation to determine whether the labeled metabolite(s) can cross the blood–brain barrier was conducted for promising ligands in a mouse model. For example, all of the radiolabeled plasma metabolites of MRB were polar species and not expected to enter the brain. To confirm this, mice were injected with (S,S)-[11C]MRB; their brains were removed at 10 min after injection and homogenized. Radiolabeled species were extracted and analyzed by radio-HPLC (Ding et al. 1996; Mathis et al. 2003; Kiyono et al. 2004).

Lipophilicity (log P)

The following method was modified from published procedures (Del Rosario et al. 1994) and was used to measure the log P of new radioligands. An aliquot of radiotracer (5–10 µCi) was added to a vial containing 2.5 mL octanol and 2.5 mL 0.1 m phosphate buffer, 7.4. The mixture was vortexed for 2 min and centrifuged at 5000 rpm for 2 min. A sample of the octanol (0.1 mL) and buffer (1 mL) layers was counted in a γ counter. The radioactivity counts were corrected for decay and the partition coefficient (P) was calculated as counts in octanol/counts in buffer. The major portion of the octanol layer (2.0 mL) was diluted with 0.5 mL octanol and mixed with a fresh portion of 2.5 mL phosphate buffer. The equilibration procedure described above was repeated until a constant value of log P was obtained; 10 separate measurements were usually performed for each tracer.

Plasma protein binding (PPB)

An aliquot of labeled compound was counted and added to 500 µL human plasma and this was incubated for 10 min at room temperature. Precisely measured volumes (20–40 µL) of the incubation mixture were counted (unspun aliquots). A portion (200–400 µL) of the incubation mixture was placed in the upper level of a Centrifree tube (Amicon Inc., Beverly, MA, USA) and this was centrifuged for 10 min. After centrifuging, the top portion of Centrifree tube containing the bound portion was removed and discarded and precisely measured aliquots (20–40 µL) of the liquid in the cup (unbound fraction) were counted. The free fraction was the ratio of the decay-corrected counts of the unbound aliquots to the decay-corrected counts of the unspun aliquots.

Region selection and data analysis

Methodologies for brain image analysis have been described previously (Ding et al. 1997, 2003). The emission data for the dynamic scans of baboon brain were corrected for attenuation and reconstructed using filtered back projection. For the purpose of drawing regions of interest (ROIs) on the baboon brain, time frames from dynamic images taken from 0 to 90 min were summed. ROIs such as thalamus (TH), midbrain, brainstem, striatum (ST) (or basal ganglia), cerebellum (CB), cortical regions (frontal, temporal, lateral-temporal, medial-temporal, parietal and occipital cortices), cingulate gyrus and a global region were drawn on these summed images and then projected to the dynamic emission scans to obtain concentration of radioactivity versus time, and expressed as a percentage of the total injected dose per millilitre. PET–magnetic resonance imaging (MRI) co-registration was used to guide and improve the reliability of ROI placement. Time–activity curves and the time course of unchanged tracer in plasma were used to estimate NET availability using a kinetic model. We calculated distribution volume (DV) in ROIs using a graphical analysis method for reversible systems (Logan plots) (Logan et al. 1990). The ratio of the DV in the NET transporter-rich region (TH) to that of the DV in the NET transporter-poor reference region (Ref) (Ding et al. 2003) was used as a model term for calculating NET availability (Logan et al. 1990). The analysis of tracer binding in terms of the DV provides a measure of binding that is a linear function of transporter (or receptor) availability given by

image(1)

for regions containing transporter characterized by an equilibrium dissociation constant Kd′ (Kd′ = Kd/fNS; fNS, free fraction of tracer in tissue) and free NET concentration, Bmax′; k1 and k2 are the plasma to tissue and tissue to plasma transport constants respectively. For transporter-rich regions the DV is given by

image(2)

A parameter proportional to free transporter concentration for the case of the TH can be obtained from equations 1 and 2 giving

image(3)

where Kd′ and k2 include the free fraction of tracer in tissue. We assumed that that the ratio of the transport constants is the same for TH and Ref (a similar assumption is made for ST and CB in the case of DAT studies). Equations 1 and 2 are based on classical compartmental analysis in which the effects of cerebral blood flow (CBF) and capillary permeability are implicitly included in K1 and k2. We examined the percentage change on drug treatment and compared DV ratios.

