Initially reported imaging probes, such as PCP and MK-801 analogs, suffered from high nonspecific in vivo accumulation, which was probably due to the high lipophilicity and lack of selectivity for NMDARs. Diarylguanidine derivatives were discovered as NMDAR open channel blockers with high potency and selectivity. These compounds showed therapeutic effect in various brain disease model animals (Reddy et al. 1994; Hu et al. 1997). Because most diarylguanidines are more hydrophilic than PCP and MK-801, several radiolabeled diarylguanidine analogs were developed for PET or SPECT imaging of NMDARs. In the first study, [¹²⁵I]CNS 1261 (Fig. 18.10) was developed as a high-affinity imaging agent for NMDARs with a K _i value of 4.2 nM for the [³H]MK-801-binding site with moderate lipophilicity (logD = 2.19). Binding assays to other neurotransmitter receptors demonstrated that CNS 1261 had a weaker binding affinity for the other 41 neurotransmitter receptors. Ex vivo autoradiography studies with normal rats showed that the [¹²⁵I]CNS 1261 uptake in the hippocampus was 2.4–2.9-fold higher than that in the cerebellum, in which the accumulation matched the distribution of NMDARs. In the ischemic rat brain, [¹²⁵I]CNS 1261 uptake was considerably increased in the neocortex and striatum, with a twofold higher uptake in the ischemic hemisphere of the caudate nucleus uptake than the same regions on the contralateral nonischemic hemisphere (Fig. 18.11). This suggested that [¹²⁵I]CNS 1261 could recognize the areas of NMDAR activation (Owens et al. 2000). Based on these results, several clinical SPECT studies of [¹²³I]CNS 1261 were conducted. SPECT imaging of [¹²³I]CNS 1261 in healthy volunteers were conducted by using both bolus and bolus plus constant infusion paradigms. Total volume of distribution (V _T) values was approximately the same in the NMDAR-rich regions (striatum, hippocampus, frontal cortex) and NMDAR-poor region (cerebellum). The V _T value in the thalamus, which is a region with considerably high expression of NMDARs, was slightly higher than in other regions (Fig. 18.12) (Erlandsson et al. 2003; Bressan et al. 2004). Bolus plus infusion of ketamine in healthy volunteers caused a global reduction in V _T of [¹²³I]CNS 1261 compared with placebo in most brain regions, including the putamen, temporal, and pericallosal regions (Stone et al. 2006). Subsequent work by the same group demonstrated that the degree of reduction in the V _T values of [¹²³I]CNS 1261 correlated with the negative symptoms of schizophrenia in all brain regions examined (Stone et al. 2008). In addition, they reported that drug-free patients with schizophrenia had reduced binding of [¹²³I]CNS 1261 in the left hippocampus relative to the whole cortex compared with healthy controls (Pilowsky et al. 2006). In contrast, the same group found that the V _T values of [¹²³I]CNS 1261 did not differ between healthy normal volunteers and drug-free or typical antipsychotic-treated schizophrenia patients. On the other hand, clozapine-treated patients showed a significant reduction of V _T in all brain regions examined, which indicated a decrease of NMDAR activation in this group. However, the regional brain distribution of [¹²³I]CNS 1261 was almost homogenous in all groups (Bressan et al. 2005). Although there was no conclusive evidence that [¹²³I]CNS 1261 could recognize NMDAR activation, these reports suggested that protein expression levels of NMDARs could be reduced in the brain of schizophrenia patients. Numerous reports indicated the involvement of NMDAR hypofunction in the pathophysiology of positive and negative symptoms of schizophrenia (Kantrowitz and Javitt 2010; Lin et al. 2012). Recently, Knol et al. reported further in vivo studies of [¹²³I]CNS 1261 using rats. The [¹²³I]CNS 1261 showed higher accumulation in the NMDAR-rich regions such as hippocampus and frontal cortex than that of NMDAR-poor cerebellum, in which the ratio was maximum (about 1.4) at 2 h postinjection. Pretreatment with the NMDAR co-agonist, d-serine, (10 mg/kg) led to no significant change in the hippocampus/cerebellar ratio. On the other hand, treatment with d-serine and MK-801 resulted in a significantly lower hippocampus/cerebellar ratio, which indicated the existence of specific trapping of a certain amount [¹²³I]CNS 1261 in the NMDAR-rich regions. However, the authors concluded that the use of [¹²³I]CNS 1261 would be limited in terms of quantification and detection of small changes in receptor availability because of a considerable amount of nonspecific binding of this tracer (Knol et al. 2009).

