Evaluation of the Novel PET Tracer [¹¹C]HACH242 for Imaging the GluN2B NMDA Receptor in Non-Human Primates

N-methyl-d-aspartate (NMDA) receptors are ionotropic glutamate receptors that mediate excitatory postsynaptic signalling and are expressed widely in the central nervous system (CNS). NMDA receptors are heterotetramers of mostly GluN1 and two GluN2(A-D) subunits in the CNS [1]. In the last decade, a major effort in NMDA receptor drug development has focused on allosteric modulation of ion channel activity via GluN2B (also known as NR2B) binding sites which are located on the extracellular N-terminal domain (NTD) of the NMDA receptor. Ligands such as ifenprodil and derivatives are noncompetitive antagonists of the GluN2B site that stabilise a closed-cleft conformation of the GluN2B NTD [2]. Immunoblotting techniques showed that in the human brain, the GluN2B subunit is predominantly expressed in the cortex, hippocampus and thalamus, consistent with findings in rats showing high expression in the forebrain. Lower levels of GluN2B are found in the midbrain, cerebellum and spinal cord [3, 4]. The interest in the therapeutic potential of GluN2B negative allosteric modulators has spurred drug development of ifenprodil derivatives in the past two decades. Second generation “prodils” such as radiprodil (RHG-896), traxoprodil (CP-101,606), besonprodil (CI-1041), CERC-301 (MK-0657), EVT-101, Ro 25-6981 and BMT-108908 have been developed for the treatment of a wide range of CNS pathologies, including acute and chronic pain, stroke, dementias and major depressive disorder.

The highest level of confidence and direct evidence that GluN2B engagement is achieved is acquired from in vivo molecular imaging experiments with positron emission tomography (PET). There are currently no PET radiotracers available that have shown reliable quantitative specific binding of GluN2B in vivo. Radiotracer development for NMDA receptors has been unsuccessful due to suboptimal physiochemical and pharmacological characteristics of ligands such as affinity, lipophilicity, stability, blood-brain barrier penetration and pharmacokinetics [5‐10]. Based on the discovery of a new class of GluN2B antagonists [11], N-((5-(4-fluoro-2-[¹¹C]methoxyphenyl)pyridin-3-yl)methyl) cyclopentanamine was synthesised and radiolabelled with carbon-11 ([¹¹C]HACH242, Fig. 1) for preclinical evaluation. Biodistribution and blocking studies with [¹¹C]HACH242 were carried out in anaesthetised mice and in non-anaesthetised rats [12]. The affinity of [¹¹C]HACH242 for GluN2B receptors was 12.4 ± 2.1 nM as shown in competition binding experiments against 5 nM [³H]ifenprodil, which is equal to ifenprodil and about 5 times lower than Ro25-6981. Brain distribution studies in mice showed a threefold increase in activity uptake in forebrain regions vs. cerebellum at 15 min, which is the highest ratio that has been reported for candidate GluN2B site radiotracers. [¹¹C]HACH242 binding in non-anaesthetised rats was reduced by Ro25-6981, showing a regional distribution that is consistent both with the expression pattern of GluN2B subunits and the density of [³H]Ro25-6981 binding sites in the adult rodent brain. However, high levels of non-specific uptake were observed in the thalamus [12].

GluN2B ligands frequently exert off-target effects. For example ifenprodil dose-dependently reduces brain uptake of the sigma-1 receptor radiotracer [¹¹C]SA4503 [13]. The selectivity of [¹¹C]HACH242 for GluN2B was investigated in an exhaustive pharmacological screen against 79 targets [12]. The radioligand showed affinity > 1.1 μM for 78 of the targets. Sigma-1 was the exception, for which the tracer showed an affinity of 20 nM. Further studies with GluN2B specific antagonists are warranted to investigate the specific binding of [¹¹C]HACH242 in vivo. Radiprodil binds to GluN2B-containing receptors irrespective of the GluN2 subunit composition, analogous to Ro25-6981 [14]. Furthermore, radiprodil binding is insensitive to 1,3-di-o-tolylguanidine (DTG), which suggests it does not bind to the sigma receptors [14]. Research on non-human primates (NHP) generates complementary data that can bridge translational imaging research from rodents to humans. This paper reports on the initial assessment of [¹¹C]HACH242 PET in NHP at baseline and following intravenous administration of a high dose of radiprodil.