Co-registration with MRI

PET–MRI co-registration can guide and improve the reliability of ROI placement. We conducted MRI scans on our baboons and used commercially available software Pixel-wise Modeling (PMOD; Mikolajczyk et al. 1998) to do the co-registration and then draw the ROIs.

Results

Chemistry

All the radioligands were synthesized in our laboratory and then subjected to initial evaluation with PET. A brief summary of the synthesis times, radiochemical yields, radiochemical purities and specific activities for all the ligands described in this manuscript is provided in Table 2.

Table 2.  Synthesis times, radiochemical yields, radiochemical purities and specific activities of 11C-labeled (R)-Nis, Lort and Oxap, MRB and 3-Cl-MRB; and 18F-labeled FRB
RadiotracerSynthesis
time (min)
Radiochemical
yield (%) EOB
Radiochemical
purity (%)
Specific activity
(Ci/µmol) EOB
  • *

    The radiochemical yield of (R)-[O-11CH3]Nis was relatively lower (23–29%, EOB) owing to the formation of the undesired N-[11C]methylated byproduct [11C]N,N-dimethyl-O-nornisoxetine, in 64–70% (EOB).

(R)-[N-11CH3]Nis3963–72> 991.7–3.7
(R)-[O-11CH3]Nis4223–29*> 992.1–2.5
[11C]Lort3285–97> 992.2–3.1
[11C]Oxap3667–78> 991.8–3.0
[11C]MRB4061–74> 981.7–2.3
[11C]3-Cl-MRB4064–72> 992.0–4.8
[18F]FRB12011–27> 980.6–1.3

Metabolite analysis

The results of the assays for the percentage unchanged for all the tracers in baboon plasma via the HPLC and solid-phase extraction methods gave similar results. As indicated in our previous publication (Ding et al. 2003), the fraction of unchanged (S,S)-[11C]MRB and (R,R)-[11C]MRB in plasma was similar, and all of the radiolabeled plasma metabolites of MRB were polar species and not expected to enter the brain. To confirm this, the chemical form of the radiolabeled species in the brain after injection of mice with (S,S)-[11C]MRB was analyzed by radio-HPLC. It was determined that > 95% of radioactivity present in the homogenates was unmetabolized parent compound, indicating very low brain entry of the radiolabeled metabolites of MRB. The results of the assay for the percentage unchanged for the other tracers in baboon plasma are tabulated in Table 3. In general, the rate of metabolism did not differ significantly between the tracers. Similar metabolism was also observed for individual (S,S) and (R,R) enantiomers of [11C]3-Cl-MRB. However, (R,R)-[18F]FRB exhibited faster metabolism than (S,S)-[18F]FRB.

Table 3.  Percentage of unchanged parent compound at various time points in baboon plasma after injection of NET radiotracers
Radiotracer% unchanged parent compund
1  min5  min10  min30  min60  min90  min120  min
  • *

    Similar metabolism for individual (S,S) and (R,R) enantiomers.

(R)-[N-11CH3]Nis9653322913  
(R)-[O-11CH3]Nis9868453725  
[11C]Lort9563543937  
[11C]Oxap99653310  
[11C]MRB*9859433530  
[11C]3-Cl-MRB*9887713326  
[18F]FRB96534438292624
(S,S)-[18F]FRB96755732252220
(R,R)-[18F]FRB95423329252323
(S,S)-[18F]FRB-D490644125221919

Evaluation of [11C]MRB, (S,S) and (R,R)-[11C]MRB

Evaluations of the racemate and individual enantiomers of [11C]MRB as radioligands for PET imaging studies of NET systems in baboons both in brain and in peripheral organs have been described previously (Ding et al. 2003). Briefly, the regional distribution of the radioactivity after injection of [11C]MRB in baboon brain is consistent with the known distribution of NET, with the highest uptake and slower washout occurring in the TH, a known NET-rich region. For a NET-poor region such as ST, there were no significant changes in the uptake with the Nis pretreatment. In contrast, a significant blocking effect by Nis was observed in NET-rich regions such as TH and CB after injection of racemic [11C]MRB, with an even greater effect after injection of (S,S)-[11C]MRB (the largest DV change occurred in TH, − 48%, which is equivalent to almost complete blockade). However, no blocking effect on the uptake of [11C]MRB was observed when baboons were pretreated with GBR12909 or citalopram. These results, along with the fact that there was no regional specificity and no blocking effect by Nis for (R,R)-[11C]MRB, suggest the enantioselectivity and specificity of MRB in vivo, which is consistent with previous in vitro and in vivo studies in rodents (Wong et al. 2000; Wilson et al. 2003). We also examined the tracer binding in baboon torso and demonstrated a blocking effect by desipramine only in the heart, a NET-rich organ, after injection of (S,S)-[11C]MRB, but not the (R,R)-isomer.

