A high molar activity 18F-labeled TAK-875 derivative for PET imaging of pancreatic β-cells

Background The free-fatty acid receptor-1 (FFA-1) is expressed by β-cells and is a promising target for molecular imaging of functional β-cell mass. Recently, the ((3-[18F]fluoropropyl)sulfonyl)propoxy-derivative of the high-affinity FFA-1 agonist TAK-875 ([18F]7) was reported. Here we describe the preparation of this tracer in high molar activity using a purification method permitting separation of [18F]7 from a structurally-related by-product and evaluation of the tracer in rats as a potential FFA-1 PET imaging agent. Results The radiotracer was produced by nucleophilic radio-fluorination of the tosylate precursor and deprotection of the methyl ester. Semi-preparative HPLC with a C18 column revealed that [18F]7 co-eluted with a non-radioactive impurity. Mass spectrometry identified the impurity as the alkene-containing elimination by-product. A pentafluorophenyl-functionalized HPLC column was found to separate the two compounds and allowed for purification of [18F]7 in high molar activity. A strong anion-exchange resin was used to reformulate [18F]7 in high concentration. Starting from 96 to 311 GBq of [18F]fluoride, 3.8–15.4 GBq of pure [18F]7 (end of synthesis (EOS)) was prepared (RCY 8.3% ± 1.1% decay-corrected, n = 4) in high molar activity (166–767 GBq/μmol at EOS). This PET agent was evaluated in rats using dynamic microPET/CT imaging, ex vivo biodistribution, and radio-metabolite studies. MicroPET/CT exhibited high uptake of the tracer in the abdominal area. There was no measurable decrease of the PET signal in the pancreatic area in rats pre-treated with saturating doses (30 mg/kg) of TAK-875. Biodistribution studies corroborated the microPET/CT results. Radiometabolism analyses revealed high compound stability with only the parent molecule detected in the pancreas. Conclusions Analysis of the crude reaction mixture and identification of the elimination by-product allowed for the development of a fully automated process to prepare the TAK-875-derived PET agent [18F]7 in high purity and high molar activity. Even though the radiotracer exhibited high in vivo stability, microPET/CT and biodistribution results confirmed recent reports demonstrating that lipophilic analogs of TAK-875 display a high degree of non-specific binding, masking any specific binding to FFA-1 in pancreatic β-cells. Future development of TAK-875-derived PET tracers should focus on reducing non-specific binding in the pancreatic tissue.


Background
Type II diabetes mellitus (T2DM) is the predominant form of diabetes. The disease is characterized by insulin resistance and, after an initial compensatory over-production of insulin by β-cells, a loss of β-cell functional mass (Kahn 2003). Current understanding of β-cell dysfunction in T2DM has largely come from post-mortem autopsy data. There have been considerable efforts made towards developing a non-invasive imaging method to monitor β-cell function for both research and clinical studies (Andralojc et al. 2012). Such research would allow better understanding of β-cell mass biology including the pathological mechanisms that lead to depletion during disease progression and could be used in longitudinal studies in the pre-clinical and clinical evaluation of new therapies.
Insulin-secreting β-cells are localized in small (20 to 600 μM in diameter) clusters of endocrine cells called the islets of Langerhans, which are dispersed throughout the pancreas and account for only 1 to 2% of the total pancreatic mass (Ionescu-Tirgoviste et al., 2015). The size of this anatomical structure and its placement deep within the body restricts potential imaging modalities to those with high sensitivity, spatial resolution, and penetration depth. Given this, nuclear molecular imaging using techniques such as positron emission tomography (PET) or single photon emission computed tomography (SPECT) hold the most promise.
The free fatty acid receptor-1 (FFA-1) is a G-coupled protein receptor that is highly expressed in human and rodent β-cells (Itoh et al. 2003;Briscoe et al. 2002). FFA-1 has attracted significant attention as a potential therapeutic target to promote glucose-stimulated insulin secretion (Mancini and Poitout 2015). The FFA-1 partial agonist TAK-875 was developed as a potent (K D = 4.8 nM and 6.3 nM for human and rat FFA-1, respectively) and selective therapeutic agent that reached phase III clinical trials before development was halted due to liver toxicity (Negoro et al. 2010;Otieno et al. 2017). Structure-activity relationship and TAK-875/FFA-1 co-crystal structure studies suggest that the 3-(methylsulfonyl)propoxy-moiety of TAK-875 could be modified to allow for radionuclide incorporation without significantly impacting target binding (Negoro et al. 2012;Srivastava et al. 2014).
Bertrand and coworkers recently reported the synthesis of TAK-875-derived fluorescent probes for β-cell imaging (Bertrand et al. 2016a). These new probes were used to track specific labeling of cells overexpressing the FFA-1 receptor. Later in 2016, while we were working on the same tracer, Bertrand and coworkers published the radiosynthesis of the ((3-[ 18 F]fluoropropyl)sulfonyl)propoxy-derivative of TAK-875 ([ 18 F]7) in 16.7% ± 5.7% radiochemical yield (RCY) in low molar activity (≥0.6 GBq/μmol) (Bertrand et al. 2016b). We report here a fully automated process to produce [ 18 F]7, including optimized HPLC purification and reformulation conditions allowing for removal of the elimination by-product to obtain a high purity, high molar activity, and concentrated formulation of [ 18 F]7 for in vivo evaluation in rats.