[¹¹C]HACH242 readily entered the brain and displayed a fairly uniform pattern of uptake (Fig. 2). Individual subject SUVs at baseline and after radiprodil administration are shown in Fig. 3 and summarised in Table 2. The peak uptake in whole brain grey matter during radiprodil scans was increased by 20 % in subject 1 but reduced by 9 %, 1 % and 20 % for subjects 2, 3 and the retest of subject 1, respectively, compared to baseline. SUV in the last 15-min PET frame was 17 to 25 % higher in the radiprodil condition for subject 1, but approximately equal (1 to 3 %) in subjects 2 and 3. Covariance of SUV_AUC between subjects for the four baseline scans was 12 %. The SUV of [¹¹C]HACH242 in whole brain, frontal cortex and cerebellum grey matter when normalised for individual peak uptake are shown in Fig. 3b. Slower washout in the radiprodil condition resulted in slight SUV increases in the last PET frame compared with peak uptake of 4 to 11 % for subjects 1, 2 and 3 and 45 % for the retest of subject 1.

This study describes the first in vivo evaluation of the radiotracer [¹¹C]HACH242 in the primate brain. [¹¹C]HACH242 TACs in the brain showed appropriate reversible pharmacokinetics within the timeframe of the PET scan. Systemic plasma clearance of the parent compound remained constant throughout the scan and was lower than clearance from brain VOIs. No consistent reduction was observed in [¹¹C]HACH242 SUVs after intravenous infusion of radiprodil, a GluN2B negative allosteric modulator. Radiprodil distribution studies in rodents showed that it has unrestricted access to the brain tissue with brain-to-plasma ratios of ~ 1 and a terminal half-life of 3.2 h in dogs (UCB Pharma, unpublished communication). Radiprodil was therefore expected to decrease the SUV ratio at late time frames in VOIs with high vs low GluN2B binding sites. This could be suggestive of lower specific binding of [¹¹C]HACH242, under the assumption that the radiotracer kinetics in plasma are equal in both conditions. The main limitation of this study is that reliable input functions for kinetic modelling could not be generated because of variability in the whole blood radioactivity measurement during the first minutes after tracer injection. However, interpretation of the [¹¹C]HACH242 kinetics in different brain VOIs in relation to plasma metabolism and clearance allows for deductions to be made about the underlying physiological processes.

