Testing PET-[¹¹C]ABP688 as a tool to quantify glutamate release in vivo

Glutamate is the primary excitatory neurotransmitter in mammalian brain where it plays a central role in the effects of drugs and alcohol, drug withdrawal, and experience-dependent neuroplasticity (Dahchour & De Witte, 2003; Fliegel et al., 2013; Saal et al., 2003; Stuber et al., 2008; Vigneault et al., 2015). Based on these observations, medications targeting glutamate neurotransmission are in development for various neuropsychiatric disorders, including schizophrenia, depression, addictions, and fragile X syndrome (Dolen & Bear, 2008; Goff & Coyle, 2001; Hashimoto et al., 2007; Kalivas, 2009; Levenga et al., 2011). Despite these advances, a lack of tools has hampered our ability to measure glutamate features that might be specific to humans and human brain-based diseases.

To date, several aspects of glutamate neurotransmission in the living human brain, including changes in glutamate levels, have only been assessed indirectly. Functional imaging techniques such as blood oxygen level dependent-functional magnetic resonance imaging (BOLD-fMRI) and [¹⁸F]fluorodeoxyglucose/positron emission tomography ([¹⁸F]FDG/PET) can identify brain regional changes in activity, but their signals are not transmitter-specific (Knutson & Gibbs, 2007; Shafiei et al., 2019). Proton magnetic resonance spectroscopy (¹H-MRS) can measure stress and pain-induced changes in glutamate turnover rate in a pre-specified region of interest (ROI) (Bryant et al., 2013; Gutzeit et al., 2013), but these spectroscopic signals are difficult to interpret and are believed to primarily indicate intracellular (neuronal and glial) levels, making the relation to neurotransmission uncertain (Ramadan et al., 2013).

A recently proposed alternative approach is to use positron emission tomography (PET) with [¹¹C]ABP688 [3-((6-methylpyridin-2-yl)ethynyl) cyclohex-2-en-1-one-O-[¹¹C]methyloxime], a tracer that binds with high selectivity to the allosteric site of metabotropic type 5 glutamate (mGlu5) receptors (Ametamey et al., 2006; Scala et al., 2021). PET/[¹¹C]ABP688 studies have identified alterations in mGlu5 receptor availability in populations with various psychiatric and neurodegenerative disorders, including a major depressive disorder, substance use disorders, epilepsy, and frontotemporal dementia (Choi et al., 2014; Cox et al., 2020; Kim et al., 2019; Leuzy et al., 2016; Martinez et al., 2014; Milella et al., 2014; Smart et al., 2017).

Preliminary evidence also indicates that, in both laboratory animals and humans, [¹¹C]ABP688 binding is affected by interventions that increase (DeLorenzo et al., 2015; Esterlis et al., 2018) and decrease (Zimmer et al., 2015) extracellular glutamate concentrations. One possibility is that the changes in tracer binding are caused by glutamate release-inducing conformational changes that alter affinity at the allosteric binding site. Alternatively, since the reductions can remain for up to 24 h, a glutamate surge-induced receptor internalization is also possible. Indeed, this mechanism is thought to account for long-lasting decreases in the binding of D₂ receptor radioligands, such as [¹¹C]raclopride and [¹²³I]IBZM, following dopamine release (Laruelle, 2012; Laruelle et al., 1997). Finally, [¹¹C]ABP688 binding values can vary by up to 70% when repeated PET scans are conducted on the same day. These changes might reflect circadian rhythm related influences on both mGlu5 receptor expression (DeLorenzo et al., 2017; Elmenhorst et al., 2016) and glutamate release (Marquez de Prado et al., 2000) or, alternatively, diminished stress-induced glutamate release during the second scan (DeLorenzo et al., 2017; Lupinsky et al., 2010, 2017). In either scenario, these findings together raise the possibility that changes in extracellular glutamate concentrations affect mGlu5 tracer binding systematically, thereby providing a non-invasive measure of glutamate release.

The gold standard for validating a PET measure of altered extracellular neurotransmitter levels is demonstrating that changes in PET tracer binding are proportional to changes in extracellular transmitter concentration using in vivo microdialysis (Breier et al., 1997; Laruelle, 2012; Laruelle et al., 1997). To apply these techniques to PET-[¹¹C]ABP688, a glutamate-release enhancer is required. Several animal studies have tentatively demonstrated that moderately high doses of aqueous ethanol (EtOH) at a 20% concentration can promote neuronal dose-dependent glutamate release in the ventral striatum in rodents (Fliegel et al., 2013). If confirmed at a dose more suitable for human ingestion, EtOH would be an easily accessible and administered glutamate trigger, fostering the translation of preclinical findings to clinical research.

Based on these observations, we first tested whether a low dose (0.5 g/kg) of 20% EtOH would lead to a glutamatergic response in the ventral striatum, as assessed with in vivo microdialysis. Second, we measured the correlation between changes in EtOH-induced glutamate concentrations in the ventral striatum and alterations in striatal [¹¹C]ABP688 non-displaceable binding potential (BP_ND) values using simultaneous in vivo microdialysis and microPET. Finally, we examined, in healthy humans, the effect of acute stress on [¹¹C]ABP688 BP_ND values, as measured with PET in mGlu5 receptor dense regions (Shigemoto et al., 1993). This latter work was combined with MRS measures of the effect of acute stress on glutamate and glutamine levels in regions with reported stress-induced alterations in glutamate–glutamine balance (Auer et al., 2000; Sheth et al., 2019; Ullmann et al., 2020). The associations between [¹¹C]ABP688 BP_ND values and brain tissue glutamate and glutamine levels were then assessed. A PET ligand sensitive to proportionate changes in endogenous glutamate release in humans would be a valuable new tool.

>99% diastereomerically pure E-isomer, (E)-[¹¹C]ABP688 was produced as previously described by our group (Bdair et al., 2019).

