PET imaging of M4 muscarinic acetylcholine receptors in rhesus macaques using [¹¹C]MK-6884: Quantification with kinetic modeling and receptor occupancy by CVL-231 (emraclidine), a novel positive allosteric modulator

Abstract

Stimulation of the M₄ muscarinic acetylcholine receptor reduces striatal hyperdopaminergia, suggesting its potential as a therapeutic target for schizophrenia. Emraclidine (CVL-231) is a novel, highly selective, positive allosteric modulator (PAM) of M4 muscarinic acetylcholine receptors i.e. acts as a modulator that increases the response of these receptors. First, we aimed to further characterize the positron emission tomography (PET) imaging and quantification performance of a recently developed M₄ PAM radiotracer, [¹¹C]MK-6884, in non-human primates (NHPs). Second, we applied these results to determine the receptor occupancy of CVL-231 as a function of dose. Using paired baseline-blocking PET scans, we quantified total volume of distribution, binding potential, and receptor occupancy. Both blood-based and reference region-based methods quantified M₄ receptor levels across brain regions. The 2-tissue 4-parameter kinetic model best fitted regional [¹¹C]MK-6884-time activity curves. Only the caudate nucleus and putamen displayed statistically significant [¹¹C]MK-6884 uptake and dose-dependent blocking by CVL-231. For binding potential and receptor occupancy quantification, the simplified reference tissue model using the grey cerebellum as a reference region was employed. CVL-231 demonstrated dose-dependent M₄ receptor occupancy in the striatum of the NHP brain and shows promise for further development in clinical trials.

Introduction

Muscarinic acetylcholine receptors (mAChRs) are G-protein-coupled receptors implicated in modulating complex behaviors such as cognition, learning, memory, motivation, motor control, and reward. These receptors are involved in signal transduction of acetylcholine (ACh), a neurotransmitter and neuromodulator of the cholinergic system signaling via three major components in the central nervous system (CNS): (1) projections from the brainstem pedunculopontine and lateral dorsal tegmental nuclei innervating the thalamus, midbrain, hindbrain, and cerebellum; (2) projections from the basal forebrain nuclei innervating neocortex, hippocampus, and thalamus; and (3) interneurons present in the dorsal and ventral striatum and to a lower extent in the nucleus accumbens.^1,2 Among the five metabotropic subtypes of mAChRs (M1-M5), the M4 mAChR is highly expressed in striatal regions, and more moderately in the cortex, thalamus, and hippocampus.^3,4 The M4 mAChR is coupled to G-protein α_i/o subunit to inhibit adenylyl cyclase and modulate ion channels upon binding a substrate (e.g., ACh) at the orthosteric binding site. Activation of the M4 mAChR can trigger various pathways depending on the neurons expressing the receptor. Notably, its activation was found to reduce striatal dopamine signaling suggesting antipsychotic-like effects. ⁵

Striatal hyperdopaminergia is thought to be an underlying cause of psychotic events in schizophrenia characterized by delusions, hallucinations, and disorganized speech and behavior.^6,7 Neuropsychiatric symptoms also occur in Alzheimer’s disease and other dementias, sometimes even in the early disease stages. ⁸ Therefore, there is a high clinical significance in developing brain penetrant M4-selective mAChR therapeutics to exploit the potential antipsychotic effects of M4 activation and thus alleviate these symptoms. Unfortunately, the orthosteric binding sites of mAChRs present a high degree of homology across receptor subtypes, ⁹ thereby significantly reducing the possibility of developing M4 agonists with sufficient selectivity to avoid adverse side effects (e.g., cardiovascular, thermoregulation, visual, and gastrointestinal disturbances) mainly attributed to the activation of M2 and M3 subtypes in the periphery.^{10 –13} On the other hand, the allosteric binding sites are significantly less structurally conserved across receptor subtypes thus offering opportunities for developing highly subtype-specific allosteric modulators. ¹⁴ In recent years, several M4-selective positive allosteric modulators (PAMs), i.e. modulators that increase the response of the receptor (either by increasing the affinity or the efficacy of an agonist acting at the orthosteric site), have undergone pre-clinical assessment.^{15 –19} Among these compounds, Emraclidine (CVL-231) is currently in development for the treatment of schizophrenia and Alzheimer’s disease psychosis. CVL-231 has shown encouraging initial results in Phase I clinical trials with a favorable safety and side-effect profile in patients with schizophrenia. ²⁰

Positron emission tomography (PET) is a powerful tool that can be used for in-vivo mapping and quantification of a specific target as well as for the assessment of target engagement and receptor occupancy (RO) of a drug candidate. The use of PET relies on the availability of a radiotracer with suitable physico-chemical properties, such as sufficient affinity and selectivity for the target of interest. Much effort has been made toward the development of subtype-selective mAChRs radiotracers, including for M4.^{21 –23} Recently, Tong et al. reported the discovery of [¹¹C]MK-6884, a promising M4 PAM PET radiotracer that showed good brain penetration, high binding affinity and good selectivity to an M4 allosteric site. ²³ Subsequently, further PET imaging in non-human primates (NHP) and then first-in-human studies revealed regional distribution consistent with M4 mAChR localization in vivo and competitive binding interactions with other M4 PAMs. Moreover, [¹¹C]MK-6884 displayed binding enhancement via pharmacological manipulations at the orthosteric site, which was consistent with the known properties and cooperative mechanism of a M4 PAM. ²⁴ While [¹¹C]MK-6884 has demonstrated promising properties for PET imaging of M4, there is still relatively limited in vivo data available regarding its pharmacokinetic characterization. A comprehensive assessment based on different kinetic modeling techniques would help find the optimal data analysis strategy, thus facilitating evaluation of novel M4 targeting drug candidates at various stages of drug development.

