Efficient Radiosynthesis of 3′-Deoxy-3′-¹⁸F-Fluorothymidine Using Electrowetting-on-Dielectric Digital Microfluidic Chip

DISCUSSION

Since the first radiosynthesis of ¹⁸F-FLT reported by Grierson et al. (18), there have been a multitude of methods developed to achieve higher and more reliable yield (19). Notably, Eisenhut et al. developed a new FLT precursor, DMTr-Boc-nosyl, to facilitate synthesis automation, and it has become the precursor of choice to achieve high radiosynthesis yield (20). Since then, multiple research groups have reported the radiosynthesis of ¹⁸F-FLT using this DMTr-Boc-nosyl FLT precursor based on the conventional no-carrier-added radiofluorination method, with radiochemical yields ranging from 23% to 50% (19,21–24). Particularly, the use of a bulky alcohol as cosolvent in assisting the no-carrier-added radiofluorination is attractive and has been reported to attain ¹⁸F-FLT in a high radiochemical yield (60%–65%) (25–28). We adapted this high-yielding, high-selectivity protic solvent chemistry to the EWOD microdevice and developed a miniature-cartridge-purification method to eliminate the expensive and complicated HPLC purification.

During this developmental procedure, most of the processes were performed manually, such as reagent loading, product collection, and cartridge purification, which limits the amount of radioactivity that was used in this work. Although this report focused on the development of a reliable microscale radiochemistry on the EWOD chip, we are also currently developing an automated reagent delivery, product collection, and purification system (29–32). We anticipate that high-activity production within microliter volume can easily be achieved based on the recent demonstration of the production of clinical dose amounts of ¹⁸F-FDG and ¹⁸F-fallypride in batch microfluidic devices (9,33). For example, a miniaturized anion exchange cartridge was able to concentrate approximately 32 GBq of ¹⁸F-fluoride/¹⁸O-H₂O into a 5-μL volume of eluent. Such volume is commensurate with the design of EWOD chip reaction site, which can accommodate up to an approximately 17-μL droplet size.

Starting from the original report of the radiosynthesis of ¹⁸F-FLT in protic solvent (26), we first performed systematic screening of various reaction parameters including reagent concentrations, precursors, precursor-to-base ratios, and phase transfer catalysts on the microscale using Teflon-coated glass substrates. In contrast to the macroscale method, we chose to use a mixture of DMSO and thexyl alcohol to improve the solubilization of the precursor throughout the entire fluorination reaction, which is critical for synthesis automation and to increase the reaction reliability. Although macroscale syntheses generally avoid the use of DMSO (boiling point, 182°C) and thexyl alcohol (boiling point, 120°C) because of the difficulty in removing these solvents after the synthesis, the much smaller volume used in the microscale synthesis (2 vs. 500–1,000 μL) facilitates the rapid removal of these solvents. At the end of the synthesis, the droplet volume has already shrunk to approximately 0.2 μL because of solvent evaporation during the fluorination reaction (Supplemental Fig. 1). Without further evaporation, we have confirmed that the level of residual solvents of the final purified product is below the recommended limits for injection. A summary of results from the optimization studies performed on Teflon-coated glass is presented in Table 1. We found that the precursor-to-base concentration ratio is one of the critical parameters affecting the fluorination efficiency, which is consistent with the work by Suehiro et al. (34). In addition to the precursor-to-base ratio, we also investigated the use of the cryptand complex as the phase transfer catalyst. As shown in Table 1, the fluorination yield using the cryptand complex is lower (58%; Table 1, condition 2) in comparison to the TBAHCO₃ (80%; Table 1, condition 4) when performed under similar conditions (i.e., precursor-to-base ratio, ∼2), which is also consistent with the report by Kim et al. (35). To confirm the catalytic effect of the alcohol in the radiosynthesis, we also performed a control experiment by replacing the DMSO/thexyl alcohol with an aprotic solvent mixture of DMSO and MeCN. The DMSO/MeCN solvent ratio was determined empirically such that the droplet size at the end of fluorination reaction was a similar size for the 2 cases. We observed significantly lower fluorination efficiency of 38% ± 4% (n = 2) when DMSO/MeCN was used versus 80% ± 7% (n = 10) when thexyl alcohol was used in the fluorination reaction (Table 1, conditions 4 and 5), confirming the advantage of thexyl alcohol in assisting the nucleophilic substitution reaction. In our attempt of using the DMTr-nosyl FLT precursor, we found that the fluorination yield was lower and less reliable in comparison to the DMTr-Boc-nosylate FLT precursor under similar conditions (Table 1, condition 1 vs. 3). The final hydrolysis step was performed at 95°C for 5 min. The completion of the hydrolysis reaction was tentatively confirmed by the emergence of a single radio peak at around 15 min, which corresponds to ¹⁸F-FLT in the radio-HPLC (Supplemental Fig. 3). This first phase of optimized method is referred to as method 1 in Table 2.

Though method 1 exhibited high fluorination and hydrolysis efficiencies, we found that the overall crude radiochemical yield of ¹⁸F-FLT at the end of the synthesis was relatively low (47% ± 19%; n = 7). To understand the discrepancy between overall efficiency and yields of individual steps, we measured the radioactivity losses after each step of the synthesis on a Teflon-coated glass substrate. The sandwiched Teflon–glass substrate was removed from the Peltier heater, and the radioactivity on the substrate was measured using the dose calibrator after each step of the method 1 synthesis (Table 2). On the basis of this study, we found that the major loss occurred after both the fluorination and the hydrolysis steps. Although the exact mechanism of the radioactivity loss is yet to be determined, our initial approach was to reduce the reaction times for both the fluorination and the hydrolysis reaction to decrease the overall radioactivity losses, while ideally not diminishing the yields.