All experimental designs were approved by McGill University Animal Care Committee (UACC; Animal Use Protocol # MNI-7914), in compliance with the guidelines of the Canadian Council on Animal Care (CCAC).

Male Lewis rats at age 60–65 postnatal days (PND) were purchased from Charles River (Saint Constant, QC, Canada) and housed at the Douglas Research Centre Animal Facility or the Center of Neurological Disease Models (CNDM) / McGill University, under 12-h/12-h light/dark cycle (lights open at 7:00 am) in a stress- and noise-free environment. Water and a protein-rich laboratory chow diet were provided ad libitum in their cages. A maximum of two rats were housed in a single standard cage, supplied with a standard environmental enrichment. After arrival, the rats were left to acclimate for at least seven days before use. Lewis strain was selected because it exhibits stronger characteristic neurochemical and behavioral aspects of drug-seeking behavior (Cadoni, 2016).

Anesthesia was induced in rats inside an induction chamber using 4–5% isoflurane in medical air admixture at a flow rate of 2 L/min. The anesthesia was maintained with 2–2.5 isoflurane at the same flow, and rats were then placed on a stereotaxic bed with a pre-warmed pad. The eyes were treated with artificial tears to prevent ocular dryness, the scalp was shaved, and its skin was injected with 0.1 mL of 5 mg/mL bupivacaine intradermally to desensitize the scalp. An incision was made to the scalp using a sharp medical blade, and the skull was cleaned with hydrogen peroxide (H₂O₂). The target location of the cannula in the left ventral striatum (anteroposterior; AP: +1.20 mm, mediolateral; ML: −1.40 mm) was determined following Paxinos coordinates (Paxinos & Watson, 2005). With a handheld drill, apertures for the cannula and three anchor screws were made and the screws were the threaded in the cranium. A 22-gauge quartz guide cannula was inserted 6.00 mm along the dorsoventral (DV) axis and fixed to the cranium using superglue followed by acrylic dental cement. The guide cannula was then capped with a stainless-steel obturator that extends 2.50 mm beyond the end of the cannula to prevent infection and cerebral spinal fluid (CSF) leakage. Post-operatively, the rats were then injected subcutaneously with 20 g/kg of 0.9% NaCl (for one day), and with 2.5 mg/kg carprofen (for 3 consecutive days) to compensate for the lost blood and provide post-operative analgesia, respectively. A topical antibiotic (Polysporin^®) was applied to the wound and the animals were left for at least a week for recovery, housed one per cage, before testing.

In vivo microdialysis was initially conducted as a standalone technique in awake, freely moving rats (n = 7). Following a 20-min habituation period, the microdialysis probe was inserted into the guide cannula and connected from the other end to a Hamilton syringe filled with artificial cerebral spinal fluid (aCSF) and secured to a computer-controlled microinfusion pump (CMA). The aCSF (26 mM NaHCO₃, 1.2 mM NaH₂PO₄, 1.3 mM MgCl₂, 2.3 mM CaCl₂, 3.0 mM KCl, 126 mM NaCl, 0.2 mM L-ascorbic acid) was pumped to the probe at a flow rate of 1 µL/min for a minimum of 1 h to stabilize neurotransmitter levels. Samples were then collected at 20-min intervals, mixed with 1 µL of 0.25 M perchloric acid, and stored in −80°C refrigerator for subsequent HPLC analyses. Following the collection of three baseline dialysate fractions, the animals received an intraperitoneal (i.p.) injection of 0.5 g/kg saline and five dialysate fractions were collected, followed by i.p. injection of 0.5 g/kg 20% EtOH with the collection of eleven dialysate fractions (Fig. 1A).

Furthermore, to study the calcium dependency of the hypothesized EtOH-induced glutamate release, the experiment was repeated with the use of a Ca²⁺-free aCSF instead (n = 4). In the latter, CaCl₂ was replaced with an equimolar concentration of MgCl₂ (final concentration: 3.6 mM). The paradigm was similar to the EtOH challenge microdialysis described above. However, after collecting four baseline samples, the perfusate was switched from aCSF to Ca²⁺-free aCSF using a liquid switch (CMA), following which four fractions were collected. After the animal received an i.p. injection of 0.5 g/kg 20% EtOH, eleven fractions were collected (Fig. 1B).

In vivo microdialysis was conducted simultaneously with microPET scanning in anesthetized animals (n = 11). Herein, animals were induced in a chamber with 4–5% isoflurane in medical air admixture and then maintained with 1.5–2.5% isoflurane at 0.8 L/min flow rate through a nose cone. The rats were placed in prone position on the prewarmed bed of the microPET scanner (CTI, Concorde Microsystems, LLC). Each rat underwent two microdialysis/microPET scans on separate days of < 7 days apart, where the animal received i.p. injection of 0.5 g/kg of either saline or 20% EtOH.

As in the standalone microdialysis study, aCSF was pumped at a flow rate of 1 µL/min and dialysate fractions were collected every 20 min. Subsequent to the collection of four baseline dialysate fractions, rats received an i.p. injection of the pharmacological intervention (i.e., saline or 20% EtOH). The collection of first post-saline or -EtOH dialysate sample started at the moment of injection. The radiotracer was injected intravenously 5 min post-saline or -EtOH injection. The simultaneous microdialysis and microPET study design is depicted in Figure 2.