To provide potential PET radioligands for the PCP-binding site, [¹¹C]-labeled diarylguanidine derivatives have been synthesized and evaluated. [¹¹C]GMOM (Fig. 18.10) with a high affinity for NMDARs (K _i  = 5.2 nM) was prepared from the corresponding hydroxyl precursor with a specific activity of 46 GBq/μmol. In rats, brain uptake of [¹¹C]GMOM was high and peaked during the early part of the study. Regional brain distribution at 10 min was 1.23, 0.93, and 0.74 % ID/g for the frontal cortex, hippocampus, and cerebellum, respectively, as regards the appearance of contrast in the NMDAR-rich and NMDAR-poor regions. Pretreatment with open channel blockers GMOM or MK-801 (1 mg/kg) or the NB2B-selective NMDAR modulator Ro 25–6981 led to a moderate and uniform decrease of [¹¹C]GMOM uptake in brain regions. On the other hand, NMDAR co-agonist d-serine increased [¹¹C]GMOM binding in all regions. PET imaging in baboons revealed that [¹¹C]GMOM exhibited high BBB penetration. However, the brain distribution of [¹¹C]GMOM was almost homogenous and pre-administration of MK801 did not reduce regional V _T ’s or regional V _T ratios. The authors hypothesized that these inconsistent results could be caused by the blocking effect of anesthesia (ketamine and isoflurane) to NMDAR activity (Waterhouse et al. 2004). Further PET investigations of [¹¹C]GMOM in other conditions, such as in a conscious state, have not been reported so far.

Another diarylguanidine derivative, CNS 5161 (Fig. 18.10), has been shown to exhibit high affinity for the PCP-binding site (K _i  = 1.9 nM) (Hu et al. 1997; Dumont et al. 2002). Clinical trials using CNS 5161 have been evaluated in the treatment of neuropathic pain (Walters et al. 2002; Forst et al. 2007). The [³H]CNS 5161 showed a heterogeneous in vivo brain distribution in rats and a cortex/cerebellum ratio of 1.4. Pretreatment of NMDA increased the ratio in the hippocampus/cerebellum to 1.6–1.9, while MK801 diminished this increase that resulted in ratios close to 1. [³H]CNS 5161 appeared to be unstable in arterial blood, as 65 % of the parent compound was metabolized to several more hydrophilic compounds even 20 min after tracer injection (Biegon et al. 2007). The [¹¹C]CNS 5161 was synthesized by [¹¹C] methylation of the desmethyl guanidine precursor with low specific activity (41GBq/mmol) (Zhao et al. 2006). In human PET studies, [¹¹C]CNS 5161 exhibited the largest uptake in the putamen and thalamus and the lowest uptake in the cerebellum, but relatively lower radioactivity was observed in the hippocampus compared with other gray matter regions. Similar to rats, the metabolism of [¹¹C]CNS 5161 was rapid in human plasma. No blocking studies were reported in the literature (Asselin et al. 2004). Recently, clinical PET studies of [¹¹C]CNS 5161 were carried out to assess the effect of levodopa on striatal and cortical NMDAR activity in patients with Parkinson’s disease with and without dyskinesias (Ahmed et al. 2011). The levodopa-treated dyskinetic patients had higher [¹¹C]CNS 5161 uptake in the caudate, putamen, and precentral gyrus compared with patients without dyskinesias. These results are consistent with reports that upregulation of NMDARs has been associated with the development of levodopa-induced dyskinesias in animal models of Parkinsonism (Oh et al. 1998). However, no significant increase in [¹¹C]CNS 5161 uptake was observed in patients with Parkinson’s disease when compared with control subjects. In addition, relatively high uptake of [¹¹C]CNS 5161 in the cerebellum was observed, which was inconsistent with the brain distribution of [¹¹C]CNS 5161 in a previous clinical PET study reported by Asselin et al. (2004). Further investigations are necessary in order to prove the usefulness of these diarylguanidines as radioligands for measuring NMDAR function by PET or SPECT.