The purpose of the present work is two-fold. First, we evaluate the pharmacokinetic properties of [¹¹C]MK-6884 in NHPs and investigate different kinetic modeling strategies for quantifying [¹¹C]MK-6884 uptake. Second, we employ [¹¹C]MK-6884 in quantifying target engagement and RO by CVL-231 on M4 receptors in the NHP brain.

Material and methods

Study design

The study diagram is depicted in Supplemental Figure 1. Briefly, a typical imaging day of testing a single CVL-231 dose consisted of a 90-minute baseline [¹¹C]MK-6884 PET/CT scan, followed by pre-treatment with CVL-231, and then another 90-minute [¹¹C]MK-6884 PET/CT scan. The two radiotracer injections were separated by 3 hours to allow for sufficient radioactive decay. The radiochemical synthesis of [¹¹C]MK-6884 is described in the supplemental material.

Animals

Two male adult Rhesus macaques (monkey 1 and monkey 2) were used in this study (ages 9 and 13 years old, respectively). Animal body weights on the day of imaging were 12.9 kg on average for monkey 1 (range 12.2–13.4 kg) and 15.1 kg for monkey 2 (consistently across studies).

Compliance

All experiments involving non-human primates were performed in accordance with the U.S. Department of Agriculture (USDA) Animal Welfare Act and Animal Welfare Regulations (Animal Care Blue Book), Code of Federal Regulations (CFR), Title 9, Chapter 1, Subchapter A, Part 2, Subpart C, §2.31. 2017. Experiments were approved by the Institutional Animal Care and Use Committee at Massachusetts General Hospital. The animal data reporting of the current study has followed the ARRIVE 2.0 guidelines (PMID: 32663096).

Animal preparation

Prior to each imaging session, animals were sedated with ketamine/xylazine (10/0.5 mg/kg IM) and were intubated for maintenance of anesthesia with isoflurane (1–2% in 100% medical-grade O₂). One venous catheter was placed in the saphenous vein for radiotracer injection, another catheter was placed in the saphenous vein of the contralateral hind limb for administration of the blocking agent, and an arterial catheter was placed in the posterior tibial artery for blood sampling. The animal was positioned on a heating pad on the bed of the scanner in the supine orientation for the duration of the study.

Magnetic resonance imaging

For both monkeys, a three-dimensional structural T1-weighted magnetization-prepared rapid gradient-echo (MEMPRAGE) magnetic resonance (MR) image was acquired for anatomical reference with acquisition parameters as described previously. ²⁵

Positron emission tomography imaging

The two animals (monkey 1 and monkey 2) were scanned on a Discovery MI (GE Healthcare) PET/CT scanner. Monkey 1 underwent three paired baseline-blocking studies, and monkey 2 underwent four. For each animal, studies were separated by more than a month. A CT scan was acquired prior to each PET acquisition for attenuation correction. Emission PET data were acquired in 3 D list mode for 90 minutes following injection of [¹¹C]MK-6884. Radiotracer was administered via the lateral saphenous vein over a 3-minute infusion followed by a 3-minute infusion of a saline flush. All injections were performed using syringe pumps (Medfusion 3500). Dynamic PET data were reconstructed using a validated fully 3 D time-of-flight iterative reconstruction algorithm using 3 iterations and 34 subsets while applying corrections for scatter, attenuation, deadtime, random coincident events, and scanner normalization. For all dynamic scans, list mode data were framed into dynamic series of 6 × 10, 8 × 15, 6 × 30, 8 × 60, 8 × 120, and 12 × 300 s frames. Final reconstructed images had voxel dimensions of 256 × 256 × 89 and voxel sizes of 1.17 ×1.17 × 2.8 mm³.

Arterial blood sampling

Arterial blood sampling was performed during each dynamic PET acquisition. Samples of 1 to 3 mL were initially drawn every 30 seconds after the radiotracer injection and decreased in frequency to every 15 minutes by the end of the scan. [¹¹C]MK-6884 in vivo metabolism was characterized from blood samples collected at 5, 8, 10, 15, 30, 60, and 90 minutes. An additional blood sample of 3 mL was drawn immediately prior to tracer injection to determine the plasma free fraction (f_p) of [¹¹C]MK-6884.

Arterial blood analysis

Radioactivity concentration (in Bq/mL) was measured in whole blood (WB) and subsequently in plasma (PL), following WB centrifugation. Radiometabolite analysis was performed using an automated column-switching radioHPLC system.^26,27 Eluent was collected at 1-minute intervals and assayed for radioactivity using a Wallac Wizard γ counter (1470 model). The procedure for these measurements was similar to what we have described earlier, except for the mobile phases. ²⁵ Injected plasma samples were initially trapped on a catch column using a mobile phase consisting of 99% water, 1% acetonitrile at 1.8 mL/min, which was backflushed with 50% 40 mM ammonium acetate, 50% acetonitrile at 1 mL/min. The sample was directed onto an Xbridge BEH C18 (Waters; 130 Å, 3.5 µm × 100 mm) analytical column. Standardized uptake value (SUV) and radioactivity time courses in WB and PL were generated by the same methods as described previously. ²⁵ The parent fraction in plasma (%PP) measurements were fitted by an inverted Hill type function such that: %PP = p – ((p-b) × tⁿ)/(kⁿ + tⁿ) where t is the time (in min) at which the blood samples were collected relative to the start of the PET acquisition and p, b, k, and n are model parameters. Individual arterial input functions were obtained by multiplying the individual plasma radioactivity concentration curves with the corresponding %PP curves. The resulting parent in plasma curves were fitted after the peak with a bi-exponential function only to characterize and quantify the plasma clearance of [¹¹C]MK-6884. In addition, the plasma free fraction (f_p) of [¹¹C]MK-6884 was measured in triplicate as described previously. ²⁵

CVL-231 dose formulation and administration

CVL-231 was received and stored in powder form in a light-protected vial at 4°C.