In previous studies in mice, [¹¹C]HACH242 showed high specific binding and heterogeneous distribution in brain slices that was similar to the expression of NMDA receptors. For example, a threefold increase in activity uptake was measured in forebrain regions vs. cerebellum at 15 min [9]. Results in the present study show that average [¹¹C]HACH242 peak uptake in the frontal cortex at baseline condition is 30 % lower than cerebellum uptake and 10 % higher at the end of the scan. This suggests a lower perfusion of the tracer into the frontal cortex and a somewhat slower washout. After radiprodil administration, the uptake in the cerebellum is more in line with the frontal cortex (Fig. 5). The lower ratio from 30 min to the end of the scan could suggest that binding of [¹¹C]HACH242 to GluN2B in the frontal cortex is reduced relative to cerebellum binding. However, the ratio change can be explained largely by the change in the cerebellar TACs after radiprodil administration. In another effort to compare the effect of radiprodil on [¹¹C]HACH242 binding, SUVs were normalised to the individual peak uptake (Fig. 3b). GluN2B saturation was expected to accelerate the washoff and reveal the degree of specific binding. However, slower washout in the radiprodil condition was observed, resulting in SUV increases in the last PET frame of 4 to 11 % for subjects 1 to 3 and 45 % for the retest of subject 1. The baseline TACs of the cerebellum were similar for subject 1 test and retest scans after normalisation to peak uptake, indicating good TAC test-retest reproducibility in a VOI with relatively low density of the target. In situ hybridisation studies of GluN2B have shown that mRNA expression patterns of the GluN2B subunit are similar between humans, monkeys, and mice in most brain regions [18]. The NR2B subunit is found predominantly in hippocampus, cerebral cortex, striatum and olfactory bulbs, as shown by immuno-cytochemical localisation studies [19]. The present baseline PET data indicate slower washout in cortex than in cerebellum, as could be expected based on the target density. Although in vitro autoradiography of [¹¹C]HACH242 in the brain of rhesus monkey has not been performed yet, the widespread distribution of GluN2B suggests that a reference region in the brain devoid of GluN2B binding sites is not available, which prevented the exploration of reference tissue compartmental models to fit to the PET data. An additional complexity in the interpretation of the current results is the affinity of [¹¹C]HACH242 to sigma-1 receptors (20 nM) in the brain. A postmortem human brain study found moderate levels of sigma-1 receptor mRNA distributed in the cortex, with the temporal cortex being particularly enriched [20]. Although experiments to confirm the affinity of [¹¹C]HACH242 to sigma-1 in primates are warranted, it is reasonable to assume that the measured [¹¹C]HACH242 PET signal constitutes binding to both GluN2B and sigma-1, which could confound interpretation of (clinical) data. The degree of specific binding to GluN2B in mice was shown to vary between 3 and 40 % [12]. Dense labelling was observed in brain regions that do not express high levels of GluN2B subunit-containing NMDA receptors, suggesting binding to sigma-1 or other proteins. Future research could employ a sigma-1 receptor compound such as the agonist fluvoxamine in combination with a GluN2B negative allosteric modulator to assess the binding potential of [¹¹C]HACH242. Importantly, however, a recent PET study in rats showed that the sigma-1 receptor agonist (1)-pentazocine abolished [¹¹C]Me-NB1-specific binding in the brain, despite the lack of direct competition in vitro and a 33-fold selectivity of [¹¹C]Me-NB1 for the human GluN2B receptor over sigma 1 [21]. It was hypothesised that an interaction between sigma-1 agonism and GluN2B in the living brain could hamper binding of [¹¹C]Me-NB1, which is structurally comparable with [¹¹C]HACH242. It would be valuable to further investigate [¹¹C]HACH242 kinetics following administration of a sigma-1 agonist to confirm the findings of an in vivo PET signal block.

The change in [¹¹C]HACH242 brain uptake and washout following radiprodil administration may be a consequence of the drug’s effect on the delivery of parent [¹¹C]HACH242 from blood to tissue. Results showed that SUV in blood and plasma, as well as the plasma-to-blood ratio were reduced in the radiprodil condition compared to baseline. A lower AUC of plasma parent fraction was also observed in the three radiprodil scans for which data was available (−11 %, −45 % and − 56 % compared to baseline). Binding in peripheral sites or to plasma proteins might have been altered by radiprodil, resulting in a different pattern of uptake in the brain. However, a closer look at the individual TACs does not show a clear relationship between SUV_AUC in plasma and brain. Fast peripheral metabolism of [¹¹C]HACH242 and rapid clearance of the parent tracer from plasma hindered prolonged measurements of [¹¹C]HACH242 from arterial plasma. Errors in the estimation of arterial blood parameters are probable. For example, the whole-blood activity at 85 min in the radiprodil retest scan of subject 1 was 40 % higher than the previous samples resulting in aberrant group averages (Fig. 4). Polar metabolites made up 50 % of measured radioactivity in plasma at 10 min p.i., compared with 75 % in previous in vivo [¹¹C]HACH242 experiments in mice [6]. In general, it is assumed that polar metabolites do not cross the blood-brain barrier, but even if they do, they will only increase the non-specific signal. The non-polar metabolite fraction remained less than 10 % throughout the time of the scan. Therefore, even if these lipophilic metabolites cross the blood-brain barrier, they would constitute only a minor fraction of total cerebral uptake. Nevertheless, future studies are needed to assess the behaviour of the labelled metabolites in the brain.

