Synthesis and Biological Evaluation of Enantiomerically Pure (R)- and (S)-[18F]OF-NB1 for Imaging the GluN2B Subunit-Containing NMDA receptors

GluN2B subunit-containing N-methyl-d-aspartate (NMDA) receptors have been implicated in various neurological disorders. Nonetheless, a validated fluorine-18 labeled positron emission tomography (PET) ligand for GluN2B imaging in the living human brain is currently lacking. As part of our PET ligand development program, we have recently reported on the preclinical evaluation of [18F]OF-NB1 – a GluN2B PET ligand with promising attributes for potential clinical translation. However, the further development of [18F]OF-NB1 is currently precluded by major limitations in the radiolabeling procedure. These limitations include the use of highly corrosive reactants and racemization during the radiosynthesis. As such, the aim of this study was to develop a synthetic approach that allows an enantiomerically pure radiosynthesis of (R)-[18F]OF-NB1 and (S)-[18F]OF-NB1, as well as to assess their in vitro and in vivo performance characteristics for imaging the GluN2B subunit-containing NMDA receptor in rodents. A two-step radiosynthesis involving radiofluorination of the boronic acid pinacol ester, followed by coupling to the 3-benzazepine core structure via reductive amination was employed. The new synthetic approach yielded enantiomerically pure (R)-[18F]OF-NB1 and (S)-[18F]OF-NB1, while concurrently circumventing the use of corrosive reactants. In vitro autoradiograms with mouse and rat brain sections revealed a higher selectivity of (R)-[18F]OF-NB1 over (S)-[18F]OFNB1 for GluN2B-rich brain regions. In concert with these observations, blockade studies with commercially available GluN2B antagonist, CP101606, showed a significant signal reduction, which was more pronounced for (R)-[18F]OF-NB1 than for (S)-[18F]OF-NB1. Conversely, blockade experiments with sigma2 ligand, FA10, did not result in a significant reduction of tracer binding for both enantiomers. PET imaging experiments with CD1 mice revealed a higher brain uptake and retention for (R)-[18F]OF-NB1, as assessed by visual inspection and volumes of distribution from Logan graphical analyses. In vivo blocking experiments with sigma2 ligand, FA10, did not result in a significant reduction of the brain signal for both enantiomers, thus corroborating the selectivity over sigma2 receptors. In conclusion, we have developed a novel synthetic approach that is suitable for upscale to human use and allows the enantiomerically pure radiosynthesis of (R)-[18F]OF-NB1 and (S)-[18F]OF-NB1. While both enantiomers were selective over sigma2 receptors in vitro and in vivo, (R)-[18F]OF-NB1 showed superior GluN2B subunit specificity by in vitro autoradiography and higher volumes of distribution in small animal PET studies.

Introduction N-methyl-d-aspartate (NMDA) receptors are ligand-gated ion channels that belong to the family of ionotropic glutamate receptors (iGluRs). Endowed with a remarkable variety of biological functions, NMDA receptors constitute heterotetrameric complexes composed of combinations of the subunits GluN1, which is processed in eight distinct splice variants, GluN2A-D, and GluN3A-B. [1][2][3][4] Typically, a functional NMDA receptor comprises two glycine-binding GluN1 subunits and at least one glutamatebinding GluN2 subunit. Simultaneous binding of glycine and glutamate initiates NMDA receptor activation, which involves voltage-dependent relief of magnesium blockade, depolarization of the postsynaptic membrane and calcium ion in ux. [5][6][7] While NMDA receptors are key players in neurophysiology, contributing to memory and learning via modulation of synaptic plasticity, the GluN2B subunit-carrying NMDA receptor has been implicated in the pathophysiology of various neurological disorders. [8][9][10][11][12][13][14][15][16] Indeed, the role of overstimulation of the excitatory GluN2B subunit in the development of several CNS-related pathologies has been corroborated, 17,18 whereas targeting GluN2B-mediated excitotoxicity has been suggested as a promising therapeutic strategy for various diseases, including Alzheimer's disease (AD), Parkinson's disease (PD), ischemic stroke, traumatic brain injury, neuropathic pain and depression. [19][20][21][22][23][24][25][26][27][28][29][30] Early efforts to develop NMDA receptor antagonists prompted the discovery of NMDA receptor channel blockers such as phencyclidine (PCP), thienylcyclohexylpiperidine (TCP), ketamine, memantine, and MK-801 (dizocilpine). Despite their well-documented therapeutic e cacy, most of these "broad-spectrum" antagonists were associated with a poor safety pro le, potentially owing to the lack of subunit-selectivity. 