Measurements of dialysate glutamate concentrations were determined with an HPLC precolumn derivatization with ultimate 3000 RS fluorescence detection (ex: 322 nm; emission: 455 nm) and further described by Lupinsky et al. (2010). The HPLC system consisted of a Dionex pump (ultimate 3000) and a Dionex RS autosampler (ultimate 3000) bundled with a Waters Xterra MS C18 3.0 × 50 mm 5 μm analytical column. The mobile phase is 3.5% MeCN, 20% methanol (CH₃OH), and 100 mmol/L sodium phosphate dibasic (Na₂HPO₄) adjusted to pH 6.7 with 85% phosphoric acid (H₃PO₄). The flow rate was set at 0.5 mL/min. Working standards (100 ng/mL) and derivatization reagents were prepared fresh daily and loaded with samples into a refrigerated (10 °C) Dionex RS autosampler (ultimate 3000). Before injection onto the analytical column, each fraction was sequentially mixed with 20 μL of o-phthaldehyde (0.0143 mol/L) diluted with 0.1 mol/L sodium tetraborate and 20 μL of 3-mercaptopropionic acid (0.071 mol/L) diluted with H₂O and allowed to react for 10 min. After each injection, the injection loop was flushed with 20% CH₃OH to prevent contamination of subsequent samples. Under these conditions, the retention time for glutamate was approximately 1 min with a total run time of 30 min/sample.

All rats (n = 11) underwent two microPET scans, baseline and challenge, conducted between 11:00 and 13:00 (to mitigate potential effects of circadian glutamate variations). The microPET procedure was conducted in the anesthetized rats using microPET R4 scanner (CTI, Concorde Microsystems, LLC; spatial resolution of approximately 1.85 mm (Knoess et al., 2003)). After collecting a minimum of 4 baseline dialysate samples, 0.5 g/kg of either saline (baseline) or 20% EtOH (challenge) was injected intraperitoneally, and the rats were placed in the scanner’s center field of view (FOV). Five minutes after EtOH or saline injection, a 0.5–1.0 mL bolus injection of (E)-[¹¹C]ABP688 at an average dose of 21.83 MBq (range: 19.94–24.98 MBq) was administered intravenously in the lateral tail vein through a pre-inserted catheter, followed by a 60-min dynamic emission acquisition. A total of 27 frames were acquired (9 × 30 s, 6 × 1 min, 5 × 2 min, 7 × 5 min), followed by a 9-min transmission scan using a rotating ⁵⁷Co source.

Magnetic resonance imaging (MRI) structural images for co-registration purposes were obtained using a 7T Bruker Pharmascan pre-clinical MRI system, with a Bruker volume resonator radiofrequency (RF) coil designed for rat brain imaging. Two-dimensional T2-weighted MRI images were obtained with multi-slice TurboRARE acquisition, in-plane resolution of 0.2 × 0.2 mm², and slice thickness of 0.5 mm. The sequence included a TR of 6000 ms, TEeff of 30 ms, RARE factor of 4, 42 slices covering the whole rat brain, 20 signal averages, and total acquisition time of 42 min. The in-plane FOV was 2.50 cm × 3.50 cm.

Images were reconstructed using Maximum a Posteriori (MAP) algorithm with scatter correction, then processed, and analyzed using MINC toolkit software (http://bic-mni.github.io). Co-registration of PET and MRI images was performed with MINC toolkit using both eyes, olfactory bulbs, temporal poles, and base of the skull as registration landmarks. Atlas-based auto-segmentation of various ROIs was performed with ITK-SNAP (v. 3.6.0; http://www.itksnap.org/pmwiki/pmwiki.php) (Yushkevich et al., 2006), based on brain region delineations reported in Waxholm Space atlas of the Sprague Dawley rat brain (Papp et al., 2014), where the MRI images were linearly co-registered to the atlas. Mean BP_ND of ROIs was calculated relative to the non-specific binding in cerebellum, using the simple reference tissue model (SRTM) (Gunn et al., 1997), as previously reported (Elmenhorst et al., 2010). Percent change in [¹¹C]ABP688 BP_ND was calculated as following: ((BP_{ND BASELINE} – BP_{ND CHALLENGE})/ BP_{ND BASELINE}) × 100.

Healthy, right-handed volunteers (5 males and 4 females) aged 25.1 ± 6.0 (Mean ± SD) years old were recruited from the general population using online advertisements on the McGill University website and through classified advertisements. Exclusion criteria included: (1) current or past DSM-5 disorders, including current or past substance use except for occasional cannabis use (< once per month), social tobacco use (< once per week), and occasional drinking (≤ seven drinks per week); (2) family history of DSM-5 disorder; (3) current or past chronic medication use, excluding birth control; (4) significant physical illness in the past 12 months; (5) any history of head injury/loss of consciousness; (6) any counterindications to MRI or PET including claustrophobia, and the presence of a medical condition that makes pain stimuli dangerous (e.g., cardiac disease, hypertension, pulmonary disease, seizure disorder, osteopenia, and anxiety syndromes). The study protocol was approved by the Research Ethics Board of the Montreal Neurological Institute and the Faculty of Medicine and was carried out in accordance with the Declaration of Helsinki. Next, physical health was evaluated by a routine examination, a standard blood work, and electrocardiogram. A urine toxicology test for illicit drugs of abuse (Triage, Biosite Diagnostics, San Diego, CA, responsive to amphetamines, methamphetamines, barbiturates, benzodiazepines, cocaine, 2-ethylene-1,5-dimethyl-3,3-diphenylpyrrolidine (EDDP), opiates, Δ⁹-tetrahydrocannabinol (Δ⁹-THC), and tricyclic antidepressants) and a urine pregnancy test for women was performed on the screening day and prior to each PET session.

All participants underwent two scanning sessions consisting of a 1-h PET scan followed by a 45 min MRI scan with MRS. The two scanning days were conducted in a counter-balanced, within-subjects cross-over design, at least 3 days apart, with each PET and MR scan conducted at the same time of the day to mitigate potential effects of circadian glutamate variations (Fig. 3).

