Imaging evaluation of 5HT2C agonists, [(11)C]WAY-163909 and [(11)C]vabicaserin, formed by Pictet-Spengler cyclization.

The serotonin subtype 2C (5HT2C) receptor is an emerging and promising drug target to treat several disorders of the human central nervous system. In this current report, two potent and selective 5HT2C full agonists, WAY-163909 (2) and vabicaserin (3), were radiolabeled with carbon-11 via Pictet-Spengler cyclization with [(11)C]formaldehyde and used in positron emission tomography (PET) imaging. Reaction conditions were optimized to exclude the major source of isotope dilution caused by the previously unknown breakdown of N,N-dimethylformamide (DMF) to formaldehyde at high temperature under mildly acid conditions. In vivo PET imaging was utilized to evaluate the pharmacokinetics and distribution of the carbon-11 labeled 5HT2C agonists. Both radiolabeled molecules exhibit high blood-brain barrier (BBB) penetration and nonspecific binding, which was unaltered by preadministration of the unlabeled agonist. Our work demonstrates that Pictet-Spengler cyclization can be used to label drugs with carbon-11 to study their pharmacokinetics and for evaluation as PET radiotracers.

Introduction

The serotonin-2C receptor subtype (5HT2C) is an important G-protein-coupled receptor that has been implicated in many central nervous system disorders including, among others, obesity, anxiety, and depression.[1−11] 5HT2C receptors are distributed heterogeneously in the mammalian brain with high densities biased to subcortical regions such as the choroid plexus. Sufficient densities for imaging exist in the hypothalamus, globus pallidus, and substantia nigra. Lower densities of the receptors have been identified in the cortex and cerebellum.[12−14] Of the diseases 5HT2C has been shown to cause or contribute to, obesity has been at the forefront of research, and new drugs targeting 5HT2C (e.g., lorcaserin, hereafter referred as 1) are the first gaining Food and Drug Administration (FDA) approval in more than a decade.[15,16] To date, a number of 5HT2C agonists (Figure 1) have been generated and used in preclinical and clinical studies.[8,17] In addition to a functional role in eating disorders, the receptor has also been implicated in mood behavior, for example, by mediating effects of selective serotonin reuptake inhibitors (SSRIs) and atypical antipsychotics.[18−20]

Figure 1

5HT2C receptor full agonists in preclinical and clinical studies.

Although many studies suggest that dysfunction of 5HT2C receptors contribute to important brain-related disorders,[1−11] direct links between these diseases and 5HT2C receptor abnormalities have been difficult to elucidate. This is in part due to the lack of a method for determining 5HT2C receptor concentration as a function of disease in humans. We have previously labeled and evaluated arylazepines[21] as ligands for positron emission tomography; however, direct carbon-11 incorporation into some of the most promising lead candidates (certain benzodiazepines) was difficult because no pendent methyl group or fluorine was available for PET isotope substitution. Two molecules of particular interest as scaffolds for ligand development were thus inaccessible because of limitations in reaction methods for carbon-11 incorporation. The potent and selective compounds 2 (WAY-163909) and 3 (vabicaserin) are agonists that exhibit high affinity (Ki of ∼10 and 3 nM, respectively) and good efficacy for 5HT2C (EC50 = 8 nM each) and have been characterized extensively in preclinical drug trials.[8,9,22−24]

In order to overcome the synthesis challenge of carbon-11 incorporation for 2 and 3, [11C]formaldehyde was used in a Pictet–Spengler cyclization. Using an in situ formation method of [11C]formaldehyde from [11C]methyl iodide developed in our lab,[25] we were able to prepare [11C]2 and [11C]3 . In the process of doing so, we determined that under the acidic conditions required for product formation, the solvent DMF degrades to generate formaldehyde, consequently reducing the specific radioactivity of the product. Through detailed isotope utilization (deuterium and carbon-13) we pinpointed and eliminated the sources of formaldehyde while simultaneously demonstrating that in situ formation of isotopically labeled formaldehyde is generalizable and can be used to install carbon and hydrogen isotopes (singly or in concert). These isotopalogs may themselves warrant preclinical drug evaluation. We further demonstrate that high specific activity [11C]2 and [11C]3, and by extension derivatives thereof, can be synthesized. Finally, we evaluated the potential of [11C]2 and [11C]3 as scaffolds for PET radiotracer development. Both molecules exhibit high brain uptake in rats and non-human primates dominated by nonspecific binding which we are now in the process of tuning through structure modification.

