Comparison of three novel radiotracers for GluN2B-containing NMDA receptors in non-human primates: (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]of-Me-NB1, and (S)-[¹⁸F]of-NB1

Abstract

The NMDA receptor GluN2B subunit is a target of interest in neuropsychiatric disorders but to date there is no selective radiotracer available to quantify its availability in vivo. Here we report direct comparisons in non-human primates of three GluN2B-targeting radioligands: (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1. Plasma free fraction, metabolism, tissue distribution and kinetics, and quantitative kinetic modeling methods and parameters were evaluated in two adult rhesus macaques. Free fraction in plasma was <2% for (R)-[¹¹C]NR2B-Me and (R)-[¹⁸F]OF-Me-NB1 and higher for (S)-[¹⁸F]OF-NB1 (15%). All radiotracers showed good brain uptake and distribution throughout grey matter, with substantial (>68%) blockade across the brain by the GluN2B-targeting drug Co-101,244 (0.25 mg/kg), including in the cerebellum. Time-activity curves were well-fitted by the one-tissue compartment model, with volume of distribution values of 20–40 mL/cm³ for (R)-[¹¹C]NR2B-Me, 8–16 mL/cm³ for (R)-[¹⁸F]OF-Me-NB1, and 15–35 mL/cm³ for (S)-[¹⁸F]OF-NB1. Estimates of regional non-displaceable binding potential were in the range of 2–3 for (R)-[¹¹C]NR2B-Me and (S)-[¹⁸F]-OF-NB1, and 0.5-1 for (R)-[¹⁸F]OF-Me-NB1. Altogether, each radiotracer showed an acceptable profile for quantitative imaging of GluN2B. (S)-[¹⁸F]OF-NB1 has particularly promising imaging characteristics for potential translation into humans. However, the source of unexpected displaceable binding in the cerebellum for each of these compounds requires further investigation.

Introduction

The glutamatergic N-methyl-D-aspartate (NMDA) receptor complex is a ligand- and voltage-gated ion channel expressed ubiquitously across the central nervous system. NMDA receptors modulate synaptic activity, plasticity, and synaptogenesis, and play a crucial role both in maintaining normal neural activity and in learning and memory processes. As such, these receptors are targets of interest for the treatment of a range of neuropsychiatric disorders. For example, mitigating excitotoxic neural damage induced by NMDA dysfunction can improve outcomes in stroke patients, while NMDA-related synaptic plasticity processes are implicated in dementia and depression.^1–3

NMDA receptors are tetrameric complexes of subunits of the types GluN1, GluN2 (subtypes GluN 2A–2D) or GluN3 (subtypes GluN3A-B).^4,5 Subunit composition determines receptor activation properties and ion channel permeability, and varies with brain region and developmental stage. GluN2B-containing receptors have distinct and well-studied effects on long-term synaptic plasticity,^5,6 learning and memory,^7,8 and have been specifically linked to a range of disorders in preclinical genetic and pharmacological studies. ⁵ In addition to the ion pore and glutamate or glycine binding sites, a number of allosteric binding sites exist on the GluN2B subunit’s amino terminal domain which can modulate the likelihood of ligand binding or channel opening. ⁹ These sites are appealing targets for therapeutic drugs as they can be leveraged to alter receptor activation patterns with fewer side effects compared to direct channel blockade or orthosteric antagonism.^4,10 For example, the GluN2B allosteric site at which the drug ifenprodil binds is particularly well-characterized. ⁹ GluN2B-targeting allosteric modulators have been tested or proposed for the treatment of neurodegenerative disorders, neuropathic pain, schizophrenia, and depression.^5,11–13

Despite extensive preclinical literature demonstrating the importance of GluN2B-containing NMDA receptors in fundamental and disease-related processes, there are few tools available to study these sites in humans. Quantification of GluN2B-containing NMDA receptors selectively in vivo using positron emission tomography (PET) can advance research into the specific role of this receptor subunit in neuropsychiatric disorders as well as the development of drugs targeting the GluN2B site. ¹⁴ Accordingly, there are long-standing efforts to develop a GluN2B-specific PET radiotracer, but to date only one radioligand, (R)-[¹¹C]Me-NB1, has advanced to human use. ¹⁵ Recently, our group and others have identified several promising candidates.^13,16–21 The compounds (R)-[¹⁸F]OF-Me-NB1 and [¹⁸F]OF-NB1 were developed to target the GluN2B ifenprodil binding site and evaluated in vitro and in vivo in rodent models, showing good brain uptake, displacement by GluN2B-specific drugs, and no evidence of brain-penetrant radiometabolites.^13,20 Sensitivity to blocking by drugs targeting the structurally similar σ receptor site was also evaluated. In parallel, Cai et al. ¹⁷ described synthesis and evaluation of (R)-[¹¹C]NR2B-Me in rodents. Here, we present a detailed comparison of the imaging properties, kinetic modeling methods, quantitative parameters, and evidence for selectivity of (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1 (Figure 1) in non-human primates.