Recently, ¹⁸F-labeled S-fluoroalkyl diarylguanidines were synthesized and used for in vitro evaluation. Among them, [¹⁸F]7 and [¹⁸F]8 showed high affinity for the PCP-binding site (K _d  = 2.4 nM for [¹⁸F]7 and K _d  = 1.4 nM for [¹⁸F]8). Further studies could not be performed on [¹⁸F]8 owing to low stability of this ligand. A broad receptor-binding assay for 60 receptors demonstrated that [¹⁸F]7 had high selectivity for the PCP-binding site of NMDARs (Robins et al. 2010). No in vivo studies of [¹⁸F]7 have been reported so far.

The 6,11-ethanobenzo[b]quinolizinium derivatives are highly hydrophilic compounds reported to exhibit high affinity and specificity to the PCP-binding site of the NMDARs (Mallamo et al. 1994). Based on the SAR studies, new fluorine-18 radioligands ([¹⁸F]9 and [¹⁸F]10) and a carbon-11 ligand ([¹¹C]11) have been developed as novel PET radioligands for the PCP-binding site of NMDARs as shown in Fig. 18.13 (Sasaki et al. 1998; Ishibashi et al. 2000). These compounds showed moderate binding affinity for the PCP-binding site (9: IC₅₀ = 47 nM, 10: IC₅₀ = 89 nM, 11: IC₅₀ = 19 nM vs [³H]TCP) (Sasaki et al. 1998). In the preliminary radiosynthesis of [¹⁸F]9 and [¹⁸F]10, nucleophilic radiofluorination of mesylate was unsuccessful probably due to trapping of ¹⁸F⁻ anion by the ammonium cation. An alternative synthetic route, including early fluorination of dienophile prior to the Diels-Alder reaction permitted successful radiosynthesis of [¹⁸F]9 and [¹⁸F]10 (Sasaki et al. 1998; Ishibashi et al. 2000). However, the specific activity of [¹⁸F]9 and [¹⁸F]10 was relatively low (>263 MBq/μmol). The [¹¹C]11 was prepared by O-methylation of the hydroxyl precursor. The specific activity of [¹¹C]11 was about 63 GBq/mmol. Values were approximately a 1,000-fold lower than the theoretical specific activity of C-11 ligands starting from [¹¹C]CO₂. In vivo biodistribution studies demonstrated that these three radioligands showed a quite low BBB permeability (0.02 − 0.05 % ID/g at 15 min) presumably because of their cationic property and extremely low lipophilicity (Ishibashi et al. 2000). To improve brain uptake and receptor specificity, the reduced derivative [¹¹C]12 was developed as a prodrug for [¹¹C]11. In the preliminary in vivo studies, cationic [¹⁴C]12 was easily metabolized to neutral [¹⁴C]11 at 15 min after injection. However, most of [¹⁴C]11 was cleared from the brain at 45 min. Although [¹¹C]12 showed a higher BBB penetration (0.7 − 0.8 % ID/g at 15 min) than that of [¹⁴C]11, no regional difference was observed between the forebrain regions and cerebellum (Sasaki et al. 2001). Although [¹¹C]12 was not confirmed to be a useful PET ligand for NMDARs, this prodrug strategy can be applied to the development of in vivo imaging agents.