Results and Discussion

Chemistry

Retrosynthetic analyses of 2 and 3 were possible provided that carbon-11 could be incorporated at the last step of the synthesis and provided that the corresponding primary amines (precursors 11 and 18) could be derived. Schemes 1 and 2 represent our routes to syntheses of precursors 11 and 18, respectively. The development of a synthesis route for preparing chiral labeling precursors 11 and 18 were more challenging than expected. For 2 synthesis, neither chiral crystallization of 6 with (−)-2,3-dibenzoyl-l-tartaric acid and ditoluyl-d-tartartic acid nor resolution of diastereomers following reaction of 6 with 1S-(+)-10-camphorsulfonyl chloride was successful. Reaction of racemic 6 with a chiral mandelic acid derivative [(R)-(−)-α-methoxyphenylacetic acid] was ultimately successful. To isolate optically active materials, racemic compound 6 was treated with an enantiopure (R)-(−)-α-methoxyphenylacetyl chloride to give two diastereomers 7 and 8. Separation of diastereomer 7 was achieved by column chromatography, and the structures of 7 and 8 were confirmed by single crystal X-ray diffraction (XRD) analysis (Figure 1 in Supporting Information). Removal of the mandelic acid derived chiral auxiliary cleavage to afford enantio-enriched 11 was not initially achieved using conditions predicted from the literature to proceed efficiently (these attempts included 1.0 M LAH in THF, saturated HCl in methanol at reflux, 5% NaOH in methanol at reflux, and lithium triethylborohydride (superhydride)). Fortunately, the use of dilute sulfuric acid at moderate temperatures (65 °C) did provide 7 albeit in moderate yield. More harsh conditions led to decomposition to unidentified products. Nonetheless, sufficient (multimilligram) quantities of the chiral precursor to 2 (enantiomeric excess of 97%) were obtained for carbon-11 labeling (Scheme 1).

Scheme 1

Synthesis of WAY-163909 Labeling Precursor (11) and Reference Standard (2)

Reagents and conditions: (i) H2 (45 psi), 5% Pd/C, EtOH, rt, 65%; (ii) (a) (S)-2-methoxy-2-phenylacetyl chloride, Et3N, DCM; (b) separation by column chromatography, 30%; (iii) dilute H2SO4, 70 °C, 8 h, 30%; (iv) 2-methyl-2-oxazoline, pTsOH (cat.), 165 °C, 3 h, 50%; (v) (a) dilute H2SO4, 70 οC, 8 h; (b) aq NaOH; (c) 2.0 M Et2O–HCl, rt, 50% (three steps), 97% ee; (vi) 37% aq HCHO, TFA/EtOH, rt, 70%.

Scheme 2

Synthesis of Vabicaserin Labeling Precursor (18) and Reference Standard (3)

Reagents and conditions: (i) cyclopentadiene, 37% aq HCHO/HCl, EtOH, 5 οC, 10 h; (ii) H2 (45 psi)/10% Pd/C, EtOH/EtOAc (1:1), rt, 10 h, 65%; (iii) (a) ditoluoyl-d-tartaric acid, IPA; (b) separation by crystallization, 30%; (iv) 5% aq NaOH, 90%, 97.5% ee; (v) 2-methyl-2-oxazoline, pTsOH (cat.), 165 οC, 3 h, 55%; (vi) (a) dilute H2SO4, 70 °C, 8 h; (b) aq NaOH; (c) 2.0 M Et2O–HCl, rt, 50% (three steps), 97% ee; (vii) 37% aq HCHO, TFA/EtOH, rt, 60%.