General
All commercially available materials were used without further purification unless otherwise noted. The synthesis of 1 was completed as previously described (Negoro Dornan et al. EJNMMI Radiopharmacy and Chemistry (2018)  temperature. Spectral data is reported in parts per million using residual solvent as a reference. High resolution and accurate mass (HR/AM) measurements were performed in positive and negative ion mode by flow injection analysis into the Thermo Scientific Q-Exactive Plus Orbitrap Mass Spectrometer (San Jose, CA, USA) interfaced with a heated electrospray ion source. An automated synthesis platform (IBA Synthera®) containing two synthesis modules and one semi-preparative radio-HPLC module was used for automated radiotracer production. The production was performed with the Synthera Nucleophilic Integrated Fluidic Processor research and development cassettes (Huayi Isotopes). The cassettes were cleaned with water, ethanol, and dried with nitrogen gas before each usage. Sep-Pak C18 Plus Light cartridges (130 mg, Waters) were pre-conditioned with 10 mL of ethanol followed by 20 mL of deionized water. Sep-Pak QMA Plus Light cartridges (130 mg, Waters) were pre-conditioned by passing through 10 mL of sodium bicarbonate solution (8.4%), followed by 10 mL of deionized water, and drying with helium gas. Semi-preparative HPLC purification was carried out with a Phenomenex Luna PFP (2) column (250 × 10 mm, 5 μm, flow rate, 8 mL/minute, 50% CH 3 CN/H 2 O + 0.1% TFA). A Phenomenex Luna C18 (2) column (250 × 4.6 mm, 10 μm, flow rate, 2 mL/minute, 35% CH 3 CN/H 2 O + 0.1% TFA) was utilized for analytical HPLC with a Waters HPLC equipped with a Raytest Gabi Star radioactivity detector. Residual solvent analyses were conducted using an Agilent 7890B gas chromatograph (headspace sampler).

Identification of the impurity 8
The radiosynthesis and HPLC purification of [ 18 F]7 was conducted as described above using a semi-preparative Phenomenex Luna C18 column. The HPLC peak containing a co-elution of [ 18 F]7 and a non-radioactive impurity was collected. The decayed fraction was diluted in methanol and analyzed by mass spectrometry. HRMS

Image acquisition
Each PET/CT session consisted of a 60 min PET emission scan, followed by a 10 min CT transmission scan. PET/CT scans were conducted under anesthesia (isoflurane 2%, oxygen 0.6 L/min) delivered by a nose cone. Temperature and heart rate were monitored throughout the procedure using the Mediso system. After the animal was placed in the scanner, with the heart positioned at the center of the field of view, emission scans were initiated immediately after a bolus injection in the tail vein. List mode data was histogrammed into 29 sequential time frames of increasing duration (6 × 10 s, 4 × 30 s, 4 × 60 s, 4 × 120 s, 5 × 180 s, 6 × 300 s frames) over 60 min. Images were reconstructed using expectation maximization (EM), ordered subset expectation maximization (OSEM), normalized, and corrected for scatter, dead time, and decay.

Image analysis
Images were analyzed with Amide software (version 1.0.5 for Linux). Regions of interest (ROI) were defined on reconstructed images in the pancreas, kidney, liver and muscle to obtain time-activity curves. A 3-D sphere with a radius of 2.5 mm was drawn within the left ventricle using an average image of very early frames (10-60 s) to sample the blood input function. The sphere was centred at the point where the maximum activity was found inside this region. The ROIs for pancreas, kidney, liver and muscle were drawn using an average image of early frames (1-20 min post injection where highest tissue to background contract can be obtained). For kidney and liver, the ROIs are 3-D spheres located roughly in the middle of the tissue with a radius of 5 mm (kidney) and 6 mm (liver), respectively. For pancreas, the ROI was ellipsoid shaped and was created based on the anatomy of rats as both CT and PET images could not help with distinguishing this tissue from other organs.. The radii of the ROI were 1.5 mm (sagittal), 3 mm (transverse), and 3 mm (coronal), respectively. For muscle, the ROI was a 3-D sphere located on the muscle near left shoulder of the animals with a radius of 3 mm.