BIII277CL (Fig. 18.13), a potent anticonvulsant and neuroprotective agent, has been reported to have high affinity (K _i  = 4.5 nM vs [³H]MK-801) at the PCP-binding site of NMDARs (Grauert et al. 1997, 1998). Kokic et al. have developed a methyl analog of BIII277CL (13, Fig. 18.13) with substantial affinity for the PCP-binding site (K _i  = 49 nM vs [³H]TCP) as a candidate imaging agent for NMDARs. [¹¹C]13 was prepared by O-[¹¹C] methylation of BIII277CL with a specific activity of 30–60 GBq/μmol. PET studies of [¹¹C]13 in pigs revealed homogenous localization in the cortex and cerebellum. The kinetic analysis using one-tissue compartment model revealed that V _T of [¹¹C]13 was approximately 1 in both the cortex and cerebellum. These results indicated a lack of specific binding of [¹¹C]13 for NMDARs (Kokic et al. 2002).

NPS 1506 (Fig. 18.13) has been reported noncompetitive NMDAR antagonist with moderate affinity (IC₅₀ = 664 nM vs [³H]MK-801). The NPS 1506 is neuroprotective in a variety of animal models of stroke and head injury with no PCP-like psychotomimetic effects at doses between 1 and 5 mg/kg i.p. in rats (Mueller et al. 1999). Considering such a pharmacological profile, which is different from the previous NMDAR antagonists, NPS 1506 was labeled with carbon-11 and evaluated as a novel PET tracer for the PCP-binding site of NMDARs. The [¹¹C]NPS 1506 was prepared by N-[¹¹C] methylation of primary amine precursor with a specific activity of around 50 GBq/μmol. Biodistribution of [¹¹C]NPS 1506 in mice and rat demonstrated that uptake into the brain was rapid and occurred at high levels. However, the regional brain distribution was fairly uniform and did not clearly reflect the known localization of NMDARs in the rodent brain. In addition, treatment with unlabeled NPS 1506 did not cause any change in uptake in the mouse brain compared with the control group. An activator of NMDARs, 3-nitropropionic acid (20 mg/kg), did not produce any change in the regional uptake in the hippocampus nor the striatum. Therefore, [¹¹C]NPS 1506 may be an unsuitable in vivo tracer for NMDARs because of its large nonspecific binding fraction and low in vitro binding affinity for NMDARs (Fuchigami et al. 2003).

7-Chloro-5-iodokynurenic acid (14, Fig. 18.15) has been identified as a potent (IC₅₀ vs [³H]glycine binding, 32 nM) and selective antagonist for the glycine site of NMDARs (Leeson et al. 1991). The [¹²³I]14 was successfully synthesized by halogen exchange reaction of a mixture of 5-bromo-7-chloro- and 7-bromo-5-chlorokynurenic acid at a specific activity of 37 GBq/μmol. In the biodistribution studies in mice, [¹²³I]14 showed fast deiodination (85 % of parent compound was changed to ¹²³I⁻) and low brain uptake (0.61 % ID/g at 5 min). The [¹²³I]14 had been demonstrated to be an unsuitable in vivo tracer (Dumont and Slegers 1997).

GV150526A (Fig. 18.14) is one of the most potent glycine site antagonists (K _i  = 3.2 nM vs [³H]glycine) with reasonable lipophilicity (log P = 2.24) (Di Fabio et al. 1997). Based on the report, Waterhouse et al. developed a meta-methoxy derivative [¹¹C]3MPICA (Fig. 18.15) with high affinity for the glycine site (K _i  = 4.8 nM vs [³H]MDL 105,519). The [¹¹C]3MPICA was prepared by O-[¹¹C] methylation of the corresponding desmethyl derivative with a specific activity of 44 GBq/μmol (Waterhouse et al. 2002a). In the biodistribution studies in rats, regional uptake of [¹¹C]3MPICA at 2 min ranged from 0.11 to 0.18 % ID/g. The radioactivity cleared rapidly from all brain regions examined. Blocking studies using unlabeled compound (1 mg/kg) did not cause a marked reduction in the ratio of the brain tissues/blood. Treatment with warfarin (100 mg/kg) led to reduction in blood radioactivity concentrations while brain uptake was not changed significantly, which indicated that factors other than binding to serum albumin could prevent BBB penetration of [¹¹C]3MPICA (Waterhouse et al. 2002b).