The CVL-231 doses lower than 1.7 mg/kg were formulated using 10% (v/v) dimethylsulfoxide (DMSO) with 90% (v/v) of 23% (w/v) 2-hydroxypropyl-beta-cyclodextrin (HPbCD) in distilled deionized water. A stock solution of 23% HPβCD was prepared by mixing 23 mg of HPβCD per 100 mL water. The quantity of CVL-231 needed for preparation of the blocking dose was weighed and dissolved in DMSO to a concentration of 1.6 mg/mL. 23% HPβCD solution was added to achieve the final concentration of 0.16 mg/mL in 10% DMSO/90% HPβCD solution (v/v). Filtration was performed with a 0.2 micron PVDF syringe filter.

To solubilize CVL-231 in higher concentrations needed for administration of larger doses, the formulation solution for 1.7 mg/kg doses was 9.1% (v/v) EtOH and 9.1% (v/v) Kolliphor EL polyethoxylated castor oil in 81.8% (v/v) of 20% (w/v) HPβCD in distilled deionized water. A stock solution of 20% HPβCD was prepared by mixing 20 mg of HPβCD per 100 mL water. The quantity of CVL-231 needed for preparation of the blocking dose was weighed and dissolved in EtOH to a concentration of 3.5 mg/mL. An equal volume of Kolliphor was added to yield a concentration of 1.75 mg/mL. 20% HPβCD solution was added to achieve the final concentration of 0.32 mg/mL in 9.1% EtOH/9.1% Kolliphor/81.8% HPβCD (v/v/v) solution. Filtration was performed using a 0.2-micron PVDF syringe filter.

The administration of all formulated CVL-231 doses was performed in two steps: (1) a loading dose (∼48% of the total dose) was administered as a slow bolus over 2–3 minutes starting 10 minutes prior to the scan start, (2) immediately following the loading dose, the maintenance dose (∼52% of the total dose) was administered as a constant infusion lasting until the end of scan. This administration protocol corresponds to K_bol = 90 min, where the bolus portion of the blocking dose contains the same amount of material as the infusion portion. The combinations of both infusions for all CVL-231 doses are given in Table 1, and we refer to the cumulative doses unless otherwise specified.

Table 1.

Injected animals, administered doses, and CVL-231 concentrations in plasma.

Animal	BW (kg)	Study	Baseline scan			Blocking scan			CVL-231 loading dose (mg/kg)	CVL-231 infusion dose (mg/kg)	CVL-231 cumulative dose (mg/kg)	CVL-231 plasma conc. at 1h (ng/mL)
			SA at EOS (GBq/µmol)	Mass dose at TOI (µg)	Injected radioactivity at TOI (MBq)	SA at EOS (GBq/µmol)	Mass dose at TOI (µg)	Injected radioactivity at TOI (MBq)
M1	13.4	M1.1	202.0	1.30	198.3	81.1	4.30	217.2	0.24	0.26	0.50	245
M2	15.1	M2.1	118.0	1.40	188.7	280.5	1.00	181.3	0.24	0.26	0.50	293
M1	13.1	M1.2	117.7	1.80	175.8	236.1	1.40	202.0	0.12	0.13	0.25	126
M2	15.1	M2.2	170.2	0.70	171.3	266.4	0.60	175.8	0.48	0.52	1.00	417
M1	12.2	M1.3	120.6	1.00	170.6	265.7	0.60	193.9	0.81	0.88	1.69	931
M2	15.1	M2.3	229.4	0.50	192.4	159.1	1.30	203.5	0.81	0.88	1.69	1040
M2	17.7	M2.4	200.9	0.70	186.5	285.3	0.60	205.7	1.62	1.75	3.37	XXX

BW: body weight; SA: specific activity; EOS: end of synthesis; TOI: time of injection.

CVL-231 plasma samples collection

Arterial blood samples of 3 mL each were collected during the blocking PET scans in tubes containing K₂EDTA and kept on ice protected from light until centrifugation. In blocking studies with injected CVL-231 doses lower than 1.7 mg/kg, samples for determining CVL-231 concentration were collected at 60 minutes post-radiotracer injection (70 minutes after start of CVL-231 administration). For CVL-231 doses of 1.7 mg/kg or higher, samples were drawn at 0-, 30-, 60-, and 90-minutes post-radiotracer injection (10, 40, 70, and 100 minutes after start of CVL-231 administration). The samples were mixed gently and centrifuged within 2 hours of collection. Centrifugation was performed at 3000 rpm (approximately 2056 g) for 10 minutes at 4 °C. The maximum amount of plasma was recovered and transferred into new light-protected polypropylene tubes. The samples were stored in a −80°C freezer until shipment on dry ice to a third-party bioanalytical facility. The obtained CVL-231 concentrations in plasma were then used in the plasma exposure-response analysis.