The acute stress stimulus consisted of unpredictable electrical stimulation of the wrist administered immediately below the individual’s pain threshold, defined as the lowest intensity at which a sensation of mild pain is felt. Participants observed 20 s countdowns followed by a blank screen during which electric stimulation occurred 67% of the time, pseudo-randomized. After a 10 s rest, the paradigm was repeated. In 6 min, participants experienced 12 × 30 s blocks, for a total of eight electrical stimulations out of 12 trials. Participants were instructed prior to the stress task that they would receive intermittent electrical stimulation at the level of their threshold. Participants were not given a distinguishing cue to identify whether a stimulation would be followed by a given countdown nor were they informed of the contingency rate. Participants provided verbal ratings of discomfort on the pain scale and visual analog scale (VAS; Fig. 4) every time the stimulus was presented, permitting adjustments to the intensity.

The pain threshold was determined outside of the PET scanner environment as follows. Electric stimulation was initiated with a duration of 200 ms and voltage of 20 V and was increased in increments of 2 V until the lowest voltage at which subjects experienced moderate discomfort was reached. This was defined as a score of 3 on the 5-point pain scale and at least 20 on the VAS. The threshold was re-established on the stress session, immediately prior to the stress task. The pain threshold determined earlier was used as a starting point to minimize the number of shocks administered before the task. The intensity of shock corresponding to the threshold was then used as the target intensity for the stimulation during the stress task.

To assess the effectiveness of the stimuli, variation of the electrical conductance in response to skin secretion was tracked continuously throughout stress/rest tasks and PET scanning sessions. First, phasic deflections in the skin conductance responses (SCRs) were analyzed (Pitman & Orr, 1986). The SCR amplitudes during the stress task were calculated by subtracting the mean skin conductance level 2 s before expectation of a shock, from the peak value obtained immediately after administration of the shocks. To avoid habituation (Davis, 1934), phasic increases occurring over the first five trials of shocks were taken into account. The same calculation was applied over the first three periods of blank screens in which shock was expected but not triggered. During the same time interval at rest, amplitudes of non-specific SCRs (occurring in the absence of stimuli) were calculated. Lastly, tonic skin conductance was compared: three time intervals were averaged, including (1) stress task or rest, (2) the first 5 to 10 min of PET, and (3) the last 5 min of PET (Fig. 3). The time interval (2) corresponds to the peak of the hypothalamus-pituitary-adrenal (HPA) axis response in response to stress, which is expected to begin 20 min after the initiation of the stimulus (t2).

The HPA axis response to stress was assessed by measuring cortisol concentrations from saliva samples collected using oral swabs (Salimetrics, LLC). In total, six saliva samples were collected over 120 min: at baseline (t₁ and t₄), at two time points after initiation of the task (t₂ and t₅: 20 min after initiation of the task), and at the end of each scan (t₃ and t₆). Area under the curve with respect to ground (AUC_G), and with respect to increase from cortisol value at t₁ (AUC_I) were calculated as described (Pruessner et al., 2003). AUC_G is the total area under the curve, whereas AUC_I is calculated with reference to the first value (cortisol value at t₁). Samples were stored at −20 °C until biochemical analysis took place.

Subjective ratings of mood, anxiety, and alertness were measured using the state-trait anxiety inventory (STAI)-State (Spielberger, 1983), and visual analog scale of alertness. Each scale was collected two times each session, before and immediately after the first stress task (Fig. 3).

All PET scans were performed using a High-Resolution Research Tomograph (HRRT; Siemens/CTI, Knoxville, TN, USA) at the Montreal Neurological Institute. Scans consisted of a 60-min dynamic acquisition collected in list-mode format, followed by a 6-min ¹³⁷Cs rotating point source transmission scan for attenuation correction. The acquisition was binned into frames, the durations of which consisted of the following sequence: 3 × 10 s, 5 × 30 s, 4 × 60 s, 4 × 120 s, 5 × 300 s, 2 × 60 s. The scan initiated concurrently with the beginning of the venous injection of an average dose of 385.54 MBq (range: 333–407 MBq) of (E)-[¹¹C]ABP688 through an intravenous catheter installed at the participant’s right arm vein (antecubital region).

For PET/MR co-registration and spectroscopy voxel placement, all subjects also underwent a high-resolution T1-weighted MRI scan after each PET session. Scans were acquired in a 3T Siemens TRIO Magneton scanner (Siemens Medical Solutions, Erlangen, Germany) using an ADNI-3D MPRAGE protocol. Images were acquired in 3D repetition time (TR) = 2300 ms, echo time (TE) = 3.42 ms, flip angle = 9°, field of view = 256 mm, and FOV = 256 × 256; 1 mm resolution isotropic resolution.

During the same session, MR spectroscopy scanning was conducted in two volumes of interests from which measures of combined glutamate (Glu) and glutamine (Gln), referred to as Glx, and Glu alone were obtained. Spectroscopic voxels were prescribed from anatomic images: a 20 × 15 × 10 mm³ voxel was placed bilaterally over the anterior cingulate cortex (ACC), immediately anterior to the rostrum of the corpus callosum, and perpendicular to the infra-callosal line. The striatum voxel was 25 × 12 × 12 mm³ in size, encompassing the right dorsal caudate-putamen. The water-suppressed proton spectra were acquired using a 90˚–180˚–180˚ (PRESS) sequence (TR = 3000 ms, TE = 40 ms), giving a total of 196 acquisitions. A water-unsuppressed reference scan to enable correction for eddy current-induced phase shifts was obtained immediately after the water-suppressed scan using the same TR, TE, voxel position, and shim settings with 16 acquisitions.