The synthesis of 3 and its precursor 18 followed the published patent procedures.[26] Resolution of the precursor enantiomers was achieved by cocrystallization of an early intermediate using ditoluyl-d-tartartic acid (Scheme 2). Using this route, we were able to produce 18 for use in carbon-11 radiochemistry and PET imaging. Reference compounds 2 and 3 were synthesized from precursors 11 and 18, respectively, using a Pictet–Spengler cyclization with formaldehyde under conditions optimal for isolated yield but not for radiochemical synthesis.

To optimize labeling conditions for carbon-11, we needed to dramatically reduce the reaction time required for formaldehyde condensation and cyclization due to the short carbon half-life (t1/2 = 20.4 min). Moreover, we needed to adapt the cyclization method to be compatible (in a one-pot strategy) with conditions for in situ generation of [11C]formaldehyde from [11C]methyl iodide. Initially we determined that 2 could be formed from 11 at a high temperature (115 °C) in DMF and trifluoroacetic acid (TFA) using nonradioactive in situ [12C]CH2O (generated from substoichiometric amounts of [12C]CH3I with an oxidant, trimethylamine N-oxide (TMAO)). Reactions with N-methylmorpholine-N-oxide (NMMNO) and quinuclidine N-oxide (QNO) were successful but provided lower yield than TMAO. TMAO is also commercially available and inexpensive.[25] To further confirm the labeling position and compound identity, we used substoichiometric [13C]CH3I and an excess of TMAO in DMF at 80 °C followed by treatment with precursor 11, TFA, and H2O at 115 °C to enrich the expected 50 ppm resonance observed by 13C NMR. Analysis of the [13C]2 by LC–-MS lead to the observation of an isotope distribution suggestive of isotopic dilution (i.e., the 99% enriched [13C]CH3I led to a product on 80–90% enriched). Dilution of the isotope when applied in the context of carbon-11 was indeed determined to provide isolated products with unacceptably low specific radioactivity (0.02 mCi/nmol) for proper evaluation as a PET radiotracer for 5HT2C receptors.

To determine the source of [12C]CH2O, we systematically replaced each reaction component starting with the two methyl sources TMAO and DMF. Previously, we had determined that TMAO under milder conditions did not decompose to form formaldehyde, but in order to completely rule out the contribution from TMAO, we synthesized an oxidant without methyl groups, QNO. Under the same reaction conditions used with TMAO and in the presence of only [13C]CH3I, we observed the formation of [12C]2 by LC–MS (Figure 2C). By use of perdeuterated N,N-dimethylformamide (DMF-d7) as solvent, deuterium incorporation was observed by LC–MS (product m/z, 217; Figure 2 in Supporting Information), indicating that decomposition of DMF-d7 to formaldehyde-d2 (CD2O) occurs at high temperatures under mild acidic conditions. We are unaware of existing data in the literature describing the formation of formaldehyde from DMF and have not yet made steps toward elucidating a potential reaction mechanism, although these steps are planned.

Figure 2

Experiments to determine mechanism of specific activity erosion. (A) Synthesis of [*C]2 using in situ generated [*C]CH2O (generated from [*C]CH3I/oxidant/solvent at 80 °C for 3 min). (B) Aliquots of the reaction mixture injected in the LC–MS instrument (UV, 254 nm). (C) Taken together, these data shown [12C]CH2O are generated from DMF and TFA at elevated temperatures. DMF should be avoided for high specific activity.

For the present goal unintended formaldehyde formation was prevented by replacing DMF with N,N-diethylformamide (DEF). Presumably, DEF can decompose (by analogue) to acetaldehyde, but we did not observe reaction products by LC–MS indicative of the subsequent condensation and cyclization of 11 with acetaldehyde. Satisfyingly, under reaction conditions with DEF no isotopic dilution was observed using [13C]CH3I with either TMAO or QNO as the oxidant to generate [13C]CH2O. Ultimately, TMAO in DEF with TFA proved to be the most efficient conditions for carbon-11 labeling, and under these conditions we were able to isolate high specific radioactivity products.