Image registration and processing

All PET processing was performed with an in-house developed MATLAB software that uses FSL. ²⁸ We have described these methods in detail previously. ²⁵

Brain kinetic analysis

Extracted brain TACs were analyzed by compartmental modeling with reversible one- (1 T) and two- (2 T) tissue compartment models using the measured metabolite-corrected arterial plasma input function. Each model was assessed with a vascular fraction V_f fixed to 5%, and likewise with V_f estimated as an additional model parameter. For quantification based on compartmental models and definition of micro- and macro- parameters, we followed the consensus nomenclature for in vivo imaging of reversibly binding radioligands. ²⁹ The primary outcome of interest was the total volume of distribution V_T, representing the equilibrium ratio of tracer concentration in tissue relative to plasma. This ratio is linearly related to the tracer binding to its target. V_T was calculated from the models microparameters as K₁/k₂ for the 1 T compartment model and as (K₁/k₂) × (1 + k₃/k₄) for the 2 T compartment model. Graphical methods with arterial input functions, such as Logan distribution volume (Logan DV) ³⁰ and multilinear analysis 1 (MA1), ³¹ were also tested as alternative techniques for V_T estimation. Binding potential (BP_ND), defined as the ratio of the specifically bound radioligand to its non-displaceable uptake (non-specifically bound and free in tissue), was also calculated. BP_ND estimates were derived using the following reference-tissue methods: simplified reference tissue model (SRTM), ³² simplified reference tissue model 2 (SRTM2), ³³ multilinear reference tissue model 2 (MRTM2) ³⁴ and the Logan graphical method using a reference region input function (Logan DVR). ³⁵ Reference tissue methods leverage a regional TAC devoid of specific binding as an input for kinetic modeling (SRTM) or graphical analysis (Logan DVR and MRTM2). Logan DVR and MRTM2 methods allowed estimation of the distribution volume ratio (DVR), where: $\mathrm{DVR}=V_T^{\mathrm{target region}}/V_T^{\mathrm{reference tissue}}$ . DVR was then converted into BP_ND using the formula: BP_ND = DVR – 1. In SRTM2, the egress rate constant from the reference region (k₂′) was calculated as k₂/R₁ from TACs extracted in high-binding target regions (caudate nucleus and putamen), where k₂ is the egress rate constant, and R₁ is the relative delivery (i.e. the ratio of the rate constants for transfer from arterial plasma to tissue in the target and reference regions, defined as R₁ = K₁/K₁′) derived by first applying SRTM. These scan-specific k₂′ values were then fixed in SRTM2 to reduce the variance of the parameter estimates.

Receptor occupancy quantification

M₄ receptor occupancy (RO) by CVL-231 was quantified using the Lassen plot approach and also by calculating the relative change in striatal BP_ND from baseline to blocking scan. The Lassen plot analysis, originally proposed by Lassen et al. in 1995 and later refined by Cunningham et al. in 2010, uses linear regression of the regional V_T, according to the formula: $V_T^{\mathrm{baseline}}–V_T^{\mathrm{blocking}}=\text{Occ}\times (V_T^{\mathit{baseline}}–V_{ND})$ .^36,37 The slope, Occ, provides an estimate of RO, and V_ND, obtained from the x-intercept, represents the non-displaceable distribution volume. Using BP_ND estimates, RO is calculated as: $\%\text{RO}=100\mathrm{\hspace{0.17em}}\%\times (1–B{P}_{ND}^{\mathit{blocking}}/B{P}_{ND}^{\mathit{baseline}})$ .

RO estimates were analyzed in a dose-response fashion against drug exposure, for both the injected drug dose and the plasma concentration at 1 hour post start of scan. The data were fitted by the function: RO =RO_max × D/(D + ID₅₀), where D is the drug dose (mg/kg), ID₅₀ is the estimated drug dose needed to achieve half-maximal occupancy, and RO_max is the estimated maximal occupancy that can be asymptotically attained by high drug levels. An analogous fitting was performed with the function: RO = RO_max × C/(C + IC₅₀), where C is the steady state plasma concentration (ng/mL) at 1 hour, and IC₅₀ is the estimated plasma concentration needed to achieve half-maximal occupancy.

Statistical analysis

The coefficient of determination R² quantitatively assessed the degree of correlation between outcomes of different methods. Non-linear least-squares regression of the RO function fit the dose-response data, with R² statistics for goodness-of-fit and 95%CI for estimates of unconstrained model parameters. Statistical comparison of regional V_T values was performed using a one-way Welch’s analysis of variances (ANOVA) test with Games-Howell post-hoc test with a set 95% level of confidence (p = 0.05). Intergroup differences characterized by p-values of 0.05 or less were considered statistically significant. Kolmogorov-Smirnov test was used to examine the normality of the model outcome estimates in all scans. Mean test-retest (TRT) reliability of V_T estimates was calculated across all permutation of baseline scans and by averaging V_T values from all brain regions: TRT = [(V_T,baseline2–V_T,baseline1)/V_T,baseline1] × 100%. Likewise TRT reliability of BP_ND estimates were calculated as TRT = [(BP_ND,baseline2–BP_ND,baseline2)/BP_ND,baseline1] × 100%. All data were expressed as mean value ± standard deviation (SD) unless otherwise specified, and 95% confidence intervals (95%CI) were presented when deemed relevant.

Results

Injection parameters

Table 1 summarizes information about the animals, specific activities of the tracer, and injected doses. No adverse events were observed under baseline conditions or at any of the different CVL-231 doses tested in this work. The mean injected radioactive dose of [¹¹C]MK-6884 at the time of injection was 190.2 ± 14.1 MBq (range 170.6–217.2, n = 14). The mean specific activity of [¹¹C]MK-6884 at the end of synthesis was 195.0 ± 68.5 GBq/μmol (range 81.0–285.3, n = 14), corresponding to a mean injected mass of 1.23 ± 0.97 mg (range 0.50–4.30, n = 14).

[¹¹C]MK-6884 arterial blood data

RadioHPLC analysis of selected [¹¹C]MK-6884 plasma samples revealed in vivo metabolic degradation of the radiotracer over the scan duration (Figure 1(a)), with 25.3 ± 3.9% remaining at 90 minutes post tracer injection (Figure 1(b)). While CVL-231 administration was not observed to affect the degradation rate of [¹¹C]MK-6884, there was a slight difference in the metabolism rate between the animals.