17,31,32 As such, more recent attempts have focused on the development of GluN2B-selective antagonists, which has become feasible since the discovery of the N-terminal domain (NTD) binding site that is located at the interface between GluN1 and GluN2B. 33 Several GluN2Bselective antagonists have been reported to date -some of which have been advanced to humans, including CP101,606 (traxoprodil) 30 , MK-0657 (CERC-301) 34 and EVT-101 (NCT01128452). Nonetheless, the development of a suitable GluN2B-selective antagonist for clinical use has proven challenging, at least in part, due to the lack of appropriate non-invasive imaging tools that allow the assessment of target engagement in the human brain.
Positron emission tomography (PET) constitutes a powerful non-invasive molecular imaging modality that allows real-time quanti cation of biochemical processes. 35 Accordingly, PET has been established as a reliable tool for CNS-targeted receptor quanti cation as well as target occupancy studies in preclinical and clinical research. 36 Given the translational relevance of visualizing drug-receptor interactions to facilitate the development of GluN2B antagonists in the pipeline, strenuous efforts have been devoted to the discovery of a suitable GluN2B PET radioligand in the past two decades. While numerous probes exhibited high in vitro speci city and selectivity towards the GluN2B subunit, the vast majority of reported ligands were plagued by unfavourable in vivo performance characteristics. 37 Major drawbacks included low brain penetration, lack of in vivo speci city and selectivity, as well as the presence of radiometabolites in the CNS. 37,38 We have recently reported on the rst successful GluN2B subunit-selective PET radioligand, (R)-[ 11 C]Me-NB1, that proved to be suitable for visualizing GluN2B in vitro and in vivo. 39 The structure of (R)-[ 11 C]Me-NB1 belongs to a class of 3-benzazepine based ligands, encompassing a series of high-a nity GluN2B antagonists, that were rst reported by Tewes et al. 40 Of note, (R)-[ 11 C]Me-NB1 was successfully translated to humans, rendering it the rst and only GluN2Btargeted PET radioligand to be clinically validated to date. 41 Despite the outstanding performance characteristics, the use of (R)-[ 11 C]Me-NB1 is limited by the short physical half-life of carbon-11 (20.3 min), which con nes the use of (R)-[ 11 C]Me-NB1 to facilities with an on-site cyclotron. As part of our efforts to develop a suitable radio uorinated analog (physical half-life of uorine-18, 109.8 min), that would allow satellite distribution to hospitals without an on-site cyclotron, we have synthesized and evaluated a series of uorinated (R)-[ 11 C]Me-NB1 derivatives 42-49 -of which [ 18 F]OF-NB1 proved to be particularly promising for translation into humans. However, the clinical translation of [ 18 F]OF-NB1 requires further optimization of the radiolabeling strategy. In particular, current radiolabeling routes utilize corrosive reactants such as boron tribromide and lead to racemization of enantiomerically pure precursors. However, based on previous observations with this class of compounds 41 , it is anticipated that (R)-and (S)-[ 18 F]OF-NB1 may exhibit distinct enantiomeric behaviors with respect to GluN2B binding speci city and selectivity over sigma receptors. Along this line of reasoning, a synthetic approach that allows the enantiomerically pure synthesis of (R)-[ 18 F]OF-NB1 and (S)-[ 18 F]OF-NB1 without racemization is warranted to enable the evaluation of the enantiomers in future human studies. Thus, the aim of this study was to develop a synthesis strategy that is devoid of corrosive reactants and provides enantiomerically pure (R)-[ 18 F]OF-NB1 and (S)-[ 18 F]OF-NB1 suitable for human use, as well as to assess their in vitro and in vivo performance characteristics for imaging the GluN2B subunits of the NMDA receptor in rodents.