CIVET pipeline (https://www.bic.mni.mcgill.ca/ServicesSoftware/CIVET) was used to preprocess the native MRI image. The resampled images were then classified into white matter (WM), grey matter (GM), and CSF; segmented in the main brain structures; and automatically labeled, using the ANIMAL probabilistic atlas-based algorithm (Collins et al., 1999). Then, the PET images were co-registered with the subject’s own MRI transformed into the Montreal Neurological Institute (MNI) template brain using transformed parameters obtained from the registration of MRI to MNI152 space. Mean BP_ND of ROI was estimated using SRTM (Gunn et al., 1997), with the cerebellar grey matter as reference region. Subcortical limbic regions amygdala and hippocampus were yielded by the segmentation generated by the ANIMAL image registration algorithm. A standard mask was used to functionally segment the striatum into ventral (VST), associative (AST), and sensorimotor (SMST) subregions, as proposed by Mawlawi et al. (2001). Remaining cortical ROIs were manually drawn on a template MRI in stereotaxic space using the software DISPLAY (http://www.bic.mni.mcgill.ca/software/Display/Display.html) and were based on the approach defined by Abi-Dargham et al. (2000). Percent change in [¹¹C]ABP688 BP_ND was calculated as following: ((BP_{ND STRESS} – BP_{ND REST})/ BP_{ND REST}) × 100.

Analyses were performed using SPSS software. (Version 29). Differences were considered statistically significant at p < 0.05. Shapiro–Wilk tests established normal distribution of all data.

For standalone microdialysis study, the effects of EtOH administration were tested using a one-way repeated-measures analysis of variance (rmANOVA) with time as the within-group factor containing 17 levels (pooled baseline and each subsequent fraction; five saline samples and eleven EtOH samples). Average values were extracted for each condition (baseline: B1-3; saline: S1-5; EtOH: E1-6) and compared using a one-way rmANOVA with condition as the within-group factor containing three levels. Both ANOVAs contained tests for violations of sphericity; Huynh-Feldt corrections were applied when necessary. Bonferroni-corrected paired t-tests were then used to compare the three experimental conditions.

In the simultaneous microdialysis and microPET study, a two-way within-subject rmANOVA test was performed to study the effect of treatment (saline vs. EtOH) over time (five levels; pooled baseline and four subsequent posttreatment fraction) on extracellular glutamate levels in the ventral striatum, as assessed with microdialysis. Regarding PET images, a one sample (i.e., paired) t-test was conducted to determine the percent difference in [¹¹C]ABP688 BP_ND in the ventral striatum. Pearson’s r test was then used to examine the correlation between the percent difference between extracellular glutamate concentrations (saline vs. EtOH) and [¹¹C]ABP688 displacement. Percent change of individual dialysate fraction (averaged across all animals) was calculated relative to pooled baseline fractions. Peak percent of baseline is defined as the dialysate fraction with the highest extracellular glutamate concentration compared to pooled baseline fractions.

For the human study, the effect of stress on subjective anxiety and physiological measurements (cortisol and IL1- β) were identified using two-way rmANOVAs or mixed-model analyses when data were missing, with sessions (rest vs. stress) and timepoints as within-subject factors. Simple-main effects analyses followed when indicated. Planned pair-wise t-tests were carried out to identify differences in the magnitude of SCR and non-stimuli SCR relative to the non-specific SCRs. Summary BP_ND values were computed as the unweighted mean of all examined regions in order to assess the effects of tracer and scan characteristics (mass of tracer injected per kilogram body weight and time of injection). Relationships between BP_ND and scans characteristics were assessed using Pearson’s r. To test the main hypothesis of differences in BP_ND between conditions, separate Condition x Region x Hemisphere repeated-measures ANOVAs were performed for (i) striatal regions (VST, AST, SMST), (ii) prefrontal regions (medial (mPFC), dorsolateral (dlPFC) prefrontal cortices, orbitofrontal cortex (OFC), and ACC), and (iii) limbic regions (amygdala and hippocampus). These were followed by planned, uncorrected two-tailed dependent measures t-tests to assess each contrast in the selected ROIs between conditions. For each ROI, percent change from scan 1 to scan 2 ((BP_{ND STRESS} – BP_{ND REST})/ BP_{ND REST} × 100%) was calculated for each participant. Parametric maps of BP_ND were compared in voxel-wise paired t-tests from scan 1 to scan 2 in each participant using RMINC with a significance threshold of p < 0.05, corrected for false discovery rate. To determine the significance of detected metabolites concentration differences due to shock administration, a Condition (rest, stress) by Region (ACC, striatum) two-way rmANOVA was applied to the MRS data. Finally, potential associations of mGlu5 receptor availability with behavioral and physiological variables were examined using Pearson’s r. Given the large number of correlations performed, the unadjusted alpha level was divided by the number of studied ROIs, which resulted in a significance threshold of p = 0.05/9 = 0.0056. In a secondary voxel-wise analysis, further exploratory correlations using mGlu5 binding across the whole brain were assessed with parameters which revealed to be significantly correlated with ROI-wise BP_ND.

The diastereomeric excess (d.e.) and radiochemical purity (RCP) of (E)-[¹¹C]ABP688 were both >99%. Mean molar activity (A_m) at time of injection was 83.24 GBq/µmol (range: 28.46–268.28 GBq/µmol, n = 22) in the animal study and 91.71 GBq/µmol (range: 23.1–163.6 GBq/µmol; n = 18) in the human study.

As depicted in Figure 5A, there were main effects of time (F(4.891,29.346) = 2.901, p = 0.035) and condition (F(1.240,7.439) = 10.621, p = 0.010) in awake rats. This reflected significant increases in extracellular glutamate concentrations following EtOH administration (E1 – E5) as compared to samples collected following the saline injection (S1 – S5: t(6) = 3.322, p = 0.048) and the pre-injection samples (B1 – B3: t(6) = 3.372, p = 0.045). As expected, no differences were seen between the baseline and saline conditions (p > 0.50). Compared to the average saline response, the largest effect occurred during the third dialysate sample collected between 40–60 min after EtOH injection (E3) (t(6) = 2.458, uncorrected p = 0.049). During this E3 fraction, extracellular glutamate concentrations reached 190.5 ± 34.7% of average baseline and 212.05 ± 47.52% of average saline. This effect of EtOH was abolished when using the Ca²⁺-free aCSF (Fig. 5B).