Statistical analysis
Results are presented as mean ± standard deviation. A two-tailed t-test was used to compare between two groups. P values < 0.05 were considered significant.

Biodistribution studies
Biodistribution studies were performed to evaluate tissue uptake and to measure specific and non-specific binding of the radiotracer to FFA-1. Briefly, rats were anesthetized (2-2.5% isofluorane) and injected via a lateral tail vein with 7.8-12.2 MBq tracer activity. Baseline group animals (n = 4) received tracer alone, whereas the blocking group animals (n = 4) received an intravenous dose of the FFA-1 agonist TAK-875 (30 mg/kg) 5 min prior to tracer administration. Rats were sacrificed by decapitation at 20 min post-injection. Trunk blood was collected in heparinized tubes. The following tissues were dissected and collected into pre-weighed tubes: pancreas, kidney, spleen, fat, bone, muscle, blood, plasma. Using a gamma counter (Per-kinElmer Wizard2), all tubes were counted along with a 1:100 dilution of the injected tracer solution as a standard. The percent of the injected dose per gram of tissue (%ID/g) was calculated from the decay-corrected counts and expressed as a ratio to blood.

Radiolabeled metabolite analysis General
Radiolabeled metabolite analysis studies were performed to evaluate the metabolic stability of [ 18 F] in plasma and the pancreas. Rats were anesthetized (2-2.5% isofluorane) and injected via the lateral tail vein with 35.2-73.3 MBq of [ 18 F]7. At 60-min post-injection, the animals were sacrificed by decapitation. Trunk blood was collected into heparinized tubes and the pancreas was dissected.

Sample preparation
Blood was centrifuged (4000 x g, 5 min, 4°C) to obtain plasma. Urea (1 g) was dissolved in the plasma (1 mL) and the sample was filtered (0.22 μm) and injected onto the column-switching HPLC. Pancreas samples were suspended in 80:20 EtOH/ H 2 O, homogenized using a polytron and centrifuged (14.8 rpm, 20 min, 4°C). The supernatant was collected and dried by rotary evaporation. The residue was reconstituted in 1 mL of 1% CH 3 CN/H 2 O containing 0.4 g of urea to reduce protein binding, filtered (0.22 μm), and injected onto the column-switching HPLC. Most of the radioactivity from the pancreas (and plasma) was present in the injected sample.

Column-switching HPLC
A modification of the column-switching HPLC procedure was used (Kenk et al. 2008) to analyse the radioactive metabolites from plasma and pancreas samples. Eluted solvents were analyzed via two detectors in series: a UV absorbance detector (Waters 2489) and a radiation detector (Raytest Gabi Star), which were connected to a chromatography data integration system (PeakSimple, using PeakSimple data analysis software, version 4.44). Using mobile solvent A (1% CH 3 CN/H 2 O + 0.1% TFA, 1 or 2 mL/minute), samples were loaded onto an in-line refillable capture column (hand-packed with 20 mg of Oasis HLB polymeric reverse phase sorbent) fitted with 2.5 μm frits. The elution of biological macromolecules was monitored by UV absorbance (254 nm) and radioactivity. When the UV signal returned to baseline, the solvent flow was switched to elute the capture column loaded sample onto the analytical column (Phenomenex Luna C18, 250 × 4.6 mm, 10 μm) using mobile Solvent B (65% CH 3 CN/ H 2 O + 0.1% TFA, 2 mL/minute flow rate). Retention times were calculated from the time of switch. Radioactivity data are expressed as the percentage of the total radioactivity signal. For standards, samples of blood and pancreas from control rats were spiked with [ 18 F]7 and processed as described above for each formulation tested. Use of such controls allowed for measurement of the proportion of radiotracer retained and not-retained by the capture column during loading. If radioactivity was eluted during loading of the capture column, this activity will be identified as authentic [ 18 F]7 and not as a hydrophilic labeled metabolite.