Results And Discussion
While [ 18 F]OF-NB1 exhibited outstanding performance characteristics as a GluN2B subunit-targeted PET radioligand in preclinical experiments, the original radiosynthesis was plagued by the harsh conditions required to cleave the hydroxyl protecting groups, hampering automatization and clinical translation of the probe (Scheme 1, previous work). 50 Moreover, when employing enantiomerically pure precursor 1, racemization of the center of chirality in benzyl position was observed during the reaction with boron tribromide, precluding the synthesis of enantiomerically pure (R)-or (S)-[ 18 F]OF-NB1. 49 We have now developed a novel synthesis strategy (Scheme 1, this work) that proceeds via a building block approach, thereby opening up several possibilities: (1) precursor 3 for the radiosynthesis does not contain any OHgroups, which alleviated the necessity of protection groups for the radio uorination. (2) Starting from enantiomerically pure 3-benzazepine building block 5 allowed us to conduct radiolabeling of enantiomerically pure of (R)-[ 18 F]OF-NB1 and (S)-[ 18 F]OF-NB1. (3) Additionally, the novel strategy was faster and more effective compared to the linear two-step synthesis, since a chiral HPLC to separate the enantiomers at the end of the radiosynthesis was no longer required. (4) Finally, given the rapid nature of reductive aminations, our proposed building block approach is adoptable to a wide scope of substrates, including those containing functional groups that are not tolerated under nucleophilic radio uorination conditions.
The concept of building block-based radiochemistry is well established in uorine-18 chemistry for a variety of reactions 51,52 , including reductive alkylations with uorinated benzaldehyde, 53 however, this is the rst successful example of a radiosynthesis that leverages a radio uorinated aliphatic aldehyde intermediate for subsequent reductive alkylation. The reductive alkylation reaction is chemoselective towards amines over alcohols, as opposed to nucleophilic substitution of an alkyl halide, where overalkylation of tertiary amines and alkylation of alcohols typically occur as side reactions. Boronic acid pinacol ester 3 was used as precursor for the radiosynthesis, as the copper-mediated radio uorination of arylboronic esters constitutes a reliable and versatile labeling strategy. 54 Arylboronic esters as precursors can be readily obtained by Miyaura-borylation of aryl halides in one step. 55 In contrast to other established radio-precursors, oxidative conditions are not required (cf. diaryliodonium salts or iodonium ylides). 56, 57 Another advantage of copper-mediated radio uorination of boronic acid esters is that regioselectivity issues during uorine-18 incorporation are not typically observed, as opposed to the radio uorination of diaryliodonium salts. 58 Boronic acid pinacol ester 3 (Scheme 1), which served as precursor for the radiolabeling, was obtained in a multistep synthesis, starting from commercially available 1iodo2halobenzenes 6a-c. A Heck reaction 59 with but-3-en-1-ol, followed by isomerization, 60, 61 gave aldehydes 7a-c in variable yields and purity (Scheme 2). Indeed, only in the case of 1,2diidodobenzene, the reaction led to pure 4(2iodophenyl)butanal (7a). In sharp contrast, the bromo and uoro analogues, 6b and 6c, yielded inseparable mixtures of 4(2halophenyl)butanal (7b and 7c) and undesired regioisomers 3(2haloophenyl)butanal (8b and8c, 10-11 %, from 1 H NMR spectra analysis). Thus, despite the low yeld of 11 %, the synthesis was continued with 7a. Aldehyde 7c was used as a non-radioactive reference for HPLC method development.