During the simultaneous microdialysis and microPET study in anesthetized rats, a two-way rmANOVA yielded a significant treatment (saline vs. EtOH) by time interaction (F(2.505,25.046) = 3.353, p = 0.042; n = 11). Post-hoc pairwise comparisons revealed a significant difference between the first two post-treatment dialysate fractions from the pooled baseline (ps < 0.05, uncorrected), and there was a significant difference between the effects of EtOH and saline in the first fraction (S1 vs. E1: (t(10) = –5.066, p < 0.001) reflecting concentrations that were 126.9 ± 5.3% of baseline (Fig. 6).

The two PET scans did not significantly differ (ps > 0.05) in the injected (E)-[¹¹C]ABP688 dose (baseline: mean 22.24 MBq, range 20.72–24.98 MBq; challenge: mean 21.43 MBq, range 19.94–24.98 MBq), A_m (baseline: mean 55.79 GBq/µmol, range 28.46–131.15 GBq/µmol; challenge: mean 107.95 GBq/µmol, range 37.72–269.28 GBq/µmol), injected volume (baseline: mean 0.62 mL, range 0.5–1.0 mL; challenge: mean 0.58 mL, range 0.5–0.75 mL), injected mass (baseline: mean 1.3 pmol/g, range 0.5–2.3 pmol/g; challenge: mean 0.9 pmol/g, range 0.3–2.0 pmol/g), start time (baseline: mean 12:22, range 11:09–13:15; challenge: mean 12:24, range 11:13–13:29), or weight of the animals (baseline: mean 306.8 g, range 276–384 g; challenge: mean 311.3 g, range 266–349 g). The average age of the animals was 83.5 ± 7.7 (Mean ± SD) at the time of the baseline scan, and 87.8 ± 7.1 days (Mean ± SD) at the time of the challenge scan.

There was a significant percent reduction in striatal [¹¹C]ABP688 BP_ND in response to EtOH administration compared to saline, corresponding to a percent change of 6.8 ± 9.6% (baseline BP_ND: mean 4.78, range 4.18–5.83; challenge BP_ND: mean 4.45, range 3.23–5.85; t(10) = 2.424, p = 0.036) (Fig. 7). The ratio of percent glutamate increase during E1 (compared to S1) to percent [¹¹C]ABP688 BP_ND reduction was 4:1. Percent changes in extracellular glutamate concentrations in the ventral striatum were not significantly correlated with percent changes in striatal [¹¹C]ABP688 BP_ND (r = 0.25, p = 0.46).

Skin conductance responses were significantly higher following exposure to the stressor (t(7) = 4.65, p = 0.0023). Numerically greater increases also occurred when shocks were expected, but not given, but this response was more variable and at the trend level only (t(7) = 2, p = 0.09) (Fig. 8). Twenty minutes following initiation of the task, skin conductance levels remained elevated (t(7) = 3, p = 0.04) before normalizing after an hour (t(7) = 0.047, p = 0.96).

AUC_I between stress and rest were significantly different (t(7) = 2, p = 0.041), reflecting increased versus decreased cortisol levels during the stress (AUC_{I STRESS} = 6.97) versus rest session (AUC_{I REST} = −6.83). The stress-induced percent increases in cortisol and skin conductance levels were significantly correlated (r = 0.838, p = 0.009).

Significant Condition × Timepoint interactions were obtained for the “alert” (F(1,7) = 9.471, p = 0.018) and “anxious” VAS measures (F(1,7) = 6.25, p = 0.041), but not “afraid” (F(1,7) = 3.3, p = 0.11). Further inspection of the data confirmed that in the stress condition, alertness and anxiety ratings were higher post-stress (t2) relative to pre-stress (t1) (alert: t(7) = 3.784, p = 0.0137; anxious: t(7) = 3.005, p = 0.0396, respectively; Fig. 9).

The two PET test sessions did not differ in injected tracer dose (rest: mean 378.88 MBq, range 355.2–392.2 MBq; stress: mean 388.87 MBq, range 333–407 MBq; t(8) = −1.16, p = 0.28), A_m (rest: mean 89.32 GBq/µmol, range 23.1–128 GBq/µmol; stress mean 94.1 GBq/µmol, range 24–163.6 GBq/µmol; t(8) = 1.204, p = 0.26), or start time (rest: mean 12:30, range 11:07–15:03; stress: mean 12:37, range 11:03–14:08; t(8) = −0.28, p = 0.79). Global BP_ND values were not related to the mass of [¹¹C]ABP688 injected (r = 0.22, p = 0.37) or time of injection (r = 0.27, p = 0.79).

Three-way Condition x Subregion x Hemisphere repeated-measures ANOVAs did not identify a significant main or interaction effect (Fs < 0.97, ps > 0.41). Controlling for test session order (stress session in the first scan vs. second scan) did not affect the results. Likewise, replacing the “Condition” factor by the “Day” factor (first scan vs. second scan) did not change the results. Percent change in BP_ND was calculated and averaged across all ROIs within a subject, which ranged from −17.5% to 18.6%. A global tendency of increase was found across regions between conditions, but this did not reach statistical significance in post-hoc pairwise comparisons (ps > 0.3, uncorrected). Voxel-wise parametric analyses were consistent with these findings, with no clusters of significant voxels emerging.

Analysis of the combined Glutamate + Glutamine (Glx) levels did not yield a significant main effect of Session (F(1,7) = 1.09, p = 0.33) but a trend level Region x Session interaction was seen (F(1,7) = 0.16, p = 0.08). Post-hoc exploratory tests yielded evidence of significant stress-induced increases in Glx concentrations in the striatum (13% increase, p = 0.048) but not in the ACC (2% increase, p = 0.5; Fig. 10).