Evaluation of [18F]FRB, (S,S) and (R,R)-[18F]FRB, and the tetradeuterated analog (S,S)-[18F]FRB-D4

We have shown that (S,S)-[11C]MRB is a potent and highly selective NET radioligand whose regional kinetics can be quantified using a graphic kinetic modeling for reversible ligands; however, it might be beneficial to extend the PET scanning time using a 18F-labeled radioligand. The preparation of its [18F]fluoromethyl analog [(S,S)-[18F]FMeNER] has been reported (Schou et al. 2003, 2004). In spite of significant uptake of (S,S)-[18F]FMeNER in the NET-rich regions of monkey brain, there was also high bone uptake due to in vivo defluorination. A di-deuterated analog (S,S)-[18F]FMeNER-D2 (Schou et al. 2004), was then developed with the intention of reducing its in vivo defluorination. PET studies indicated that the extent of defluorination was significantly reduced, although not totally inhibited, as shown by the continuous increase of bone uptake.

With the hope of minimizing defluorination, we prepared two new 18F-labeled reboxetine derivatives (S,S)-2-[α-2-(2-[18F]fluoroethoxyphenoxy)benzyl]-morpholine [(S,S)-[18F]FRB-H4] and its tetradeuterated analog (S,S)-[18F]FRB-D4, and evaluated their potential as PET tracers.

The racemate and (S,S) and (R,R) enantiomers of [18F]FRB and [18F]FRB-D4 were obtained in 11–27% (end of bombardment, EOB) in 120 min with a radiochemical purity of > 98% and specific activities of 0.57–1.29 Ci/µm (EOB). The racemate and (S,S) enantiomer of each tracer displayed regional specificities that were consistent with the known NET distribution, and their uptakes could be blocked by NET inhibitors. In contrast, no regional specificity or blocking effect were observed for the (R,R)-enantiomers. Figure 2 shows the significant difference in regional specificity between the (S,S) and (R,R) enantiomers of [18F]FRB. Similar signal to noise (S/N) ratios were obtained for both (S,S) enantiomers of [18F]FRB and [18F]FRB-D4(Fig. 3). However, the bone uptake was low and did not increase with time, whereas the clearance rate was faster for [18F]FRB-D4 than for [18F]FRB-H4(Fig. 4). These findings were also supported by a biodistribution study in mice (Lin and Ding 2004). These results are quite different from those of previous studies with (S,S)-[18F]FMeNER and (S,S)-[18F]FMeNER-D2, in that the bone uptake increased with time. In terms of S/N ratio, (S,S)-[18F]FRB was not as good as (S,S)-[11C]MRB (Table 4). However, in addition to the advantage of the longer half-life of 18F (110 min) compared with 11C (20 min), (S,S)-[18F]FRB displayed faster kinetics in NET-rich regions, which may facilitate its kinetic analysis. Thus (S,S)-[18F]FRB or (S,S)-[18F]FRB-D4 may have potential as ligands for NET studies.

Figure 2.

Comparative studies of (a) (S,S)-[18F]FRB and (b) (R,R)-[18F]FRB in the same baboon.

Figure 3.

Comparative studies of (S,S)-[18F]FRB (open symbols) and (S,S)-[18F]FRB-D4 (solid symbols) in the same baboon.

Figure 4.

Bone uptake and kinetics of (S,S)-[18F]FRB-H4 (open symbols) and (S,S)-[18F]FRB-D4 (solid symbols) in the same baboon.

Table 4.  Log P, PPB, peak brain uptake and S/N ratio of NET radioligands
RadiotracerAffinityLog P
(measured)
CSlog P
(calculated)
PPB
(% unbound)
Peak brain uptake†
(% injected dose)
S/N ratio
[DVR(TH/Ref)]
  1. †Peak brain uptake is based on percentage injected dose per millilitre  × 200 g (average brain weight). ‡0.48 is our measurement with (R)-[11C]Nis; – 0.02 is value measured with [3H]Nis by Kiyono et al. (2004). §1.74 and 3.4 are calculated values for Nis when different software was used.¶In rats, 0.45% injected dose/brain for (R)-[123I]iodoNis compared with 0.53% injected dose/brain for (S,S)-[11C]MRB (Kiyono et al. 2004).