Figure 1.

Structural diagrams and brain distribution in non-human primate of GluN2B-targeting radiotracers. Top, chemical structures of (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1. Middle and bottom rows, representative mean PET SUV images in template space from 60–90 minutes at baseline or following pre-treatment with the GluN2B antagonist Co-101,244 (0.25 mg/kg).

Methods

Overview

Detailed descriptions of radiochemistry, metabolite analysis, and initial non-human primate imaging results for [¹⁸F]OF-Me-NB1 and [¹⁸F]OF-NB1 are described in separate manuscripts.^22,23 Here, we present direct comparisons of imaging characteristics, kinetic modeling methods, and blocking studies for (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1. A minimum of three bolus injection scans were performed in each of two animals for each of the three radiotracers, comprising scans at baseline and after pre-blocking with the GluN2B antagonist Co-101,244 ²⁴ (0.25 mg/kg) or the σ₁ receptor antagonist FTC-146 (0.027 or 0.125 mg/kg). ²⁵ An additional scan was performed during bolus plus constant infusion of (R)-[¹⁸F]OF-Me-NB1 with sequential displacement with both Co-101,244 and FTC-146. Further, a ketamine challenge scan was performed with (S)-[¹⁸F]OF-NB1, for a total of 20 scans. All procedures were approved by the Yale University Institutional Animal Care and Use Committee under guidelines consistent with the U.S. Government Principles for the Utilization and Care of Vertebrate Animals Used in Testing, Research and Training, the Animal Welfare Act, and Animal Welfare Regulations. Animal data are reported in accordance with the ARRIVE 2.0 guidelines. ²⁶

PET scan procedures

Two rhesus monkeys (Macaca mulatta) were used in this study (animal 1: male, 11 years; animal 2: female, 12 years). Prior to each scan, animals were sedated with a combination of alfaxalone (2 mg/kg), midazolam (0.3 mg/kg) and dexmedetomidine (0.01 mg/kg) and maintained in an anesthetized state with 1.5–2.5% isoflurane. Heart rate, blood pressure, respiration rate, oxygen saturation and respiration rate were monitored continuously. A catheter was placed in the radial artery for blood sampling.

Radiosyntheses were performed as previously described for (R)-[¹¹C]NR2B-Me,^17,27 (R)-[¹⁸F]OF-Me-NB1,^20,22 and (S)-[¹⁸F]OF-NB1.^13,23 All PET scans were performed on the Focus 220 scanner (Siemens Medical Solutions, Knoxville, TN) after radiotracer injection. A transmission scan was collected before tracer injection for attenuation correction. For bolus injection scans, the tracer was then injected in 10 mL over 3 minutes by an infusion pump (PHD 22/2000; Harvard Apparatus). Dynamic PET data were acquired in list mode for at least 120 min and binned into frames of increasing durations (bolus injections, 6 × 30 s, 3 × 1 min, 2 × 2 min, 22 × 5 min; bolus plus infusion, 6 × 30 s, 3 × 1 min, 2 × 2 min, 61 × 5 min).

Blocking and challenge studies

For GluN2B and σ₁ receptor pre-blocking studies with all radiotracers, Co-101,244 (0.25 mg/kg) or FTC-146 (0.027 or 0.125 mg/kg) was administered as a 10-minute IV slow bolus beginning 10-15 min prior to radiotracer injection, followed by a 120 min scan as described above. An additional single-scan sequential displacement experiment was performed using (R)-[¹⁸F]OF-Me-NB1 administered as a bolus plus constant infusion. A 315 min scan was acquired beginning concurrent with the start of infusion of 330.8 MBq (R)-[¹⁸F]OF-Me-NB1 with a K_bol of 80 min (that is, a bolus dose equivalent to 80 min of infusion), ²⁸ with the remaining dose administered as a constant infusion over the rest of the scan. Timing of infusion and competitor drug administration were determined based on simulations using (R)-[¹⁸F]OF-Me-NB1 bolus injection data from the same animal. When tissue concentration reached a steady state in most brain regions (75 min after the beginning of tracer infusion), Co-101,244 was administered (0.75 mg/kg, 10-min IV infusion). After a new steady-state concentration was established under the blocking condition, FTC-146 was added (0.125 mg/kg, 10-min IV infusion) beginning at 225 min after scan start (150 min after Co-101,244 administration) and scanning continued until 315 min.