Chemistry and radiochemistry
The monoprotected diol 2 was synthesized following a literature procedure (Wilke et al. 2014). A Mitsunobu coupling reaction was used to synthesize 3 in 70% yield. Treatment of 3 with Oxone oxidized the thioether to a sulfone and deprotected the terminal alcohol protecting group to afford 4 in 93% yield (Scheme 1). Tosylation of 4 provided the radiochemical precursor 5 in 92% yield.
To produce the non-radioactive standard 7, the compound 4 was treated with Deoxo-Fluor (bis(2-methoxyethyl)aminosulfur trifluoride) followed by base-mediated ester hydrolysis, as described before (Bertrand et al. 2016b). The radiosynthesis of [ 18 F]7 was performed with an automated one-pot, two-step procedure. [ 18 F]fluoride nucleophilic displacement of the tosylate leaving group was followed by base-mediated hydrolysis of the methyl ester (Scheme 2). The quenched Scheme 1 Synthesis of the radiochemical precursor 5 and the non-radioactive standard 7 Dornan et al. EJNMMI Radiopharmacy and Chemistry (2018) 3:14 reaction mixture was purified by C18 SPE cartridge followed by semi-preparative HPLC (Fig. 1). The peak corresponding to [ 18 F]7 was collected into a vessel containing aqueous NaOH, loaded onto a weak anionic exchange resin, then eluted with acidified ethanol into isotonic bicarbonate-buffered saline. This procedure allowed for rapid reformulation in high concentration. Starting from 96 to 311 GBq of [ 18 F]fluoride, 3.8-15.4 GBq of [ 18 F]7 (at EOS) was prepared (decay-corrected RCY 8.3% ± 1.1%, n = 4) in 75-89 min. The molar activity ranged from 166 to 767 GBq/μmol at EOS.

Ex vivo biodistribution
The average % ID/g tissue-to-blood ratios for the control and blocked groups are presented in Fig. 3. In the control group, the highest average activity concentrations were found in the plasma, kidney, and pancreas (average tissue-to-blood ratios 1.60 ± 0.03, 0.98 ± 0.02, and 0.54 ± 0.04, respectively; n = 3). Lower activity concentrations were detected in the muscle, spleen, bone, and fat tissue samples (average tissue-to-blood ratios 0.30 ± 0.1, 0.27 ± 0.02, 0.19 ± 0.01, and 0.16 ± 0.01, respectively; n = 3). Blocking the FFA-1 receptor did not result in any significant difference in the tissue average activity concentrations except for the kidney, which was reduced by 20% (P < 0.05).

Radiolabeled metabolites analysis
Column-switching HPLC analysis of the control plasma and pancreas samples spiked with authentic [ 18 F]7 revealed only one peak, corresponding to unchanged tracer (R t = 5.5 min) ( Fig. 4a-b). This demonstrated that the tissue processing method did not degrade or affect the radiotracer. No other peaks were detected in the control samples. Sixty minutes following intravenous administration of [ 18 F]7 in rats (n = 4), one labeled hydrophilic metabolite was observed in rat plasma after the switch (R t = 3.1 min) on analytical HPLC, while 88% ± 4% of the total radioactivity was attributed to unchanged radiotracer (Fig. 4c). In the pancreas homogenate, only the unchanged radiotracer was detected (n = 4) (Fig. 4d).