(R)-[11C]Nis10.48 (−0.02)1.74 (3.4)§ 2Poor
(S,S)-[11C]MRB2.5*1.171.13143.2–41.8–2.2
(S,S)-[18F]FRB0.911.85131.81.3–1.6
(S,S)-[11C]3-Cl-MRB3.3*1.911.8563.21.4–1.6
[11C]Oxap4.92.13.509.16.0Poor
[11C]Lort0.21.352.08442.6Poor
(R)-[123I]IodoNis0.061.312.05Poor

Evaluation of (R)-[O-11CH3]Nis and (R)-[N-11CH3]Nis

Comparative PET studies showed similar uptakes in three brain regions (TH, CB and ST) and relatively low uptake in frontal cortex for both tracers. Plasma metabolite assays indicated a similar metabolism profile for both tracers, with a slightly slower metabolism for (R)-[O-11CH3]Nis than (R)-[N-11CH3]Nis (Table 4). Pretreatment with unlabeled Nis (1 mg/kg i.v., 10 min before tracer injection) did not reduce the before tracer binding; instead, increased uptakes were observed (Fig. 6), suggesting high non-specific binding in vivo; consequently, both tracers are not suitable for PET imaging of NET.

Figure 6.

[11C]Oxap in baboon brain. MB, midbrain; Occi, occipital cortex; Cing, cingulate gyrus; GL, global.

Evaluation of [11C]Oxap

Racemic [11C]Oxap was synthesized and subjected to initial evaluation with PET (Fig. 7). Uptake of the tracer in baboon brain, in general, was high; however, the distribution did not match the known NET distribution. The fact that Oxap has the highest uptake in ST and low S/N ratio suggested strongly that an analog with this type of molecular structure is probably not desirable for in vivo imaging of NET, despite the fact that its measured log P is ideal (2.1).

Figure 7.

[11C]Lort in baboon brain under baseline conditions (a, solid symbols) and after pretreatment with Nis (b, open symbols).

Evaluation of [11C]Lort

Racemic [11C]Lort was synthesized and subjected to initial evaluation with PET in baboon. In terms of regional specificity and S/N ratio, [11C]Lort was slightly better than [11C]Oxap with similar uptakes in both TH and ST in the baseline study. Pretreatment with unlabeled Nis did reduce the tracer binding in TH and CB, but less blocking effect was observed in ST as expected (Fig. 7 shows the time–activity curves for a control study and a blocking study; Figs 10(d) and (e) are the corresponding PET images). However, these positive characteristics, which indicate the specific binding of the tracer to NET, were diminished by the fact that [11C]Lort still showed high non-specific uptake in ST (ST had higher uptake than TH).

Figure 10.

(a) A representative PET image for (S,S)-[11C]MRB in baboon brain co-registered with the magnetic resonance image; (b) and (c) are PET images in the same baboon for (S,S)-[11C]MRB from a control study and from a study with pre-administration of Nis (1 mg/kg, i.v.) respectively. Note the high S/N ratio with much higher uptake in the TH than the ST in the control study (b), and a significant blocking effect by Nis on uptake in the TH, but not the ST (c). (d) and (e) PET images in the same baboon for [11C]Lort from a control study and from a study with pre-administration of nisoxetine (1 mg/kg, i.v.) respectively. Note the poor S/N ratio with similar uptakes in both the TH and ST in the control study (d); less blocking was observed in the ST than TH (e). All images were generated via PMOD after ‘add volumes’ and dose correction.

Evaluation of 3-Cl-[11C]MRB

We used the same chiral resolution strategy, followed by radiosynthesis to obtain pure (S,S) and (R,R)-3-Cl-[11C]MRB. Comparative PET studies were then carried out in baboons. Figure 8 shows that (S,S)-3-Cl-[11C]MRB displayed appropriate regional specificities and S/N ratio, with uptake in TH significantly higher than that in ST; in contrast, high uptakes in both the ST and TH were observed for the (R,R)-enantiomer. Comparative studies of (S,S)-3-Cl-[11C]MRB versus (S,S)-[11C]MRB (Fig. 9) in the same baboon indicated that both tracers have similar high uptakes in the TH; however, (S,S)-[11C]MRB appeared to be superior because it had a faster clearance from the ST.

Figure 8.