Figure 1.

Arterial whole-blood and plasma measurements and in vivo metabolism. (a) Representative radiochromatograms showing peaks corresponding to [¹¹C]MK-6884 parent and radiometabolites in plasma. Data are shown for the baseline M2.3 scan. (b) Time dependency of the %PP for each [¹¹C]MK-6884 scan. HPLC data for M1.3 were not acquired due to technical issues. (c) Representative whole-blood and parent-in-plasma time courses in SUV. The data for baseline and blocking scans are shown for the low (0.25 mg/kg) and high (1.7 mg/kg) doses of CVL-231 injected in monkey 1 and (d) Whole blood to plasma ratio of radioactivity concentrations. Abbreviations: BSLN-baseline, BLK-blocking, STD-standard, WB-whole blood, PP-parent in plasma, SUV-standard uptake value.

Time courses of total radioactivity concentration in whole blood and [¹¹C]MK-6884 in plasma were characterized by a peak between 2 and 3.5 minutes after the start of the 3-minute tracer infusion (Figure 1(c)). After the initial peak, the parent in plasma curves were characterized by bi-exponential clearance. The half-lives of the rapid and slow components at baseline were 2.3 ± 0.5 min and 68.0 ± 13.0 min, respectively. No significant changes of both half-lives at blocking scans were observed except at the highest CVL-231 doses of 1.7 and 3.4 mg/kg, for which the half-lives of the rapid component increased to 5.2 min (M1.3 study) and 4.1 min (M2.4 study), and the half-lives of the slow component increased to 143.3 min (M1.3 study), 263.8 min (M2.3 study), and 653.9 min (M2.4 study). This dramatic decrease in blood clearance may be explained by the blocking effect that high doses of CVL-231 may exert on peripheral M₄ receptors leading in turn to more [¹¹C]MK-6884 available in the vascular compartment. The ratio of radioactivity concentrations in whole blood to plasma reached equilibrium within 1 minute and remained stable thereafter (Figure 1(d)). f_p measurements averaged 32 ± 10%, indicating moderate binding to plasma proteins with no statistically significant differences between animals or scan conditions.

[¹¹C]MK-6884 brain uptake and kinetics

PET SUV images (Figure 2(a)) demonstrated high brain uptake shortly after radiotracer injection (0–10 minutes), with the highest baseline SUV_0–10min exceeding 4.2 in the striatum. SUV_0–10min values in the other brain regions ranged from 2 to 3.5. At the later 30 to 90 minutes time window, which is more representative of the tracer binding to its target, the highest baseline SUV_30–90min levels in the striatum were 1.4. This resulted in good contrast with other brain regions, where SUV values were roughly 0.5. Late SUV images (30–90 minutes) exhibited a clear dose-dependent blocking effect of the CVL-231 drug on [¹¹C]MK-6884 binding in the striatum (Figure 2(a)). At high CVL-231 doses (1.7 mg/kg and 3.4 mg/kg), highly vascularized areas outside the brain exhibited a noticeable increase in SUV, consistent with the observed reduced blood clearance discussed above. SUV TACs in the hippocampus, cortex, cerebellum, white matter, and striatum (Figure 2(b)) demonstrated rapid uptake of [¹¹C]MK-6884 (peak at 3.75–5.25 minutes), followed by relatively fast clearance. The highest uptake was found in the striatum, and the lowest was in white matter.

Figure 2.

PET SUV data. (a) PET SUV images of [¹¹C]MK-6884 at baseline and after each blocking dose of CVL-231 for monkey 1. The PET and MRI MEMPRAGE images are co-registered in the MRI NIMH macaque template (NMT) space using affine transformations. The red crosshairs are placed on the striatum area (characterized by the highest tracer uptake). PET SUV images are generated by summing dynamic frames over the 0–10 min (early phase reflecting tracer delivery) and 30–90 min (later phase more representative of tracer binding) time intervals and (b) SUV TACs in the sampled brain regions fitted with either the reversible one-tissue compartment model (1T2k) or the reversible two-tissue compartment model (2T4k) with the vascular contribution (V_f) fixed to a nominal value.

Kinetic modeling and quantification of [¹¹C]MK-6884 kinetics in the NHP brain

Blood-based methods

A reversible one-tissue compartment model (1T2k), with and without the vascular contribution V_f expressed as a model parameter, failed to adequately describe the brain TACs (Figure 2(b), top graph) except at the highest CVL-231 dose tested. A reversible two-tissue compartment model (2T4k) more accurately described the regional TACs throughout the entire scan duration of 90 min, regardless of whether V_f was fixed to 5% or unconstrained (Figure 2(b), bottom graph). However, in the latter case, the estimated V_f values were unrealistically high (well above >5–7%), and therefore the 2T4k model with fixed V_f was selected for further compartmental analysis. Despite relatively good model fits over the 90-min scan duration, V_T values derived from the 2T4k model using the entire scan duration were unstable in some brain regions due to extremely low values of the estimated receptor dissociation rate constant (k₄). Truncating the TACs down to 60 or 40 minutes improved the overall stability of V_T values compared to using the full scan duration (Supplemental Figure 2). At the highest CVL-231 dose tested (3.4 mg/kg), regardless of the scan duration used, the 2T4k model provided unstable V_T estimates and the 1T2k model was preferred. Therefore, in this work the data acquired at the 3.4 mg/kg CVL-231 blocking dose were excluded for all remaining analyses involving comparisons of 2T4k model outcomes.