It is possible that changes in cerebral blood flow (CBF) contributed to the difference between baseline and radiprodil [¹¹C]HACH242 SUV. There are currently no scientific reports available that have investigated with PET or MRI the effect of “prodils” or other GluN2B negative allosteric modulators on CBF in the healthy brain. One study on the protective effects of ifenprodil on ischemia used laser doppler flowmetry and showed that there was no significant change in cerebral blood flow [22]. In the present study, the anaesthestics medetomidine and midazolam were administered intramuscularly approximately 2 h before the baseline PET scan, equating to ~ 6 h before the radiprodil scan. Considering medetomidine’s time to peak concentration and half-life of ~ 30 min following i.m. administration [23], the plasma and brain concentrations may still have been high enough to exert pharmacodynamic effects in the baseline scan. Reports of the effect of anaesthesia on CBF are important in this regard. A SPECT study in dogs showed that medetomidine increased the regional CBF in all brain regions bar the subcortical (thalamic) area [24]. Consistent with increased CBF, [¹¹C]HACH242 TACs showed higher uptake and faster washout in the baseline scan. However, midazolam, which was co-administered with medetomidine, has been shown to reduce global CBF dose-dependently in humans [25]. Propofol was injected intravenously at a constant rate during the entire scan day for maintenance of anaesthesia and therefore should not have affected CBF differently between the two [¹¹C]HACH242 PET scans on the same day. Polypharmacological effects, which potentially modulate multiple targets in the brain, complicate predictions of regional CBF effects. Ketamine is commonly used as a sole anaesthetic agent in NHP studies. However, this compound acts on the NMDA receptor [26] and, therefore, it was not suitable for implementation in the current study. Opioidergic compounds such as fentanyl can affect NMDA receptor function [17], and hence propofol was deemed the most appropriate anaesthetic in the present study design. Future attempts at visualising the GluN2B binding site of the NMDA receptor are preferably performed in awake primates or humans to exclude potential CBF effects of anaesthetics. Simultaneous [¹¹C]HACH242 PET and fMRI could elucidate the relationship between NMDA receptor blockers, blood flow and radiotracer binding [27].

Preclinical experiments with [¹¹C]HACH242 have shown that anaesthesia in mice, administered as a mixture of fentanyl, fluanisone and midazolam, resulted in inconsistent effects on brain uptake of [¹¹C]HACH242 following pretreatment with Ro25-6981 [6]. Specifically, Ro25-6981 induced either a mean 22 % decrease or a mean 29 % increase in radioactivity uptake across all brain regions analysed in ex vivo brains 20 min after radiotracer injection. It is striking that in the present study there was considerable test-retest variability of subject 1 SUV in brain and plasma. PET experiments performed in NHP under awake conditions with the C-11 labelled analogue of the GluN2B-selective ligand CP-101,606 showed no appreciable specific binding [28]. Endogenous ligands of the NMDA receptor such as spermine and spermidine, as well as divalent cations, strongly inhibited radiolabelled CP-101,606 binding in the order of Zn²⁺ > Mg²⁺ > Ca²⁺ at their physiological concentrations. l-glutamate and NMDA tended to slightly increase the binding, while glycine and d-serine showed no effect [28]. Discrepancies between in vitro and in vivo binding properties of radiotracers developed for the GluN2B binding site have been reported for other NMDA receptor radiotracers [21, 29]. Ifenprodil and spermine inhibited the in vitro binding of C-11 labelled bis(phenylalkyl)amines with potencies similar to those of non-radioactive ligand. However, in vivo pre-treatment with non-radioactive tracer caused an increase of binding in the cerebral cortex and hippocampus compared with the control group [29]. Estimating the [¹¹C]HACH242 binding potential in vivo is complicated by the atypical mode of action of GluN2B-containing NMDA receptors. For example, binding is context-dependent [30], such that strongly activated receptors are preferentially inhibited. In addition, GluN2B-selective antagonists can divide into two distinct classes according to binding pose, resulting in strikingly different ligand orientation and receptor interactions [31]. Taken together, there are various receptor interactions and biophysical factors that affect in vivo binding properties of NMDA receptor radioligands, which remain to be tested, for example using bolus-plus-infusion PET experiments in awake primates while monitoring hemodynamic changes [27].

The present results show that [¹¹C]HACH242 has a suitable kinetic profile in the brain and low accumulation of lipophilic radiometabolites. Under the tested conditions, however, the specific binding of [¹¹C]HACH242 in vivo using PET could not be visualised. The divergent preclinical behaviour in mice and primates could be a consequence of variations in cerebral blood flow, a low fraction of specifically bound tracer, or interactions of the radiotracer with the sigma-1 receptor or endogenous ligands at the binding site. The typical study design for neuroreceptor PET tracers executed here may not be applicable to such a dynamic and complex receptor system as the GluN2B site of the NMDA receptor. Further experiments of ligand interactions are necessary in order to facilitate the development of radiotracers for in vivo imaging of the ionotropic NMDA receptor.