Comparative studies: (S,S)-3-Cl-[11C]MRB (a, solid symbols) versus (R,R)-(3-Cl-[11C]MRB (b, open symbols).

Figure 9.

Comparative studies: (S,S)-3-Cl-[11C]MRB (3-Cl, solid symbols) versus (S,S)-[11C]MRB (MRB, open symbols) in the same baboon.

Kinetic analysis

Our PET studies in baboon with the lead compound (S,S)-[11C]MRB showed the highest binding in TH, and also high binding in midbrain and brainstem, consistent with the known NET distribution. Locus coeruleus (located in the brainstem area) is known to have the highest NET concentration; however, its volume is too small to be reliably identified with PET. We therefore used the TH as the region of interest to determine the S/N ratio, by comparing the uptake of the ligand in the TH to that in a reference region. We noted that the identification of a reference region in vivo for quantification of the NET ligands has been problematic. However, since our last report (Ding et al. 2003) we have made advances toward the identification of reference regions for NET tracers. Briefly, we found that the ST and the occipital cortex were two brain regions that showed negligible response to drug challenge, such as studies with Nis or cocaine (which binds to the DAT, NET and SERT), and that the use of an average of these two brain regions as a composite reference region (Ref) appeared to provide a reasonable approach to quantification. The S/N ratios presented in Table 3 were calculated based on this strategy as the DV ratio (DVR) (the ratio of the DV from TH to that of ST + occipital cortex) for individual ligands. Details regarding the kinetic analysis using this method for a series of studies will be published elsewhere (Logan et al. 2005).

To summarize the above preliminary studies on our tracer development, the measured log P, PPB, peak brain uptake (e.g. percentage injected dose in the whole brain) and S/N (DVRTH/Ref, see below for details) are tabulated in Table 3 along with their reported affinity and calculated log P. For comparison, data for (R)-[123I]iodoNis is also included (Kiyono et al. 2004; Kung et al. 2004). These results clearly indicate the superiority of (S,S)-[11C]MRB and the suitability of the MRB analogs [(S,S)-[11C]MRB > (S,S)-[11C]3-Cl-MRB > (S,S)-[18F]FRB] as NET ligands for PET. In addition to high uptake in ST (higher than in TH), Nis, Oxap and Lort displayed high non-specific binding and poor S/N ratios. According to the in vitro mapping of the NET by quantitative autoradiography, lowest binding was found in the CA1 layer of the hippocampus and the caudate–putamen (Tejani-Butt 1992; Charnay et al. 1995). The mechanism underlying the relatively high in vivo uptake in the ST for all the NET radioligands that have been investigated is not known. It may reflect the presence of non-specific or low affinity non-NET binding sites in the ST, as has been seen for other NET ligands (Yavin et al. 1978; Biegon and Rainbow 1983). Our comparative studies indicate that although there is uptake in the ST after i.v. injection of (S,S)-[11C]MRB, its striatal uptake is, by far, the least significant among all the tracers. A representative PET image for (S,S)-[11C]MRB in baboon brain co-registered with the magnetic resonance image using PMOD is shown in Fig. 10(a). PET images generated via PMOD from a comparative study in the same baboon for (S,S)-[11C]MRB from a control study and from a study with pre-administration of Nis (1 mg/kg, i.v.) is shown in Figs 10(b) and (c) respectively. Figs 10(b) and (d) show control studies after injection of (S,S)-[11C]MRB and [11C]lor, respectively; note the high S/N ratio with much higher uptake in TH than in ST for (S,S)-[11C]MRB, compared with a poor S/N ratio with similar uptake in both TH and ST for [11C]lort. Furthermore, this uptake in ST appears to be non-specific because it could not be blocked by pretreatment with Nis. There was a significant blocking effect of Nis on the uptake in the TH, but not the ST, for (S,S)-[11C]MRB (Fig. 10c). In the case of [11C]lort, a more significant blocking effect was observed in the TH than the ST (Fig. 10e).

(R)-[123I]iodoNis, which has high affinity as a promising in vitro ligand (Kung et al. 2004), also has high non-specific binding and lack of selectivity in vivo, as indicated by Kiyono et al. (2004). Thus (S,S)-[11C]MRB remains by far the most promising in vivo NET radiotracer with adequate pharmacokinetic and metabolism, and is expected to provide specific and functional maps of the NET in the brain.