Radiolabeling of [11C]2 and [11C]3

After optimization, with a good chemical synthesis method in hand to prevent isotope dilution while affecting the reaction in reasonable reaction times, carbon-11 radiolabeling procedures was performed as follows for imaging: an amount of 4 mg of TMAO was dissolved in DEF and cooled to −10 °C in a reactor, and [11C]CH3I was trapped. Then the mixture was heated to 80 °C for 3 min. In a separate vial 1.2 ± 0.3 mg of the alkylamines, 11/18, was dissolved in a mixture of DEF, water, and TFA. This solution was then transferred into the reaction vial, and heating was continued at 115 °C for 6 min (Scheme 3). To quench the reaction, 1 mL of 0.1% formic acid was added, and the entire reaction mixture was purified by reversed-phase semipreparative HPLC. The final product was reformulated into 10% EtOH with saline using C-18 solid-phase exchange (SPE). The chemical and radiochemical purity of the final products, [11C]2 and [11C]3, were tested by analytical HPLC. The identities of the respective products were confirmed by analytical HPLC with additional co-injection of corresponding reference standards (Figures 3 and 4 in Supporting Information). The average time required for the synthesis from end of cyclotron bombardment to end of synthesis and purification was 50 min [11C]2 and [11C]3. The average radiochemical yield was [11C]2 (1.5 ± 0.5%; decay-corrected to trapped [11C]CH3I; n = 3) and [11C]3 (1.3 ± 0.2%; decay-corrected to trapped [11C]CH3I; n = 3). Chemical and radiochemical purities were >97% in all instances. Specific activities of [11C]2 and [11C]3 were 0.85 ± 0.2 mCi/nmol (n = 3) and 1.18 ± 0.20 mCi/nmol (n = 3), respectively, at the end of synthesis (EOS).

Scheme 3

Radiosynthesis of [11C]2 and [11C]3 Using in Situ [11C]CH2O (Generated from [11C]CH3I/Me3N+O–/DEF at 80 °C for 3 min)

RCY is an isolated, formulated, and decay corrected yield.

Preliminary Imaging Evaluation of [11C]2 and [11C]3

Using a small animal PET-CT, we determined that both [11C]2 and [11C]3 exhibited high initial blood–brain barrier (BBB) penetration and good retention in the brain over the scanning time (60 min) when administered intravenously to Sprague–Dawley rats (Figure 3). Although whole-brain time–activity curves generated from imaging data are presented, heterogeneous binding was observed for both labeled compounds. In order to determine the sensitivity of uptake to pretreatment (and potential saturation of the 5HT2C receptor sites), rodents were pretreated in separate experiments with the corresponding unlabeled ligand (e.g., 3, 1 mg/kg iv, was administered 5 min prior to [11C]3 imaging). Pharmacokinetics of [11C]2 were altered by drug treatment; pharmacokinetics of [11C]3 were not. The more rapid washout of [11C]2 can be attributed to several phenomena including receptor saturation, changes in arterial plasma concentration, or changes in metabolism. We were encouraged by the data to further evaluate specificity (saturability) of [11C]2 binding and thus moved to non-human primates where we have extensive experience performing MR-PET imaging with arterial-derived and metabolite corrected blood data.[21,27]

Figure 3

Top: Summed PET images (0–60 min) following injection of [11C]2: baseline scan (0.265 mCi); image acquired following pretreatment (5 min prior) with 2 (1 mg/kg, 0.303 mCi). Images are dose corrected. Bottom: Summed PET images (0–60 min) following injection of [11C]3: baseline scan (0.96 mCi); image acquired following pretreatment (5 min prior) with 3 (1 mg/kg, 0.99 mCi). Images are dose corrected. Whole-brain time–activity curves were generated from imaging data.