In humans, there were no significant associations between stress-induced changes in BP_ND values and stress-induced changes in Glx/Glu ratios or between BP_ND values and Glx/Glu ratios at rest. However, BP_ND values on the stress session (BP_{ND STRESS}) in limbic, sensorimotor, and associative striatum, OFC, and left amygdala were all negatively correlated with stress-induced changes in Glx/Glu levels in the ACC (rs > −0.71, ps < 0.044, uncorrected). BP_{ND STRESS} values were also associated with Glx/Glu levels in the ACC at stress in the ACC and both associative and sensorimotor striatum (rs > −0.43, ps < −0.71, ps < 0.048, uncorrected), and to a lesser extent in the hippocampus (r = −0.7, p = 0.054, uncorrected). Stress-induced changes in the cortisol AUC_G were also correlated with BP_{ND STRESS} values in the striatum, OFC, amygdale, and hippocampus (rs < −0.72, ps < 0.045, uncorrected). Correlations that survived at p = 0.0056 are shown in Figure 11.

The present series of experiments yielded four novel findings. First, as predicted, administration of 0.5 g/kg of 20% EtOH led to a doubling of extracellular glutamate concentrations within the ventral striatum. This contrasts with smaller glutamate effects produced by higher EtOH doses, as predicted by Fliegel et al. (2013). By extrapolating the trend line constructed from the ventral striatal glutamatergic responses to higher EtOH doses (1, 2, and 3 g/kg) (Fliegel et al., 2013), they hypothesized that a lower dose (i.e., 0.5 g/kg) would lead to a glutamatergic response of about 200%. Second, this glutamatergic effect of EtOH relies on Ca²⁺-dependent exocytotic release, implicating a neuronal source of the transmitter. Third, the low-dose EtOH challenge induced significant decreases in striatal [¹¹C]ABP688 BP_ND values in rodents. Fourth, changes in BP_ND values did not systematically covary with the changes in dialysate glutamate concentrations in rats or MRS measured indices of glutamate turnover in humans.

The mechanism by which low-dose EtOH induces a glutamatergic response is not fully understood. One possibility is that EtOH acts as a negative modulator of the ionotropic glutamatergic receptors, N-methyl-D-aspartate (NMDA). Drawing such a conclusion is plausible since ketamine, a non-competitive NMDA receptor antagonist, was found to elicit a similar dose-dependent glutamatergic effect to that proposed with EtOH (Moghaddam et al., 1997). Furthermore, there is evidence that NMDA receptor antagonism is linked to the activation of the glutamatergic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (Maeng et al., 2008) via predominately decreased activity of inhibitory γ-aminobutyric acid (GABA) interneurons (Homayoun & Moghaddam, 2007).

Contrary to what was observed in awake animals, the glutamatergic responses in anesthetized rats peaked at approximately 127% of saline during the E1 fraction instead of continuing to grow through to the E3 fraction. This curtailed response likely reflects an effect of the anesthetic, isoflurane. Isoflurane is believed to abate excitatory glutamatergic transmission via decreasing synaptic glutamate release (Larsen et al., 1998) and increasing uptake by isolated nerve terminals (Larsen et al., 1997) and astrocytes (Miyazaki et al., 1997). The glutamate-suppressing effect of isoflurane could also explain why ketamine administration causes a significant reduction in [¹¹C]ABP688 binding in humans (DeLorenzo et al., 2015; Esterlis et al., 2018), but not in anesthetized rodents (Kosten et al., 2018). Nevertheless, despite the relatively limited increase in EtOH-induced glutamate release in anesthetized rodents, we observed a significant EtOH-induced reduction in striatal [¹¹C]ABP688 BP_ND. This adds to the evidence that striatal [¹¹C]ABP688 binding responds to changes in glutamate concentrations with a microdialysis to PET ratio of 4:1. This change compares quite favorably to the 44:1 and 64:1 ratios seen for [¹¹C]raclopride and dopamine release responses induced by 0.2 and 0.4 mg/kg of d-amphetamine, respectively (Breier et al., 1997). Despite this, the percent change in striatal [¹¹C]ABP688 BP_ND did not correlate with changes in extracellular glutamate concentrations (r = 0.25, p = 0.46). This might be attributed to the small increase in EtOH-induced glutamate release (~27%), which in turn could have a limited effect on [¹¹C]ABP688 displacement from mGlu5 receptors. Indeed, the percent change in EtOH-induced decreases in [¹¹C]ABP688 BP_ND is in the range of test-retest variability for the striatum previously reported in rodents (Elmenhorst et al., 2012) (6.8% vs. 4.8%). In comparison, d-amphetamine (0.2 and 0.4 mg/kg) induces larger effects on both [¹¹C]raclopride binding (10.5% and 21.3%) and dopamine release (459% and 1365%), respectively (Breier et al., 1997).

Since [¹¹C]ABP688 and glutamate do not share the same binding sites, our microdialysis/microPET findings suggest that glutamate binding to the active orthosteric site causes indirect displacement of the radiotracer from the allosteric site. It is hypothesized that glutamate binding to the orthosteric site causes either alteration in the affinity of the allosteric site to [¹¹C]ABP688 or internalization of mGlu5 receptors, precluding the radiotracer from binding to the allosteric site.

A limitation of our study is that the microdialysis measurements represent changes in ventral striatum glutamate concentrations, whereas the PET signal represents the full striatum. This noted, Fliegel et al. (2013) hypothesized that a low dose of EtOH (i.e., 0.5 g/kg) would increase glutamate release throughout the entire striatum (i.e., ventral striatum and caudate-putamen). Future studies could also address whether the EtOH-induced glutamate response from neurons reflects impulse-dependent activity; this could be tested by measuring the effect of tetrodotoxin (TTX), an Na⁺ channel blocker that inhibits the firing of action potentials (Lupinsky et al., 2010). A second limitation to our animal studies is that only male rats were used. Future studies will be needed to examine [¹¹C]ABP688 sensitivity to glutamate release in females.