In addition to compartmental modeling, regional V_T values were also quantified using graphical analysis methods, such as Logan DV and MA1. Using a t* value of 30 minutes provided good linearization of the Logan plot (Supplemental Figure 3 A) and led to a very good correlation between V_T values estimated by Logan DV and MA1 when using 60 minutes of data (V_T,MA1,60min = 0.99 × V_{T,Logan DV,60min} + 0.06; R² = 0.99) as shown in Supplemental Figure 3 C. Both Logan DV and MA1 methods demonstrated good time-stability of V_T when comparing estimates obtained using 90 minutes to those obtained using 60 minutes, but further truncation (to 40 minutes) led to outliers (Supplemental Figure 3B), likely because the end time of the TAC duration was too close to the t* value of 30 minutes. Altogether, the results obtained from our compartmental and graphical analyses led us to select an optimal scan duration of 60 minutes for the remaining kinetic analyses performed in this manuscript. Using 60 minutes of data, V_T values estimated by Logan DV correlated well with those estimated from the 2T4k model (V_{T,Logan DV,60min} =0.96 × V_T,2T4k,60min + 0.14; R² = 0.95) as shown on Figure 3(a). Likewise, V_T estimated by MA1 correlated well with V_T obtained from the 2T4k model (V_T,MA1,60min = 0.95 × V_T,2T4k,60min + 0.19; R² = 0.96). Since Logan DV and MA1 are not relying on a particular model structure, these graphical methods were deemed more generalizable than compartmental models to the present datasets and Logan DV was selected as the method of choice for the remaining blood-based kinetic analyses.

Figure 3.

Quantification of [¹¹C]MK-6884 binding by Logan DV analysis. (a) Scatter plot showing correlation between V_T estimates obtained by Logan DV and by the 2T4k kinetic model. A scan duration of 60 minutes was used for both methods. (b) Lassen plot based on V_T values for monkey 1 (top) and monkey 2 (bottom). Linear regressions are shown for each paired baseline-blocking study. (c) Bar plots showing regional V_T values for each paired baseline-blocking study in monkey 1 (top graph) and monkey 2 (bottom graph). Shaded areas indicate the 95%CI corresponding to the estimated V_ND, which were [1.18, 2.36] for monkey 1 and [1.46, 1.77] for monkey 2 and (d) Logan V_T parametric images of [¹¹C]MK-6884 binding at baseline and at each CVL-231 blocking dose in monkey 2. V_T and MEMPRAGE images were aligned to the MRI NIMH macaque template (NMT) space using affine transformations. The red crosshair is positioned on the striatum.

For each animal, we performed Lassen plot analyses of each baseline-blocking pairs to determine the non-displaceable volume of distribution (V_ND) (Figure 3(b)). The average V_ND was 1.8 ± 0.2 mL/cc for monkey 1, and 1.6 ± 0.1 mL/cc for monkey 2. Regional V_T values derived by Logan DV are listed in Supplemental Table 1 and plotted in Figure 3(c) for each baseline and blocking scan. A clear dose-dependent blocking effect was observed on V_T values only for the caudate nucleus and putamen. In other brain regions, the baseline and blocking V_T values fell within the 95% CI of the V _ND (Figure 3(c)) suggesting no statistically significant differences, which was further validated by Welch’s ANOVA test with Games-Howell post-hoc test. Difference effects and their 95% confidence intervals for both monkeys are listed in Supplemental Table 2. Based on this analysis, we also concluded that the cerebellar grey matter seems to behave as a suitable reference region for subsequent reference tissue-based kinetic modeling.

Lastly, the mean TRT of V_T estimates across all brain regions and baseline scans was 15.6 ± 2.5% when using individual metabolite-corrected input functions. Grouping metabolite data from all scans within each animal to derive averaged metabolite corrections did not significantly improve TRT (14.0 ± 2.5%). Figure 3(d) shows Logan DV V_T parametric images computed for monkey 2 at baseline, as well as at each dose of CVL-231.

Reference tissue-based methods

SRTM, MRTM2, and Logan DVR using cerebellar grey matter as a reference region were assessed for quantifying BP_ND. SRTM provided good model fits for all regional TACs over the whole scan duration except in the striatum, where fit quality deteriorated after 60 minutes (Figure 4(a)). This observation led us to retain a 60-minute scan duration for quantifying BP_ND. However, despite this apparent poor model fit after 60 minutes, BP_ND values estimated using 60 or 90 minutes of data were nonetheless stable (Supplemental Figure 4 A). The TRT of the striatal BP_ND averaged across all baseline scans and animals was 8.3 ± 2.9%. Figure 4(b) shows bar plots of striatal BP_ND values measured for each paired baseline-blocking study at baseline and at different CVL-231 blocking doses. The average baseline striatal BP_ND value measured was 1.13 ± 0.07 in monkey 1 and 0.83 ± 0.05 in monkey 2. Individual BP_ND values in the caudate nucleus, putamen, and whole striatum are shown in Supplemental Table 3. Despite strong correlations, BP_ND values estimated from SRTM using 60 minutes of data overestimated the BP_ND values derived from the 2T4k model and the Logan DV graphical method (Supplemental Figure 4B,C). SRTM2 provided BP_ND estimates in very good agreement with those obtained from SRTM (Supplemental Figure 4 D), and consequently, presented similar agreement with blood-based methods (Supplemental Figure 4E,F). BP_ND parametric images computed with SRTM2 (Figure 4(c)) were characterized by improved signal-to-noise ratio as compared to the SRTM-based maps (not shown). The binding pattern was consistent with the ROI-based analysis (Figures 3(c) and 4(b)), and the images demonstrated a robust dose-dependent blocking effect in the striatum with increasing CVL-231 doses.

Figure 4.