Protection of the aldehyde as an acetal (9) 62 was necessary, as the Miyaura-borylation gave only trace amounts of precursor 3, when the reaction was performed directly with aldehyde 7a (Scheme 3). With acetal 9, the borylation 63 proceeded in 48% yield, followed by hydrolysis 64 to the pinacol precursor 3 in 91%. With an open-chain diethyl acetal, instead of cyclic ethylene acetal, the hydrolysis of the diethyl acetal required only very mild conditions using catalytic iodine in acetone. Thus, the diethyl acetal 9 ensured the successful Miyaura-borylation and allowed the facile release of the aldehyde functionality, while leaving the boronic ester intact. Accordingly, precursor 3 for the radiosynthesis was synthesized from 1,2diiodobenzene in four steps in an overall yield of %.
The racemic 3-benzazepine building block (rac)-5 was obtained from commercially available 3benzazepine 11 by cleavage of the benzyl ether via catalytic hydrogenation in 91% yield (Scheme 3). Chiral resolution of 3-benzazepine (rac)-11 or (rac)-5 to obtain enantiomerically pure 3-benzazepines (R)-5 and (S)-5 by chiral HPLC was not successful, using both normal and reversed stationary phases and various eluent combinations. Based on previous experience with this class of compounds, chiral resolution was generally possible for tertiary amines bearing a bulky lipophilic substituent. 39 Therefore, 3-benzazepine 11 was reductively N-alkylated with benzaldehyde and NaBH(OAc) 3 to give the tertiary amine (rac)-12 in 99% yield. It is worthwhile mentioning that the benzyl group was selected to enable simultaneous cleavage of both, the benzyl ether and benzylamine after chiral resolution. Indeed, benzylated 3-benzazepines (R)-12 and (S)12 were successfully separated by chiral HPLC and the enantiomerically pure benzazepine building blocks (R)-5 and (S)-5 were obtained by catalytic hydrogenation of (R)-12 and (S)-12, respectively.
The radiolabeling was performed using a two-step procedure involving copper-mediated nucleophilic radio uorination of precursor 3, 54 Table 1. It should be noted that attempts to perform the reductive amination prior to the copper-mediated uorine-18 labeling did not yield the desired product, potentially owing to the interference of the two alcohol groups (Supporting Information). In vitro autoradiography with rodent brain tissue GluN2B subunit-speci city and selectivity of (R)-[ 18 F]OF-NB1 and (S)-[ 18 F]OF-NB1 were assessed by in vitro autoradiography using mouse and rat brain tissue sections. In accordance with reported GluN2B expression patterns in the adult mammalian brain, 65 a high tracer binding was observed in GluN2B-rich forebrain regions such as the hippocampus, striatum, thalamus and cortex, whereas tracer binding was relatively low in the GluN2B-de cient cerebellum ( Fig. 1). Although the latter observations were generally made for both enantiomers, (R)-[ 18 F]OF-NB1 exhibited a more favorable binding pattern in the rodent brain ( Fig. 1A), which was superior to that of (S)-[ 18 F]OF-NB1 ( Fig. 1B) with respect to selectivity for GluN2B-rich brain areas. In concert with these observations, blocking studies with the commercially available GluN2B antagonist, CP101,606 (K D of 10 nM towards GluN2B), revealed a more pronounced reduction of tracer binding for (R)-[ 18 F]OF-NB1 as compared to (S)-[ 18 F]OF-NB1 on mouse and rat brain sections.
One of the major drawbacks of previously reported GluN2B PET radioligands was off-target binding towards sigma receptors. 39 While we have previously demonstrated the in vitro and in vivo selectivity of (rac)-[ 18 F]OF-NB1 over sigma1 receptors; 44 it remains unclear whether off-target activity towards sigma2 receptors can also be excluded. As such, we performed additional autoradiography studies by with the previously reported sigma2 ligand, 11b 66 (here codenamed FA10, K i of 1.4 nM towards sigma2). Overall, no considerable signal reduction was observed for either (R)-[ 18 F]OF-NB1 or (S)-[ 18 F]OF-NB1 when employing an excess of FA10 (10 µM), indicating that both enantiomers exhibited selectivity over sigma2 receptors (Fig. 1).