Discussion
A novel automated method was developed to produce the ((3-[ 18 F]fluoropropyl)sulfonyl)propoxy-derivative of TAK-875 ([ 18 F]7) in high molar activity and purity, allowing for in vivo evaluation in rats as a potential FFA-1 probe for β-cell PET imaging. The tosylate precursor 5 and the non-radioactive standard 7 were synthesized starting from the aromatic alcohol 1. The thioether intermediate 3 was produced by coupling the mono-protected diol 2 with the aromatic alcohol 1 using Mitsunobu-coupling conditions. Removal of the silyl protecting group and oxidation of the thioether was achieved in a single step by treatment with Oxone to give 4. The radiochemical precursor 5 was generated by tosylation of 4 as described previously (Bertrand et al. 2016b). The non-radioactive standard was synthesized by fluorination of 4 using Deoxo-Fluor, followed by base-mediated ester hydrolysis (Bertrand et al. 2016b). The radiosynthesis of [ 18 F]7 was completed with a nucleophilic radio-fluorination reaction followed by base-mediated ester hydrolysis in a one-pot, two-step process. Semi-preparative HPLC purification of [ 18 F]7 was initially explored using a C18 column. Co-elution of the radiotracer with a non-radioactive by-product was observed. Despite efforts, chromatographic resolution of the tracer from this by-product was not achieved by adjusting the mobile phase composition. High resolution mass spectrometry indicated that the identity of the impurity was the alkene-containing elimination by-product 8 (Fig. 1). Resolution of [ 18 F]7 from 8 was achieved employing a pentafluorophenyl-functionalized reverse phase column which enabled preparation of [ 18 F]7 in high purity. Inspection of previously described structure-activity relationships and the TAK-875/FFA-1 co-crystal structure predicted that 8 would potentially compete with [ 18 F]7 for binding to FFA-1 (Negoro et al. 2012;Srivastava et al. 2014). Thus, removal of 8 will increase the apparent molar activity of the final product and improve the potential for quantitative PET imaging of pancreatic FFA-1 and β-cell mass.
Reformulation of [ 18 F]7 was performed using an anion exchange resin. The purified [ 18 F]7 HPLC peak was collected into a vessel containing aqueous base to ensure ionization of the carboxylic acid functional group. From this solution, the carboxylate form of [ 18 F]7 was efficiently captured by the resin. The residual HPLC solvent was removed by washing with water. The radiotracer was then eluted from the resin with acidified ethanol, which was filtered and diluted into an isotonic bicarbonate and saline solution. This efficiently buffered the pH of the final injectable solution and provided [ 18 F]7 as a concentrated, injectable solution for animal PET studies. Dynamic microPET/CT imaging (with or without a saturating dose of TAK-875) was conducted in rats to evaluate tracer kinetics and the potential of [ 18 F]7 as an FFA-1-targeting agent. The images displayed high uptake of the tracer in abdominal organs. While tracer uptake could be detected in the pancreas ROI, there was no reduction in tracer retention in the TAK-875 pre-treatment group. This suggests that the signal from the pancreas is confounded by non-specific binding in the pancreatic tissue. To further understand the in vivo value of this tracer and confirm these results, the focus of investigation was shifted to the analysis of tissue samples ex vivo.
Ex vivo biodistribution studies have the added benefit of eliminating error introduced when drawing the microPET/CT pancreatic ROI, which is indistinguishable from surrounding organs and tissues in rodents by CT and could only be estimated based on known anatomy (Yin et al. 2015). Theses studies indicated very little uptake of the radiotracer in the fat, bone, and muscle tissues. [ 18 F]7 was detected in the pancreas (0.54 ± 0.04 tissue-to-blood ratio). However, saturating FFA-1 by pre-treatment with TAK-875 did not lead to a decrease in tracer accumulation in the pancreas, which corroborated the microPET results. This led to an examination of the formation and presence of potential radiolabeled metabolites to characterize the radioactive signal that accumulated in the pancreas. Interestingly, HPLC radio-metabolite analysis exhibited only the presence of unchanged [ 18 F]7 in the pancreas at 60-min post-injection. The formation of one minor metabolite was also detected in the plasma. Together with the absence of radioactivity accumulation in the bone in the biodistribution studies, this result confirmed the relatively high metabolic stability of [ 18 F]7 in rats.
During the course of this work, a study found that [ 3 H]TAK-875 exhibits a high level of off-target binding, likely related to the high degree of lipophilicity of the molecule (cLogP = 4.2) (Hellström-Lindahl et al. 2017). Given the similarity of [ 18 F]7 with TAK-875 in terms of chemical structure and lipophilicity (cLogP = 4.58), and with the results described here, it can be concluded that non-specific binding of [ 18 F]7 in the pancreatic tissue invalidates this tracer as an in vivo or ex vivo quantitative FFA-1 imaging agent. Future development of TAK-875-derived FFA-1 targeting radiotracers should be approached with a focus on decreasing off-target labeling in the pancreas and surrounding tissue by increasing the hydrophilicity.

Conclusions
A one-pot, two-step fully automated process for the preparation of [ 18 F]7 was developed. The use of a pentafluorophenyl-functionalized HPLC column in this process allowed for the removal of the potential FFA-1 competitive impurity 8 and improved the purity and apparent molar activity of the final product. A strong anion-exchange resin was used to conduct rapid and efficient reformulation of the radiotracer in high concentration, enabling in vivo evaluation of [ 18 F]7 as a quantitative PET agent. MicroPET/CT imaging and biodistribution results exhibited activity accumulation in the pancreas although no specific binding could be detected following saturation studies of FFA-1 receptor. Radio-metabolite analysis indicated that [ 18 F]7 is sufficiently stable, ruling out the possibility that degradation was a contributing factor to the high degree of non-specific binding observed. While these results allow for the conclusion that non-specific binding of [ 18 F]7 in the pancreas impedes the ability to measure specific binding, they are instructive and provide direction for the development of future PET tracers for FFA-1-based β-cell PET imaging.