Finally, a ketamine pre-blocking study was performed to assess sensitivity to this NMDA- and glutamate-modulating drug. The animal was anesthetized as described and then administered ketamine HCl at a dose of 1 mg/kg as a constant infusion over 40 minutes. This subanesthetic dose was chosen given its clinical relevance in depression treatment^29–31 and has been shown to modulate neural function in anesthetized non-human primates. ³² At the end of the ketamine infusion, 168.7 MBq (S)-[¹⁸F]OF-NB1 was injected as a slow bolus and a 120 min scan was performed as described above. Additional blood samples were taken at intervals during ketamine infusion and throughout the PET scan for measurement of drug concentration in plasma by mass spectrometry.

Plasma radiometabolite analysis and input function measurement

Arterial blood samples were collected at intervals post-injection to measure plasma input functions. Additional arterial blood samples were collected at 3, 8, 15, 30, 60 and 90 min after tracer injection with additional samples at 120, 180, 240, and 300 min during the bolus plus infusion scan. Samples were processed and analyzed for unmetabolized parent fraction using an automatic column-switching high-performance liquid chromatography system. ³³ The plasma free fraction (f_p) for each scan was measured in triplicate using the ultrafiltration method. Briefly, the tracer solution was added to 3.0 mL of whole blood. After 10 min incubation at ambient temperature, the sample was centrifuged at 2930 g for 5 min and supernatant plasma (0.3 mL) was loaded onto the reservoir of a micropartition device (MilliporeSigma Centrifree Ultrafiltration, Burlington, MA) and centrifuged at 1228 g for 20 min. The f_p value was calculated as the ratio of radioactivity in the filtrate to that in the plasma. To assess lipophilicity, the log D was determined at pH 7.4 for each compound by modification of the previously published procedure. ³⁴

Image processing

PET emission data were attenuation-corrected using the transmission scan then reconstructed using a Fourier rebinning and filtered back-projection algorithm. PET images summed from the first 10 minutes of each scan were registered to a rhesus monkey anatomical reference MR image. Inverted transformations were applied to register an atlas region of interest (ROI) mask to the PET image and time-activity curves (TACs) were extracted from frontal, occipital, temporal, cingulate, and insular cortex, caudate, putamen, globus pallidus, nucleus accumbens, amygdala, hippocampus, thalamus, pons, substantia nigra, cerebellum, and centrum semiovale.

Kinetic modeling and quantitative analysis

Kinetic parameters were determined from regional TACs and radiometabolite-corrected arterial input functions, calculated as the product of the fitted total plasma curve and the parent radiotracer fraction curve. The one-tissue compartment model (1TCM), two-tissue compartment model (2TCM), and the multilinear analysis 1 (MA1) method ³⁵ with starting time (t*) of 30 min were each assessed. Regional volume of distribution (V_T, mL/cm³) and, for compartment models, tissue uptake constant K₁ were derived for each radiotracer and ROI. Quality of fit for each method was assessed visually and precision of V_T estimates was assessed using relative standard error (rSE) to identify the optimal method for kinetic analysis in each case.

From each blocking study, an occupancy plot was constructed showing difference in V_T between baseline and blocking scans relative to baseline V_T values across ROIs. Percent target occupancy across the brain and estimated nondisplaceable volume of distribution (V_ND) were determined from the resulting linear relationship. ³⁶ From baseline scans, Guo plots ³⁷ were constructed to directly compare binding parameters within each animal, providing estimates of the ratios of K_D and BP_ND values between each pair of radiotracers. Finally, regional BP_ND values were determined for each radiotracer using V_ND estimates derived from the GluN2B blocking scans (BP_ND = [V_T – V_ND]/V_ND).

Results

Injection parameters

Injection parameters for the 3 tracers are summarized in Table 1. (R)-[¹¹C]NR2B-Me was produced with the highest molar activity, followed by (R)-[¹⁸F]OF-Me-NB1 and (S)-[¹⁸F]OF-NB1.

Table 1.

Injection parameters and plasma analysis results for bolus injections of GluN2B-targeting radiotracers in rhesus macaque.