Quantification of the striatal BP_ND by the simplified reference tissue (SRTM) models. (a) Representative baseline TACs fitted by the SRTM model using cerebellar grey matter as a reference tissue. (b) Striatal BP_ND values measured for each paired baseline-blocking study at baseline and at different blocking doses of CVL-231. BP_ND estimates were derived by SRTM using 60 minutes of data. (c) BP_ND parametric images of [¹¹C]MK-6884 at baseline and at each blocking dose of CVL-231 in monkey 2. Voxel-based SRTM2 modeling was performed using 60 minutes of data and k₂′ was constrained to values derived by ROI-based analysis within each individual scan. BP_ND and MEMPRAGE images were aligned to the MRI NIMH macaque template (NMT) space using affine transformations. The red crosshair is positioned on the striatum.

Supplemental Figure 5 A shows representative model fits for the graphical analysis methods Logan DVR and MRTM2. Both methods provided BP_ND estimates that were highly correlated with those derived from SRTM (Supplemental Figure 5B). For the remaining analyses, SRTM was selected for RO quantification based on changes in the striatal BP_ND between baseline and blocking conditions.

Quantification of M₄ occupancy by CVL-231

RO values in the striatum for each injected CVL-231 dose and plasma concentration are given in Table 2. RO calculations using SRTM BP_ND were in good agreement with those derived from the Lassen plots shown in Figure 3(b) (RO_{SRTM BPND} = 0.99 ×RO_{Lassen-Logan DV} + 0.03; R² = 0.90). The agreement between RO estimates diminished when BP_ND was calculated from the Logan V_T estimates with the cerebellar grey matter as the reference region: BP_{ND-Logan DV} = DVR_{Logan DV}–1 (RO_SRTM = 0.66 × RO_{BPND-Logan DV} + 0.22; R² = 0.72). When V_ND estimates from the Lassen plots were used in lieu of the cerebellar Logan V_T values for calculating BP _ND -Logan DV, the agreement between RO estimates significantly improved (RO_SRTM = 0.98 ×RO_{BPND-Logan DV} + 0.04; R² = 0.89).

Table 2.

CVL-231 RO values in the striatum as calculated from SRTM BP_ND estimates using cerebellar grey matter as a reference region and a 60-min fitting interval.

CVL-231 injected dose (mg/kg)	Monkey 1		Monkey 2
	CVL-231 plasma concentration at 1 h (ng/mL)	Receptor occupancy (%)	CVL-231 plasma concentration at 1 h (ng/mL)	Receptor occupancy (%)
0.25	126	18.3	–	–
0.5	245	31.6	293	29.7
1	–	–	417	42.4
1.7	931	67.1	1040	59.3
3.4	–	–	–	74

CVL-231 mass dose-response and plasma concentration-response relationships

Estimates of RO derived from SRTM BP_ND values were analyzed in a dose-response fashion. The functional forms RO = RO_max × D/(D + ID₅₀) and RO = RO_max × C/(C + IC₅₀) were used to fit CVL-231 dose-response and CVL-231 plasma exposure-response curves respectively. The resulting estimates of RO_max were 1.0 ± 0.3 (95%CI = [0.8, 1.3]) for the dose-response function and 1.0 ± 0.4 (95%CI = [0.6, 1.4]) for the plasma exposure-response function (Supplemental Figure 6). Considering that the expected 100% RO (RO_max = 1) was within the statistical boundaries, we then constrained the RO_max parameter to 1 for the final estimation of the ID₅₀ (Figure 5(a)) and IC₅₀ (Figure 5(b)) parameters. The resulting estimated ID₅₀ value was 1.1 ± 0.2 mg/kg (95%CI = [1.0, 1.3]), and the estimated IC₅₀ value was 581.4 ± 102.6 ng/mL (95%CI = [473.8, 689.0]).

Figure 5.

Relationship between striatal M₄ mAChR occupancy by CVL-231 and CVL-231 mass dose (a) and CVL-231 plasma concentration (b) with data averaged across animals. Non-linear regression was performed to fit the data using a one-parameter E_max model based on a Hill function assuming a 100% maximal RO. The goodness of fit is indicated by R² statistics. The regression models provided estimates of ID₅₀ and IC₅₀, respectively. Corresponding standard errors and values for 95% confidence intervals (95%CI) are also indicated. The 95% confidence intervals are plotted using dashed lines.

Discussion

Pharmacokinetic evaluation of [¹¹C]MK-6884

The first aim of this work was to analyze the PET imaging and quantification properties of the recently developed M₄-selective radiotracer [¹¹C]MK-6884 in order to subsequently employ this tracer, along with the appropriate modeling strategy, for investigating target engagement and RO of the novel PAM drug, CVL-231.

In arterial plasma, [¹¹C]MK6884 displayed favorable characteristics for PET imaging with relatively fast washout and good bioavailability. Radio-HPLC analysis of selected plasma samples demonstrated moderately fast in-vivo degradation (Figure 1). Inspection of the radiochromatograms revealed the presence of radio-metabolites, which were more polar than the parent compound and consequently less likely to penetrate the blood brain barrier (Figure 1(a)). However, we cannot completely rule out the possibility of a brain penetrant metabolite, especially considering the relatively moderate time stability of V_T measurements (Supplemental Figure 7). Pre-treatment with the CVL-231 drug did not affect the rate of [¹¹C]MK-6884 metabolism nor its plasma free fraction. Likewise, CVL-231 did not affect the time course and washout of [¹¹C]MK-6884 in arterial plasma except at the highest drug doses (1.7–3.4 mg/kg), for which a decrease in plasma clearance was observed. This effect may be due to cumulative blocking of M₄ allosteric sites in the periphery, causing an increase in tracer availability in the vascular compartment. Similar peripheral blocking effects on the concentration of parent compounds in plasma have been previously reported for other tracers and targets.^38,39