Trans-5,7-dichloro-4-substituted-2-carboxytetrahydroquinolines such as 15 (Fig. 18.15) showed high affinities (K _i  = 6.0 nM vs [³H]5,7-dichlorokynurenic acid) for the glycine site of NMDARs (Leeson et al. 1992). Based on the SAR studies, fluoroethyl derivative [¹⁸F]16 (Fig. 18.15) was designed, synthesized, and evaluated as a PET radioligand for glycine site. The 16 had a relatively low lipophilicity (logD = 1.3) and high affinity (K _i  = 12 nM vs [³H]MDL 105,519) for the glycine-binding site. The radiosynthesis of [¹⁸F]16 was achieved by N-[¹⁸F]fluoroethylation of a piperidine precursor with 2-[¹⁸F]fluoroethyltosylate followed by hydrolysis of methyl ester. Biodistribution studies revealed that [¹⁸F]16 showed homogenous low levels of accumulation in all considered brain regions. Because the difference in regional brain distribution of [¹⁸F]16 could not be detected, no blocking experiments were performed (Piel et al. 2003). Recently, hydantoin-substituted indole carboxylic acids were developed as novel PET ligands for the glycine-binding site with a higher lipophilicity than 16. Compound 17 and 18 showed moderate affinity for the glycine site (K _i  = 53 nM for 17 and K _i  = 31 for 18 in [³H]MDL 105,519 binding assay). The logD values of 17 and 18 were 2.07 and 2.26, respectively. [¹⁸F]17 and [¹¹C]18 were successfully synthesized with overall decay-corrected radiochemical yields of 5–7 and 6–9 % and specific activities of 24–67 GBq/μmol and 8–26 GBq/μmol, respectively (Bauman et al. 2011). No further in vivo evaluations of these radioligands have been reported so far.

(_D)-7-Bromo-N-(1-phosphonoethyl)-5-aminomethylquinoxaline-2,3-dione (PAMQX, Fig. 18.16) has been discovered as a selective and high-affinity glycine site antagonist with an IC₅₀ value of 6 nM against [³H]MDL-105519 binding (Auberson et al. 1999). The 7-iodo derivative of PAMQX (19, Fig. 18.16) exhibited comparable affinity (IC₅₀ = 8 nM vs [³H]MDL-105,519) for the glycine site with PAMQX (Ametamey et al. 2000). Radiosynthesis of [¹³¹I]19 was achieved by iododestannylation of N-CBZ-protected tributylstannyl precursor followed by deprotection of CBZ group. Biodistribution studies of [¹²³I]19 in mice revealed an extremely low BBB permeability of [¹²³I]19 (0.03 % ID/g at 30 min) probably due to its high polarity and low log D value (logD _7.4 = −2.5) (Ametamey et al. 2000).