	(R)-[11C]NR2B-Me n = 6mean ± SD	(R)-[18F]OF-Me-NB1 n = 6 mean ± SD	(S)-[18F]OF-NB1 n = 7mean ± SD
Molar activity (MBq/nmol)	550 ± 248	277 ± 151	124 ± 153
Injected mass (ng/kg)	17.0 ± 7.1	25.0 ± 15.6	75.9 ± 58.4
fP (%)	0.64% ± 0.24%	1.33% ± 0.32%	15.6% ± 1.2%
% parent compound
15 min post-injection	38% ± 11%	39% ± 14%	53% ± 5.3%
60 min post-injection	11% ± 3.7%	18% ± 4.5%	22% ± 2.0%

Blood measurements

Emergence of radiometabolites from plasma was similar for (R)-[¹¹C]NR2B-Me and (R)-[¹⁸F]OF-Me-NB1 (38.4% ± 11.2% and 39.0% ± 13.5% of parent compound remaining at 15 minutes post-injection, respectively) and slower for (S)-[¹⁸F]OF-NB1 (52.3% ± 5.3% parent at 15 minutes; Table 1). In one animal, the parent fraction of (R)-[¹¹C]NR2B-Me and (R)-[¹⁸F]OF-Me-NB1 appeared to decrease more quickly in blocking scans relative to baseline, while there were no noted differences between the baseline and blocking scans for (S)-[¹⁸F]OF-NB1 (Supplemental Figure S1). Consistent with their lipophilicity profiles, plasma free fraction was lower for (R)-[¹¹C]NR2B-Me (0.64% ± 0.24%; logD = 2.89 ± 0.07) and (R)-[¹⁸F]OF-Me-NB1 (1.29% ± 0.28%; logD = 2.76 ± 0.13) compared to (S)-[¹⁸F]OF-NB1 (15.3% ± 2.0%; logD = 2.05 ± 0.08).

Brain distribution and kinetics

Representative images of brain tissue concentrations in template space are shown in Figure 1 (bottom). Each radiotracer showed high uptake throughout cortical and subcortical grey matter and cerebellum, with lower concentrations in white matter regions. Regional brain TACs at baseline showed fast uptake and reversible kinetics, with peak SUV ranging from approximately 1.5 to 7 across regions and tracers (Figure 2(a)). (R)-[¹¹C]NR2B-Me and (S)-[¹⁸F]OF-NB1 each reached peak uptake values within 20–30 minutes post-injection and then cleared from the brain slowly, while the concentrations of (R)-[¹⁸F]OF-Me-NB1 in grey matter regions peaked within 20 minutes then decreased more rapidly to approximately the level in the white matter (centrum semiovale) by the end of the scan.

Figure 2.

Time-activity curves and kinetic modeling results. a, example regional time-activity curves for each radiotracer with 1TCM fits. b, comparison of V_T values derived from 2TCM vs. 1TCM (top row) and MA1 vs. 1TCM (bottom row).

Kinetic modeling

For each radiotracer, TACs were well-fitted by the 1TCM (Figure 2(a)). Relative standard error (rSE) of V_T values in grey matter regions were <4% for (R)-[¹⁸F]OF-Me-NB1, 3–20% for (R)-[¹¹C]NR2B-Me, and <5% for (S)-[¹⁸F]OF-NB1. The 2TCM frequently produced unstable estimates of V_T (rSE >25%). For 2TCM V_T estimates with rSE < 20%, agreement with 1TCM values was good (Figure 2(b), top row). The MA1 method (t* = 30 min) produced good model fits, stable estimates of V_T, and good to excellent agreement with 1TCM values (Figure 2B, bottom row). Parameter values derived from the 1TCM are therefore presented for further comparison here, but the simplified MA1 method is also appropriate for quantitative analysis of (R)-[¹⁸F]OF-Me-NB1 and (S)-[¹⁸F]OF-NB1.

Regional V_T and K₁ values are shown in Table 2 (full ROI list in Supplemental Tables S1 and S2). V_T values in the examined ROIs ranged from 23.5–54.6 mL/cm³ for (R)-[¹¹C]NR2B-Me, 8.2–12.8 mL/cm³ for (R)-[¹⁸F]OF-Me-NB1, and 16.9–38.8 mL/cm³ for (S)-[¹⁸F]OF-NB1. Values were highest in cingulate cortex in all cases and lowest in white matter for (R)-[¹¹C]NR2B-Me and (S)-[¹⁸F]OF-NB1, or occipital cortex for (R)-[¹⁸F]OF-Me-NB1. There was good agreement in regional V_T estimates between the two animals in each case (mean difference between animals across ROIs: −1.9% ± 18.4% for (R)-[¹¹C]NR2B-Me; 2.3 ± 5.8% for (R)-[¹⁸F]OF-Me-NB1; 11.7 ± 5.4% for (S)-[¹⁸F]OF-NB1). K₁ estimates in grey matter regions were 0.26–0.47 mL·cm⁻³·min⁻¹ for (R)-[¹¹C]NR2B-Me, 0.16–0.36 mL·cm⁻³·min⁻¹ for (R)-[¹⁸F]OF-Me-NB1, and 0.30–0.66 mL·cm⁻³·min⁻¹ for (S)-[¹⁸F]OF-NB1 (Table 2).