Evaluation of [11C]2 in Papio anubis Baboon

Prior to non-human primate imaging, we experimentally determined that the partition coefficient (log D) between pH 7.4 phosphate buffered saline (PBS) and octanol was 1.53 ± 0.13 (n = 3). In addition, the predisposition of [11C]2 to bind plasma protein (ppb) was determined with a baboon plasma sample after 10 min of incubation time at room temperature. Relatively high (26.5 ± 0.30%) amounts of [11C]2 were free at equilibrium which was encouraging given previously reported correlations between plasma protein binding and nonspecific brain homogenate binding.[28,29]

To further investigate the brain permeability of [11C]2 and specific binding to 5HT2C receptors, we administered the labeled compound in a paired baseline/pretreatment protocol to Papio anubis baboon. As in rodents, [11C]2 exhibited very good BBB penetration and high brain uptake over the scanning time (80 min) when the [11C]2 (2.08 mCi) was administered intravenously (Figure 4A). Unlike in rodents, pretreatment with 2 (1 mg/kg, 2.33 mCi) 10 min prior to injection of [11C]2 did not alter the pharmacokinetics of radioactivity in the whole brain. Time–activity curves generated from PET imaging data provided that maximal uptake occurs at 20 min after injection. Neither a reduction in uptake nor a change in the distribution or kinetics was observed between the baseline and blocking dose curves (Figure 4B). Arterial blood samples indicated that there was no change in blood half-life or in compound metabolism. The differences in TACs for rodents and NHPs (as well as other species) are not fully understood but also not uncommonly observed. Current understanding of the expression of 5HT2C receptors in rat and non-human primate brain indicates an overall similarity, with differences noted in subregions including the neocortex, hippocampal subfields, and substantia nigra,[30,31] and range in density from ∼50 to 150 fmol/mg protein.[32,33] Little is described about 5HT2C receptors outside the brain. We noticed in rat a robust uptake of [11C]2 in regions outside the brain but within the skull (Figure 3 top, baseline, blue). This accumulation may comprise olfactory epithelium or lacrimal glands, two components modulated by serotonin. Pretreatment with unlabeled 2 decreased binding in these peripheral target sites, resulting in altered pharmacokinetics of the tracer. This was evidenced by a vertical shift in early time points of the time–activity curve, to a large extent in rat (Figure 3 top) and less so in baboon (Figure 4). Further studies are warranted to quantify species differences in non-CNS 5HT2C receptors and understand their role in 5-HT signaling within the brain. Combined, these data suggest that the signal from [11C]2 and/or its metabolites in the non-human primate brain is nonspecific and not an indicator of 5HT2C distribution and density. Nonetheless, given the favorable properties of low PPB, good metabolic stability, appropriate lipophilicity, and high BBB penetration, 2 is a good scaffold for further derivation.

Figure 4

(A) Summed PET images (0–80 min) following injection of [11C]2: (left) baseline scan (2.08 mCi); (middle) image acquired following pretreatment (10 min prior) with 2 (1 mg/kg, 2.33 mCi). Images are dose corrected. (B) Whole-brain time–activity curves generated from PET imaging data. Maximum uptake is observed at 20 min after injection. Neither a reduction in uptake nor a change in the distribution or kinetics was observed between the baseline and blocking curves.

Conclusions

We have developed a one-pot, two-step synthesis of [11C]2 and [11C]3 by using a Pictet–Spengler type cyclization with in situ generated high specific activity [11C]formaldehyde. We isolated sources of isotopic dilution and have provided a general set of reaction conditions for labeling certain benzodiazepines isotopically (both stable and unstable) at benzylic position. We accomplished the first PET imaging evaluation of two important 5HT2C receptor agonists in both rat and baboon and have provided insight into CNS pharmacokinetics. Both [11C]2 and [11C]3 exhibit rapid and high brain uptake but with high nonspecific binding (not necessarily a negative feature for use of these compounds as preclinical pharmacology tools). Our radiolabeling strategy is adaptable, and future work will be well positioned to radiolabel additional 5HT2C drugs or any tracers to investigate the serotonergic system and others in using PET imaging.