Discussion

The NET has long been recognized as an important molecular target for both stimulant and therapeutic drugs to treat depression, ADHD and other CNS disorders. However, despite widespread abuse and therapeutic use, the mechanisms for addictive and therapeutic properties of stimulant drugs such as cocaine and methylphenidate (ritalin, the most prescribed drug for treatment of ADHD) are not well understood. Cocaine binds to all three monoamine transporters (DAT, NET, SERT) with comparable affinities, and methylphenidate binds to the DAT and NET with even higher affinity for the NET. Their effects on the brain dopaminergic system have been well characterized in living humans, but our knowledge of their effects on the NET has been limited to post-mortem studies owing to a lack of suitable radiotracers. This places a sense of urgency in developing radiotracers that can characterize their binding to different molecular targets and the relationship to behavioral and therapeutic properties in living humans. We and other researchers are making progress in developing suitable ligands for mapping NET in vivo, in the hope that we will soon be able to better understand the role of the NET in various CNS disorders. In this study, the suitability of several new PET radioligands for the NET, including (R)-[11C]Nis, [11C]oxap and [11C]lort, and (S,S) and (R,R) enantiomers of [11C]MRB, [11C]3-Cl-MRB and [18F]FRB, were evaluated and compared in baboons. The results indicate that reboxetine derivatives are by far the best candidates for providing specific and functional maps of the NET in the human brain.

Tracer kinetics

Based on our baboon studies (S,S)-[11C]MRB exhibits high brain uptake with reasonable kinetics and suitable clearance rate from the binding sites; however, a study by the Karolinska group found that the same tracer [(S,S)-[11C]MeNER] has slow kinetics in the brain (Schou et al., 2003). The most likely explanation for the contrasting results lies in the fact that these two independent studies were carried out in two different primate species (baboons vs. cynomolgus monkeys). Different anesthesia procedures may also affect the tracer kinetics; these have been well documented (Dolle et al. 1999). In our previous paper (Ding et al. 2003), we not only reported time–activity curves but also carried out kinetic modeling and provided DV data showing that the uptake of (S,S)-[11C]MRB in ST is lower than that in TH and higher than that in most cortical regions (TH/ST 1.63, TH/occipital cortex 2.06, and TH/frontal cortex 1.94). In fact, a high ratio of binding in the TH to that in the ST is an important criterion for a NET tracer, considering that the NET concentration in ST is low. Almost all of the known NET ligands (in vitro and in vivo) displayed a significantly higher striatal uptake (higher than TH) than reboxetine analogs, as indicated in our studies of (R)-[11C]Nis, [11C]Oxap and [11C]Lort.

18F-labeled analogs of reboxetine, such as (S,S)-[18F]FRB and (S,S)-[18F]FRB-D4 that have been evaluated in our laboratory, as well as (S,S)-[18F]FMeNER and (S,S)-[18F]FMeNER-D2 studied by Schou et al., (2004) displayed relatively fast kinetics in NET-rich regions which, in principle, should facilitate kinetic modeling. However, their characteristics, such as defluorination and a relatively poor S/N ratio compared with (S,S)-[11C]MRB, make them less desirable as in vivo NET ligands.

(S,S)-3-Cl-[11C]MRB is a promising ligand for imaging the brain NET. However, comparative studies of (S,S)-3-Cl-[11C]MRB and (S,S)-[11C]MRB (Fig. 9) in the same baboon indicated that (S,S)-[11C]MRB is still the best as it has a faster clearance from the ST.

In addition to high uptake in ST (higher than that in TH), non-specific binding and poor S/N ratio, Nis, Oxap and Lort displayed undesirable slow kinetics. Thus, the kinetics of (S,S)-[11C]MRB remain by far the most promising for PET studies.