Table 2.

Summary of kinetic parameters in select brain regions.

	(R)-[11C]NR2B-Me		(R)-[18F]OF-Me-NB1		(S)-[18F]OF-NB1
Animal	1	2	1	2	1	2
VT, mL/cm3
Neocortex	40.8	45.7	10.4	10.5	31.5	28.2
Basal ganglia	36.4	32.2	10.7	10.2	29.6	24.4
Hippocampus	39.0	37.8	10.9	10.1	32.1	26.3
Cerebellum	32.5	34.6	8.7	9.1	27.3	25.5
Centrum semiovale	23.5	35.5	8.7	9.0	20.3	16.9
K1, mL · cm−3·min−1
Neocortex	0.350	0.340	0.249	0.313	0.393	0.536
Basal ganglia	0.392	0.336	0.278	0.312	0.417	0.517
Hippocampus	0.355	0.322	0.222	0.280	0.386	0.510
Cerebellum	0.482	0.459	0.308	0.371	0.495	0.744
Centrum semiovale	0.176	0.141	0.088	0.105	0.166	0.190
VND, mL/cm3	11.21	10.46	6.36	6.42	9.32	7.28
BPND
Neocortex	2.64	3.36	0.64	0.63	2.38	2.88
Basal ganglia	2.25	2.08	0.68	0.59	2.17	2.35
Hippocampus	2.48	2.61	0.71	0.58	2.44	2.61
Cerebellum	1.90	2.31	0.37	0.41	1.93	2.50
Centrum semiovale	1.10	2.39	0.37	0.40	1.18	1.32

V_T: volume of distribution from 1TCM; V_ND: nondisplaceable volume of distribution estimated from Lassen plots with GluN2B blocking (see Figure 3); BP_ND: binding potential calculated as V_T/V_ND – 1.

Blocking studies

GluN2B. Pre-treatment with the GluN2B antagonist Co-101,244 (0.25 mg/kg) decreased radioactivity uptake across the brain (Figure 1, bottom row, and Table 3). V_T was reduced across all regions, demonstrating specific binding of each radiotracer to the GluN2B site and suggesting absence of a reference region. Occupancy estimates were consistent between the two animals and were lowest for (R)-[¹⁸F]OF-Me-NB1 (74% and 69%) followed by (S)-[¹⁸F]OF-NB1 (84% and 79%) and (R)-[¹¹C]NR2B-Me (95% and 90%) (Figure 3(a) and Table 3). From occupancy plots, estimated V_ND was 10.5–11.2 mL/cm³ for (R)-[¹¹C]NR2B-Me (23-27% of average V_T in neocortex); 6.4 mL/cm³ for (R)-[¹⁸F]OF-Me-NB1 (61% of neocortex V_T) and 7.3–9.3 mL/cm³ for (S)-[¹⁸F]OF-NB1 (26–30% of neocortex V_T) (Table 2).

Table 3.

Percent binding reduction with GluN2B and sigma-1 receptor-targeting competitor drugs calculated from occupancy plots.

	(R)-[11C]NR2B-Me		(R)-[18F]OF-Me-NB1		(S)-[18F]OF-NB1
Animal	1	2	1	2	1	2
GluN2B:
Co-101,244, 0.25 mg/kg	95%	90%	74%	69%	84%	79%
σ1: FTC-146, 0.027 mg/kg		n.d.	30%		49%
FTC-146, 0.125 mg/kg	56%			33%		48%

n.d.: not determined.

Figure 3.

Blocking and displacement studies. a, Occupancy plots showing reduction in specific binding of GluN2B-targeting radiotracers following pre-treatment with Co-101,244 (0.25 mg/kg) relative to baseline in each animal. b, Time-activity curve showing bolus plus constant infusion study of (R)-[¹⁸F]OF-Me-NB1 with sequential administration of Co-101,244 (0.75 mg/kg, +75 min) and FTC-146 (0.125 mg/kg, +225 min). Reduced signal in tissue is seen following administration of Co-101,244 but no further displacement is apparent following the addition of FTC-146. The fitted input function is shown in red.