Quanti cation of the autoradiographic data corroborated that highest (R)-[ 18 F]OF-NB1 binding was observed in the hippocampus, followed by the cortex, striatum and thalamus, whereas the cerebellum showed lowest (R)-[ 18 F]OF-NB1 binding ( Fig. 2A) -with a hippocampus-to-cerebellum ratio of ≈ 20.
Although similar quanti cation patterns were obtained for (S)-[ 18 F]OF-NB1 (Fig. 2B), the hippocampus-tocerebellum ratio was ≈ 5, indicating that the selectivity for the GluN2B-rich forebrain was less accentuated. While GluN2B blockade studies with CP101606 showed a signi cant signal reduction, the whole brain, as well as in GluN2B-rich brain regions (Fig. 4A). These ndings indicated a higher retention of the R-enantiomer in the rodent brain. Notable, blockade experiments with sigma2 ligand, FA10 (1 mg/kg), did not result in a signi cant reduction of the PET signal in the hippocampus (Fig. 4B), or in any other brain region (data not shown), indicating that both enantiomers were selective over sigma2 receptors in vivo.
In vitro autoradiography with non-human primate brain sections Due to the superior performance characteristics of (R)-[ 18 F]OF-NB1 in rodent studies, we sought to assess its utility for imaging GluN2B subunit-containing NMDA receptors in higher species. As such, post-mortem brain tissue sections from non-human primates (NHPs) were used for autoradiographic testing of (R)- In accordance with observations from rodent studies, we found a heterogenous binding pattern with preferential accumulation of (R)-[ 18 F]OF-NB1 in GluN2B-rich brain regions (Fig. 5). Employing GluN2B antagonist, CP101,606, a high degree of speci city (80.7% signal reduction) was corroborated. In contrast, blockade studies with FA10 did not reveal a signi cant reduction of the signal, implying that (R)-

Chemistry
General procedure for the synthesis of 4-(2halophenyl)butanal 7a-7c A ame dried Schlenk ask was charged with Pd(OAc) 2 (6 mol%), tetrabutylammonium bromide (1.00 eq.), NaHCO 3 (2.50 eq.) and molecular sieves (4 Å). The air atmosphere was exchanged by nitrogen in three cycles of evacuation and ushing with N 2 . The reagents were suspended in DMF (dry), the iodohaloaryl compound (6a-6c, 1.00 eq.) was dissolved in DMF (dry) and added to the mixture. But-3-en1ol was added and the mixture was stirred at 70°C for 4 h. After cooling down, ethyl acetate was added, and the mixture was ltered through a pad of Celite®. It was washed with water and the aqueous layer was extracted with ethyl acetate, the organic layers were combined and concentrated in vacuo. This process was repeated two more times. The organic layer was dried with Na 2 SO 4 and evaporated in vacuo. The crude product was puri ed via ash column chromatography to yield product 7a-c.
The mixture was diluted with ethyl acetate (20 mL) and washed with half concentrated aqueous Na 2 SO 3 solution (20 mL). The aqueous layer was extracted with ethyl acetate (20 mL) and the combined organic layers were dried over Na 2 SO 4 and the solvent was removed in vacuo. The residue was puri ed via ash column chromatography (hexanes/ethyl acetate = 97:3 → 94:6). Yellow oil, yield 172 mg (628 µmol, 91%).