To do so, we first investigated the reaction kinetics on the Teflon–glass substrate and found that a reduction of the fluorination reaction time from 5 to 3 min resulted in a significant decrease in the fluorination efficiency. Through a series of experiments, we found that the fluorination reaction kinetics can be improved by increasing the reagent concentration and the reaction temperature. Interestingly, we also found that the fluorination yield was further increased (from 79% to 94%) by simply reducing the droplet size from 4 to 2 μL, while maintaining the 2-fold increase in the concentration of the reagent. The enhanced fluorination yield could be speculated to be attributable to the increase in the ¹⁸F-fluoride ion concentration as the reaction volume was decreased. Although consistent with speculations that this is an advantage of microfluidic synthesis platforms (15), further investigation of this effect is necessary.

On the basis of new optimized synthesis (method 2, Table 3), we indeed observed a reduction in radioactivity loss from 33% to only about 15% (Table 2). The new microscale ¹⁸F-FLT conditions also yielded higher fluorination efficiency (91% ± 4%; n = 10). Overall, the crude radiochemical yield of ¹⁸F-FLT increased from 49% ± 11% (Table 2) to 82% ± 10% (Table 3).

With the exception of ¹⁸F-FDG and ¹⁸F-NaF, other PET probes generally require final purification via HPLC to remove excess precursor or other radiolabeled or toxic side products that cannot be easily removed via solid-phase extraction. However, this technique is not easily miniaturized as would be needed for a benchtop radiosynthesis platform. These shortcomings of HPLC have led to the emergence of several HPLC-free radiochemistry methods (36–38).

From the point of view of the cartridge purification, the most desirable solvent to collect the product from the chip is water. However, in the microscale, where the surface-to-volume ratio is large, we found that the collection efficiency (ratio of radioactivity recovered from the chip vs. the total radioactivity that was measured on-chip after the synthesis) of the final crude ¹⁸F-FLT from the EWOD chip was below 15% when only water was used as the extraction solvent. With 50:50 v/v MeCN/H₂O as the extraction solvent, the collection efficiency improved to 47% ± 10% (n = 5). Finally, we attempted a mixture of DMSO, MeCN, and water (18, 50, and 200 μL, respectively) and obtained 84% ± 2% (n = 5) collection efficiency. However, in this solvent mixture, we found that the ¹⁸F-FLT was not able to be separated from the other side products when the conventional reversed-phase C18 resins were used in the cartridge purification. This observation can be explained by the increasing composition of nonpolar solvent during the cartridge purification, which is sufficient to disrupt the van der Waals forces between the ¹⁸F-FLT analyte and the reversed-phase resin (39). To address this issue, we investigated a new type of sorbent known as the Oasis HLB. On the basis of systematic cartridge-purification studies, we determined the optimal HLB resin to be 150 mg to efficiently retain ¹⁸F-FLT from a solvent mixture of MeCN, water, and DMSO, while enabling elution of the other side products from the cartridge. This resin was combined in a miniaturized cartridge with additional resins for removal of other impurities. On the cartridge purification of ¹⁸F-FLT, the final purified ¹⁸F-FLT product was collected in 500 μL of 100% ethanol. On evaporation and reformulation in saline, 52 MBq of ¹⁸F-FLT was obtained for small-animal PET imaging studies of several A431 tumor–bearing mice. As expected, the small-animal PET images showed a high accumulation of ¹⁸F-FLT in the tumor due to the high level of expression of the thymidine kinase-1 enzyme in rapidly proliferating cells (Supplemental Fig. 7). No adverse effect on the physiology of the mice was observed after injection.

On the basis of several ¹⁸F-FLT samples that were subjected to a standard quality control procedure, the final ¹⁸F-FLT solutions were found to have a negligible amount of impurities on formulation in approximately 0.2 mL of saline. The only radioactive component present in the final compound was ¹⁸F-FLT on successful removal of the unreacted ¹⁸F-fluoride ion using the custom-made cartridge (Fig. 3B; radio-TLC in Supplemental Fig. 6). Because of the minute amount of reagent used on the EWOD microfluidic device for radiosynthesis, the absolute amounts of impurities and residual solvent reported here and in our previous report (12) were extremely small. Additionally, the level of impurities reported here would be approximately 50 times lower on diluting the single dose of ¹⁸F-FLT from the EWOD chip in 10 mL of saline for clinical PET imaging. For clinical production, we anticipate to increase only the amount of radioactivity loaded onto the EWOD chip while keeping all other aspects of the synthesis process the same. Therefore, there will be a significant reduction in the level of impurities, bacterial endotoxins, and residual solvents that are present in a single dose of PET radiopharmaceuticals for clinical studies.

The specific activity of ¹⁸F-FLT synthesized on the EWOD chip was measured to be more than 10 times higher than literature reports using conventional macroscale radiosynthesizers. The high specific activity of ¹⁸F-FLT synthesized on the EWOD chip suggests that the Teflon layer on the chip does not lead to a significant amount of carrier fluoride contamination through radiolysis or other mechanisms. This result is consistent with the recent report by Rensch et al. (40) on the reduction of radiolysis on microfluidic devices due to the geometric confinement. The simulation and experimental studies conducted by Rensch suggested that most of the energy of positrons is deposited into the walls of microfluidic chips rather than in the reaction mixture when the dimension of the microfluidic channel or gap is smaller than the positron range (∼400 μm). The high specific activity suggests an advantage of microfluidic platforms for the production of high-specific-activity radiopharmaceuticals for imaging low-abundance receptors. We are currently investigating key contributing factors that lead to this high specific activity.