σ₁ receptor. Pre-treatment with the σ₁-selective antagonist FTC-146 (0.027 or 0.125 mg/kg) also reduced uptake and V_T values of all three tracers, with 30-56% reduction in specific binding (Supplemental Figure S2). However, this reduction was not observed consistently and did not show a clear relationship to dose (Table 3). An additional experiment was then performed to investigate possible off-target binding at the σ₁ site. Given its faster kinetics (comparable K₁ and lower V_T), an infusion and double displacement study was performed using (R)-[¹⁸F]OF-Me-NB1. After reaching steady tracer concentration in tissue, Co-101,244 (0.75 mg/kg) and FTC-146 (0.125 mg/kg) were administered in sequence separated by 150 minutes to determine whether further displacement would be seen following administration of the σ₁-targeting drug when the GluN2B sites were occupied. Reductions in tissue concentrations and V_T values were observed following injection of the GluN2B ligand, but no further change was apparent following administration of the σ₁ ligand (Figure 3(b) and Supplemental Figure S3). Administering the two competitor drugs together therefore did not result in greater occupancy of binding sites than the GluN2B antagonist alone, suggesting minimal direct binding of (R)-[¹⁸F]OF-Me-NB1 at σ₁ receptors.

Ketamine challenge

Pre-treatment with 1 mg/kg ketamine produced plasma concentrations during the scan of 95-382 ng/mL (Supplemental Figure S4), consistent with therapeutic levels in humans treated for Major Depressive Disorder.^30,38 There was no detectable change in regional V_T values of (S)-[¹⁸F]OF-NB1 compared to the baseline scan (percent decrease 2.4% ± 9.8%).

Comparison of binding parameters

Baseline V_T values were lowest for (R)-[¹⁸F]OF-Me-NB1 and highest for (R)-[¹¹C]NR2B-Me (Table 2). Guo plots ³⁷ were constructed to directly compare V_T values between radiotracers (Figure 4). Assuming passive diffusion of all tracers across the blood-brain barrier, slopes of these plots indicate that in vivo K_D values are highest for (S)-[¹⁸F]OF-NB1 (K_D ^{(S) -[18F]OF-NB1/}K_D ^{(R) -[18F]OF-Me-NB1} = 2.8–3.1; K_D ^{(S) -[18F]OF-NB1}/K_D ^{(R) -[11C]NR2B-Me} = 25.9-45.0) and lowest for (R)-[¹¹C]NR2B-Me (K_D ^{(R) -[18F]OF-Me-NB1/}K_D ^{(R) -[11C]NR2B-Me} = 6.7–12.7). Higher affinity for [¹¹C]NR2B-Me and (R)-[¹⁸F]-OF-Me-NB1 compared to (S)-[¹⁸F]-OF-NB1 is comparable to the pattern seen with in vitro K_i estimates (for [¹¹C]NR2B-Me, K_i = 5.4 nM; ³⁹ for (R)-OF-Me-NB1, ²⁰ K_i = 4 nM; for OF-NB1, K_i = 10.4 nM ¹³ ) From the y-intercepts of these plots, BP_ND values followed the rank order of (R)-[¹¹C]NR2B-Me > (S)-[¹⁸F]OF-NB1 > (R)-[¹⁸F]OF-Me-NB1. Consistent with this ranking, BP_ND estimates determined using V_ND values from the GluN2B blocking scans were higher for (R)-¹¹C-NR2B-Me (1.90–3.36 in grey matter regions) and (S)-¹⁸F-OF-NB1 (1.93–2.88), and lower for (R)-¹⁸F-OF-Me-NB1 (0.37–0.71) (Table 2). BP_ND values for (R)-[¹¹C]NR2B-Me and (S)-[¹⁸F]OF-NB1 were similar across regions (mean difference 8.7% ± 8.7%).

Figure 4.

Guo plots comparing V_T across ROIs during baseline scans for two animals. In these plots, y-intercept > 0 indicates that the tracer on the x-axis is expected to have higher binding potential values. Thus, BP_ND values follow the rank order (R)-[¹¹C]NR2B-Me > (S)-[¹⁸F]OF-NB1 > (R)-[¹⁸F]OF-Me-NB1.