TLC: 0.38 (hexanes/ethyl acetate = 9:1). 1  ( rac )-2,3,4,5-Tetrahydro-1 H -3-benzazepine-1,7-diol (( rac )-5) A ask was charged with 3-benzazepine 11 (500 mg, 1.86 mmol, 1.00 eq.) and Pd/C (200 mg, 10 wt%) and THF (25 mL) was added. The air atmosphere was exchanged with a hydrogen atmosphere, by ushing the ask with H 2 for 10 min. A balloon with H 2 was connected to the ask and the reaction mixture was stirred at 60°C overnight. After cooling down, the mixture was ltered over Celite® and the lter was extracted with MeOH (6 × 30 mL). The solvent was removed in vacuo. Beige solid, mp 173°C (R)-12 (39.8 mg, 111 µmol) was dissolved in THF (5 mL, dry) and Pd/C (10 mg, 10 wt%) was added. The air atmosphere was exchanged with a hydrogen atmosphere, by ushing the ask with H 2 for 10 min. A balloon with H 2 was connected to the ask and the reaction mixture was stirred at 60°C overnight. After cooling down, the solvent was removed in vacuo and the residue was suspended in CH 3 OH (10 mL) and passed through a syringe lter. The solvent was removed under reduced pressure and the residue was suspended in ethyl acetate (5 mL) and passed through a pad of cotton and washed with ethyl acetate.
The residue on the lter was dissolved in CH 3 OH and passed through the lter. The solvent was removed in vacuo. Colorless solid, yield 9.7 mg (54 µmol, 49%). The analytical data are in agreement with the data of (rac)-5.   Table 1.
Conditions for the single step radiosynthesis approach can be found in the Supporting Information.

I n vitro autoradiography
In vitro autoradiography studies were performed as previously described, however, with minor modi cations. 42 Tissue-TEK (O.C.T.) was utilized to embed rodent and NHP postmortem brain tissue, which was subsequently prepared as 20 µm thick tissue sections on a cryostat and mounted on glass slides. The slides were then stored at -80°C until the time of utilization. Brain sections were initially thawed for 10 min on ice prior to in vitro autoradiography experiments. They were then preconditioned for 10 min in the assay buffer (pH 7.4) composed of 30 mM HEPES, 0.56 mM MgCl 2 , 110 mM NaCl, 5 mM KCl, 3.3 mM CaCl 2 and 1% fatty acid-free bovine serum albumin (BSA) at ambient temperature. The tissue sections were then dried and subsequently incubated for 30 min at room temperature with (R)-or (S)-[ 18 F]OF-NB1 solution, respectively. For blockade conditions, 10 µM of the respective blocker were added. These blockers included CP101606 (GluN2B ligand) and FA10 (sigma2 ligand). After incubation, the brain sections were washed in assay buffer for 5 minutes followed by washing buffer (same as assay buffer but without BSA) for 2 × 2 min. They were then dipped twice in distilled water for 5 seconds, subsequently dried and exposed to a phosphor imager plate for 180 min. The plates were scanned and on an Amersham Typhoon scanner, whereas ImageQuant TL 8.1 and ImageJ v1.53e were utilized for image analyses.
PET imaging PET imaging studies were performed under Institutional Animal Care and Use Committee (IACUC) guidelines. Female CD-1 mice (10-12 weeks of age) were kept under a 12-h light/12-h dark cycle, with ad libitum access to food and water. On the day of experiment, mice were scanned using a G8 PET scanner (So e) under 1-2% iso urane in air/oxygen 1:1 anesthesia. Body temperature was monitored and maintained by a heating pad installed in the scanner bed. The tracer solution containing 0.6-1. GluN2B antagonist, CP101,606 (10 µM). FA10 (10 µM) was used to assess selectivity over sigma2 receptors.

Figure 2
Quanti cation of autoradiographic data from rat brain sections. A. Distribution of (R)-[ 18 F]OF-NB1 throughout distinct brain regions. GluN2B subunit-speci city was derived from the extent of signal reduction from baseline to CP (CP101,606) blockade. Selectivity over sigma2 receptors was assessed by the extent of signal reduction from baseline to FA10 blockade. B. Distribution of (S)-[ 18 F]OF-NB1 throughout distinct brain regions. GluN2B subunit-speci city was derived from the extent of signal reduction from baseline to CP (CP101,606) blockade. Selectivity over sigma2 receptors was assessed by the extent of signal reduction from baseline to FA10 blockade.