The radioligands consisting of carboxylic acids showed poor brain penetration due to the highly polar nature of the charged carboxylate. The 4-hydroxyquinolones (4-HQs)-devised carboxylic bioisostere had been developed as one of the most potent antagonists for the glycine-binding site of NMDARs. As expected, this series showed in vivo anticonvulsant activity, which suggested an ability to penetrate the BBB (Kulagowski et al. 1994; Rowley et al. 1997). In line with this, Morris et al. synthesized [¹¹C]L-703,717 (Fig. 18.17), one of the most potent glycine site antagonists (IC₅₀ = 4.5 nM vs [³H]L-689,560), by O-[¹¹C] methylation of phenol precursor using [¹¹C]CH₃I containing K₂CO₃/Kryptofix 2.2.2 as a base, but the radiochemical yield was very low (7 %) (Morris et al. 1997). The labeling precursor has three possible sites (amide, enol, and phenol) which could be reacted with [¹¹C]CH₃I. Therefore, regioselective methylation at the phenol position was required to prepare [¹¹C]L-703,717 for high yield and purity. Haradahira et al. succeeded in the synthesis of [¹¹C]L-703,717 by using 5 eq of NaH as a base in high yields (87 %) with a specific activity of 47–53 GBq/μmol. The excess amount of NaH can give the triply deprotonated species, then the most accessible terminal phenol can be preferentially methylated (Haradahira et al. 1999). In the biodistribution study in mice, [¹¹C]L-703,717 showed low BBB permeability (0.32 − 0.36 % ID/g at 1 min) and extremely high blood concentration presumably due to high affinity for the warfarin-binding site of plasma protein. Indeed, warfarin led to a dose-dependent enhancement of the initial brain uptake of [¹¹C]L-703,717 that resulted in a fivefold increase in the brain radioactivity at the dose of 200 mg of warfarin. However, the radioactivity in the cerebrum was rapidly decreased (0.2 % ID/g) with some retention observed in the cerebellum (0.65 % ID/g) at 30 min after injection. This suggested inconsistent binding with known distribution of NMDARs. In addition, a significant blocking effect by nonradioactive L-703,717 (2 mg/kg) was observed only in the cerebellum (Haradahira et al. 2000). To improve brain penetration of [¹¹C] L-703,717, acetyl derivative of L-703,717 ([¹¹C]20, Fig. 18.17) was prepared and evaluated as prodrug of [¹¹C]L-703,717. Initial brain uptake of [¹¹C]20 (0.71 % ID/g) was twofold higher than that of [¹¹C]L-703,717 (0.36 %ID/g). Metabolism experiments of [¹¹C]20 in rat brain homogenates confirmed that about 80 % of [¹¹C]20 was converted to [¹¹C]L-703,717 at 20 min after injection. Ex vivo autoradiography demonstrated that [¹¹C]20 exhibited similar cerebellar localization of radioactivity to [¹¹C]20 (Haradahira et al. 2001). Further in vivo evaluation revealed that the specific cerebellar binding of [¹¹C]L-703,717 was abolished in the NR2C-deficient mice as shown in Fig. 18.18a–c. This result suggested that the in vivo cerebellar localization of [¹¹C]L-703,717 could be caused by preferential binding to the NR2C subunit-containing NMDARs predominantly expressed in the cerebellum. However, [¹¹C]L-703,717 exhibited high specific localization in the hippocampus and cerebral cortex under in vitro condition as shown in Fig. 18.18d–f, indicative of the absence of intrinsic selectivity for the NR2C subunit-containing NMDARs in the cerebellum (Haradahira et al. 2002b). Thus, the authors next hypothesized that endogenous NMDAR co-agonists might play a key role in the unusual localization in the cerebellum of [¹¹C]L-703,717. Both glycine and d-serine are known as co-agonists of NMDARs and exist in micromolar level. Glycine is distributed throughout the brain, while d-serine is localized only in the forebrain regions (Hashimoto et al. 1993; Schell et al. 1997). d-serine is reported to be nearly undetectable in cerebellum due to the highest level of d-amino acid oxidase (DAO) activity (Schell et al. 1995). Haradahira et al. demonstrated that the cerebellar localization of [¹¹C]L-703,717 was abolished in mutant ddY/DAO^– mice, in which the d-serine level in cerebellum was markedly increased (Fig. 18.19). Hence, the authors proposed that the low level of [¹¹C]L-703,717 binding in forebrain regions could be explained by the strong inhibition of high amount of endogenous d-serine; whereas, specific localization in the cerebellum might be due to the lack of d-serine (Haradahira et al. 2003). Considering that [¹¹C]L-703,717 has been demonstrated to recognize cerebellar NMDARs to some extent, clinical PET studies of [¹¹C]20 (a prodrug of [¹¹C] L-703,717) were performed for the visualization of NMDARs in the cerebellum (Matsumoto et al. 2007). Binding potential (BP) calculated from white matter as a reference region in the cerebellar cortex was twofold higher than in the cerebral cortices (cerebellar cortex; BP = 2.2, cerebral cortices; BP = 1.1). Regional brain distributions were consistent with previous reports of rodents (Haradahira et al. 2001). However, the brain uptake of [¹¹C]20 was too low for the visualization of the cerebellar NMDAs by PET (Matsumoto et al. 2007).