Kinetic analysis

The cell bodies of norepinephrinergic neurons are located in the brainstem. Projections from these cell bodies are widespread – in TH, cortex, hippocampus and CB. Based on in vitro studies, the highest binding of [3H]desipramine was found in the locus coeruleus (888 fmol per mg protein), with moderately high binding in the cingulate (240 fmol per mg protein) and lower binding in cerebellar cortex (172 fmol per mg protein); binding in the basal ganglia (45 fmol per mg protein) practically indistinguishable from non-specific binding (Biegon and Rainbow 1983). Autoradiographic studies in rat with [3H]Nis also revealed that the highest specific binding was in brainstem (locus coeruleus) and TH (1526 and 1444 fmol per mg protein respectively), with 54 fmol per mg protein in caudate–putamen (Tejani-Butt 1992). Our PET studies in baboon showed high uptake of (S,S)-[11C]MRB and its analogs in TH, midbrain and brainstem, which is consistent with the known NET distribution. We also demonstrated their binding to the transporter to be sufficiently reversible that it can be characterized by a DV. The DV is a measure of the capacity of the tissue to bind the tracer and is therefore a function of the number of binding sites. However, there are also other factors such as non-specific binding, the presence of endogenous neurotransmitter, and the presence of low affinity sites that can also alter the amount of binding and affect the DV measurement. In our previous paper (Ding et al. 2003) we reported the DV rather than the DVR because we had not yet identified a reference region. It is not unusual to see variations in absolute DV values in the same baboon on different days owing to different physiological states of the animals probably related to anesthesia and other variables. That is why it is preferable to use DVR instead of the absolute DV when comparing studies. Normally this is done with a reference region, and the DVs are reported as the ratio of the DV from the region with specific binding to that of a region with little or no specific binding. Because the NET is present to some extent in many brain regions it is challenging to identify a reference region; furthermore, some of the ligands may bind to sites other than the NET. In order to identify a potential reference region, we used paired studies with the tracer (S,S)-[11C]MRB, in which a baseline scan was followed by a scan after pretreatment with a dose of cocaine. Results of these studies have been submitted as a manuscript (Logan et al. 2005). The advanced graphical analysis methods for quantification of the NET by choosing an average of the occipital and striatal regions for the reference region should in principle minimize the effect of fluctuations in the non-specific binding and statistical variations. In fact, using the data from our previous paper (Ding et al. 2003), we calculated the DVR as the ratio of the DV from TH to that of ST + occipital cortex for (S,S), racemic and (R,R) compounds; the value for the racemic compound (1.38) fell between those of the (S,S) (1.83) and the (R,R) (1.13) enantiomers, which is consistent with expectations. Nevertheless, a critical examination of the kinetic properties of all new radioligands will continue to be a crucial part of our radiotracer development strategy.

Lipophilicity (Log P)

The finding of a very low log P value (0.48) for (R)-[11C]Nis was unexpected because it was very different from the calculated values (1.74 or 3.4). We repeated the measurement several times on different days with different batches of (R)-[11C]Nis and each time obtained consistent results (log P =0.48 ± 0.01; n = 20). In fact, Kiyono et al. 2004 also recently reported a very low log P value of −0.02 for [3H]Nis, which was similar to what we obtained for (R)-[11C]Nis. As we and others have pointed out, calculated log P values are not always identical to measured values; Oxap is another example: the calculated value was 3.50 and the measured value 2.1. Furthermore, calculated log P values are often different when different software is used. The rationale for computational methods to generate the log P values for each structure is based on the information contained in the program library. It is believed that most estimates reflect only partitioning of the neutral species, and therefore represent more a log P value than a log D determination (which includes a partitioning value obtained by measurement of all species present in solution and therefore accounts for solubility affects associated with hydrogen bonding and ionization) (Avdeef 2001; van de Waterbeemd et al. 2001; Avdeef and Testa 2002; Waterhouse 2003). The discrepancy in the log P measurement of Nis and other ligands further supports our point that, although calculated log P values provide initial guidance about the lipophilicity of the molecule, they should be used with caution when making predictions.

Can good Kd guarantee its suitability as an in vivo ligand?

Measurements of density of the NET in the rat brain by various radioligands using autoradiography (Tejani-Butt 1992; Kung et al. 2004) showed the Bmax ranged from 50 to 1500 fmol per mg protein in various brain regions. The DAT density in the rat ST measured by [125I]IPT (N-(3-iodopropen-2-yl)-2 beta-carbomethoxy-3 beta-(4-chlorophenyl)tropane), a tropane derivative, was about 2000 fmol per mg protein (Kung et al. 1995), and a Bmax value of 100 fmol per mg protein for the SERT in rat cortical homogenates was obtained. Thus, in vivo mapping of the NET by PET, although challenging, is feasible. There are many NET inhibitors such as Nis (Kd 1 nm) and desipramine (Kd 0.83 nm) that have higher affinity than MRB (Kd 2.5 nm); however, they did not survive the in vivo test. A recent study reported that [125I]2-iodonisoxetine, which has a very high affinity (Kd = 0.06 nm), displayed extremely promising in vitro properties (Kung et al. 2004). Unfortunately, this tracer also failed the in vivo tests; i.e. it displayed high non-specific binding and/or binding to secondary sites resulting in a high background uptake (Kiyono et al. 2004). These disappointing results were similar to those obtained in our studies with (R)-[11C]Nis, [11C]Oxap and [11C]Lort. Thus, even though (S,S)-[11C]MRB has lower affinity than the abovementioned compounds, it displays more desirable selectivity and specificity in vivo than any existing NET radioligand. Similarly, [11C]raclopride, which has an affinity (Ki 1 nm) lower than that of many existing dopaminergic D2 ligands, is a superior in vivo ligand. [11C]DASB is another good example; it has an affinity (Ki 1.77 nm) that is lower than that of [11C]McN 5652 (Ki 0.26 nm); however, [11C]DASB is a much better tracer than [11C]McN 5652 for in vivo imaging of the SERT (Huang et al. 2002). Therefore, it is not strictly an affinity issue, and the high in vitro affinity of a ligand does not guarantee its suitability as an in vivo ligand.