Discussion

Here we compared three candidate radiotracers, (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1, for the quantification of the GluN2B subunits of the NMDA receptor in non-human primate brain. Each had good imaging characteristics, with (S)-[¹⁸F]OF-NB1 showing particular promise for future translation into humans based on its appropriate kinetics, BP_ND values in the range of 2–3, and high free fraction in plasma. Carbon-11-labeled (R)-NR2B-Me showed similar tissue kinetics and BP_ND estimates but lower free fraction in plasma, whereas ¹⁸F-labeled (R)-OF-Me-NB1 had faster kinetics and lower V_T and BP_ND values across the brain. Each of these ligands has a similar brain distribution pattern and selectivity profile in vivo consistent with good selectivity for GluN2B-containing NMDA receptors, though further work is necessary to characterize the nature of displaceable binding in cerebellum.

Imaging characteristics of the three radiotracers were consistent with the high brain uptake, displaceable binding, and favorable metabolite profile seen in rodent experiments.^13,17,18 Parent fraction in plasma was somewhat higher early in the scan for (S)-[¹⁸F]OF-NB1 compared to (R)-[¹¹C]NR2B-Me and (R)-[¹⁸F]OF-Me-NB1, but in each case 35-55% parent compound remained at 15 minutes post-injection. There was a marked difference in plasma free fraction, with (S)-[¹⁸F]OF-NB1 showing higher free fraction (15%) compared to (R)-[¹¹C]NR2B-Me and (R)-[¹⁸F]OF-Me-NB1 (each <2%), reflecting the additional hydroxyl group (Figure 1, top) and lower lipophilicity of (S)-[¹⁸F]OF-NB1 (logD of 2.1 compared to logD > 2.7 for the other two compounds). Free fraction values in this higher range can be measured with greater precision, improving accuracy of the quantitative PET measurements, so that V_T/f_p could be a useful outcome measure in ultimate clinical studies.

The tissue distribution pattern was similar between the compounds, with high uptake throughout grey matter and the highest V_T values in cingulate cortex in all cases. As in rodent studies, binding was displaceable by the GluN2B antagonist across all brain regions and therefore a reference region is unlikely to be available for these radioligands and reference tissue quantification was not assessed. However, tissue kinetics further distinguished the three radiotracers, with fast uptake to peak SUV values of 3–6 in grey matter and relatively slow clearance over the two-hour scan time for (R)-[¹¹C]NR2B-Me and (S)-[¹⁸F]OF-NB1, but more rapid clearance for (R)-[¹⁸F]OF-Me-NB1. This faster kinetics allowed for a scan using bolus plus constant infusion administration of (R)-[¹⁸F]OF-Me-NB1, in which stable concentration in tissue was reached within approximately one hour, as in the displacement study presented here. If similar kinetics are observed in humans, (R)-[¹⁸F]OF-Me-NB1 may be a suitable choice in applications where radiotracer infusion can allow for simplified quantification and interpretation, despite its substantially lower BP_ND values.

In kinetic modeling comparisons, regional TACs for each compound were well-fitted by the 1TCM, producing reliable estimates of V_T and K₁ (Table 2). The MA1 method also produced excellent results for (R)-[¹⁸F]OF-Me-NB1 and (S)-[¹⁸F]OF-NB1 (Figure 2(b)). This simplified quantification method can improve the reliability and precision of quantification and may simplify voxel-level kinetic modeling. In direct comparisons, in vivo K_D was higher for (S)-[¹⁸F]OF-NB1 compared to (R)-[¹¹C]NR2B-Me, consistent with in vitro K_i estimates and underlining the effect of higher free fraction of (S)-[¹⁸F]OF-NB1 to produce V_T and BP_ND values comparable to (R)-[¹¹C]NR2B-Me despite the latter’s higher target affinity. V_T estimates for (S)-[¹⁸F]OF-NB1 were in the range of 16–32 mL/cm³ and V_ND estimates were 7.3 and 9.3 mL/cm³, giving BP_ND values of 2–3. Free fraction in tissue (f_ND) values, calculated as f_p/V_ND, were 0.067–0.051% for (R)-[¹¹C]NR2B-Me, 0.18% for (R)-[¹⁸F]OF-Me-NB1, and 1.6–2.1% for (S)-[¹⁸F]OF-NB1, giving f_ND ratios for (S)-[¹⁸F]OF-NB1 relative to (R)-[¹¹C]NR2B-Me of 24.4–41.7, very similar to the estimated K_D ratio (see Results). Thus, given very similar f_ND/K_D ratios for the two compounds, estimated BP_ND values (BP_ND = f_NDB_max/K_D) are also similar (Table 2). Altogether, these analyses suggest that (S)-[¹⁸F]OF-NB1 has excellent imaging properties that allow for precise quantification of GluN2B-containing NMDA receptors. (R)-[¹¹C]NR2B-Me had similarly high BP_ND values and may be a suitable option in applications for which a carbon-11 radiotracer is preferred.