In our study in humans, a series of electric shocks was sufficient to increase striatal Glx:Glu ratios, putatively reflecting elevated glutamine metabolism (Yuksel & Ongur, 2010) following stress-induced glutamate release, enhanced glial glutamate reuptake, and subsequent conversion to glutamine. However, these effects were not large. Potentially related to this, exposure to the laboratory stressor did not lead to significant changes in [¹¹C]ABP688 BP_ND values. This might reflect marked individual differences in the responses rather than statistical noise. Indeed, exploratory analyses identified negative correlations between stress-induced increases in salivary cortisol and BP_ND values in the amygdala, ACC, OFC, and limbic striatum on the stress session. Since, however, correlations were not seen with stress-induced changes in BP_ND values, the above associations should be interpreted cautiously. Both men and women were tested, but the modest sample size precluded sufficient statistical power to test for a possible effect of sex.

Although our study reduced time-related scan variability as much as possible, both across sessions and between individual participants, one participant underwent a control scan much later than the stress scan due to a tracer production failure. This same person exhibited higher binding at rest compared to other participants, which could be attributed to circadian rhythm effects, causing diurnal variations in [¹¹C]ABP688 binding as previously reported in animal and human studies (DeLorenzo, Kumar, et al., 2011; Elmenhorst et al., 2016).

We found associations between [¹¹C]ABP688 BP_ND values on the stress session with ACC Glx:Glu values at stress, as well as with stress-induced changes in ACC Glx:Glu values. However, these correlations did not survive correction for multiple comparisons, somewhat limiting the generalizability of the findings. Supporting this caution, mGlu5 receptor availability and MRS measures of glutamate have been measured in cocaine-dependent and healthy subjects (Martinez et al., 2014) and correlations between the two modalities were not found. In comparison, a second study identified a correlation between MRS measured glutamate turnover in the ACC and [¹⁸F]FPEB binding, another PET tracer that binds to the same site as [¹¹C]ABP688, in patients with major depression (Abdallah et al., 2017). However, these results likely reflected long-term effects of elevated glutamate levels on receptor availability, an interpretation supported by post-mortem tissue evidence that patients with a history of depression have reduced mGlu5 protein expression (Deschwanden et al., 2011).

Lastly, BP_ND values calculated using the cerebellum as a reference region correlate highly with values derived with arterial input (Michele et al., 2024) but low levels of mGlu5 receptor specific binding have been identified in the human cerebellum (Michele et al., 2024), raising the possibility that this could produce systematic biases (DeLorenzo, Milak, et al., 2011; Kagedal et al., 2013), potentially decreasing the ability to identify associations. This hypothesized effect would not influence our rodent data since their cerebellum has negligible mGlu5 receptors (Ametamey et al., 2006; Kessler, 2004; Wyss et al., 2007).

The present study has demonstrated that a low dose (0.5g/kg) of 20% EtOH causes a significant calcium-dependent increase in ventral striatum glutamate release in awake animals. Although the EtOH-induced glutamatergic response was less pronounced in anesthetized rats, this smaller increase in glutamate release was accompanied by a significant reduction in striatal [¹¹C]ABP688 BP_ND with a ratio of 4:1. Nevertheless, a significant correlation was not observed between changes in ventral striatum glutamate release and striatal [¹¹C]ABP688 binding. Lastly, no correlation was observed between stress-induced changes in Glx/Glu and [¹¹C]ABP688 BP_ND in humans. Together, these findings suggest that [¹¹C]ABP688 binding could be affected by moderate fluctuations in extracellular glutamate release but does not provide a proportional measure. Future PET studies of mGlu5 receptor availability should use highly controlled testing conditions to avoid these effects.

Data are available upon request from the corresponding author.

M.L. conceived of the studies in laboratory animals; H.B. wrote the ethics protocol of the animal study and conducted the simultaneous microdialysis–microPET study in rodents, including performing animal surgeries, PET and MRI scanning, and analyzing PET and MRI images; S.P. carried out the standalone microdialysis study in rodents; C.B. conceived of the experiment in humans; M.S.-F. conducted the human study; L.M. built the microdialysis probes inhouse and analysed the dialysate samples using resources and equipment from the laboratory of A.G.; M.S.K. assisted in analysing human and animal PET and MRI images using resources from the laboratory of P.R.-N.; A.A. assisted in animal microPET and MRI scanning; A.N.-S. assisted in writing the ethics protocol for the human study, recruiting human participants and analyzing human PET images; K.S. wrote the ethics protocol for the human study; J.N. provided assistance in analyzing MRS images; A.A.-V. analyzed human saliva samples; S.C. contributed to the writing of ethics protocol for the human study and analyzed human PET and MRI images; G.M., J.-P.S., and A.K. provided access to the radiotracer, [¹¹C]ABP688; H.B. drafted the manuscript for the findings of both animal studies, including creation of images and figures; M.S.-F. contributed to the writing of the human study section in the manuscript; M.L. contributed to the writing of the manuscript and provided critical feedback; and all co-authors critically reviewed the manuscript.

Authors have nothing to declare.

We extend our thanks to the PET Radiochemistry team at the Montreal Neurological Institute for providing us access to the tracer, and to the human and small animal PET scanners. We would also like to thank Dr. David Rudko and Marius Tuznik for their assistance with the animal MR imaging. The study benefited from the financial support of Health Canada, through the Canada Brain Research Fund, an innovative partnership between the Government of Canada (through Health Canada) and Brain Canada, and the Montreal Neurological Institute.