Indeed, the poor ability to predict the behavior of chemical compounds in vivo based on their log P values and affinities emphasizes the need for more knowledge in this area (Fowler 2003). We and others have pointed this out, and the NET system is an outstanding example of the fact that generalizations do not always apply. Designing a ligand without the high non-NET uptake in the ST is the next challenge; however, the limitations of (S,S)-[11C]MRB are no more limiting than were initial studies with [11C]McN 5652 for the SERT. One might speculate that human studies of this preclinically well characterized ligand, coupled with advances in kinetic modeling, may provide extremely important information to guide us towards the development of a new generation of NET ligands.

Conclusions

We successfully developed new NET radioligands, including (S,S) and (R,R) enantiomers of [11C]MRB, [11C]3-Cl-MRB and [18F]FRB; and (R)-[11C]Nis, [11C]Oxap and [11C]Lort. Tracer evaluation in baboons indicated that (1) (S,S)-[11C]3-Cl-MRB displayed desired properties, similar to those of our lead compound (S,S)-[11C]MRB; (2) although high brain uptake is desirable, the high non-specific uptake in ST and low S/N ratio led us to conclude that [11C]Oxap is not a suitable ligand for in vivo NET imaging studies; and (3) [11C]Lort is better than [11C]Oxap, but still suffers from high non-specific uptake in the ST. Our tracer metabolism study in mice demonstrated that radiolabeled metabolites of (S,S)-[11C]MRB do not enter the brain and are not likely to contribute to the PET images. We identified the superiority of (S,S)-[11C]MRB and the suitability of the MRB analogs [(S,S)-[11C]MRB > (S,S)-[11C]3-Cl-MRB > (S,S)-[18F]FRB] as NET ligands for PET. PET–MRI co-registration in baboons improved the reliability of ROI placement. Our improved kinetic analysis method using DVR calculated by using an average of occipital cortex and ST as the reference region provided more reliable quantification in our occupancy studies (Ding et al. 2004).

In addition, we also wish to point out that after decades of effort in searching for a suitable in vivo NET radioligand for PET imaging, finally, for the first time there is an opportunity to look at this important molecular target. The lead compound (S,S)-[11C]MRB should be suitable for further in vivo evaluation in humans, as it has been well characterized by three research groups in three different species (rat, monkeys and baboons) and it displays much more desirable selectivity and specificity in vivo than any existing NET radioligand, even those with far higher affinity. With our new advances in modeling and quantification of the NET in vivo, we believe that this lead compound will facilitate better understanding of the role of the NET in substance abuse and other neuropsychiatric disorders. Furthermore, the knowledge and expertise gained from such human studies will, in principle, guide further investigation into the search for optimal tracers for imaging studies of the NET in human subjects.

Acknowledgements

This research was carried out at Brookhaven National Laboratory under contract DE-AC02-98CH10886 with the US Department of Energy and Office of Biological Environmental Research, and was also supported by the National Institutes of Health (National Institute for Biomedical Imaging and Bioengineering EB002630 and National Institute on Drug Abuse DA-06278) and Office of National Drug Control Policy. The authors are grateful to V. Garza, C. Shea, Y. Xu, D. Alexoff, M. J. Schueller, D. J. Schlyer and R. A. Ferrieri for their assistance in cyclotron operation, radiosynthesis and metabolite analysis; to T. Betzel and G. Quandt for their assistance in measurement of log P and PPB; to Wynne Schiffer for helping with co-registration using PMOD; and to J. S. Fowler for reviewing this manuscript.

Ancillary