In a series of blocking experiments, specific binding was reduced 69–95% by the GluN2B antagonist Co-101,244. Some reduction was also observed by the σ₁-selective drug FTC-146. As chaperone molecules, σ1 receptors regulate the plasma membrane expression of many targets, including NMDA receptors,^18,40 and also have a structurally similar binding site to that seen on GluN2B. Blocking or displacement of GluN2B-targeting compounds by σ₁-targeting drugs have been investigated previously in rodents, with variable results.^13,17 In one study, this effect of σ₁ blocking drugs was also seen in σ₁ genetic knockout mice, suggesting these may not reflect direct interaction of the radiotracer with the σ₁ receptor. ¹³ The infusion and sequential displacement experiment performed with (R)-[¹⁸F]OF-Me-NB1 (Figure 3(b)) found that no additional reduction in binding was seen when adding a σ₁-targeting drug after administration of a GluN2B-targeting drug. This is consistent with the results of a previous blocking scan in which Co-101,244 (0.25 mg/kg) and FTC-146 (0.125 mg/kg) were administered together prior to bolus injection of (S)-[¹⁸F]OF-NB1. ²³ This produced an 82% reduction in specific binding relative to baseline, similar to the reduction seen here with Co-101,244 alone (Table 3). Together with the high K_i values at σ₁ receptors of (R)-[¹⁸F]-OF-Me-NB1 (100 nM) ²⁰ and [¹⁸F]-OF-NB1 (410 nM), ¹³ these experiments suggest that it is unlikely that these radiotracers bind directly to σ₁ receptors since such an interaction would be expected to produce additional displaceable binding beyond that observed with a GluN2B antagonist. Instead, σ₁ drug actions could reflect an effect of these drugs at the GluN2B site, at another allosteric binding site on GluN2B-containing NMDA receptors, or an effect of σ₁ receptors themselves on the receptors. Thus, studies with these GluN2B tracers may also be useful to assess these apparently indirect actions of σ₁ drugs.

As in previous studies performed in rodents using these and similar GluN2B-targeting radioligands,^13,16–20 displaceable binding was observed in the cerebellum. In situ hybridization studies performed show negligible expression of the mRNA encoding GluN2B subunits in this region during adulthood in rodents^41,42 and humans.^43,44 Nevertheless, each of the three radiotracers had substantial uptake in cerebellum, with V_T values similar to those in cortical regions. In all cases this was reduced by the GluN2B antagonist Co-101,244, suggesting that rather than representing nonspecific binding, an off-target binding site may be present in the cerebellum and potentially elsewhere. This is in contrast to in vitro assays showing high target selectivity and autoradiography experiments in rodent brain that showed negligible binding in cerebellum.^13,20 Given this discrepancy, there could be a potential unknown off-target binding site or sites, though structurally similar proteins such as the NMDA receptor GluN2A subunit might be a possibility. While the high levels of blocking (>80% for (S)-[¹⁸F]OF-NB1) with the GluN2B-selective antagonist Co-101,244 suggest a generally favorable selectivity profile in vivo, further experiments to characterize the source of binding in the cerebellum will be necessary to rule out off-target binding for all three radiotracers.

The distinct structural character of the NMDA receptor complex provides several binding targets relevant to drug development. The three benzazepine-based radiotracers described here are presumed to bind to the allosteric ifenprodil binding site on the GluN2B subunit. As such, they may be useful for quantification of GluN2B-containing NMDA receptors and for occupancy studies of drugs targeting that site. We assessed whether (S)-[¹⁸F]OF-NB1 may also be sensitive to the acute NMDA- and glutamate-modulating agent ketamine, which binds at a separate site within the ion pore, and found no evidence of a change in target availability at plasma ketamine levels comparable to those seen in antidepressant studies in humans.^38,45 If these initial data that ketamine has no direct interaction at the GluN2B site are confirmed, then further work can explore whether ketamine treatment may produce sustained effects on GluN2B-containing receptors, as has been observed in rodent studies. ⁴⁶

Altogether, (R)-[¹¹C]NR2B-Me, (R)-[¹⁸F]OF-Me-NB1, and (S)-[¹⁸F]OF-NB1 each showed suitable imaging characteristics as selective radiotracers for quantification of the NMDA receptor GluN2B site in primates. Given its high BP_ND values, free fraction in plasma, and ¹⁸F label, (S)-[¹⁸F]OF-NB1 is a particularly promising candidate for evaluation in humans. If validated, this will provide a key tool to advance the study of a range of neuropsychiatric disorders linked to GluN2B dysfunction.