In the NHP brain, the kinetics of [¹¹C]MK-6884 were relatively fast in all regions. The highest regional uptake was observed in the striatum, which is consistent with regional M4 receptor density. Late [¹¹C]MK-6884 images displayed high contrast between the striatum and the rest of the brain, however the other brain regions showed relatively low level of heterogeneity. While this observation is consistent with previous imaging findings in NHP, Li et al. recently reported more heterogeneity across brain regions of human subjects with notable tracer binding in cortical areas, possibly resulting from species differences in M4 expression or distribution. ²⁴

Using blood-based compartmental modeling methods, the regional brain kinetics of [¹¹C]MK-6884 were best described by a reversible two-tissue compartment model (2T4k) with a fixed parameter for the vascular fraction contributing to the measured PET signal (Figure 2(b)). This model provided robust V_T estimates for all scans except for the blocking scan performed at the highest CVL-231 dose (3.4 mg/kg), where a simple one-tissue model (1T2k) was preferred across brain regions. This shift in model preference may be explained by the limited access of [¹¹C]MK-6884 to M4 allosteric binding sites due to CVL-231 occupying a large fraction of those binding sites at 3.4 mg/kg. Since blood-based graphical methods such as Logan DV and MA1 are independent of any model structure, those techniques were also investigated for direct estimation of V_T and were found to perform equally well across all studies and CVL-231 doses. We found that cerebellar V_T values calculated by Logan DV were statistically similar to the non-displaceable volume of distribution (V_ND) estimated by Lassen plot analysis. In addition, we did not observe any statistically significant reduction in cerebellar V_T at different exposures of CVL-231 compared to baseline scans. These findings further support the use of the cerebellum as a reference region (i.e. devoid of specific binding) for non-invasive, blood-free, quantification methods as suggested by previous work. ²⁴

This finding prompted us to investigate reference tissue-based methods for estimating DVR or BP_ND. Overall, estimates from the reference tissue-based methods agreed with each other and were highly correlated with the blood-based methods, despite some level of overestimation. The overestimation of BP_ND values by SRTM compared to those obtained from blood-based methods could be attributed to possible violation(s) of the SRTM model assumptions. The effect of violating these assumptions varies on a tracer-by-tracer basis. Although it is not straightforward to assess the exact reason of the discrepancy between SRTM-based and blood-based BP_ND measurements, one possible explanation may be that both reference and target regions were best described by a two-tissue compartmental model whereas a one-tissue compartment model would be required to strictly fit the SRTM framework. ⁴⁰ Nevertheless, reference tissue-based methods produced robust estimates and as expected, provided slightly higher TRT performance compared to blood-based techniques. We selected SRTM for quantifying RO of M4 by CVL-231 at the ROI level and its variant SRTM2 for generating parametric maps of BP_ND because of the robustness of these models.

Quantification of M4 occupancy by CVL-231

The second aim of this work was to apply our findings to measure the RO of M4 by the novel PAM drug candidate CVL-231 in NHP. Pre-treatment at different doses of CVL-231 led to a clear dose-dependent decrease in [¹¹C]MK-6884 uptake as observed from the late SUV images (Figure 2(a)), confirming that these compounds bind to the same allosteric site on M4. This finding was confirmed quantitatively using the blood-based Logan DV graphical method for direct estimation of V_T and generation of parametric maps (Figure 3), as well as using the reference-tissue methods SRTM and SRTM2 for calculating BP_ND (Figure 4). Calculations of RO using BP_ND values derived from SRTM were in very good agreement with the ROs given by the slopes of the Lassen plots using V_T values from Logan DV. Dose-response curves (Figure 5) were well described by standard Emax models and suggested saturable binding of the M₄ receptors by CVL-231 with a highest theoretical RO of 100%. Because a first pass analysis using a two-parameter Emax model (RO_max and either ID₅₀ or IC₅₀ as model parameters) led to RO_max estimates of 100%, we fixed this parameter for the final estimation of the ID₅₀ and IC₅₀ parameters in the mass dose-response and plasma exposure-response relationships respectively. The final ID₅₀ was estimated to be 1.1 ± 0.2 mg/kg, and the IC₅₀ was estimated to be 581.4 ± 102.6 ng/mL. Altogether, the absence of adverse effects and results presented herein show promise for further development and evaluation of CVL-231 in clinical trials, which are currently underway. ²⁰ One limitation of our study is the possible interaction of isoflurane, used for anesthesia, and ACh. Indeed, previous studies in rodents have shown that isoflurane can modulate the release of ACh in the striatum and cortex despite providing contradictory information on the direction of this effect.^41,42 While we kept the level of isoflurane constant across scans and animals in the present study, it is nonetheless difficult to evaluate its effect on the findings reported here.

In summary, our results suggest that kinetic modeling of [¹¹C]MK-6884 using arterial input function and reference region methods can reliably and reproducibly quantify M4 receptor in the brain. Logan DV and MA1 were the preferred blood-based methods in this work because their performance was more generalizable at all doses of CVL-231 compared to compartmental models. Due to the robustness of these methods, the reference-region based models SRTM and SRTM2 were selected for the non-invasive quantification of binding potential and receptor occupancy using the cerebellar grey matter as a reference region. Second, [¹¹C]MK-6884 was employed to quantify M4 RO by CVL-231, a novel M4 receptor positive allosteric modulator drug. Our data indicated a dose-dependent relationship of RO well fitted by a classical Hill function model which led to the determination of ID₅₀ and IC₅₀. Our work further supports the use of [¹¹C]MK-6884 as a PET tracer for imaging M4 mAChR and also provides valuable information for future development of CVL-231 as a mAChR PAM drug in clinical trials.